Update CHANGELOG for v0.9.2 release

Change-Id: I184e927987544e9f34f890249b589ea13a93a330

.mailmap

@@ -0,0 +1,2 @@
+Adrian Grange <agrange@google.com>
+Johann Koenig <johannkoenig@google.com>

AUTHORS

@@ -1,16 +1,26 @@
 # This file is automatically generated from the git commit history
 # by tools/gen_authors.sh.
 
+Adrian Grange <agrange@google.com>
 Alex Converse <alex.converse@gmail.com>
 Andres Mejia <mcitadel@gmail.com>
 Fabio Pedretti <fabio.ped@libero.it>
 Frank Galligan <fgalligan@google.com>
+Fredrik Söderquist <fs@opera.com>
+Fritz Koenig <frkoenig@google.com>
+Giuseppe Scrivano <gscrivano@gnu.org>
 Guillermo Ballester Valor <gbvalor@gmail.com>
 James Zern <jzern@google.com>
+Jan Kratochvil <jan.kratochvil@redhat.com>
+Jeff Muizelaar <jmuizelaar@mozilla.com>
+Jim Bankoski <jimbankoski@google.com>
+Johann Koenig <johannkoenig@google.com>
 John Koleszar <jkoleszar@google.com>
 Justin Clift <justin@salasaga.org>
+Justin Lebar <justin.lebar@gmail.com>
 Luca Barbato <lu_zero@gentoo.org>
 Makoto Kato <makoto.kt@gmail.com>
+Michael Kohler <michaelkohler@live.com>
 Paul Wilkins <paulwilkins@google.com>
 Pavol Rusnak <stick@gk2.sk>
 Philip Jägenstedt <philipj@opera.com>

CHANGELOG

@@ -1,3 +1,53 @@
+2010-09-02 v0.9.2
+  - Enhancements:
+      Disable frame dropping by default
+      Improved multithreaded performance
+      Improved Force Key Frame Behaviour
+      Increased rate control buffer level precision
+      Fix bug in 1st pass motion compensation
+      ivfenc: correct fixed kf interval, --disable-kf
+  - Speed:
+      Changed above and left context data layout
+      Rework idct calling structure.
+      Removed unnecessary MB_MODE_INFO copies
+      x86: SSSE3 sixtap prediction
+      Reworked IDCT to include reconstruction (add) step
+      Swap alt/gold/new/last frame buffer ptrs instead of copying.
+      Improve SSE2 loopfilter functions
+      Change bitreader to use a larger window.
+      Avoid loopfilter reinitialization when possible
+  - Quality:
+      Normalize quantizer's zero bin and rounding factors
+      Add trellis quantization.
+      Make the quantizer exact.
+      Updates to ARNR filtering algorithm
+      Fix breakout thresh computation for golden & AltRef frames
+      Redo the forward 4x4 dct
+      Improve the accuracy of forward walsh-hadamard transform
+      Further adjustment of RD behaviour with Q and Zbin.
+  - Build System:
+      Allow linking of libs built with MinGW to MSVC
+      Fix target auto-detection on mingw32
+      Allow --cpu= to work for x86.
+      configure: pass original arguments through to make dist
+      Fix builds without runtime CPU detection
+      msvs: fix install of codec sources
+      msvs: Change devenv.com command line for better msys support
+      msvs: Add vs9 targets.
+      Add x86_64-linux-icc target
+  - Bugs:
+      Potential crashes on older MinGW builds
+      Fix two-pass framrate for Y4M input.
+      Fixed simple loop filter, other crashes on ARM v6
+      arm: fix missing dependency with --enable-shared
+      configure: support directories containing .o
+      Replace pinsrw (SSE) with MMX instructions
+      apple: include proper mach primatives
+      Fixed rate control bug with long key frame interval.
+      Fix DSO link errors on x86-64 when not using a version script
+      Fixed buffer selection for UV in AltRef filtering
+
+
 2010-06-17 v0.9.1
   - Enhancements:
       * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O

build/arm-wince-vs8/vpx.sln

@@ -8,7 +8,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "example", "example.vcproj",
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "obj_int_extract", "obj_int_extract.vcproj", "{E1360C65-D375-4335-8057-7ED99CC3F9B2}"
 EndProject
-Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "vpx_decoder", "vpx_decoder.vcproj", "{DCE19DAF-69AC-46DB-B14A-39F0FAA5DB74}"
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "vpx", "vpx.vcproj", "{DCE19DAF-69AC-46DB-B14A-39F0FAA5DB74}"
 	ProjectSection(ProjectDependencies) = postProject
 		{E1360C65-D375-4335-8057-7ED99CC3F9B2} = {E1360C65-D375-4335-8057-7ED99CC3F9B2}
 	EndProjectSection

build/make/Makefile

@@ -39,13 +39,8 @@ dist:
 	@if [ -d "$(DIST_DIR)/src" ]; then \
             mkdir -p "$(DIST_DIR)/build"; \
             cd "$(DIST_DIR)/build"; \
-            if [ "$(TGT_CC)" = "rvct" ] ; then \
+            echo "Rerunning configure $(CONFIGURE_ARGS)"; \
-				echo "../src/configure --target=$(TOOLCHAIN) --libc=$(ALT_LIBC)"; \
+            ../src/configure $(CONFIGURE_ARGS); \
-				../src/configure --target=$(TOOLCHAIN) --libc=$(ALT_LIBC); \
-			else \
-				echo "../src/configure --target=$(TOOLCHAIN)"; \
-				../src/configure --target=$(TOOLCHAIN); \
-			fi; \
             $(if $(filter vs%,$(TGT_CC)),make NO_LAUNCH_DEVENV=1;) \
         fi
 	@if [ -d "$(DIST_DIR)" ]; then \
@@ -334,6 +329,7 @@ ifneq ($(call enabled,DIST-SRCS),)
     DIST-SRCS-$(CONFIG_MSVS)  += build/make/gen_msvs_def.sh
     DIST-SRCS-$(CONFIG_MSVS)  += build/make/gen_msvs_proj.sh
     DIST-SRCS-$(CONFIG_MSVS)  += build/make/gen_msvs_sln.sh
+    DIST-SRCS-$(CONFIG_MSVS)  += build/x86-msvs/yasm.rules
     DIST-SRCS-$(CONFIG_RVCT) += build/make/armlink_adapter.sh
     #
     # This isn't really ARCH_ARM dependent, it's dependant on whether we're

build/make/configure.sh

@@ -378,7 +378,7 @@ EOF
 fmt_deps = sed -e 's;^__image.axf;\$(dir \$@)\$(notdir \$<).o \$@;' #hide
 EOF
     else cat >> $1 << EOF
-fmt_deps = sed -e 's;^\(.*\)\.o;\$(dir \$@)\1\$(suffix \$<).o \$@;' #hide
+fmt_deps = sed -e 's;^\([a-zA-Z0-9_]*\)\.o;\$(dir \$@)\1\$(suffix \$<).o \$@;'
 EOF
     fi
 
@@ -395,8 +395,6 @@ EOF
 write_common_target_config_h() {
     cat > ${TMP_H} << EOF
 /* This file automatically generated by configure. Do not edit! */
-#define INLINE      ${INLINE}
-#define FORCEINLINE ${FORCEINLINE:-${INLINE}}
 #define RESTRICT    ${RESTRICT}
 EOF
     print_config_h ARCH   "${TMP_H}" ${ARCH_LIST}
@@ -520,6 +518,7 @@ process_common_toolchain() {
                 tgt_os=darwin9
                 ;;
             *mingw32*|*cygwin*)
+                [ -z "$tgt_isa" ] && tgt_isa=x86
                 tgt_os=win32
                 ;;
             *linux*|*bsd*)
@@ -783,6 +782,9 @@ process_common_toolchain() {
         soft_enable ssse3
 
         case  ${tgt_os} in
+            win*)
+                enabled gcc && add_cflags -fno-common
+                ;;
             solaris*)
                 CC=${CC:-${CROSS}gcc}
                 LD=${LD:-${CROSS}gcc}
@@ -799,11 +801,21 @@ process_common_toolchain() {
                 add_ldflags -i-static
                 enabled x86_64 && add_cflags -ipo -no-prec-div -static -xSSE3 -axSSE3
                 enabled x86_64 && AR=xiar
+                case ${tune_cpu} in
+                    atom*)
+                        tune_cflags="-x"
+                        tune_cpu="SSE3_ATOM"
+                    ;;
+                    *)
+                        tune_cflags="-march="
+                    ;;
+                esac
                 ;;
             gcc*)
                 add_cflags  -m${bits}
                 add_ldflags -m${bits}
                 link_with_cc=gcc
+                tune_cflags="-march="
             setup_gnu_toolchain
                 ;;
         esac

build/make/gen_msvs_proj.sh

@@ -206,7 +206,7 @@ for opt in "$@"; do
     ;;
     --ver=*) vs_ver="$optval"
              case $optval in
-             [78])
+             [789])
              ;;
              *) die Unrecognized Visual Studio Version in $opt
              ;;
@@ -248,6 +248,8 @@ case "${vs_ver:-8}" in
     ;;
     8) vs_ver_id="8.00"
     ;;
+    9) vs_ver_id="9.00"
+    ;;
 esac
 
 [ -n "$name" ] || die "Project name (--name) must be specified!"
@@ -347,7 +349,7 @@ generate_vcproj() {
         x86*) $uses_asm && tag ToolFile RelativePath="$self_dirname/../x86-msvs/yasm.rules"
         ;;
         arm*|iwmmx*)
-            if [ "$name" == "vpx_decoder" ];then
+            if [ "$name" == "vpx" ];then
             case "$target" in
                 armv5*)
                     tag ToolFile RelativePath="$self_dirname/../arm-wince-vs8/armasmv5.rules"
@@ -376,7 +378,7 @@ generate_vcproj() {
 
         if [ "$target" == "armv6-wince-vs8" ] || [ "$target" == "armv5te-wince-vs8" ] || [ "$target" == "iwmmxt-wince-vs8" ] || [ "$target" == "iwmmxt2-wince-vs8" ];then
             case "$name" in
-                vpx_decoder) tag Tool \
+                vpx)         tag Tool \
                              Name="VCPreBuildEventTool" \
                              CommandLine="call obj_int_extract.bat \$(ConfigurationName)"
                              tag Tool \
@@ -437,7 +439,7 @@ generate_vcproj() {
                 Name="VCCLCompilerTool" \
                 Optimization="0" \
                 AdditionalIncludeDirectories="$incs" \
-                PreprocessorDefinitions="WIN32;_DEBUG;_CRT_SECURE_NO_WARNINGS;$defines" \
+                PreprocessorDefinitions="WIN32;_DEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
                 RuntimeLibrary="$debug_runtime" \
                 UsePrecompiledHeader="0" \
                 WarningLevel="3" \
@@ -510,7 +512,7 @@ generate_vcproj() {
 
         if [ "$target" == "armv6-wince-vs8" ] || [ "$target" == "armv5te-wince-vs8" ] || [ "$target" == "iwmmxt-wince-vs8" ] || [ "$target" == "iwmmxt2-wince-vs8" ];then
             case "$name" in
-                vpx_decoder) tag DeploymentTool \
+                vpx)         tag DeploymentTool \
                              ForceDirty="-1" \
                              RegisterOutput="0"
                                 ;;
@@ -534,7 +536,7 @@ generate_vcproj() {
 
         if [ "$target" == "armv6-wince-vs8" ] || [ "$target" == "armv5te-wince-vs8" ] || [ "$target" == "iwmmxt-wince-vs8" ] || [ "$target" == "iwmmxt2-wince-vs8" ];then
             case "$name" in
-                vpx_decoder) tag Tool \
+                vpx)         tag Tool \
                                      Name="VCPreBuildEventTool" \
                                      CommandLine="call obj_int_extract.bat \$(ConfigurationName)"
                              tag Tool \
@@ -595,7 +597,7 @@ generate_vcproj() {
         x86*) tag       Tool \
                       Name="VCCLCompilerTool" \
                       AdditionalIncludeDirectories="$incs" \
-                      PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;$defines" \
+                      PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
                       RuntimeLibrary="$release_runtime" \
                       UsePrecompiledHeader="0" \
                       WarningLevel="3" \
@@ -672,7 +674,7 @@ generate_vcproj() {
 
         if [ "$target" == "armv6-wince-vs8" ] || [ "$target" == "armv5te-wince-vs8" ] || [ "$target" == "iwmmxt-wince-vs8" ] || [ "$target" == "iwmmxt2-wince-vs8" ];then
             case "$name" in
-                vpx_decoder) tag DeploymentTool \
+                vpx)         tag DeploymentTool \
                              ForceDirty="-1" \
                              RegisterOutput="0"
                 ;;

build/make/gen_msvs_sln.sh

@@ -25,7 +25,7 @@ files.
 Options:
     --help                      Print this message
     --out=outfile               Redirect output to a file
-    --ver=version               Version (7,8) of visual studio to generate for
+    --ver=version               Version (7,8,9) of visual studio to generate for
     --target=isa-os-cc          Target specifier
 EOF
     exit 1
@@ -193,11 +193,11 @@ ${TAB}rm -rf "$platform"/"$config"
 ifneq (\$(found_devenv),)
   ifeq (\$(CONFIG_VS_VERSION),7)
 $nows_sln_config: $outfile
-${TAB}devenv.com $outfile /build "$config"
+${TAB}devenv.com $outfile -build "$config"
 
   else
 $nows_sln_config: $outfile
-${TAB}devenv.com $outfile /build "$sln_config"
+${TAB}devenv.com $outfile -build "$sln_config"
 
   endif
 else
@@ -224,7 +224,7 @@ for opt in "$@"; do
     ;;
     --ver=*) vs_ver="$optval"
              case $optval in
-             [78])
+             [789])
              ;;
              *) die Unrecognized Visual Studio Version in $opt
              ;;
@@ -235,7 +235,7 @@ for opt in "$@"; do
              7) sln_vers="8.00"
                 sln_vers_str="Visual Studio .NET 2003"
              ;;
-             8)
+             [89])
              ;;
              *) die "Unrecognized Visual Studio Version '$optval' in $opt"
              ;;
@@ -257,6 +257,9 @@ case "${vs_ver:-8}" in
     8) sln_vers="9.00"
        sln_vers_str="Visual Studio 2005"
     ;;
+    9) sln_vers="10.00"
+       sln_vers_str="Visual Studio 2008"
+    ;;
 esac
 
 for f in "${file_list[@]}"; do

configure

@@ -38,6 +38,7 @@ Advanced options:
   ${toggle_realtime_only}         enable this option while building for real-time encoding
   ${toggle_runtime_cpu_detect}    runtime cpu detection
   ${toggle_shared}                shared library support
+  ${toggle_arm_asm_detok}         assembly version of the detokenizer (ARM platforms only)
 
 Codecs:
   Codecs can be selectively enabled or disabled individually, or by family:
@@ -107,11 +108,13 @@ all_platforms="${all_platforms} x86-solaris-gcc"
 all_platforms="${all_platforms} x86-win32-gcc"
 all_platforms="${all_platforms} x86-win32-vs7"
 all_platforms="${all_platforms} x86-win32-vs8"
+all_platforms="${all_platforms} x86-win32-vs9"
 all_platforms="${all_platforms} x86_64-darwin9-gcc"
 all_platforms="${all_platforms} x86_64-linux-gcc"
 all_platforms="${all_platforms} x86_64-linux-icc"
 all_platforms="${all_platforms} x86_64-solaris-gcc"
 all_platforms="${all_platforms} x86_64-win64-vs8"
+all_platforms="${all_platforms} x86_64-win64-vs9"
 all_platforms="${all_platforms} universal-darwin8-gcc"
 all_platforms="${all_platforms} universal-darwin9-gcc"
 all_platforms="${all_platforms} generic-gnu"
@@ -228,10 +231,8 @@ CONFIG_LIST="
 
     dequant_tokens
     dc_recon
-    new_tokens
     runtime_cpu_detect
     postproc
-    postproc_generic
     multithread
     psnr
     ${CODECS}
@@ -242,7 +243,7 @@ CONFIG_LIST="
     spatial_resampling
     realtime_only
     shared
-    dixie
+    arm_asm_detok
 "
 CMDLINE_SELECT="
     extra_warnings
@@ -269,9 +270,7 @@ CMDLINE_SELECT="
 
     dequant_tokens
     dc_recon
-    new_tokens
     postproc
-    postproc_generic
     multithread
     psnr
     ${CODECS}
@@ -281,7 +280,7 @@ CMDLINE_SELECT="
     spatial_resampling
     realtime_only
     shared
-    dixie
+    arm_asm_detok
 "
 
 process_cmdline() {
@@ -389,6 +388,7 @@ VERSION_MAJOR=${VERSION_MAJOR}
 VERSION_MINOR=${VERSION_MINOR}
 VERSION_PATCH=${VERSION_PATCH}
 
+CONFIGURE_ARGS=${CONFIGURE_ARGS}
 EOF
     enabled child || echo "CONFIGURE_ARGS?=${CONFIGURE_ARGS}" >> config.mk
 
@@ -523,8 +523,6 @@ process_toolchain() {
              enable solution
              vs_version=${tgt_cc##vs}
              all_targets="${all_targets} solution"
-             INLINE=__inline
-             FORCEINLINE=__forceinline
         ;;
     esac
 

ivfdec.c

@@ -29,10 +29,6 @@
 
 static const char *exec_name;
 
-#if CONFIG_DIXIE
-extern vpx_codec_iface_t vpx_codec_vp8_dixie_algo;
-#endif
-
 static const struct
 {
     char const *name;
@@ -44,9 +40,6 @@ static const struct
 #if CONFIG_VP8_DECODER
     {"vp8",  &vpx_codec_vp8_dx_algo,   0x00385056, 0x00FFFFFF},
 #endif
-#if CONFIG_DIXIE
-    {"dixie", &vpx_codec_vp8_dixie_algo, 0x00385056, 0x00FFFFFF},
-#endif
 };
 
 #include "args.h"

ivfenc.c

@@ -505,9 +505,11 @@ static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
                                      "Minimum keyframe interval (frames)");
 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
                                      "Maximum keyframe interval (frames)");
+static const arg_def_t kf_disabled = ARG_DEF(NULL, "disable-kf", 0,
+                                     "Disable keyframe placement");
 static const arg_def_t *kf_args[] =
 {
-    &kf_min_dist, &kf_max_dist, NULL
+    &kf_min_dist, &kf_max_dist, &kf_disabled, NULL
 };
 
 
@@ -800,6 +802,8 @@ int main(int argc, const char **argv_)
             cfg.kf_min_dist = arg_parse_uint(&arg);
         else if (arg_match(&arg, &kf_max_dist, argi))
             cfg.kf_max_dist = arg_parse_uint(&arg);
+        else if (arg_match(&arg, &kf_disabled, argi))
+            cfg.kf_mode = VPX_KF_DISABLED;
         else
             argj++;
     }
@@ -886,6 +890,8 @@ int main(int argc, const char **argv_)
                 {
                     cfg.g_timebase.num = y4m.fps_d;
                     cfg.g_timebase.den = y4m.fps_n;
+                    /* And don't reset it in the second pass.*/
+                    arg_have_timebase = 1;
                 }
                 arg_use_i420 = 0;
             }
@@ -1016,9 +1022,6 @@ int main(int argc, const char **argv_)
 
 
         /* Construct Encoder Context */
-        if (cfg.kf_min_dist == cfg.kf_max_dist)
-            cfg.kf_mode = VPX_KF_FIXED;
-
         vpx_codec_enc_init(&encoder, codec->iface, &cfg,
                            show_psnr ? VPX_CODEC_USE_PSNR : 0);
         ctx_exit_on_error(&encoder, "Failed to initialize encoder");

libs.mk

@@ -11,6 +11,8 @@
 
 ASM:=$(if $(filter yes,$(CONFIG_GCC)),.asm.s,.asm)
 
+CODEC_SRCS-yes += libs.mk
+
 include $(SRC_PATH_BARE)/vpx/vpx_codec.mk
 CODEC_SRCS-yes += $(addprefix vpx/,$(call enabled,API_SRCS))
 
@@ -59,7 +61,6 @@ CODEC_LIB=$(if $(CONFIG_STATIC_MSVCRT),vpxmt,vpxmd)
 # This variable uses deferred expansion intentionally, since the results of
 # $(wildcard) may change during the course of the Make.
 VS_PLATFORMS = $(foreach d,$(wildcard */Release/$(CODEC_LIB).lib),$(word 1,$(subst /, ,$(d))))
-CODEC_SRCS-yes += $(SRC_PATH_BARE)/libs.mk # to show up in the msvs workspace
 endif
 
 # The following pairs define a mapping of locations in the distribution
@@ -144,7 +145,6 @@ obj_int_extract.vcproj: $(SRC_PATH_BARE)/build/make/obj_int_extract.c
 
 PROJECTS-$(BUILD_LIBVPX) += obj_int_extract.vcproj
 PROJECTS-$(BUILD_LIBVPX) += obj_int_extract.bat
-PROJECTS-$(BUILD_LIBVPX) += armasm$(ARM_ARCH).rules
 endif
 
 vpx.def: $(call enabled,CODEC_EXPORTS)

vp8/common/alloccommon.c

@@ -24,43 +24,36 @@ extern  void vp8_init_scan_order_mask();
 void vp8_update_mode_info_border(MODE_INFO *mi, int rows, int cols)
 {
     int i;
-    vpx_memset(mi - cols - 1, 0, sizeof(MODE_INFO) * cols + 1);
+    vpx_memset(mi - cols - 2, 0, sizeof(MODE_INFO) * (cols + 1));
 
     for (i = 0; i < rows; i++)
     {
         vpx_memset(&mi[i*cols-1], 0, sizeof(MODE_INFO));
     }
 }
+
 void vp8_de_alloc_frame_buffers(VP8_COMMON *oci)
 {
+    int i;
+
+    for (i = 0; i < NUM_YV12_BUFFERS; i++)
+        vp8_yv12_de_alloc_frame_buffer(&oci->yv12_fb[i]);
+
     vp8_yv12_de_alloc_frame_buffer(&oci->temp_scale_frame);
-    vp8_yv12_de_alloc_frame_buffer(&oci->new_frame);
-    vp8_yv12_de_alloc_frame_buffer(&oci->last_frame);
-    vp8_yv12_de_alloc_frame_buffer(&oci->golden_frame);
-    vp8_yv12_de_alloc_frame_buffer(&oci->alt_ref_frame);
     vp8_yv12_de_alloc_frame_buffer(&oci->post_proc_buffer);
 
-    vpx_free(oci->above_context[Y1CONTEXT]);
+    vpx_free(oci->above_context);
-    vpx_free(oci->above_context[UCONTEXT]);
-    vpx_free(oci->above_context[VCONTEXT]);
-    vpx_free(oci->above_context[Y2CONTEXT]);
     vpx_free(oci->mip);
 
-    oci->above_context[Y1CONTEXT] = 0;
+    oci->above_context = 0;
-    oci->above_context[UCONTEXT]  = 0;
-    oci->above_context[VCONTEXT]  = 0;
-    oci->above_context[Y2CONTEXT] = 0;
     oci->mip = 0;
 
-    // Structure used to minitor GF useage
-    if (oci->gf_active_flags != 0)
-        vpx_free(oci->gf_active_flags);
-
-    oci->gf_active_flags = 0;
 }
 
 int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
 {
+    int i;
+
     vp8_de_alloc_frame_buffers(oci);
 
     // our internal buffers are always multiples of 16
@@ -71,37 +64,33 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
         height += 16 - (height & 0xf);
 
 
+    for (i = 0; i < NUM_YV12_BUFFERS; i++)
+    {
+      oci->fb_idx_ref_cnt[0] = 0;
+
+      if (vp8_yv12_alloc_frame_buffer(&oci->yv12_fb[i],  width, height, VP8BORDERINPIXELS) < 0)
+        {
+            vp8_de_alloc_frame_buffers(oci);
+            return ALLOC_FAILURE;
+        }
+    }
+
+    oci->new_fb_idx = 0;
+    oci->lst_fb_idx = 1;
+    oci->gld_fb_idx = 2;
+    oci->alt_fb_idx = 3;
+
+    oci->fb_idx_ref_cnt[0] = 1;
+    oci->fb_idx_ref_cnt[1] = 1;
+    oci->fb_idx_ref_cnt[2] = 1;
+    oci->fb_idx_ref_cnt[3] = 1;
+
     if (vp8_yv12_alloc_frame_buffer(&oci->temp_scale_frame,   width, 16, VP8BORDERINPIXELS) < 0)
     {
         vp8_de_alloc_frame_buffers(oci);
         return ALLOC_FAILURE;
     }
 
-
-    if (vp8_yv12_alloc_frame_buffer(&oci->new_frame,   width, height, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    if (vp8_yv12_alloc_frame_buffer(&oci->last_frame,  width, height, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    if (vp8_yv12_alloc_frame_buffer(&oci->golden_frame, width, height, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    if (vp8_yv12_alloc_frame_buffer(&oci->alt_ref_frame, width, height, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
     if (vp8_yv12_alloc_frame_buffer(&oci->post_proc_buffer, width, height, VP8BORDERINPIXELS) < 0)
     {
         vp8_de_alloc_frame_buffers(oci);
@@ -123,33 +112,9 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
     oci->mi = oci->mip + oci->mode_info_stride + 1;
 
 
-    oci->above_context[Y1CONTEXT] = vpx_calloc(sizeof(ENTROPY_CONTEXT) * oci->mb_cols * 4 , 1);
+    oci->above_context = vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * oci->mb_cols, 1);
 
-    if (!oci->above_context[Y1CONTEXT])
+    if (!oci->above_context)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    oci->above_context[UCONTEXT]  = vpx_calloc(sizeof(ENTROPY_CONTEXT) * oci->mb_cols * 2 , 1);
-
-    if (!oci->above_context[UCONTEXT])
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    oci->above_context[VCONTEXT]  = vpx_calloc(sizeof(ENTROPY_CONTEXT) * oci->mb_cols * 2 , 1);
-
-    if (!oci->above_context[VCONTEXT])
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    oci->above_context[Y2CONTEXT] = vpx_calloc(sizeof(ENTROPY_CONTEXT) * oci->mb_cols     , 1);
-
-    if (!oci->above_context[Y2CONTEXT])
     {
         vp8_de_alloc_frame_buffers(oci);
         return ALLOC_FAILURE;
@@ -157,20 +122,6 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
 
     vp8_update_mode_info_border(oci->mi, oci->mb_rows, oci->mb_cols);
 
-    // Structures used to minitor GF usage
-    if (oci->gf_active_flags != 0)
-        vpx_free(oci->gf_active_flags);
-
-    oci->gf_active_flags = (unsigned char *)vpx_calloc(oci->mb_rows * oci->mb_cols, 1);
-
-    if (!oci->gf_active_flags)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return ALLOC_FAILURE;
-    }
-
-    oci->gf_active_count = oci->mb_rows * oci->mb_cols;
-
     return 0;
 }
 void vp8_setup_version(VP8_COMMON *cm)

vp8/common/arm/armv6/dc_only_idct_add_v6.asm

@@ -0,0 +1,67 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license and patent
+;  grant that can be found in the LICENSE file in the root of the source
+;  tree. All contributing project authors may be found in the AUTHORS
+;  file in the root of the source tree.
+;
+
+    EXPORT  |vp8_dc_only_idct_add_v6|
+
+    AREA    |.text|, CODE, READONLY
+
+;void vp8_dc_only_idct_add_v6(short input_dc, unsigned char *pred_ptr,
+;                             unsigned char *dst_ptr, int pitch, int stride)
+; r0  input_dc
+; r1  pred_ptr
+; r2  dest_ptr
+; r3  pitch
+; sp  stride
+
+|vp8_dc_only_idct_add_v6| PROC
+    stmdb       sp!, {r4 - r7, lr}
+
+    add         r0, r0, #4                ; input_dc += 4
+    ldr         r12, c0x0000FFFF
+    ldr         r4, [r1], r3
+    ldr         r6, [r1], r3
+    and         r0, r12, r0, asr #3       ; input_dc >> 3 + mask
+    ldr         lr, [sp, #20]
+    orr         r0, r0, r0, lsl #16       ; a1 | a1
+
+    uxtab16     r5, r0, r4                ; a1+2 | a1+0
+    uxtab16     r4, r0, r4, ror #8        ; a1+3 | a1+1
+    uxtab16     r7, r0, r6
+    uxtab16     r6, r0, r6, ror #8
+    usat16      r5, #8, r5
+    usat16      r4, #8, r4
+    usat16      r7, #8, r7
+    usat16      r6, #8, r6
+    orr         r5, r5, r4, lsl #8
+    orr         r7, r7, r6, lsl #8
+    ldr         r4, [r1], r3
+    ldr         r6, [r1]
+    str         r5, [r2], lr
+    str         r7, [r2], lr
+
+    uxtab16     r5, r0, r4
+    uxtab16     r4, r0, r4, ror #8
+    uxtab16     r7, r0, r6
+    uxtab16     r6, r0, r6, ror #8
+    usat16      r5, #8, r5
+    usat16      r4, #8, r4
+    usat16      r7, #8, r7
+    usat16      r6, #8, r6
+    orr         r5, r5, r4, lsl #8
+    orr         r7, r7, r6, lsl #8
+    str         r5, [r2], lr
+    str         r7, [r2]
+
+    ldmia       sp!, {r4 - r7, pc}
+
+    ENDP  ; |vp8_dc_only_idct_add_v6|
+
+; Constant Pool
+c0x0000FFFF DCD 0x0000FFFF
+    END

vp8/common/arm/armv6/idct_v6.asm

@@ -15,8 +15,6 @@
     EXPORT  |vp8_short_idct4x4llm_v6_scott|
     EXPORT  |vp8_short_idct4x4llm_v6_dual|
 
-    EXPORT  |vp8_dc_only_idct_armv6|
-
     AREA    |.text|, CODE, READONLY
 
 ;********************************************************************************
@@ -344,34 +342,4 @@ loop2_dual
     ldmia   sp!, {r4 - r11, pc} ; replace vars, return                      restore
     ENDP
 
-
-; sjl added 10/17/08
-;void dc_only_idct_armv6(short input_dc, short *output, int pitch)
-|vp8_dc_only_idct_armv6| PROC
-    stmdb       sp!, {r4 - r6, lr}
-
-    add         r0, r0, #0x4
-    add         r4, r1, r2                      ; output + shortpitch
-    mov         r0, r0, ASR #0x3    ;aka a1
-    add         r5, r1, r2, LSL #1              ; output + shortpitch * 2
-    pkhbt       r0, r0, r0, lsl #16             ; a1 | a1
-    add         r6, r5, r2                      ; output + shortpitch * 3
-
-    str         r0, [r1, #0]
-    str         r0, [r1, #4]
-
-    str         r0, [r4, #0]
-    str         r0, [r4, #4]
-
-    str         r0, [r5, #0]
-    str         r0, [r5, #4]
-
-    str         r0, [r6, #0]
-    str         r0, [r6, #4]
-
-
-    ldmia       sp!, {r4 - r6, pc}
-
-    ENDP  ; |vp8_dc_only_idct_armv6|
-
     END

vp8/common/arm/armv6/iwalsh_v6.asm

@@ -8,8 +8,8 @@
 ;  be found in the AUTHORS file in the root of the source tree.
 ;
 
-    EXPORT |vp8_short_inv_walsh4x4_armv6|
+    EXPORT |vp8_short_inv_walsh4x4_v6|
-    EXPORT |vp8_short_inv_walsh4x4_1_armv6|
+    EXPORT |vp8_short_inv_walsh4x4_1_v6|
 
     ARM
     REQUIRE8
@@ -17,8 +17,8 @@
 
     AREA    |.text|, CODE, READONLY  ; name this block of code
 
-;short vp8_short_inv_walsh4x4_armv6(short *input, short *output)
+;short vp8_short_inv_walsh4x4_v6(short *input, short *output)
-|vp8_short_inv_walsh4x4_armv6| PROC
+|vp8_short_inv_walsh4x4_v6| PROC
 
     stmdb       sp!, {r4 - r11, lr}
 
@@ -123,11 +123,11 @@
     str         r5, [r1]
 
     ldmia       sp!, {r4 - r11, pc}
-    ENDP        ; |vp8_short_inv_walsh4x4_armv6|
+    ENDP        ; |vp8_short_inv_walsh4x4_v6|
 
 
-;short vp8_short_inv_walsh4x4_1_armv6(short *input, short *output)
+;short vp8_short_inv_walsh4x4_1_v6(short *input, short *output)
-|vp8_short_inv_walsh4x4_1_armv6| PROC
+|vp8_short_inv_walsh4x4_1_v6| PROC
 
     ldrsh       r2, [r0]             ; [0]
     add         r2, r2, #3           ; [0] + 3
@@ -145,7 +145,7 @@
     str         r2, [r1]
 
     bx          lr
-    ENDP        ; |vp8_short_inv_walsh4x4_1_armv6|
+    ENDP        ; |vp8_short_inv_walsh4x4_1_v6|
 
 ; Constant Pool
 c0x00030003 DCD 0x00030003

vp8/common/arm/armv6/simpleloopfilter_v6.asm

@@ -55,113 +55,87 @@ pstep       RN  r1
 ;stack  const char *thresh,
 ;stack  int  count
 
-;Note: All 16 elements in flimit are equal. So, in the code, only one load is needed
+; All 16 elements in flimit are equal. So, in the code, only one load is needed
-;for flimit. Same way applies to limit and thresh.
+; for flimit. Same applies to limit. thresh is not used in simple looopfilter
 
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 |vp8_loop_filter_simple_horizontal_edge_armv6| PROC
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    sub         src, src, pstep, lsl #1     ; move src pointer down by 2 lines
+    ldr         r12, [r3]                   ; limit
-
+    ldr         r3, [src, -pstep, lsl #1]   ; p1
-    ldr         r12, [r3], #4               ; limit
+    ldr         r4, [src, -pstep]           ; p0
-    ldr         r3, [src], pstep            ; p1
+    ldr         r5, [src]                   ; q0
-
+    ldr         r6, [src, pstep]            ; q1
-    ldr         r9, [sp, #36]               ; count for 8-in-parallel
+    ldr         r7, [r2]                    ; flimit
-    ldr         r4, [src], pstep            ; p0
-
-    ldr         r7, [r2], #4                ; flimit
-    ldr         r5, [src], pstep            ; q0
     ldr         r2, c0x80808080
-
+    ldr         r9, [sp, #40]               ; count for 8-in-parallel
-    ldr         r6, [src]                   ; q1
-
     uadd8       r7, r7, r7                  ; flimit * 2
-    mov         r9, r9, lsl #1              ; 4-in-parallel
+    mov         r9, r9, lsl #1              ; double the count. we're doing 4 at a time
     uadd8       r12, r7, r12                ; flimit * 2 + limit
+    mov         lr, #0                      ; need 0 in a couple places
 
 |simple_hnext8|
-    ; vp8_simple_filter_mask() function
+    ; vp8_simple_filter_mask()
 
     uqsub8      r7, r3, r6                  ; p1 - q1
     uqsub8      r8, r6, r3                  ; q1 - p1
     uqsub8      r10, r4, r5                 ; p0 - q0
     uqsub8      r11, r5, r4                 ; q0 - p0
     orr         r8, r8, r7                  ; abs(p1 - q1)
-    ldr         lr, c0x7F7F7F7F             ; 01111111 mask
     orr         r10, r10, r11               ; abs(p0 - q0)
-    and         r8, lr, r8, lsr #1          ; abs(p1 - q1) / 2
     uqadd8      r10, r10, r10               ; abs(p0 - q0) * 2
-    mvn         lr, #0                      ; r10 == -1
+    uhadd8      r8, r8, lr                  ; abs(p1 - q2) >> 1
     uqadd8      r10, r10, r8                ; abs(p0 - q0)*2 + abs(p1 - q1)/2
-    ; STALL waiting on r10 :(
+    mvn         r8, #0
-    uqsub8      r10, r10, r12               ; compare to flimit
+    usub8       r10, r12, r10               ; compare to flimit. usub8 sets GE flags
-    mov         r8, #0
+    sel         r10, r8, lr                 ; filter mask: F or 0
-
-    usub8       r10, r8, r10                ; use usub8 instead of ssub8
-    ; STALL (maybe?) when are flags set? :/
-    sel         r10, lr, r8                 ; filter mask: lr
-
     cmp         r10, #0
-    beq         simple_hskip_filter         ; skip filtering
+    beq         simple_hskip_filter         ; skip filtering if all masks are 0x00
 
-    ;vp8_simple_filter() function
+    ;vp8_simple_filter()
 
     eor         r3, r3, r2                  ; p1 offset to convert to a signed value
     eor         r6, r6, r2                  ; q1 offset to convert to a signed value
     eor         r4, r4, r2                  ; p0 offset to convert to a signed value
     eor         r5, r5, r2                  ; q0 offset to convert to a signed value
 
-    qsub8       r3, r3, r6                  ; vp8_filter (r3) = vp8_signed_char_clamp(p1-q1)
+    qsub8       r3, r3, r6                  ; vp8_filter = p1 - q1
-    qsub8       r6, r5, r4                  ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( q0 - p0))
+    qsub8       r6, r5, r4                  ; q0 - p0
-
+    qadd8       r3, r3, r6                  ; += q0 - p0
-    qadd8       r3, r3, r6
-    ldr         r8, c0x03030303             ; r8 = 3
-
-    qadd8       r3, r3, r6
     ldr         r7, c0x04040404
+    qadd8       r3, r3, r6                  ; += q0 - p0
+    ldr         r8, c0x03030303
+    qadd8       r3, r3, r6                  ; vp8_filter = p1-q1 + 3*(q0-p0))
+    ;STALL
+    and         r3, r3, r10                 ; vp8_filter &= mask
 
-    qadd8       r3, r3, r6
+    qadd8       r7 , r3 , r7                ; Filter1 = vp8_filter + 4
-    and         r3, r3, lr                  ; vp8_filter &= mask;
+    qadd8       r8 , r3 , r8                ; Filter2 = vp8_filter + 3
 
-    ;save bottom 3 bits so that we round one side +4 and the other +3
+    shadd8      r7 , r7 , lr
-    qadd8       r8 , r3 , r8                ; Filter2 (r8) = vp8_signed_char_clamp(vp8_filter+3)
+    shadd8      r8 , r8 , lr
-    qadd8       r3 , r3 , r7                ; Filter1 (r3) = vp8_signed_char_clamp(vp8_filter+4)
+    shadd8      r7 , r7 , lr
+    shadd8      r8 , r8 , lr
+    shadd8      r7 , r7 , lr                ; Filter1 >>= 3
+    shadd8      r8 , r8 , lr                ; Filter2 >>= 3
 
-    mov         r7, #0
+    qsub8       r5 ,r5, r7                  ; u = q0 - Filter1
-    shadd8      r8 , r8 , r7                ; Filter2 >>= 3
+    qadd8       r4, r4, r8                  ; u = p0 + Filter2
-    shadd8      r3 , r3 , r7                ; Filter1 >>= 3
-    shadd8      r8 , r8 , r7
-    shadd8      r3 , r3 , r7
-    shadd8      r8 , r8 , r7                ; r8: Filter2
-    shadd8      r3 , r3 , r7                ; r7: filter1
-
-    ;calculate output
-    sub         src, src, pstep, lsl #1
-
-    qadd8       r4, r4, r8                  ; u = vp8_signed_char_clamp(p0 + Filter2)
-    qsub8       r5 ,r5, r3                  ; u = vp8_signed_char_clamp(q0 - Filter1)
-    eor         r4, r4, r2                  ; *op0 = u^0x80
-    str         r4, [src], pstep            ; store op0 result
     eor         r5, r5, r2                  ; *oq0 = u^0x80
-    str         r5, [src], pstep            ; store oq0 result
+    str         r5, [src]                   ; store oq0 result
+    eor         r4, r4, r2                  ; *op0 = u^0x80
+    str         r4, [src, -pstep]           ; store op0 result
 
 |simple_hskip_filter|
-    add         src, src, #4
-    sub         src, src, pstep
-    sub         src, src, pstep, lsl #1
-
     subs        r9, r9, #1
+    addne       src, src, #4                ; next row
 
-    ;pld            [src]
+    ldrne       r3, [src, -pstep, lsl #1]   ; p1
-    ;pld            [src, pstep]
+    ldrne       r4, [src, -pstep]           ; p0
-    ;pld            [src, pstep, lsl #1]
+    ldrne       r5, [src]                   ; q0
-
+    ldrne       r6, [src, pstep]            ; q1
-    ldrne           r3, [src], pstep            ; p1
-    ldrne           r4, [src], pstep            ; p0
-    ldrne           r5, [src], pstep            ; q0
-    ldrne           r6, [src]                   ; q1
 
     bne         simple_hnext8
 
@@ -174,9 +148,9 @@ pstep       RN  r1
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    ldr         r12, [r2], #4               ; r12: flimit
+    ldr         r12, [r2]                   ; r12: flimit
     ldr         r2, c0x80808080
-    ldr         r7, [r3], #4                ; limit
+    ldr         r7, [r3]                    ; limit
 
     ; load soure data to r7, r8, r9, r10
     ldrh        r3, [src, #-2]
@@ -213,16 +187,14 @@ pstep       RN  r1
     uqsub8      r10, r5, r4                 ; q0 - p0
     orr         r7, r7, r8                  ; abs(p1 - q1)
     orr         r9, r9, r10                 ; abs(p0 - q0)
-    ldr         lr, c0x7F7F7F7F             ; 0111 1111 mask
-    uqadd8      r9, r9, r9                  ; abs(p0 - q0) * 2
-    and         r7, lr, r7, lsr #1          ; abs(p1 - q1) / 2
     mov         r8, #0
+    uqadd8      r9, r9, r9                  ; abs(p0 - q0) * 2
+    uhadd8      r7, r7, r8                  ; abs(p1 - q1) / 2
     uqadd8      r7, r7, r9                  ; abs(p0 - q0)*2 + abs(p1 - q1)/2
     mvn         r10, #0                     ; r10 == -1
-    uqsub8      r7, r7, r12                 ; compare to flimit
 
-    usub8       r7, r8, r7
+    usub8       r7, r12, r7                 ; compare to flimit
-    sel         r7, r10, r8                 ; filter mask: lr
+    sel         lr, r10, r8                 ; filter mask
 
     cmp         lr, #0
     beq         simple_vskip_filter         ; skip filtering
@@ -233,35 +205,34 @@ pstep       RN  r1
     eor         r4, r4, r2                  ; p0 offset to convert to a signed value
     eor         r5, r5, r2                  ; q0 offset to convert to a signed value
 
-    qsub8       r3, r3, r6                  ; vp8_filter (r3) = vp8_signed_char_clamp(p1-q1)
+    qsub8       r3, r3, r6                  ; vp8_filter = p1 - q1
-    qsub8       r6, r5, r4                  ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( q0 - p0))
+    qsub8       r6, r5, r4                  ; q0 - p0
 
-    qadd8       r3, r3, r6
+    qadd8       r3, r3, r6                  ; vp8_filter += q0 - p0
-    ldr         r8, c0x03030303             ; r8 = 3
+    ldr         r9, c0x03030303             ; r9 = 3
 
-    qadd8       r3, r3, r6
+    qadd8       r3, r3, r6                  ; vp8_filter += q0 - p0
     ldr         r7, c0x04040404
 
-    qadd8       r3, r3, r6
+    qadd8       r3, r3, r6                  ; vp8_filter = p1-q1 + 3*(q0-p0))
+    ;STALL
     and         r3, r3, lr                  ; vp8_filter &= mask
 
-    ;save bottom 3 bits so that we round one side +4 and the other +3
+    qadd8       r9 , r3 , r9                ; Filter2 = vp8_filter + 3
-    qadd8       r8 , r3 , r8                ; Filter2 (r8) = vp8_signed_char_clamp(vp8_filter+3)
+    qadd8       r3 , r3 , r7                ; Filter1 = vp8_filter + 4
-    qadd8       r3 , r3 , r7                ; Filter1 (r3) = vp8_signed_char_clamp(vp8_filter+4)
 
-    mov         r7, #0
+    shadd8      r9 , r9 , r8
-    shadd8      r8 , r8 , r7                ; Filter2 >>= 3
+    shadd8      r3 , r3 , r8
-    shadd8      r3 , r3 , r7                ; Filter1 >>= 3
+    shadd8      r9 , r9 , r8
-    shadd8      r8 , r8 , r7
+    shadd8      r3 , r3 , r8
-    shadd8      r3 , r3 , r7
+    shadd8      r9 , r9 , r8                ; Filter2 >>= 3
-    shadd8      r8 , r8 , r7                ; r8: filter2
+    shadd8      r3 , r3 , r8                ; Filter1 >>= 3
-    shadd8      r3 , r3 , r7                ; r7: filter1
 
     ;calculate output
     sub         src, src, pstep, lsl #2
 
-    qadd8       r4, r4, r8                  ; u = vp8_signed_char_clamp(p0 + Filter2)
+    qadd8       r4, r4, r9                  ; u = p0 + Filter2
-    qsub8       r5, r5, r3                  ; u = vp8_signed_char_clamp(q0 - Filter1)
+    qsub8       r5, r5, r3                  ; u = q0 - Filter1
     eor         r4, r4, r2                  ; *op0 = u^0x80
     eor         r5, r5, r2                  ; *oq0 = u^0x80
 
@@ -286,10 +257,6 @@ pstep       RN  r1
 |simple_vskip_filter|
     subs        r11, r11, #1
 
-    ;pld            [src]
-    ;pld            [src, pstep]
-    ;pld            [src, pstep, lsl #1]
-
     ; load soure data to r7, r8, r9, r10
     ldrneh      r3, [src, #-2]
     ldrneh      r4, [src], pstep
@@ -309,14 +276,12 @@ pstep       RN  r1
 
     bne         simple_vnext8
 
-    ldmia       sp!, {r4 - r12, pc}
+    ldmia       sp!, {r4 - r11, pc}
     ENDP        ; |vp8_loop_filter_simple_vertical_edge_armv6|
 
 ; Constant Pool
 c0x80808080 DCD     0x80808080
 c0x03030303 DCD     0x03030303
 c0x04040404 DCD     0x04040404
-c0x01010101 DCD     0x01010101
-c0x7F7F7F7F DCD     0x7F7F7F7F
 
     END

vp8/common/arm/idct_arm.h

@@ -15,9 +15,9 @@
 #if HAVE_ARMV6
 extern prototype_idct(vp8_short_idct4x4llm_1_v6);
 extern prototype_idct(vp8_short_idct4x4llm_v6_dual);
-extern prototype_idct_scalar(vp8_dc_only_idct_armv6);
+extern prototype_idct_scalar_add(vp8_dc_only_idct_add_v6);
-extern prototype_second_order(vp8_short_inv_walsh4x4_1_armv6);
+extern prototype_second_order(vp8_short_inv_walsh4x4_1_v6);
-extern prototype_second_order(vp8_short_inv_walsh4x4_armv6);
+extern prototype_second_order(vp8_short_inv_walsh4x4_v6);
 
 #undef  vp8_idct_idct1
 #define vp8_idct_idct1 vp8_short_idct4x4llm_1_v6
@@ -25,20 +25,20 @@ extern prototype_second_order(vp8_short_inv_walsh4x4_armv6);
 #undef  vp8_idct_idct16
 #define vp8_idct_idct16 vp8_short_idct4x4llm_v6_dual
 
-#undef  vp8_idct_idct1_scalar
+#undef  vp8_idct_idct1_scalar_add
-#define vp8_idct_idct1_scalar vp8_dc_only_idct_armv6
+#define vp8_idct_idct1_scalar_add vp8_dc_only_idct_add_v6
 
 #undef  vp8_idct_iwalsh1
-#define vp8_idct_iwalsh1 vp8_short_inv_walsh4x4_1_armv6
+#define vp8_idct_iwalsh1 vp8_short_inv_walsh4x4_1_v6
 
 #undef  vp8_idct_iwalsh16
-#define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_armv6
+#define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_v6
 #endif
 
 #if HAVE_ARMV7
 extern prototype_idct(vp8_short_idct4x4llm_1_neon);
 extern prototype_idct(vp8_short_idct4x4llm_neon);
-extern prototype_idct_scalar(vp8_dc_only_idct_neon);
+extern prototype_idct_scalar_add(vp8_dc_only_idct_add_neon);
 extern prototype_second_order(vp8_short_inv_walsh4x4_1_neon);
 extern prototype_second_order(vp8_short_inv_walsh4x4_neon);
 
@@ -48,8 +48,8 @@ extern prototype_second_order(vp8_short_inv_walsh4x4_neon);
 #undef  vp8_idct_idct16
 #define vp8_idct_idct16 vp8_short_idct4x4llm_neon
 
-#undef  vp8_idct_idct1_scalar
+#undef  vp8_idct_idct1_scalar_add
-#define vp8_idct_idct1_scalar vp8_dc_only_idct_neon
+#define vp8_idct_idct1_scalar_add vp8_dc_only_idct_add_neon
 
 #undef  vp8_idct_iwalsh1
 #define vp8_idct_iwalsh1 vp8_short_inv_walsh4x4_1_neon

vp8/common/arm/loopfilter_arm.c

@@ -14,16 +14,6 @@
 #include "loopfilter.h"
 #include "onyxc_int.h"
 
-typedef void loop_filter_uvfunction
-(
-    unsigned char *u,   // source pointer
-    int p,              // pitch
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
-    unsigned char *v
-);
-
 extern prototype_loopfilter(vp8_loop_filter_horizontal_edge_armv6);
 extern prototype_loopfilter(vp8_loop_filter_vertical_edge_armv6);
 extern prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_armv6);

vp8/common/arm/neon/dc_only_idct_add_neon.asm

@@ -0,0 +1,49 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license and patent
+;  grant that can be found in the LICENSE file in the root of the source
+;  tree. All contributing project authors may be found in the AUTHORS
+;  file in the root of the source tree.
+;
+
+
+    EXPORT  |vp8_dc_only_idct_add_neon|
+    ARM
+    REQUIRE8
+    PRESERVE8
+
+    AREA ||.text||, CODE, READONLY, ALIGN=2
+;void vp8_dc_only_idct_add_neon(short input_dc, unsigned char *pred_ptr,
+;                               unsigned char *dst_ptr, int pitch, int stride)
+; r0  input_dc
+; r1  pred_ptr
+; r2  dst_ptr
+; r3  pitch
+; sp  stride
+|vp8_dc_only_idct_add_neon| PROC
+    add             r0, r0, #4
+    asr             r0, r0, #3
+    ldr             r12, [sp]
+    vdup.16         q0, r0
+
+    vld1.32         {d2[0]}, [r1], r3
+    vld1.32         {d2[1]}, [r1], r3
+    vld1.32         {d4[0]}, [r1], r3
+    vld1.32         {d4[1]}, [r1]
+
+    vaddw.u8        q1, q0, d2
+    vaddw.u8        q2, q0, d4
+
+    vqmovun.s16     d2, q1
+    vqmovun.s16     d4, q2
+
+    vst1.32         {d2[0]}, [r2], r12
+    vst1.32         {d2[1]}, [r2], r12
+    vst1.32         {d4[0]}, [r2], r12
+    vst1.32         {d4[1]}, [r2]
+
+    bx             lr
+
+    ENDP
+    END

vp8/common/arm/reconintra_arm.c

@@ -29,7 +29,7 @@ void vp8_build_intra_predictors_mby_neon(MACROBLOCKD *x)
     unsigned char *y_buffer = x->dst.y_buffer;
     unsigned char *ypred_ptr = x->predictor;
     int y_stride = x->dst.y_stride;
-    int mode = x->mbmi.mode;
+    int mode = x->mode_info_context->mbmi.mode;
     int Up = x->up_available;
     int Left = x->left_available;
 
@@ -52,7 +52,7 @@ void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x)
     unsigned char *y_buffer = x->dst.y_buffer;
     unsigned char *ypred_ptr = x->predictor;
     int y_stride = x->dst.y_stride;
-    int mode = x->mbmi.mode;
+    int mode = x->mode_info_context->mbmi.mode;
     int Up = x->up_available;
     int Left = x->left_available;
 

vp8/common/arm/systemdependent.c

@@ -43,7 +43,6 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
 
     rtcd->idct.idct1        = vp8_short_idct4x4llm_1_neon;
     rtcd->idct.idct16       = vp8_short_idct4x4llm_neon;
-    rtcd->idct.idct1_scalar = vp8_dc_only_idct_neon;
     rtcd->idct.iwalsh1      = vp8_short_inv_walsh4x4_1_neon;
     rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_neon;
 
@@ -75,7 +74,6 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
 
     rtcd->idct.idct1        = vp8_short_idct4x4llm_1_v6;
     rtcd->idct.idct16       = vp8_short_idct4x4llm_v6_dual;
-    rtcd->idct.idct1_scalar = vp8_dc_only_idct_armv6;
     rtcd->idct.iwalsh1      = vp8_short_inv_walsh4x4_1_armv6;
     rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_armv6;
 
@@ -128,11 +126,13 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_c;
 #endif
 
+#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_PSNR)
     rtcd->postproc.down        = vp8_mbpost_proc_down_c;
     rtcd->postproc.across      = vp8_mbpost_proc_across_ip_c;
     rtcd->postproc.downacross  = vp8_post_proc_down_and_across_c;
     rtcd->postproc.addnoise    = vp8_plane_add_noise_c;
 #endif
+#endif
 
 #if HAVE_ARMV7
     vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby_neon;

vp8/common/arm/vpx_asm_offsets.c

@@ -51,19 +51,18 @@ DEFINE(mb_dst_y_stride,                          offsetof(MACROBLOCKD, dst.y_str
 DEFINE(mb_dst_y_buffer,                         offsetof(MACROBLOCKD, dst.y_buffer));
 DEFINE(mb_dst_u_buffer,                         offsetof(MACROBLOCKD, dst.u_buffer));
 DEFINE(mb_dst_v_buffer,                         offsetof(MACROBLOCKD, dst.v_buffer));
-DEFINE(mb_mbmi_mode,                            offsetof(MACROBLOCKD, mbmi.mode));
 DEFINE(mb_up_available,                         offsetof(MACROBLOCKD, up_available));
 DEFINE(mb_left_available,                       offsetof(MACROBLOCKD, left_available));
 
 DEFINE(detok_scan,                              offsetof(DETOK, scan));
-DEFINE(detok_ptr_onyxblock2context_leftabove,    offsetof(DETOK, ptr_onyxblock2context_leftabove));
+DEFINE(detok_ptr_block2leftabove,               offsetof(DETOK, ptr_block2leftabove));
-DEFINE(detok_onyx_coef_tree_ptr,                  offsetof(DETOK, vp8_coef_tree_ptr));
+DEFINE(detok_coef_tree_ptr,                     offsetof(DETOK, vp8_coef_tree_ptr));
 DEFINE(detok_teb_base_ptr,                      offsetof(DETOK, teb_base_ptr));
 DEFINE(detok_norm_ptr,                          offsetof(DETOK, norm_ptr));
-DEFINE(detok_ptr_onyx_coef_bands_x,                offsetof(DETOK, ptr_onyx_coef_bands_x));
+DEFINE(detok_ptr_coef_bands_x,                  offsetof(DETOK, ptr_coef_bands_x));
 
-DEFINE(DETOK_A,                                 offsetof(DETOK, A));
+DEFINE(detok_A,                                 offsetof(DETOK, A));
-DEFINE(DETOK_L,                                 offsetof(DETOK, L));
+DEFINE(detok_L,                                 offsetof(DETOK, L));
 
 DEFINE(detok_qcoeff_start_ptr,                  offsetof(DETOK, qcoeff_start_ptr));
 DEFINE(detok_current_bc,                        offsetof(DETOK, current_bc));

vp8/common/blockd.c

@@ -12,13 +12,13 @@
 #include "blockd.h"
 #include "vpx_mem/vpx_mem.h"
 
-void vp8_setup_temp_context(TEMP_CONTEXT *t, ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l, int count)
-{
-    vpx_memcpy(t->l, l, sizeof(ENTROPY_CONTEXT) * count);
-    vpx_memcpy(t->a, a, sizeof(ENTROPY_CONTEXT) * count);
-}
-
-const int vp8_block2left[25] = { 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 1, 1, 0, 0, 1, 1, 0};
-const int vp8_block2above[25] = { 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 0, 1, 0, 1, 0, 1, 0};
 const int vp8_block2type[25] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2, 1};
-const int vp8_block2context[25] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3};
+
+const unsigned char vp8_block2left[25] =
+{
+    0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8
+};
+const unsigned char vp8_block2above[25] =
+{
+    0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8
+};

vp8/common/blockd.h

@@ -49,19 +49,19 @@ typedef struct
 } POS;
 
 
-typedef int ENTROPY_CONTEXT;
+typedef char ENTROPY_CONTEXT;
-
 typedef struct
 {
-    ENTROPY_CONTEXT l[4];
+    ENTROPY_CONTEXT y1[4];
-    ENTROPY_CONTEXT a[4];
+    ENTROPY_CONTEXT u[2];
-} TEMP_CONTEXT;
+    ENTROPY_CONTEXT v[2];
+    ENTROPY_CONTEXT y2;
+} ENTROPY_CONTEXT_PLANES;
 
-extern void vp8_setup_temp_context(TEMP_CONTEXT *t, ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l, int count);
-extern const int vp8_block2left[25];
-extern const int vp8_block2above[25];
 extern const int vp8_block2type[25];
-extern const int vp8_block2context[25];
+
+extern const unsigned char vp8_block2left[25];
+extern const unsigned char vp8_block2above[25];
 
 #define VP8_COMBINEENTROPYCONTEXTS( Dest, A, B) \
     Dest = ((A)!=0) + ((B)!=0);
@@ -215,9 +215,10 @@ typedef struct
 {
     DECLARE_ALIGNED(16, short, diff[400]);      // from idct diff
     DECLARE_ALIGNED(16, unsigned char,  predictor[384]);
-    DECLARE_ALIGNED(16, short, reference[384]);
+//not used    DECLARE_ALIGNED(16, short, reference[384]);
     DECLARE_ALIGNED(16, short, qcoeff[400]);
     DECLARE_ALIGNED(16, short, dqcoeff[400]);
+    DECLARE_ALIGNED(16, char,  eobs[25]);
 
     // 16 Y blocks, 4 U, 4 V, 1 DC 2nd order block, each with 16 entries.
     BLOCKD block[25];
@@ -232,14 +233,12 @@ typedef struct
 
     FRAME_TYPE frame_type;
 
-    MB_MODE_INFO mbmi;
-
     int up_available;
     int left_available;
 
     // Y,U,V,Y2
-    ENTROPY_CONTEXT *above_context[4];   // row of context for each plane
+    ENTROPY_CONTEXT_PLANES *above_context;
-    ENTROPY_CONTEXT(*left_context)[4];   // (up to) 4 contexts ""
+    ENTROPY_CONTEXT_PLANES *left_context;
 
     // 0 indicates segmentation at MB level is not enabled. Otherwise the individual bits indicate which features are active.
     unsigned char segmentation_enabled;
@@ -275,9 +274,6 @@ typedef struct
     int mb_to_top_edge;
     int mb_to_bottom_edge;
 
-    //char * gf_active_ptr;
-    signed char *gf_active_ptr;
-
     unsigned int frames_since_golden;
     unsigned int frames_till_alt_ref_frame;
     vp8_subpix_fn_t  subpixel_predict;

vp8/common/entropymode.c

@@ -264,8 +264,10 @@ void vp8_entropy_mode_init()
     vp8_tokens_from_tree(vp8_uv_mode_encodings,  vp8_uv_mode_tree);
     vp8_tokens_from_tree(vp8_mbsplit_encodings, vp8_mbsplit_tree);
 
-    vp8_tokens_from_tree(VP8_MVREFENCODINGS,   vp8_mv_ref_tree);
+    vp8_tokens_from_tree_offset(vp8_mv_ref_encoding_array,
-    vp8_tokens_from_tree(VP8_SUBMVREFENCODINGS, vp8_sub_mv_ref_tree);
+                                vp8_mv_ref_tree, NEARESTMV);
+    vp8_tokens_from_tree_offset(vp8_sub_mv_ref_encoding_array,
+                                vp8_sub_mv_ref_tree, LEFT4X4);
 
     vp8_tokens_from_tree(vp8_small_mvencodings, vp8_small_mvtree);
 }

vp8/common/entropymode.h

@@ -54,10 +54,6 @@ extern struct vp8_token_struct vp8_mbsplit_encodings  [VP8_NUMMBSPLITS];
 extern struct vp8_token_struct vp8_mv_ref_encoding_array    [VP8_MVREFS];
 extern struct vp8_token_struct vp8_sub_mv_ref_encoding_array [VP8_SUBMVREFS];
 
-#define VP8_MVREFENCODINGS      (vp8_mv_ref_encoding_array - NEARESTMV)
-#define VP8_SUBMVREFENCODINGS   (vp8_sub_mv_ref_encoding_array - LEFT4X4)
-
-
 extern const vp8_tree_index vp8_small_mvtree[];
 
 extern struct vp8_token_struct vp8_small_mvencodings [8];

vp8/common/extend.c

@@ -39,6 +39,9 @@ static void extend_plane_borders
 
     for (i = 0; i < h - 0 + 1; i++)
     {
+        // Some linkers will complain if we call vpx_memset with el set to a
+        // constant 0.
+        if (el)
             vpx_memset(dest_ptr1, src_ptr1[0], el);
         vpx_memset(dest_ptr2, src_ptr2[0], er);
         src_ptr1  += sp;

vp8/common/generic/systemdependent.c

@@ -32,7 +32,7 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
 
     rtcd->idct.idct1        = vp8_short_idct4x4llm_1_c;
     rtcd->idct.idct16       = vp8_short_idct4x4llm_c;
-    rtcd->idct.idct1_scalar = vp8_dc_only_idct_c;
+    rtcd->idct.idct1_scalar_add = vp8_dc_only_idct_add_c;
     rtcd->idct.iwalsh1      = vp8_short_inv_walsh4x4_1_c;
     rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_c;
 
@@ -61,7 +61,7 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_c;
     rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_c;
 
-#if CONFIG_POSTPROC || CONFIG_VP8_ENCODER
+#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_PSNR)
     rtcd->postproc.down        = vp8_mbpost_proc_down_c;
     rtcd->postproc.across      = vp8_mbpost_proc_across_ip_c;
     rtcd->postproc.downacross  = vp8_post_proc_down_and_across_c;

vp8/common/idct.h

@@ -18,8 +18,10 @@
 #define prototype_idct(sym) \
     void sym(short *input, short *output, int pitch)
 
-#define prototype_idct_scalar(sym) \
+#define prototype_idct_scalar_add(sym) \
-    void sym(short input, short *output, int pitch)
+    void sym(short input, \
+             unsigned char *pred, unsigned char *output, \
+             int pitch, int stride)
 
 #if ARCH_X86 || ARCH_X86_64
 #include "x86/idct_x86.h"
@@ -39,10 +41,10 @@ extern prototype_idct(vp8_idct_idct1);
 #endif
 extern prototype_idct(vp8_idct_idct16);
 
-#ifndef vp8_idct_idct1_scalar
+#ifndef vp8_idct_idct1_scalar_add
-#define vp8_idct_idct1_scalar vp8_dc_only_idct_c
+#define vp8_idct_idct1_scalar_add vp8_dc_only_idct_add_c
 #endif
-extern prototype_idct_scalar(vp8_idct_idct1_scalar);
+extern prototype_idct_scalar_add(vp8_idct_idct1_scalar_add);
 
 
 #ifndef vp8_idct_iwalsh1
@@ -56,14 +58,14 @@ extern prototype_second_order(vp8_idct_iwalsh1);
 extern prototype_second_order(vp8_idct_iwalsh16);
 
 typedef prototype_idct((*vp8_idct_fn_t));
-typedef prototype_idct_scalar((*vp8_idct_scalar_fn_t));
+typedef prototype_idct_scalar_add((*vp8_idct_scalar_add_fn_t));
 typedef prototype_second_order((*vp8_second_order_fn_t));
 
 typedef struct
 {
     vp8_idct_fn_t            idct1;
     vp8_idct_fn_t            idct16;
-    vp8_idct_scalar_fn_t  idct1_scalar;
+    vp8_idct_scalar_add_fn_t idct1_scalar_add;
 
     vp8_second_order_fn_t iwalsh1;
     vp8_second_order_fn_t iwalsh16;

vp8/common/idctllm.c

@@ -104,23 +104,30 @@ void vp8_short_idct4x4llm_1_c(short *input, short *output, int pitch)
     }
 }
 
-
+void vp8_dc_only_idct_add_c(short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride)
-void vp8_dc_only_idct_c(short input_dc, short *output, int pitch)
 {
-    int i;
+    int a1 = ((input_dc + 4) >> 3);
-    int a1;
+    int r, c;
-    short *op = output;
-    int shortpitch = pitch >> 1;
-    a1 = ((input_dc + 4) >> 3);
 
-    for (i = 0; i < 4; i++)
+    for (r = 0; r < 4; r++)
     {
-        op[0] = a1;
+        for (c = 0; c < 4; c++)
-        op[1] = a1;
+        {
-        op[2] = a1;
+            int a = a1 + pred_ptr[c] ;
-        op[3] = a1;
+
-        op += shortpitch;
+            if (a < 0)
+                a = 0;
+
+            if (a > 255)
+                a = 255;
+
+            dst_ptr[c] = (unsigned char) a ;
         }
+
+        dst_ptr += stride;
+        pred_ptr += pitch;
+    }
+
 }
 
 void vp8_short_inv_walsh4x4_c(short *input, short *output)

vp8/common/invtrans.c

@@ -65,7 +65,8 @@ void vp8_inverse_transform_mb(const vp8_idct_rtcd_vtable_t *rtcd, MACROBLOCKD *x
 {
     int i;
 
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != B_PRED &&
+        x->mode_info_context->mbmi.mode != SPLITMV)
     {
         // do 2nd order transform on the dc block
 

vp8/common/loopfilter.h

@@ -117,5 +117,14 @@ typedef struct
 #define LF_INVOKE(ctx,fn) vp8_lf_##fn
 #endif
 
+typedef void loop_filter_uvfunction
+(
+    unsigned char *u,   // source pointer
+    int p,              // pitch
+    const signed char *flimit,
+    const signed char *limit,
+    const signed char *thresh,
+    unsigned char *v
+);
 
 #endif

vp8/common/loopfilter_filters.c

@@ -18,7 +18,7 @@
 
 typedef unsigned char uc;
 
-__inline signed char vp8_signed_char_clamp(int t)
+static __inline signed char vp8_signed_char_clamp(int t)
 {
     t = (t < -128 ? -128 : t);
     t = (t > 127 ? 127 : t);
@@ -27,7 +27,7 @@ __inline signed char vp8_signed_char_clamp(int t)
 
 
 // should we apply any filter at all ( 11111111 yes, 00000000 no)
-__inline signed char vp8_filter_mask(signed char limit, signed char flimit,
+static __inline signed char vp8_filter_mask(signed char limit, signed char flimit,
                                      uc p3, uc p2, uc p1, uc p0, uc q0, uc q1, uc q2, uc q3)
 {
     signed char mask = 0;
@@ -47,7 +47,7 @@ __inline signed char vp8_filter_mask(signed char limit, signed char flimit,
 }
 
 // is there high variance internal edge ( 11111111 yes, 00000000 no)
-__inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0, uc q1)
+static __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0, uc q1)
 {
     signed char hev = 0;
     hev  |= (abs(p1 - p0) > thresh) * -1;
@@ -55,7 +55,7 @@ __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0, uc q1)
     return hev;
 }
 
-__inline void vp8_filter(signed char mask, signed char hev, uc *op1, uc *op0, uc *oq0, uc *oq1)
+static __inline void vp8_filter(signed char mask, signed char hev, uc *op1, uc *op0, uc *oq0, uc *oq1)
 
 {
     signed char ps0, qs0;
@@ -161,7 +161,7 @@ void vp8_loop_filter_vertical_edge_c
     while (++i < count * 8);
 }
 
-__inline void vp8_mbfilter(signed char mask, signed char hev,
+static __inline void vp8_mbfilter(signed char mask, signed char hev,
                            uc *op2, uc *op1, uc *op0, uc *oq0, uc *oq1, uc *oq2)
 {
     signed char s, u;
@@ -281,7 +281,7 @@ void vp8_mbloop_filter_vertical_edge_c
 }
 
 // should we apply any filter at all ( 11111111 yes, 00000000 no)
-__inline signed char vp8_simple_filter_mask(signed char limit, signed char flimit, uc p1, uc p0, uc q0, uc q1)
+static __inline signed char vp8_simple_filter_mask(signed char limit, signed char flimit, uc p1, uc p0, uc q0, uc q1)
 {
 // Why does this cause problems for win32?
 // error C2143: syntax error : missing ';' before 'type'
@@ -294,7 +294,7 @@ __inline signed char vp8_simple_filter_mask(signed char limit, signed char flimi
     return mask;
 }
 
-__inline void vp8_simple_filter(signed char mask, uc *op1, uc *op0, uc *oq0, uc *oq1)
+static __inline void vp8_simple_filter(signed char mask, uc *op1, uc *op0, uc *oq0, uc *oq1)
 {
     signed char vp8_filter, Filter1, Filter2;
     signed char p1 = (signed char) * op1 ^ 0x80;

vp8/common/onyxc_int.h

@@ -33,6 +33,7 @@ void vp8_initialize_common(void);
 #define MAXQ 127
 #define QINDEX_RANGE (MAXQ + 1)
 
+#define NUM_YV12_BUFFERS 4
 
 typedef struct frame_contexts
 {
@@ -94,15 +95,16 @@ typedef struct VP8Common
     YUV_TYPE clr_type;
     CLAMP_TYPE  clamp_type;
 
-    YV12_BUFFER_CONFIG last_frame;
-    YV12_BUFFER_CONFIG golden_frame;
-    YV12_BUFFER_CONFIG alt_ref_frame;
-    YV12_BUFFER_CONFIG new_frame;
     YV12_BUFFER_CONFIG *frame_to_show;
+
+    YV12_BUFFER_CONFIG yv12_fb[NUM_YV12_BUFFERS];
+    int fb_idx_ref_cnt[NUM_YV12_BUFFERS];
+    int new_fb_idx, lst_fb_idx, gld_fb_idx, alt_fb_idx;
+
     YV12_BUFFER_CONFIG post_proc_buffer;
     YV12_BUFFER_CONFIG temp_scale_frame;
 
-    FRAME_TYPE last_frame_type;  //Add to check if vp8_frame_init_loop_filter() can be skiped.
+    FRAME_TYPE last_frame_type;  //Add to check if vp8_frame_init_loop_filter() can be skipped.
     FRAME_TYPE frame_type;
 
     int show_frame;
@@ -131,8 +133,6 @@ typedef struct VP8Common
 
     unsigned int frames_since_golden;
     unsigned int frames_till_alt_ref_frame;
-    unsigned char *gf_active_flags;   // Record of which MBs still refer to last golden frame either directly or through 0,0
-    int gf_active_count;
 
     /* We allocate a MODE_INFO struct for each macroblock, together with
        an extra row on top and column on the left to simplify prediction. */
@@ -165,8 +165,8 @@ typedef struct VP8Common
     int ref_frame_sign_bias[MAX_REF_FRAMES];    // Two state 0, 1
 
     // Y,U,V,Y2
-    ENTROPY_CONTEXT *above_context[4];   // row of context for each plane
+    ENTROPY_CONTEXT_PLANES *above_context;   // row of context for each plane
-    ENTROPY_CONTEXT left_context[4][4];  // (up to) 4 contexts ""
+    ENTROPY_CONTEXT_PLANES left_context;  // (up to) 4 contexts ""
 
 
     // keyframe block modes are predicted by their above, left neighbors
@@ -201,6 +201,7 @@ typedef struct VP8Common
 
 void vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int *filter_level);
 void vp8_init_loop_filter(VP8_COMMON *cm);
+void vp8_frame_init_loop_filter(loop_filter_info *lfi, int frame_type);
 extern void vp8_loop_filter_frame(VP8_COMMON *cm,    MACROBLOCKD *mbd,  int filt_val);
 
 #endif

vp8/common/postproc.c

@@ -330,13 +330,6 @@ void vp8_de_noise(YV12_BUFFER_CONFIG         *source,
 
 }
 
-
-//Notes: It is better to change CHAR to unsigned or signed to
-//avoid error on ARM platform.
-char vp8_an[8][64][3072];
-int vp8_cd[8][64];
-
-
 double vp8_gaussian(double sigma, double mu, double x)
 {
     return 1 / (sigma * sqrt(2.0 * 3.14159265)) *

vp8/common/reconinter.c

@@ -210,7 +210,8 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 {
     int i;
 
-    if (x->mbmi.ref_frame != INTRA_FRAME && x->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
+        x->mode_info_context->mbmi.mode != SPLITMV)
     {
         unsigned char *uptr, *vptr;
         unsigned char *upred_ptr = &x->predictor[256];
@@ -254,16 +255,18 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
     }
 }
 
-
+//encoder only
 void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
 {
-    if (x->mbmi.ref_frame != INTRA_FRAME && x->mbmi.mode != SPLITMV)
+
+  if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
+      x->mode_info_context->mbmi.mode != SPLITMV)
     {
         unsigned char *ptr_base;
         unsigned char *ptr;
         unsigned char *pred_ptr = x->predictor;
-        int mv_row = x->mbmi.mv.as_mv.row;
+        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mbmi.mv.as_mv.col;
+        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
         int pre_stride = x->block[0].pre_stride;
 
         ptr_base = x->pre.y_buffer;
@@ -282,7 +285,7 @@ void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
     {
         int i;
 
-        if (x->mbmi.partitioning < 3)
+        if (x->mode_info_context->mbmi.partitioning < 3)
         {
             for (i = 0; i < 4; i++)
             {
@@ -313,7 +316,9 @@ void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
 
 void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
 {
-    if (x->mbmi.ref_frame != INTRA_FRAME && x->mbmi.mode != SPLITMV)
+
+    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
+        x->mode_info_context->mbmi.mode != SPLITMV)
     {
         int offset;
         unsigned char *ptr_base;
@@ -323,8 +328,8 @@ void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
         unsigned char *upred_ptr = &x->predictor[256];
         unsigned char *vpred_ptr = &x->predictor[320];
 
-        int mv_row = x->mbmi.mv.as_mv.row;
+        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mbmi.mv.as_mv.col;
+        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
         int pre_stride = x->block[0].pre_stride;
 
         ptr_base = x->pre.y_buffer;
@@ -361,7 +366,7 @@ void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
     {
         int i;
 
-        if (x->mbmi.partitioning < 3)
+        if (x->mode_info_context->mbmi.partitioning < 3)
         {
             for (i = 0; i < 4; i++)
             {
@@ -410,7 +415,7 @@ void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel)
 {
     int i, j;
 
-    if (x->mbmi.mode == SPLITMV)
+    if (x->mode_info_context->mbmi.mode == SPLITMV)
     {
         for (i = 0; i < 2; i++)
         {
@@ -455,8 +460,8 @@ void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel)
     }
     else
     {
-        int mvrow = x->mbmi.mv.as_mv.row;
+        int mvrow = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mvcol = x->mbmi.mv.as_mv.col;
+        int mvcol = x->mode_info_context->mbmi.mv.as_mv.col;
 
         if (mvrow < 0)
             mvrow -= 1;
@@ -535,7 +540,7 @@ void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
     unsigned char *pred_ptr = x->predictor;
     unsigned char *dst_ptr = x->dst.y_buffer;
 
-    if (x->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
         int offset;
         unsigned char *ptr_base;
@@ -547,8 +552,8 @@ void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
         unsigned char *udst_ptr = x->dst.u_buffer;
         unsigned char *vdst_ptr = x->dst.v_buffer;
 
-        int mv_row = x->mbmi.mv.as_mv.row;
+        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mbmi.mv.as_mv.col;
+        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
         int pre_stride = x->dst.y_stride; //x->block[0].pre_stride;
 
         ptr_base = x->pre.y_buffer;
@@ -587,7 +592,7 @@ void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
         //if sth is wrong, go back to what it is in build_inter_predictors_mb.
         int i;
 
-        if (x->mbmi.partitioning < 3)
+        if (x->mode_info_context->mbmi.partitioning < 3)
         {
             for (i = 0; i < 4; i++)
             {

vp8/common/reconintra.c

@@ -43,7 +43,7 @@ void vp8_build_intra_predictors_mby(MACROBLOCKD *x)
     }
 
     // for Y
-    switch (x->mbmi.mode)
+    switch (x->mode_info_context->mbmi.mode)
     {
     case DC_PRED:
     {
@@ -164,7 +164,7 @@ void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x)
     }
 
     // for Y
-    switch (x->mbmi.mode)
+    switch (x->mode_info_context->mbmi.mode)
     {
     case DC_PRED:
     {
@@ -290,7 +290,7 @@ void vp8_build_intra_predictors_mbuv(MACROBLOCKD *x)
         vleft_col[i] = x->dst.v_buffer [i* x->dst.uv_stride -1];
     }
 
-    switch (x->mbmi.uv_mode)
+    switch (x->mode_info_context->mbmi.uv_mode)
     {
     case DC_PRED:
     {
@@ -430,7 +430,7 @@ void vp8_build_intra_predictors_mbuv_s(MACROBLOCKD *x)
         vleft_col[i] = x->dst.v_buffer [i* x->dst.uv_stride -1];
     }
 
-    switch (x->mbmi.uv_mode)
+    switch (x->mode_info_context->mbmi.uv_mode)
     {
     case DC_PRED:
     {

vp8/common/threading.h

@@ -75,7 +75,8 @@
 #define thread_sleep(nms) // { struct timespec ts;ts.tv_sec=0; ts.tv_nsec = 1000*nms;nanosleep(&ts, NULL);}
 #else
 #include <unistd.h>
-#define thread_sleep(nms) usleep(nms*1000);// {struct timespec ts;ts.tv_sec=0; ts.tv_nsec = 1000*nms;nanosleep(&ts, NULL);}
+#include <sched.h>
+#define thread_sleep(nms) sched_yield();// {struct timespec ts;ts.tv_sec=0; ts.tv_nsec = 1000*nms;nanosleep(&ts, NULL);}
 #endif
 /* Not Windows. Assume pthreads */
 

vp8/common/treecoder.c

@@ -47,6 +47,12 @@ void vp8_tokens_from_tree(struct vp8_token_struct *p, vp8_tree t)
     tree2tok(p, t, 0, 0, 0);
 }
 
+void vp8_tokens_from_tree_offset(struct vp8_token_struct *p, vp8_tree t,
+                                 int offset)
+{
+    tree2tok(p - offset, t, 0, 0, 0);
+}
+
 static void branch_counts(
     int n,                      /* n = size of alphabet */
     vp8_token tok               [ /* n */ ],

vp8/common/treecoder.h

@@ -54,6 +54,8 @@ typedef const struct vp8_token_struct
 /* Construct encoding array from tree. */
 
 void vp8_tokens_from_tree(struct vp8_token_struct *, vp8_tree);
+void vp8_tokens_from_tree_offset(struct vp8_token_struct *, vp8_tree,
+                                 int offset);
 
 
 /* Convert array of token occurrence counts into a table of probabilities

vp8/common/x86/idct_x86.h

@@ -22,7 +22,7 @@
 #if HAVE_MMX
 extern prototype_idct(vp8_short_idct4x4llm_1_mmx);
 extern prototype_idct(vp8_short_idct4x4llm_mmx);
-extern prototype_idct_scalar(vp8_dc_only_idct_mmx);
+extern prototype_idct_scalar_add(vp8_dc_only_idct_add_mmx);
 
 extern prototype_second_order(vp8_short_inv_walsh4x4_mmx);
 extern prototype_second_order(vp8_short_inv_walsh4x4_1_mmx);
@@ -34,8 +34,8 @@ extern prototype_second_order(vp8_short_inv_walsh4x4_1_mmx);
 #undef  vp8_idct_idct16
 #define vp8_idct_idct16 vp8_short_idct4x4llm_mmx
 
-#undef  vp8_idct_idct1_scalar
+#undef  vp8_idct_idct1_scalar_add
-#define vp8_idct_idct1_scalar vp8_dc_only_idct_mmx
+#define vp8_idct_idct1_scalar_add vp8_dc_only_idct_add_mmx
 
 #undef vp8_idct_iwalsh16
 #define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_mmx

vp8/common/x86/idctllm_mmx.asm

@@ -220,35 +220,61 @@ sym(vp8_short_idct4x4llm_1_mmx):
     pop         rbp
     ret
 
-;void dc_only_idct_mmx(short input_dc, short *output, int pitch)
+;void vp8_dc_only_idct_add_mmx(short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride)
-global sym(vp8_dc_only_idct_mmx)
+global sym(vp8_dc_only_idct_add_mmx)
-sym(vp8_dc_only_idct_mmx):
+sym(vp8_dc_only_idct_add_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 3
+    SHADOW_ARGS_TO_STACK 5
     GET_GOT     rbx
+    push        rsi
+    push        rdi
     ; end prolog
 
-        movd        mm0,            arg(0) ;input_dc
+        mov         rsi,            arg(1) ;s -- prediction
+        mov         rdi,            arg(2) ;d -- destination
+        movsxd      rax,            dword ptr arg(4) ;stride
+        movsxd      rdx,            dword ptr arg(3) ;pitch
+        pxor        mm0,            mm0
 
-        paddw       mm0,            [fours GLOBAL]
+        movd        mm5,            arg(0) ;input_dc
-        mov         rdx,            arg(1) ;output
 
-        psraw       mm0,            3
+        paddw       mm5,            [fours GLOBAL]
-        movsxd      rax,            dword ptr arg(2) ;pitch
 
-        punpcklwd   mm0,            mm0
+        psraw       mm5,            3
-        punpckldq   mm0,            mm0
 
-        movq        [rdx],          mm0
+        punpcklwd   mm5,            mm5
-        movq        [rdx+rax],      mm0
+        punpckldq   mm5,            mm5
 
-        movq        [rdx+rax*2],    mm0
+        movd        mm1,            [rsi]
-        add         rdx,            rax
+        punpcklbw   mm1,            mm0
+        paddsw      mm1,            mm5
+        packuswb    mm1,            mm0              ; pack and unpack to saturate
+        movd        [rdi],          mm1
 
-        movq        [rdx+rax*2],    mm0
+        movd        mm2,            [rsi+rdx]
+        punpcklbw   mm2,            mm0
+        paddsw      mm2,            mm5
+        packuswb    mm2,            mm0              ; pack and unpack to saturate
+        movd        [rdi+rax],      mm2
+
+        movd        mm3,            [rsi+2*rdx]
+        punpcklbw   mm3,            mm0
+        paddsw      mm3,            mm5
+        packuswb    mm3,            mm0              ; pack and unpack to saturate
+        movd        [rdi+2*rax],    mm3
+
+        add         rdi,            rax
+        add         rsi,            rdx
+        movd        mm4,            [rsi+2*rdx]
+        punpcklbw   mm4,            mm0
+        paddsw      mm4,            mm5
+        packuswb    mm4,            mm0              ; pack and unpack to saturate
+        movd        [rdi+2*rax],    mm4
 
     ; begin epilog
+    pop rdi
+    pop rsi
     RESTORE_GOT
     UNSHADOW_ARGS
     pop         rbp

vp8/common/x86/idctllm_sse2.asm

@@ -0,0 +1,708 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license
+;  that can be found in the LICENSE file in the root of the source
+;  tree. An additional intellectual property rights grant can be found
+;  in the file PATENTS.  All contributing project authors may
+;  be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+;void idct_dequant_0_2x_sse2
+; (
+;   short *qcoeff       - 0
+;   short *dequant      - 1
+;   unsigned char *pre  - 2
+;   unsigned char *dst  - 3
+;   int dst_stride      - 4
+;   int blk_stride      - 5
+; )
+
+global sym(idct_dequant_0_2x_sse2)
+sym(idct_dequant_0_2x_sse2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    ; end prolog
+
+        mov         rdx,            arg(1) ; dequant
+        mov         rax,            arg(0) ; qcoeff
+
+    ; Zero out xmm7, for use unpacking
+        pxor        xmm7,           xmm7
+
+        movd        xmm4,           [rax]
+        movd        xmm5,           [rdx]
+
+        pinsrw      xmm4,           [rax+32],   4
+        pinsrw      xmm5,           [rdx],      4
+
+        pmullw      xmm4,           xmm5
+
+    ; clear coeffs
+        movd        [rax],          xmm7
+        movd        [rax+32],       xmm7
+;pshufb
+        pshuflw     xmm4,           xmm4,       00000000b
+        pshufhw     xmm4,           xmm4,       00000000b
+
+        mov         rax,            arg(2) ; pre
+        paddw       xmm4,           [fours GLOBAL]
+
+        movsxd      rcx,            dword ptr arg(5) ; blk_stride
+        psraw       xmm4,           3
+
+        movq        xmm0,           [rax]
+        movq        xmm1,           [rax+rcx]
+        movq        xmm2,           [rax+2*rcx]
+        lea         rcx,            [3*rcx]
+        movq        xmm3,           [rax+rcx]
+
+        punpcklbw   xmm0,           xmm7
+        punpcklbw   xmm1,           xmm7
+        punpcklbw   xmm2,           xmm7
+        punpcklbw   xmm3,           xmm7
+
+        mov         rax,            arg(3) ; dst
+        movsxd      rdx,            dword ptr arg(4) ; dst_stride
+
+    ; Add to predict buffer
+        paddw       xmm0,           xmm4
+        paddw       xmm1,           xmm4
+        paddw       xmm2,           xmm4
+        paddw       xmm3,           xmm4
+
+    ; pack up before storing
+        packuswb    xmm0,           xmm7
+        packuswb    xmm1,           xmm7
+        packuswb    xmm2,           xmm7
+        packuswb    xmm3,           xmm7
+
+    ; store blocks back out
+        movq        [rax],          xmm0
+        movq        [rax + rdx],    xmm1
+
+        lea         rax,            [rax + 2*rdx]
+
+        movq        [rax],          xmm2
+        movq        [rax + rdx],    xmm3
+
+    ; begin epilog
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+global sym(idct_dequant_full_2x_sse2)
+sym(idct_dequant_full_2x_sse2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 7
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; special case when 2 blocks have 0 or 1 coeffs
+    ; dc is set as first coeff, so no need to load qcoeff
+        mov         rax,            arg(0) ; qcoeff
+        mov         rsi,            arg(2) ; pre
+        mov         rdi,            arg(3) ; dst
+        movsxd      rcx,            dword ptr arg(5) ; blk_stride
+
+    ; Zero out xmm7, for use unpacking
+        pxor        xmm7,           xmm7
+
+        mov         rdx,            arg(1)  ; dequant
+
+    ; note the transpose of xmm1 and xmm2, necessary for shuffle
+    ;   to spit out sensicle data
+        movdqa      xmm0,           [rax]
+        movdqa      xmm2,           [rax+16]
+        movdqa      xmm1,           [rax+32]
+        movdqa      xmm3,           [rax+48]
+
+    ; Clear out coeffs
+        movdqa      [rax],          xmm7
+        movdqa      [rax+16],       xmm7
+        movdqa      [rax+32],       xmm7
+        movdqa      [rax+48],       xmm7
+
+    ; dequantize qcoeff buffer
+        pmullw      xmm0,           [rdx]
+        pmullw      xmm2,           [rdx+16]
+        pmullw      xmm1,           [rdx]
+        pmullw      xmm3,           [rdx+16]
+
+    ; repack so block 0 row x and block 1 row x are together
+        movdqa      xmm4,           xmm0
+        punpckldq   xmm0,           xmm1
+        punpckhdq   xmm4,           xmm1
+
+        pshufd      xmm0,           xmm0,       11011000b
+        pshufd      xmm1,           xmm4,       11011000b
+
+        movdqa      xmm4,           xmm2
+        punpckldq   xmm2,           xmm3
+        punpckhdq   xmm4,           xmm3
+
+        pshufd      xmm2,           xmm2,       11011000b
+        pshufd      xmm3,           xmm4,       11011000b
+
+    ; first pass
+        psubw       xmm0,           xmm2        ; b1 = 0-2
+        paddw       xmm2,           xmm2        ;
+
+        movdqa      xmm5,           xmm1
+        paddw       xmm2,           xmm0        ; a1 = 0+2
+
+        pmulhw      xmm5,           [x_s1sqr2 GLOBAL]
+        paddw       xmm5,           xmm1        ; ip1 * sin(pi/8) * sqrt(2)
+
+        movdqa      xmm7,           xmm3
+        pmulhw      xmm7,           [x_c1sqr2less1 GLOBAL]
+
+        paddw       xmm7,           xmm3        ; ip3 * cos(pi/8) * sqrt(2)
+        psubw       xmm7,           xmm5        ; c1
+
+        movdqa      xmm5,           xmm1
+        movdqa      xmm4,           xmm3
+
+        pmulhw      xmm5,           [x_c1sqr2less1 GLOBAL]
+        paddw       xmm5,           xmm1
+
+        pmulhw      xmm3,           [x_s1sqr2 GLOBAL]
+        paddw       xmm3,           xmm4
+
+        paddw       xmm3,           xmm5        ; d1
+        movdqa      xmm6,           xmm2        ; a1
+
+        movdqa      xmm4,           xmm0        ; b1
+        paddw       xmm2,           xmm3        ;0
+
+        paddw       xmm4,           xmm7        ;1
+        psubw       xmm0,           xmm7        ;2
+
+        psubw       xmm6,           xmm3        ;3
+
+    ; transpose for the second pass
+        movdqa      xmm7,           xmm2        ; 103 102 101 100 003 002 001 000
+        punpcklwd   xmm2,           xmm0        ; 007 003 006 002 005 001 004 000
+        punpckhwd   xmm7,           xmm0        ; 107 103 106 102 105 101 104 100
+
+        movdqa      xmm5,           xmm4        ; 111 110 109 108 011 010 009 008
+        punpcklwd   xmm4,           xmm6        ; 015 011 014 010 013 009 012 008
+        punpckhwd   xmm5,           xmm6        ; 115 111 114 110 113 109 112 108
+
+
+        movdqa      xmm1,           xmm2        ; 007 003 006 002 005 001 004 000
+        punpckldq   xmm2,           xmm4        ; 013 009 005 001 012 008 004 000
+        punpckhdq   xmm1,           xmm4        ; 015 011 007 003 014 010 006 002
+
+        movdqa      xmm6,           xmm7        ; 107 103 106 102 105 101 104 100
+        punpckldq   xmm7,           xmm5        ; 113 109 105 101 112 108 104 100
+        punpckhdq   xmm6,           xmm5        ; 115 111 107 103 114 110 106 102
+
+
+        movdqa      xmm5,           xmm2        ; 013 009 005 001 012 008 004 000
+        punpckldq   xmm2,           xmm7        ; 112 108 012 008 104 100 004 000
+        punpckhdq   xmm5,           xmm7        ; 113 109 013 009 105 101 005 001
+
+        movdqa      xmm7,           xmm1        ; 015 011 007 003 014 010 006 002
+        punpckldq   xmm1,           xmm6        ; 114 110 014 010 106 102 006 002
+        punpckhdq   xmm7,           xmm6        ; 115 111 015 011 107 103 007 003
+
+        pshufd      xmm0,           xmm2,       11011000b
+        pshufd      xmm2,           xmm1,       11011000b
+
+        pshufd      xmm1,           xmm5,       11011000b
+        pshufd      xmm3,           xmm7,       11011000b
+
+    ; second pass
+        psubw       xmm0,           xmm2            ; b1 = 0-2
+        paddw       xmm2,           xmm2
+
+        movdqa      xmm5,           xmm1
+        paddw       xmm2,           xmm0            ; a1 = 0+2
+
+        pmulhw      xmm5,           [x_s1sqr2 GLOBAL]
+        paddw       xmm5,           xmm1            ; ip1 * sin(pi/8) * sqrt(2)
+
+        movdqa      xmm7,           xmm3
+        pmulhw      xmm7,           [x_c1sqr2less1 GLOBAL]
+
+        paddw       xmm7,           xmm3            ; ip3 * cos(pi/8) * sqrt(2)
+        psubw       xmm7,           xmm5            ; c1
+
+        movdqa      xmm5,           xmm1
+        movdqa      xmm4,           xmm3
+
+        pmulhw      xmm5,           [x_c1sqr2less1 GLOBAL]
+        paddw       xmm5,           xmm1
+
+        pmulhw      xmm3,           [x_s1sqr2 GLOBAL]
+        paddw       xmm3,           xmm4
+
+        paddw       xmm3,           xmm5            ; d1
+        paddw       xmm0,           [fours GLOBAL]
+
+        paddw       xmm2,           [fours GLOBAL]
+        movdqa      xmm6,           xmm2            ; a1
+
+        movdqa      xmm4,           xmm0            ; b1
+        paddw       xmm2,           xmm3            ;0
+
+        paddw       xmm4,           xmm7            ;1
+        psubw       xmm0,           xmm7            ;2
+
+        psubw       xmm6,           xmm3            ;3
+        psraw       xmm2,           3
+
+        psraw       xmm0,           3
+        psraw       xmm4,           3
+
+        psraw       xmm6,           3
+
+    ; transpose to save
+        movdqa      xmm7,           xmm2        ; 103 102 101 100 003 002 001 000
+        punpcklwd   xmm2,           xmm0        ; 007 003 006 002 005 001 004 000
+        punpckhwd   xmm7,           xmm0        ; 107 103 106 102 105 101 104 100
+
+        movdqa      xmm5,           xmm4        ; 111 110 109 108 011 010 009 008
+        punpcklwd   xmm4,           xmm6        ; 015 011 014 010 013 009 012 008
+        punpckhwd   xmm5,           xmm6        ; 115 111 114 110 113 109 112 108
+
+
+        movdqa      xmm1,           xmm2        ; 007 003 006 002 005 001 004 000
+        punpckldq   xmm2,           xmm4        ; 013 009 005 001 012 008 004 000
+        punpckhdq   xmm1,           xmm4        ; 015 011 007 003 014 010 006 002
+
+        movdqa      xmm6,           xmm7        ; 107 103 106 102 105 101 104 100
+        punpckldq   xmm7,           xmm5        ; 113 109 105 101 112 108 104 100
+        punpckhdq   xmm6,           xmm5        ; 115 111 107 103 114 110 106 102
+
+
+        movdqa      xmm5,           xmm2        ; 013 009 005 001 012 008 004 000
+        punpckldq   xmm2,           xmm7        ; 112 108 012 008 104 100 004 000
+        punpckhdq   xmm5,           xmm7        ; 113 109 013 009 105 101 005 001
+
+        movdqa      xmm7,           xmm1        ; 015 011 007 003 014 010 006 002
+        punpckldq   xmm1,           xmm6        ; 114 110 014 010 106 102 006 002
+        punpckhdq   xmm7,           xmm6        ; 115 111 015 011 107 103 007 003
+
+        pshufd      xmm0,           xmm2,       11011000b
+        pshufd      xmm2,           xmm1,       11011000b
+
+        pshufd      xmm1,           xmm5,       11011000b
+        pshufd      xmm3,           xmm7,       11011000b
+
+        pxor        xmm7,           xmm7
+
+    ; Load up predict blocks
+        movq        xmm4,           [rsi]
+        movq        xmm5,           [rsi+rcx]
+
+        punpcklbw   xmm4,           xmm7
+        punpcklbw   xmm5,           xmm7
+
+        paddw       xmm0,           xmm4
+        paddw       xmm1,           xmm5
+
+        movq        xmm4,           [rsi+2*rcx]
+        lea         rcx,            [3*rcx]
+        movq        xmm5,           [rsi+rcx]
+
+        punpcklbw   xmm4,           xmm7
+        punpcklbw   xmm5,           xmm7
+
+        paddw       xmm2,           xmm4
+        paddw       xmm3,           xmm5
+
+.finish:
+
+    ; pack up before storing
+        packuswb    xmm0,           xmm7
+        packuswb    xmm1,           xmm7
+        packuswb    xmm2,           xmm7
+        packuswb    xmm3,           xmm7
+
+    ; Load destination stride before writing out,
+    ;   doesn't need to persist
+        movsxd      rdx,            dword ptr arg(4) ; dst_stride
+
+    ; store blocks back out
+        movq        [rdi],          xmm0
+        movq        [rdi + rdx],    xmm1
+
+        lea         rdi,            [rdi + 2*rdx]
+
+        movq        [rdi],          xmm2
+        movq        [rdi + rdx],    xmm3
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void idct_dequant_dc_0_2x_sse2
+; (
+;   short *qcoeff       - 0
+;   short *dequant      - 1
+;   unsigned char *pre  - 2
+;   unsigned char *dst  - 3
+;   int dst_stride      - 4
+;   short *dc           - 5
+; )
+global sym(idct_dequant_dc_0_2x_sse2)
+sym(idct_dequant_dc_0_2x_sse2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 7
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; special case when 2 blocks have 0 or 1 coeffs
+    ; dc is set as first coeff, so no need to load qcoeff
+        mov         rax,            arg(0) ; qcoeff
+        mov         rsi,            arg(2) ; pre
+        mov         rdi,            arg(3) ; dst
+        mov         rdx,            arg(5) ; dc
+
+    ; Zero out xmm7, for use unpacking
+        pxor        xmm7,           xmm7
+
+    ; load up 2 dc words here == 2*16 = doubleword
+        movd        xmm4,           [rdx]
+
+    ; Load up predict blocks
+        movq        xmm0,           [rsi]
+        movq        xmm1,           [rsi+16]
+        movq        xmm2,           [rsi+32]
+        movq        xmm3,           [rsi+48]
+
+    ; Duplicate and expand dc across
+        punpcklwd   xmm4,           xmm4
+        punpckldq   xmm4,           xmm4
+
+    ; Rounding to dequant and downshift
+        paddw       xmm4,           [fours GLOBAL]
+        psraw       xmm4,           3
+
+    ; Predict buffer needs to be expanded from bytes to words
+        punpcklbw   xmm0,           xmm7
+        punpcklbw   xmm1,           xmm7
+        punpcklbw   xmm2,           xmm7
+        punpcklbw   xmm3,           xmm7
+
+    ; Add to predict buffer
+        paddw       xmm0,           xmm4
+        paddw       xmm1,           xmm4
+        paddw       xmm2,           xmm4
+        paddw       xmm3,           xmm4
+
+    ; pack up before storing
+        packuswb    xmm0,           xmm7
+        packuswb    xmm1,           xmm7
+        packuswb    xmm2,           xmm7
+        packuswb    xmm3,           xmm7
+
+    ; Load destination stride before writing out,
+    ;   doesn't need to persist
+        movsxd      rdx,            dword ptr arg(4) ; dst_stride
+
+    ; store blocks back out
+        movq        [rdi],          xmm0
+        movq        [rdi + rdx],    xmm1
+
+        lea         rdi,            [rdi + 2*rdx]
+
+        movq        [rdi],          xmm2
+        movq        [rdi + rdx],    xmm3
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+global sym(idct_dequant_dc_full_2x_sse2)
+sym(idct_dequant_dc_full_2x_sse2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 7
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; special case when 2 blocks have 0 or 1 coeffs
+    ; dc is set as first coeff, so no need to load qcoeff
+        mov         rax,            arg(0) ; qcoeff
+        mov         rsi,            arg(2) ; pre
+        mov         rdi,            arg(3) ; dst
+
+    ; Zero out xmm7, for use unpacking
+        pxor        xmm7,           xmm7
+
+        mov         rdx,            arg(1)  ; dequant
+
+    ; note the transpose of xmm1 and xmm2, necessary for shuffle
+    ;   to spit out sensicle data
+        movdqa      xmm0,           [rax]
+        movdqa      xmm2,           [rax+16]
+        movdqa      xmm1,           [rax+32]
+        movdqa      xmm3,           [rax+48]
+
+    ; Clear out coeffs
+        movdqa      [rax],          xmm7
+        movdqa      [rax+16],       xmm7
+        movdqa      [rax+32],       xmm7
+        movdqa      [rax+48],       xmm7
+
+    ; dequantize qcoeff buffer
+        pmullw      xmm0,           [rdx]
+        pmullw      xmm2,           [rdx+16]
+        pmullw      xmm1,           [rdx]
+        pmullw      xmm3,           [rdx+16]
+
+    ; DC component
+        mov         rdx,            arg(5)
+
+    ; repack so block 0 row x and block 1 row x are together
+        movdqa      xmm4,           xmm0
+        punpckldq   xmm0,           xmm1
+        punpckhdq   xmm4,           xmm1
+
+        pshufd      xmm0,           xmm0,       11011000b
+        pshufd      xmm1,           xmm4,       11011000b
+
+        movdqa      xmm4,           xmm2
+        punpckldq   xmm2,           xmm3
+        punpckhdq   xmm4,           xmm3
+
+        pshufd      xmm2,           xmm2,       11011000b
+        pshufd      xmm3,           xmm4,       11011000b
+
+    ; insert DC component
+        pinsrw      xmm0,           [rdx],      0
+        pinsrw      xmm0,           [rdx+2],    4
+
+    ; first pass
+        psubw       xmm0,           xmm2        ; b1 = 0-2
+        paddw       xmm2,           xmm2        ;
+
+        movdqa      xmm5,           xmm1
+        paddw       xmm2,           xmm0        ; a1 = 0+2
+
+        pmulhw      xmm5,           [x_s1sqr2 GLOBAL]
+        paddw       xmm5,           xmm1        ; ip1 * sin(pi/8) * sqrt(2)
+
+        movdqa      xmm7,           xmm3
+        pmulhw      xmm7,           [x_c1sqr2less1 GLOBAL]
+
+        paddw       xmm7,           xmm3        ; ip3 * cos(pi/8) * sqrt(2)
+        psubw       xmm7,           xmm5        ; c1
+
+        movdqa      xmm5,           xmm1
+        movdqa      xmm4,           xmm3
+
+        pmulhw      xmm5,           [x_c1sqr2less1 GLOBAL]
+        paddw       xmm5,           xmm1
+
+        pmulhw      xmm3,           [x_s1sqr2 GLOBAL]
+        paddw       xmm3,           xmm4
+
+        paddw       xmm3,           xmm5        ; d1
+        movdqa      xmm6,           xmm2        ; a1
+
+        movdqa      xmm4,           xmm0        ; b1
+        paddw       xmm2,           xmm3        ;0
+
+        paddw       xmm4,           xmm7        ;1
+        psubw       xmm0,           xmm7        ;2
+
+        psubw       xmm6,           xmm3        ;3
+
+    ; transpose for the second pass
+        movdqa      xmm7,           xmm2        ; 103 102 101 100 003 002 001 000
+        punpcklwd   xmm2,           xmm0        ; 007 003 006 002 005 001 004 000
+        punpckhwd   xmm7,           xmm0        ; 107 103 106 102 105 101 104 100
+
+        movdqa      xmm5,           xmm4        ; 111 110 109 108 011 010 009 008
+        punpcklwd   xmm4,           xmm6        ; 015 011 014 010 013 009 012 008
+        punpckhwd   xmm5,           xmm6        ; 115 111 114 110 113 109 112 108
+
+
+        movdqa      xmm1,           xmm2        ; 007 003 006 002 005 001 004 000
+        punpckldq   xmm2,           xmm4        ; 013 009 005 001 012 008 004 000
+        punpckhdq   xmm1,           xmm4        ; 015 011 007 003 014 010 006 002
+
+        movdqa      xmm6,           xmm7        ; 107 103 106 102 105 101 104 100
+        punpckldq   xmm7,           xmm5        ; 113 109 105 101 112 108 104 100
+        punpckhdq   xmm6,           xmm5        ; 115 111 107 103 114 110 106 102
+
+
+        movdqa      xmm5,           xmm2        ; 013 009 005 001 012 008 004 000
+        punpckldq   xmm2,           xmm7        ; 112 108 012 008 104 100 004 000
+        punpckhdq   xmm5,           xmm7        ; 113 109 013 009 105 101 005 001
+
+        movdqa      xmm7,           xmm1        ; 015 011 007 003 014 010 006 002
+        punpckldq   xmm1,           xmm6        ; 114 110 014 010 106 102 006 002
+        punpckhdq   xmm7,           xmm6        ; 115 111 015 011 107 103 007 003
+
+        pshufd      xmm0,           xmm2,       11011000b
+        pshufd      xmm2,           xmm1,       11011000b
+
+        pshufd      xmm1,           xmm5,       11011000b
+        pshufd      xmm3,           xmm7,       11011000b
+
+    ; second pass
+        psubw       xmm0,           xmm2            ; b1 = 0-2
+        paddw       xmm2,           xmm2
+
+        movdqa      xmm5,           xmm1
+        paddw       xmm2,           xmm0            ; a1 = 0+2
+
+        pmulhw      xmm5,           [x_s1sqr2 GLOBAL]
+        paddw       xmm5,           xmm1            ; ip1 * sin(pi/8) * sqrt(2)
+
+        movdqa      xmm7,           xmm3
+        pmulhw      xmm7,           [x_c1sqr2less1 GLOBAL]
+
+        paddw       xmm7,           xmm3            ; ip3 * cos(pi/8) * sqrt(2)
+        psubw       xmm7,           xmm5            ; c1
+
+        movdqa      xmm5,           xmm1
+        movdqa      xmm4,           xmm3
+
+        pmulhw      xmm5,           [x_c1sqr2less1 GLOBAL]
+        paddw       xmm5,           xmm1
+
+        pmulhw      xmm3,           [x_s1sqr2 GLOBAL]
+        paddw       xmm3,           xmm4
+
+        paddw       xmm3,           xmm5            ; d1
+        paddw       xmm0,           [fours GLOBAL]
+
+        paddw       xmm2,           [fours GLOBAL]
+        movdqa      xmm6,           xmm2            ; a1
+
+        movdqa      xmm4,           xmm0            ; b1
+        paddw       xmm2,           xmm3            ;0
+
+        paddw       xmm4,           xmm7            ;1
+        psubw       xmm0,           xmm7            ;2
+
+        psubw       xmm6,           xmm3            ;3
+        psraw       xmm2,           3
+
+        psraw       xmm0,           3
+        psraw       xmm4,           3
+
+        psraw       xmm6,           3
+
+    ; transpose to save
+        movdqa      xmm7,           xmm2        ; 103 102 101 100 003 002 001 000
+        punpcklwd   xmm2,           xmm0        ; 007 003 006 002 005 001 004 000
+        punpckhwd   xmm7,           xmm0        ; 107 103 106 102 105 101 104 100
+
+        movdqa      xmm5,           xmm4        ; 111 110 109 108 011 010 009 008
+        punpcklwd   xmm4,           xmm6        ; 015 011 014 010 013 009 012 008
+        punpckhwd   xmm5,           xmm6        ; 115 111 114 110 113 109 112 108
+
+
+        movdqa      xmm1,           xmm2        ; 007 003 006 002 005 001 004 000
+        punpckldq   xmm2,           xmm4        ; 013 009 005 001 012 008 004 000
+        punpckhdq   xmm1,           xmm4        ; 015 011 007 003 014 010 006 002
+
+        movdqa      xmm6,           xmm7        ; 107 103 106 102 105 101 104 100
+        punpckldq   xmm7,           xmm5        ; 113 109 105 101 112 108 104 100
+        punpckhdq   xmm6,           xmm5        ; 115 111 107 103 114 110 106 102
+
+
+        movdqa      xmm5,           xmm2        ; 013 009 005 001 012 008 004 000
+        punpckldq   xmm2,           xmm7        ; 112 108 012 008 104 100 004 000
+        punpckhdq   xmm5,           xmm7        ; 113 109 013 009 105 101 005 001
+
+        movdqa      xmm7,           xmm1        ; 015 011 007 003 014 010 006 002
+        punpckldq   xmm1,           xmm6        ; 114 110 014 010 106 102 006 002
+        punpckhdq   xmm7,           xmm6        ; 115 111 015 011 107 103 007 003
+
+        pshufd      xmm0,           xmm2,       11011000b
+        pshufd      xmm2,           xmm1,       11011000b
+
+        pshufd      xmm1,           xmm5,       11011000b
+        pshufd      xmm3,           xmm7,       11011000b
+
+        pxor        xmm7,           xmm7
+
+    ; Load up predict blocks
+        movq        xmm4,           [rsi]
+        movq        xmm5,           [rsi+16]
+
+        punpcklbw   xmm4,           xmm7
+        punpcklbw   xmm5,           xmm7
+
+        paddw       xmm0,           xmm4
+        paddw       xmm1,           xmm5
+
+        movq        xmm4,           [rsi+32]
+        movq        xmm5,           [rsi+48]
+
+        punpcklbw   xmm4,           xmm7
+        punpcklbw   xmm5,           xmm7
+
+        paddw       xmm2,           xmm4
+        paddw       xmm3,           xmm5
+
+.finish:
+
+    ; pack up before storing
+        packuswb    xmm0,           xmm7
+        packuswb    xmm1,           xmm7
+        packuswb    xmm2,           xmm7
+        packuswb    xmm3,           xmm7
+
+    ; Load destination stride before writing out,
+    ;   doesn't need to persist
+        movsxd      rdx,            dword ptr arg(4) ; dst_stride
+
+    ; store blocks back out
+        movq        [rdi],          xmm0
+        movq        [rdi + rdx],    xmm1
+
+        lea         rdi,            [rdi + 2*rdx]
+
+        movq        [rdi],          xmm2
+        movq        [rdi + rdx],    xmm3
+
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+SECTION_RODATA
+align 16
+fours:
+    times 8 dw 0x0004
+align 16
+x_s1sqr2:
+    times 8 dw 0x8A8C
+align 16
+x_c1sqr2less1:
+    times 8 dw 0x4E7B

vp8/common/x86/loopfilter_x86.c

@@ -34,6 +34,11 @@ prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_sse2);
 prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_sse2);
 prototype_loopfilter(vp8_fast_loop_filter_vertical_edges_sse2);
 
+extern loop_filter_uvfunction vp8_loop_filter_horizontal_edge_uv_sse2;
+extern loop_filter_uvfunction vp8_loop_filter_vertical_edge_uv_sse2;
+extern loop_filter_uvfunction vp8_mbloop_filter_horizontal_edge_uv_sse2;
+extern loop_filter_uvfunction vp8_mbloop_filter_vertical_edge_uv_sse2;
+
 #if HAVE_MMX
 // Horizontal MB filtering
 void vp8_loop_filter_mbh_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -157,10 +162,7 @@ void vp8_loop_filter_mbh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsign
     vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-
-    if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
 }
 
 
@@ -183,10 +185,7 @@ void vp8_loop_filter_mbv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsign
     vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+        vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
-
-    if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
 }
 
 
@@ -211,10 +210,7 @@ void vp8_loop_filter_bh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigne
     vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4 * uv_stride);
-
-    if (v_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
 }
 
 
@@ -241,10 +237,7 @@ void vp8_loop_filter_bv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigne
     vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+        vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4);
-
-    if (v_ptr)
-        vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
 }
 
 

vp8/common/x86/subpixel_mmx.asm

@@ -731,7 +731,7 @@ rd:
     times 4 dw 0x40
 
 align 16
-global sym(vp8_six_tap_mmx) HIDDEN_DATA
+global HIDDEN_DATA(sym(vp8_six_tap_mmx))
 sym(vp8_six_tap_mmx):
     times 8 dw 0
     times 8 dw 0
@@ -791,7 +791,7 @@ sym(vp8_six_tap_mmx):
 
 
 align 16
-global sym(vp8_bilinear_filters_mmx) HIDDEN_DATA
+global HIDDEN_DATA(sym(vp8_bilinear_filters_mmx))
 sym(vp8_bilinear_filters_mmx):
     times 8 dw 128
     times 8 dw 0

vp8/common/x86/subpixel_ssse3.asm

@@ -0,0 +1,931 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license
+;  that can be found in the LICENSE file in the root of the source
+;  tree. An additional intellectual property rights grant can be found
+;  in the file PATENTS.  All contributing project authors may
+;  be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%define BLOCK_HEIGHT_WIDTH 4
+%define VP8_FILTER_WEIGHT 128
+%define VP8_FILTER_SHIFT  7
+
+
+;/************************************************************************************
+; Notes: filter_block1d_h6 applies a 6 tap filter horizontally to the input pixels. The
+; input pixel array has output_height rows. This routine assumes that output_height is an
+; even number. This function handles 8 pixels in horizontal direction, calculating ONE
+; rows each iteration to take advantage of the 128 bits operations.
+;
+; This is an implementation of some of the SSE optimizations first seen in ffvp8
+;
+;*************************************************************************************/
+;void vp8_filter_block1d8_h6_ssse3
+;(
+;    unsigned char  *src_ptr,
+;    unsigned int    src_pixels_per_line,
+;    unsigned char *output_ptr,
+;    unsigned int    output_pitch,
+;    unsigned int    output_height,
+;    unsigned int    vp8_filter_index
+;)
+global sym(vp8_filter_block1d8_h6_ssse3)
+sym(vp8_filter_block1d8_h6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4
+
+    movdqa      xmm7, [rd GLOBAL]
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+    mov         rdi, arg(2)             ;output_ptr
+
+    cmp         esi, DWORD PTR [rax]
+    je          vp8_filter_block1d8_h4_ssse3
+
+    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+    sub         rdi, rdx
+;xmm3 free
+filter_block1d8_h6_rowloop_ssse3:
+    movdqu      xmm0,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm1, xmm0
+    pshufb      xmm0, [shuf1b GLOBAL]
+
+    movdqa      xmm2, xmm1
+    pshufb      xmm1, [shuf2b GLOBAL]
+    pmaddubsw   xmm0, xmm4
+    pmaddubsw   xmm1, xmm5
+
+    pshufb      xmm2, [shuf3b GLOBAL]
+    add         rdi, rdx
+    pmaddubsw   xmm2, xmm6
+
+    lea         rsi,    [rsi + rax]
+    dec         rcx
+    paddsw      xmm0, xmm1
+    paddsw      xmm0, xmm7
+    paddsw      xmm0, xmm2
+    psraw       xmm0, 7
+    packuswb    xmm0, xmm0
+
+    movq        MMWORD Ptr [rdi], xmm0
+    jnz         filter_block1d8_h6_rowloop_ssse3
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d8_h4_ssse3:
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+
+    movdqa      xmm3, XMMWORD PTR [shuf2b GLOBAL]
+    movdqa      xmm4, XMMWORD PTR [shuf3b GLOBAL]
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+    sub         rdi, rdx
+;xmm3 free
+filter_block1d8_h4_rowloop_ssse3:
+    movdqu      xmm0,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm2, xmm0
+    pshufb      xmm0, xmm3 ;[shuf2b GLOBAL]
+    pshufb      xmm2, xmm4 ;[shuf3b GLOBAL]
+
+    pmaddubsw   xmm0, xmm5
+    add         rdi, rdx
+    pmaddubsw   xmm2, xmm6
+
+    lea         rsi,    [rsi + rax]
+    dec         rcx
+    paddsw      xmm0, xmm7
+    paddsw      xmm0, xmm2
+    psraw       xmm0, 7
+    packuswb    xmm0, xmm0
+
+    movq        MMWORD Ptr [rdi], xmm0
+
+    jnz         filter_block1d8_h4_rowloop_ssse3
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+;void vp8_filter_block1d16_h6_ssse3
+;(
+;    unsigned char  *src_ptr,
+;    unsigned int    src_pixels_per_line,
+;    unsigned char  *output_ptr,
+;    unsigned int    output_pitch,
+;    unsigned int    output_height,
+;    unsigned int    vp8_filter_index
+;)
+global sym(vp8_filter_block1d16_h6_ssse3)
+sym(vp8_filter_block1d16_h6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4      ;
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+
+    mov         rdi, arg(2)             ;output_ptr
+    movdqa      xmm7, [rd GLOBAL]
+
+;;
+;;    cmp         esi, DWORD PTR [rax]
+;;    je          vp8_filter_block1d16_h4_ssse3
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+filter_block1d16_h6_rowloop_ssse3:
+    movdqu      xmm0,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm1, xmm0
+    pshufb      xmm0, [shuf1b GLOBAL]
+    movdqa      xmm2, xmm1
+    pmaddubsw   xmm0, xmm4
+    pshufb      xmm1, [shuf2b GLOBAL]
+    pshufb      xmm2, [shuf3b GLOBAL]
+    pmaddubsw   xmm1, xmm5
+
+    movdqu      xmm3,   XMMWORD PTR [rsi + 6]
+
+    pmaddubsw   xmm2, xmm6
+    paddsw      xmm0, xmm1
+    movdqa      xmm1, xmm3
+    pshufb      xmm3, [shuf1b GLOBAL]
+    paddsw      xmm0, xmm7
+    pmaddubsw   xmm3, xmm4
+    paddsw      xmm0, xmm2
+    movdqa      xmm2, xmm1
+    pshufb      xmm1, [shuf2b GLOBAL]
+    pshufb      xmm2, [shuf3b GLOBAL]
+    pmaddubsw   xmm1, xmm5
+    pmaddubsw   xmm2, xmm6
+
+    psraw       xmm0, 7
+    packuswb    xmm0, xmm0
+    lea         rsi,    [rsi + rax]
+    paddsw      xmm3, xmm1
+    paddsw      xmm3, xmm7
+    paddsw      xmm3, xmm2
+    psraw       xmm3, 7
+    packuswb    xmm3, xmm3
+
+    punpcklqdq  xmm0, xmm3
+
+    movdqa      XMMWORD Ptr [rdi], xmm0
+
+    add         rdi, rdx
+    dec         rcx
+    jnz         filter_block1d16_h6_rowloop_ssse3
+
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d16_h4_ssse3:
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+filter_block1d16_h4_rowloop_ssse3:
+    movdqu      xmm1,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm2, xmm1
+    pshufb      xmm1, [shuf2b GLOBAL]
+    pshufb      xmm2, [shuf3b GLOBAL]
+    pmaddubsw   xmm1, xmm5
+
+    movdqu      xmm3,   XMMWORD PTR [rsi + 6]
+
+    pmaddubsw   xmm2, xmm6
+    movdqa      xmm0, xmm3
+    pshufb      xmm3, [shuf3b GLOBAL]
+    pshufb      xmm0, [shuf2b GLOBAL]
+
+    paddsw      xmm1, xmm7
+    paddsw      xmm1, xmm2
+
+    pmaddubsw   xmm0, xmm5
+    pmaddubsw   xmm3, xmm6
+
+    psraw       xmm1, 7
+    packuswb    xmm1, xmm1
+    lea         rsi,    [rsi + rax]
+    paddsw      xmm3, xmm0
+    paddsw      xmm3, xmm7
+    psraw       xmm3, 7
+    packuswb    xmm3, xmm3
+
+    punpcklqdq  xmm1, xmm3
+
+    movdqa      XMMWORD Ptr [rdi], xmm1
+
+    add         rdi, rdx
+    dec         rcx
+    jnz         filter_block1d16_h4_rowloop_ssse3
+
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_filter_block1d4_h6_ssse3
+;(
+;    unsigned char  *src_ptr,
+;    unsigned int    src_pixels_per_line,
+;    unsigned char  *output_ptr,
+;    unsigned int    output_pitch,
+;    unsigned int    output_height,
+;    unsigned int    vp8_filter_index
+;)
+global sym(vp8_filter_block1d4_h6_ssse3)
+sym(vp8_filter_block1d4_h6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4      ;
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+    movdqa      xmm7, [rd GLOBAL]
+
+    cmp         esi, DWORD PTR [rax]
+    je          vp8_filter_block1d4_h4_ssse3
+
+    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+    mov         rdi, arg(2)             ;output_ptr
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+;xmm3 free
+filter_block1d4_h6_rowloop_ssse3:
+    movdqu      xmm0,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm1, xmm0
+    pshufb      xmm0, [shuf1b GLOBAL]
+
+    movdqa      xmm2, xmm1
+    pshufb      xmm1, [shuf2b GLOBAL]
+    pmaddubsw   xmm0, xmm4
+    pshufb      xmm2, [shuf3b GLOBAL]
+    pmaddubsw   xmm1, xmm5
+
+;--
+    pmaddubsw   xmm2, xmm6
+
+    lea         rsi,    [rsi + rax]
+;--
+    paddsw      xmm0, xmm1
+    paddsw      xmm0, xmm7
+    pxor        xmm1, xmm1
+    paddsw      xmm0, xmm2
+    psraw       xmm0, 7
+    packuswb    xmm0, xmm0
+
+    movd        DWORD PTR [rdi], xmm0
+
+    add         rdi, rdx
+    dec         rcx
+    jnz         filter_block1d4_h6_rowloop_ssse3
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d4_h4_ssse3:
+    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
+    movdqa      xmm0, XMMWORD PTR [shuf2b GLOBAL]
+    movdqa      xmm3, XMMWORD PTR [shuf3b GLOBAL]
+
+    mov         rsi, arg(0)             ;src_ptr
+    mov         rdi, arg(2)             ;output_ptr
+    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
+    movsxd      rcx, dword ptr arg(4)   ;output_height
+
+    movsxd      rdx, dword ptr arg(3)   ;output_pitch
+
+filter_block1d4_h4_rowloop_ssse3:
+    movdqu      xmm1,   XMMWORD PTR [rsi - 2]
+
+    movdqa      xmm2, xmm1
+    pshufb      xmm1, xmm0 ;;[shuf2b GLOBAL]
+    pshufb      xmm2, xmm3 ;;[shuf3b GLOBAL]
+    pmaddubsw   xmm1, xmm5
+
+;--
+    pmaddubsw   xmm2, xmm6
+
+    lea         rsi,    [rsi + rax]
+;--
+    paddsw      xmm1, xmm7
+    paddsw      xmm1, xmm2
+    psraw       xmm1, 7
+    packuswb    xmm1, xmm1
+
+    movd        DWORD PTR [rdi], xmm1
+
+    add         rdi, rdx
+    dec         rcx
+    jnz         filter_block1d4_h4_rowloop_ssse3
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+
+
+;void vp8_filter_block1d16_v6_ssse3
+;(
+;    unsigned char *src_ptr,
+;    unsigned int   src_pitch,
+;    unsigned char *output_ptr,
+;    unsigned int   out_pitch,
+;    unsigned int   output_height,
+;    unsigned int   vp8_filter_index
+;)
+global sym(vp8_filter_block1d16_v6_ssse3)
+sym(vp8_filter_block1d16_v6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4      ;
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+
+    cmp         esi, DWORD PTR [rax]
+    je          vp8_filter_block1d16_v4_ssse3
+
+    movdqa      xmm5, XMMWORD PTR [rax]         ;k0_k5
+    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
+    mov         rdi, arg(2)             ;output_ptr
+
+%if ABI_IS_32BIT=0
+    movsxd      r8, DWORD PTR arg(3)    ;out_pitch
+%endif
+    mov         rax, rsi
+    movsxd      rcx, DWORD PTR arg(4)   ;output_height
+    add         rax, rdx
+
+
+vp8_filter_block1d16_v6_ssse3_loop:
+    movq        xmm1, MMWORD PTR [rsi]                  ;A
+    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
+    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   xmm2, xmm4                  ;B D
+    punpcklbw   xmm3, xmm0                  ;C E
+
+    movq        xmm0, MMWORD PTR [rax + rdx * 4]        ;F
+
+    pmaddubsw   xmm3, xmm6
+    punpcklbw   xmm1, xmm0                  ;A F
+    pmaddubsw   xmm2, xmm7
+    pmaddubsw   xmm1, xmm5
+
+    paddsw      xmm2, xmm3
+    paddsw      xmm2, xmm1
+    paddsw      xmm2, [rd GLOBAL]
+    psraw       xmm2, 7
+    packuswb    xmm2, xmm2
+
+    movq        MMWORD PTR [rdi], xmm2          ;store the results
+
+    movq        xmm1, MMWORD PTR [rsi + 8]                  ;A
+    movq        xmm2, MMWORD PTR [rsi + rdx + 8]            ;B
+    movq        xmm3, MMWORD PTR [rsi + rdx * 2 + 8]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2 + 8]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4 + 8]        ;E
+
+    punpcklbw   xmm2, xmm4                  ;B D
+    punpcklbw   xmm3, xmm0                  ;C E
+
+    movq        xmm0, MMWORD PTR [rax + rdx * 4 + 8]        ;F
+    pmaddubsw   xmm3, xmm6
+    punpcklbw   xmm1, xmm0                  ;A F
+    pmaddubsw   xmm2, xmm7
+    pmaddubsw   xmm1, xmm5
+
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      xmm2, xmm3
+    paddsw      xmm2, xmm1
+    paddsw      xmm2, [rd GLOBAL]
+    psraw       xmm2, 7
+    packuswb    xmm2, xmm2
+
+    movq        MMWORD PTR [rdi+8], xmm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;out_pitch
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d16_v6_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d16_v4_ssse3:
+    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
+    mov         rdi, arg(2)             ;output_ptr
+
+%if ABI_IS_32BIT=0
+    movsxd      r8, DWORD PTR arg(3)    ;out_pitch
+%endif
+    mov         rax, rsi
+    movsxd      rcx, DWORD PTR arg(4)   ;output_height
+    add         rax, rdx
+
+vp8_filter_block1d16_v4_ssse3_loop:
+    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
+    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   xmm2, xmm4                  ;B D
+    punpcklbw   xmm3, xmm0                  ;C E
+
+    pmaddubsw   xmm3, xmm6
+    pmaddubsw   xmm2, xmm7
+    movq        xmm5, MMWORD PTR [rsi + rdx + 8]            ;B
+    movq        xmm1, MMWORD PTR [rsi + rdx * 2 + 8]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2 + 8]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4 + 8]        ;E
+
+    paddsw      xmm2, [rd GLOBAL]
+    paddsw      xmm2, xmm3
+    psraw       xmm2, 7
+    packuswb    xmm2, xmm2
+
+    punpcklbw   xmm5, xmm4                  ;B D
+    punpcklbw   xmm1, xmm0                  ;C E
+
+    pmaddubsw   xmm1, xmm6
+    pmaddubsw   xmm5, xmm7
+
+    movdqa      xmm4, [rd GLOBAL]
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      xmm5, xmm1
+    paddsw      xmm5, xmm4
+    psraw       xmm5, 7
+    packuswb    xmm5, xmm5
+
+    punpcklqdq  xmm2, xmm5
+
+    movdqa       XMMWORD PTR [rdi], xmm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;out_pitch
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d16_v4_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_filter_block1d8_v6_ssse3
+;(
+;    unsigned char *src_ptr,
+;    unsigned int   src_pitch,
+;    unsigned char *output_ptr,
+;    unsigned int   out_pitch,
+;    unsigned int   output_height,
+;    unsigned int   vp8_filter_index
+;)
+global sym(vp8_filter_block1d8_v6_ssse3)
+sym(vp8_filter_block1d8_v6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4      ;
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+
+    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
+    mov         rdi, arg(2)             ;output_ptr
+%if ABI_IS_32BIT=0
+    movsxd      r8, DWORD PTR arg(3)    ; out_pitch
+%endif
+    movsxd      rcx, DWORD PTR arg(4)   ;[output_height]
+
+    cmp         esi, DWORD PTR [rax]
+    je          vp8_filter_block1d8_v4_ssse3
+
+    movdqa      xmm5, XMMWORD PTR [rax]         ;k0_k5
+    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    mov         rax, rsi
+    add         rax, rdx
+
+vp8_filter_block1d8_v6_ssse3_loop:
+    movq        xmm1, MMWORD PTR [rsi]                  ;A
+    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
+    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   xmm2, xmm4                  ;B D
+    punpcklbw   xmm3, xmm0                  ;C E
+
+    movq        xmm0, MMWORD PTR [rax + rdx * 4]        ;F
+    movdqa      xmm4, [rd GLOBAL]
+
+    pmaddubsw   xmm3, xmm6
+    punpcklbw   xmm1, xmm0                  ;A F
+    pmaddubsw   xmm2, xmm7
+    pmaddubsw   xmm1, xmm5
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      xmm2, xmm3
+    paddsw      xmm2, xmm1
+    paddsw      xmm2, xmm4
+    psraw       xmm2, 7
+    packuswb    xmm2, xmm2
+
+    movq        MMWORD PTR [rdi], xmm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d8_v6_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d8_v4_ssse3:
+    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
+    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
+    movdqa      xmm5, [rd GLOBAL]
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    mov         rax, rsi
+    add         rax, rdx
+
+vp8_filter_block1d8_v4_ssse3_loop:
+    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
+    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
+    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
+    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   xmm2, xmm4                  ;B D
+    punpcklbw   xmm3, xmm0                  ;C E
+
+    pmaddubsw   xmm3, xmm6
+    pmaddubsw   xmm2, xmm7
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      xmm2, xmm3
+    paddsw      xmm2, xmm5
+    psraw       xmm2, 7
+    packuswb    xmm2, xmm2
+
+    movq        MMWORD PTR [rdi], xmm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d8_v4_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+;void vp8_filter_block1d4_v6_ssse3
+;(
+;    unsigned char *src_ptr,
+;    unsigned int   src_pitch,
+;    unsigned char *output_ptr,
+;    unsigned int   out_pitch,
+;    unsigned int   output_height,
+;    unsigned int   vp8_filter_index
+;)
+global sym(vp8_filter_block1d4_v6_ssse3)
+sym(vp8_filter_block1d4_v6_ssse3):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 6
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    movsxd      rdx, DWORD PTR arg(5)   ;table index
+    xor         rsi, rsi
+    shl         rdx, 4      ;
+
+    lea         rax, [k0_k5 GLOBAL]
+    add         rax, rdx
+
+    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
+    mov         rdi, arg(2)             ;output_ptr
+%if ABI_IS_32BIT=0
+    movsxd      r8, DWORD PTR arg(3)    ; out_pitch
+%endif
+    movsxd      rcx, DWORD PTR arg(4)   ;[output_height]
+
+    cmp         esi, DWORD PTR [rax]
+    je          vp8_filter_block1d4_v4_ssse3
+
+    movq        mm5, MMWORD PTR [rax]         ;k0_k5
+    movq        mm6, MMWORD PTR [rax+256]     ;k2_k4
+    movq        mm7, MMWORD PTR [rax+128]     ;k1_k3
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    mov         rax, rsi
+    add         rax, rdx
+
+vp8_filter_block1d4_v6_ssse3_loop:
+    movd        mm1, DWORD PTR [rsi]                  ;A
+    movd        mm2, DWORD PTR [rsi + rdx]            ;B
+    movd        mm3, DWORD PTR [rsi + rdx * 2]        ;C
+    movd        mm4, DWORD PTR [rax + rdx * 2]        ;D
+    movd        mm0, DWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   mm2, mm4                  ;B D
+    punpcklbw   mm3, mm0                  ;C E
+
+    movd        mm0, DWORD PTR [rax + rdx * 4]        ;F
+
+    movq        mm4, [rd GLOBAL]
+
+    pmaddubsw   mm3, mm6
+    punpcklbw   mm1, mm0                  ;A F
+    pmaddubsw   mm2, mm7
+    pmaddubsw   mm1, mm5
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      mm2, mm3
+    paddsw      mm2, mm1
+    paddsw      mm2, mm4
+    psraw       mm2, 7
+    packuswb    mm2, mm2
+
+    movd        DWORD PTR [rdi], mm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d4_v6_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+vp8_filter_block1d4_v4_ssse3:
+    movq        mm6, MMWORD PTR [rax+256]     ;k2_k4
+    movq        mm7, MMWORD PTR [rax+128]     ;k1_k3
+    movq        mm5, MMWORD PTR [rd GLOBAL]
+
+    mov         rsi, arg(0)             ;src_ptr
+
+    mov         rax, rsi
+    add         rax, rdx
+
+vp8_filter_block1d4_v4_ssse3_loop:
+    movd        mm2, DWORD PTR [rsi + rdx]            ;B
+    movd        mm3, DWORD PTR [rsi + rdx * 2]        ;C
+    movd        mm4, DWORD PTR [rax + rdx * 2]        ;D
+    movd        mm0, DWORD PTR [rsi + rdx * 4]        ;E
+
+    punpcklbw   mm2, mm4                  ;B D
+    punpcklbw   mm3, mm0                  ;C E
+
+    pmaddubsw   mm3, mm6
+    pmaddubsw   mm2, mm7
+    add         rsi,  rdx
+    add         rax,  rdx
+;--
+;--
+    paddsw      mm2, mm3
+    paddsw      mm2, mm5
+    psraw       mm2, 7
+    packuswb    mm2, mm2
+
+    movd        DWORD PTR [rdi], mm2
+
+%if ABI_IS_32BIT
+    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
+%else
+    add         rdi,        r8
+%endif
+    dec         rcx
+    jnz         vp8_filter_block1d4_v4_ssse3_loop
+
+    ; begin epilog
+    pop rdi
+    pop rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+SECTION_RODATA
+align 16
+shuf1b:
+    db 0, 5, 1, 6, 2, 7, 3, 8, 4, 9, 5, 10, 6, 11, 7, 12
+shuf2b:
+    db 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10, 9, 11
+shuf3b:
+    db 1, 3, 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10
+
+align 16
+rd:
+    times 8 dw 0x40
+
+align 16
+k0_k5:
+    times 8 db 0, 0             ;placeholder
+    times 8 db 0, 0
+    times 8 db 2, 1
+    times 8 db 0, 0
+    times 8 db 3, 3
+    times 8 db 0, 0
+    times 8 db 1, 2
+    times 8 db 0, 0
+k1_k3:
+    times 8 db  0,    0         ;placeholder
+    times 8 db  -6,  12
+    times 8 db -11,  36
+    times 8 db  -9,  50
+    times 8 db -16,  77
+    times 8 db  -6,  93
+    times 8 db  -8, 108
+    times 8 db  -1, 123
+k2_k4:
+    times 8 db 128,    0        ;placeholder
+    times 8 db 123,   -1
+    times 8 db 108,   -8
+    times 8 db  93,   -6
+    times 8 db  77,  -16
+    times 8 db  50,   -9
+    times 8 db  36,  -11
+    times 8 db  12,   -6
+

vp8/common/x86/subpixel_x86.h

@@ -86,4 +86,37 @@ extern prototype_subpixel_predict(vp8_bilinear_predict8x8_sse2);
 #endif
 #endif
 
+#if HAVE_SSSE3
+extern prototype_subpixel_predict(vp8_sixtap_predict16x16_ssse3);
+extern prototype_subpixel_predict(vp8_sixtap_predict8x8_ssse3);
+extern prototype_subpixel_predict(vp8_sixtap_predict8x4_ssse3);
+extern prototype_subpixel_predict(vp8_sixtap_predict4x4_ssse3);
+//extern prototype_subpixel_predict(vp8_bilinear_predict16x16_sse2);
+//extern prototype_subpixel_predict(vp8_bilinear_predict8x8_sse2);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+#undef  vp8_subpix_sixtap16x16
+#define vp8_subpix_sixtap16x16 vp8_sixtap_predict16x16_ssse3
+
+#undef  vp8_subpix_sixtap8x8
+#define vp8_subpix_sixtap8x8 vp8_sixtap_predict8x8_ssse3
+
+#undef  vp8_subpix_sixtap8x4
+#define vp8_subpix_sixtap8x4 vp8_sixtap_predict8x4_ssse3
+
+#undef  vp8_subpix_sixtap4x4
+#define vp8_subpix_sixtap4x4 vp8_sixtap_predict4x4_ssse3
+
+
+//#undef  vp8_subpix_bilinear16x16
+//#define vp8_subpix_bilinear16x16 vp8_bilinear_predict16x16_sse2
+
+//#undef  vp8_subpix_bilinear8x8
+//#define vp8_subpix_bilinear8x8 vp8_bilinear_predict8x8_sse2
+
+#endif
+#endif
+
+
+
 #endif

vp8/common/x86/vp8_asm_stubs.c

@@ -359,3 +359,195 @@ void vp8_sixtap_predict8x4_sse2
 }
 
 #endif
+
+#if HAVE_SSSE3
+
+extern void vp8_filter_block1d8_h6_ssse3
+(
+    unsigned char  *src_ptr,
+    unsigned int    src_pixels_per_line,
+    unsigned char  *output_ptr,
+    unsigned int    output_pitch,
+    unsigned int    output_height,
+    unsigned int    vp8_filter_index
+);
+
+extern void vp8_filter_block1d16_h6_ssse3
+(
+    unsigned char  *src_ptr,
+    unsigned int    src_pixels_per_line,
+    unsigned char  *output_ptr,
+    unsigned int    output_pitch,
+    unsigned int    output_height,
+    unsigned int    vp8_filter_index
+);
+
+extern void vp8_filter_block1d16_v6_ssse3
+(
+    unsigned char *src_ptr,
+    unsigned int   src_pitch,
+    unsigned char *output_ptr,
+    unsigned int   out_pitch,
+    unsigned int   output_height,
+    unsigned int   vp8_filter_index
+);
+
+extern void vp8_filter_block1d8_v6_ssse3
+(
+    unsigned char *src_ptr,
+    unsigned int   src_pitch,
+    unsigned char *output_ptr,
+    unsigned int   out_pitch,
+    unsigned int   output_height,
+    unsigned int   vp8_filter_index
+);
+
+extern void vp8_filter_block1d4_h6_ssse3
+(
+    unsigned char  *src_ptr,
+    unsigned int    src_pixels_per_line,
+    unsigned char  *output_ptr,
+    unsigned int    output_pitch,
+    unsigned int    output_height,
+    unsigned int    vp8_filter_index
+);
+
+extern void vp8_filter_block1d4_v6_ssse3
+(
+    unsigned char *src_ptr,
+    unsigned int   src_pitch,
+    unsigned char *output_ptr,
+    unsigned int   out_pitch,
+    unsigned int   output_height,
+    unsigned int   vp8_filter_index
+);
+
+void vp8_sixtap_predict16x16_ssse3
+(
+    unsigned char  *src_ptr,
+    int   src_pixels_per_line,
+    int  xoffset,
+    int  yoffset,
+    unsigned char *dst_ptr,
+    int dst_pitch
+
+)
+{
+    DECLARE_ALIGNED_ARRAY(16, unsigned char, FData2, 24*24);
+
+    if (xoffset)
+    {
+        if (yoffset)
+        {
+            vp8_filter_block1d16_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 16, 21, xoffset);
+            vp8_filter_block1d16_v6_ssse3(FData2 , 16, dst_ptr, dst_pitch, 16, yoffset);
+        }
+        else
+        {
+            // First-pass only
+            vp8_filter_block1d16_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 16, xoffset);
+        }
+    }
+    else
+    {
+        // Second-pass only
+        vp8_filter_block1d16_v6_ssse3(src_ptr - (2 * src_pixels_per_line) , src_pixels_per_line, dst_ptr, dst_pitch, 16, yoffset);
+    }
+}
+
+void vp8_sixtap_predict8x8_ssse3
+(
+    unsigned char  *src_ptr,
+    int   src_pixels_per_line,
+    int  xoffset,
+    int  yoffset,
+    unsigned char *dst_ptr,
+    int dst_pitch
+)
+{
+    DECLARE_ALIGNED_ARRAY(16, unsigned char, FData2, 256);
+
+    if (xoffset)
+    {
+        if (yoffset)
+        {
+            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 8, 13, xoffset);
+            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch, 8, yoffset);
+        }
+        else
+        {
+            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 8, xoffset);
+        }
+    }
+    else
+    {
+        // Second-pass only
+        vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 8, yoffset);
+    }
+}
+
+
+void vp8_sixtap_predict8x4_ssse3
+(
+    unsigned char  *src_ptr,
+    int   src_pixels_per_line,
+    int  xoffset,
+    int  yoffset,
+    unsigned char *dst_ptr,
+    int dst_pitch
+)
+{
+    DECLARE_ALIGNED_ARRAY(16, unsigned char, FData2, 256);
+
+    if (xoffset)
+    {
+        if (yoffset)
+        {
+            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 8, 9, xoffset);
+            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch, 4, yoffset);
+        }
+        else
+        {
+            // First-pass only
+            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 4, xoffset);
+        }
+    }
+    else
+    {
+        // Second-pass only
+        vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 4, yoffset);
+    }
+}
+
+void vp8_sixtap_predict4x4_ssse3
+(
+    unsigned char  *src_ptr,
+    int   src_pixels_per_line,
+    int  xoffset,
+    int  yoffset,
+    unsigned char *dst_ptr,
+    int dst_pitch
+)
+{
+  DECLARE_ALIGNED_ARRAY(16, unsigned char, FData2, 4*9);
+
+  if (xoffset)
+  {
+      if (yoffset)
+      {
+          vp8_filter_block1d4_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 4, 9, xoffset);
+          vp8_filter_block1d4_v6_ssse3(FData2, 4, dst_ptr, dst_pitch, 4, yoffset);
+      }
+      else
+      {
+          vp8_filter_block1d4_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 4, xoffset);
+      }
+  }
+  else
+  {
+      vp8_filter_block1d4_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 4, yoffset);
+  }
+
+}
+
+#endif

vp8/common/x86/x86_systemdependent.c

@@ -27,6 +27,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
     int mmx_enabled = flags & HAS_MMX;
     int xmm_enabled = flags & HAS_SSE;
     int wmt_enabled = flags & HAS_SSE2;
+    int SSSE3Enabled = flags & HAS_SSSE3;
 
     /* Note:
      *
@@ -42,7 +43,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
     {
         rtcd->idct.idct1        = vp8_short_idct4x4llm_1_mmx;
         rtcd->idct.idct16       = vp8_short_idct4x4llm_mmx;
-        rtcd->idct.idct1_scalar = vp8_dc_only_idct_mmx;
+        rtcd->idct.idct1_scalar_add = vp8_dc_only_idct_add_mmx;
         rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_mmx;
         rtcd->idct.iwalsh1     = vp8_short_inv_walsh4x4_1_mmx;
 
@@ -114,5 +115,17 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
     }
 
 #endif
+
+#if HAVE_SSSE3
+
+    if (SSSE3Enabled)
+    {
+        rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_ssse3;
+        rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_ssse3;
+        rtcd->subpix.sixtap8x4     = vp8_sixtap_predict8x4_ssse3;
+        rtcd->subpix.sixtap4x4     = vp8_sixtap_predict4x4_ssse3;
+    }
+#endif
+
 #endif
 }

vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm

@@ -0,0 +1,218 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license and patent
+;  grant that can be found in the LICENSE file in the root of the source
+;  tree. All contributing project authors may be found in the AUTHORS
+;  file in the root of the source tree.
+;
+
+
+    EXPORT |vp8_dequant_dc_idct_add_v6|
+
+    AREA |.text|, CODE, READONLY
+
+;void vp8_dequant_dc_idct_v6(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride, int Dc)
+; r0 = input
+; r1 = dq
+; r2 = pred
+; r3 = dest
+; sp + 36 = pitch  ; +4 = 40
+; sp + 40 = stride  ; +4 = 44
+; sp + 44 = Dc  ; +4 = 48
+
+
+|vp8_dequant_dc_idct_add_v6| PROC
+    stmdb   sp!, {r4-r11, lr}
+
+    ldr     r6, [sp, #44]
+
+    ldr     r4, [r0]                ;input
+    ldr     r5, [r1], #4            ;dq
+
+    sub     sp, sp, #4
+    str     r3, [sp]
+
+    smultt  r7, r4, r5
+
+    ldr     r4, [r0, #4]            ;input
+    ldr     r5, [r1], #4            ;dq
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    smulbb  r6, r4, r5
+    smultt  r7, r4, r5
+
+    ldr     r4, [r0, #4]            ;input
+    ldr     r5, [r1], #4            ;dq
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    mov     r12, #3
+
+vp8_dequant_dc_add_loop
+    smulbb  r6, r4, r5
+    smultt  r7, r4, r5
+
+    ldr     r4, [r0, #4]            ;input
+    ldr     r5, [r1], #4            ;dq
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    smulbb  r6, r4, r5
+    smultt  r7, r4, r5
+
+    subs    r12, r12, #1
+
+    ldrne   r4, [r0, #4]
+    ldrne   r5, [r1], #4
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    bne     vp8_dequant_dc_add_loop
+
+    sub     r0, r0, #32
+    mov     r1, r0
+
+; short_idct4x4llm_v6_dual
+    ldr     r3, cospi8sqrt2minus1
+    ldr     r4, sinpi8sqrt2
+    ldr     r6, [r0, #8]
+    mov     r5, #2
+vp8_dequant_dc_idct_loop1_v6
+    ldr     r12, [r0, #24]
+    ldr     r14, [r0, #16]
+    smulwt  r9, r3, r6
+    smulwb  r7, r3, r6
+    smulwt  r10, r4, r6
+    smulwb  r8, r4, r6
+    pkhbt   r7, r7, r9, lsl #16
+    smulwt  r11, r3, r12
+    pkhbt   r8, r8, r10, lsl #16
+    uadd16  r6, r6, r7
+    smulwt  r7, r4, r12
+    smulwb  r9, r3, r12
+    smulwb  r10, r4, r12
+    subs    r5, r5, #1
+    pkhbt   r9, r9, r11, lsl #16
+    ldr     r11, [r0], #4
+    pkhbt   r10, r10, r7, lsl #16
+    uadd16  r7, r12, r9
+    usub16  r7, r8, r7
+    uadd16  r6, r6, r10
+    uadd16  r10, r11, r14
+    usub16  r8, r11, r14
+    uadd16  r9, r10, r6
+    usub16  r10, r10, r6
+    uadd16  r6, r8, r7
+    usub16  r7, r8, r7
+    str     r6, [r1, #8]
+    ldrne   r6, [r0, #8]
+    str     r7, [r1, #16]
+    str     r10, [r1, #24]
+    str     r9, [r1], #4
+    bne     vp8_dequant_dc_idct_loop1_v6
+
+    mov     r5, #2
+    sub     r0, r1, #8
+vp8_dequant_dc_idct_loop2_v6
+    ldr     r6, [r0], #4
+    ldr     r7, [r0], #4
+    ldr     r8, [r0], #4
+    ldr     r9, [r0], #4
+    smulwt  r1, r3, r6
+    smulwt  r12, r4, r6
+    smulwt  lr, r3, r8
+    smulwt  r10, r4, r8
+    pkhbt   r11, r8, r6, lsl #16
+    pkhbt   r1, lr, r1, lsl #16
+    pkhbt   r12, r10, r12, lsl #16
+    pkhtb   r6, r6, r8, asr #16
+    uadd16  r6, r1, r6
+    pkhbt   lr, r9, r7, lsl #16
+    uadd16  r10, r11, lr
+    usub16  lr, r11, lr
+    pkhtb   r8, r7, r9, asr #16
+    subs    r5, r5, #1
+    smulwt  r1, r3, r8
+    smulwb  r7, r3, r8
+    smulwt  r11, r4, r8
+    smulwb  r9, r4, r8
+    pkhbt   r1, r7, r1, lsl #16
+    uadd16  r8, r1, r8
+    pkhbt   r11, r9, r11, lsl #16
+    usub16  r1, r12, r8
+    uadd16  r8, r11, r6
+    ldr     r9, c0x00040004
+    ldr     r12, [sp, #40]
+    uadd16  r6, r10, r8
+    usub16  r7, r10, r8
+    uadd16  r7, r7, r9
+    uadd16  r6, r6, r9
+    uadd16  r10, r14, r1
+    usub16  r1, r14, r1
+    uadd16  r10, r10, r9
+    uadd16  r1, r1, r9
+    ldr     r11, [r2], r12
+    mov     r8, r7, asr #3
+    pkhtb   r9, r8, r10, asr #19
+    mov     r8, r1, asr #3
+    pkhtb   r8, r8, r6, asr #19
+    uxtb16  lr, r11, ror #8
+    qadd16  r9, r9, lr
+    uxtb16  lr, r11
+    qadd16  r8, r8, lr
+    usat16  r9, #8, r9
+    usat16  r8, #8, r8
+    orr     r9, r8, r9, lsl #8
+    ldr     r11, [r2], r12
+    ldr     lr, [sp]
+    ldr     r12, [sp, #44]
+    mov     r7, r7, lsl #16
+    mov     r1, r1, lsl #16
+    mov     r10, r10, lsl #16
+    mov     r6, r6, lsl #16
+    mov     r7, r7, asr #3
+    pkhtb   r7, r7, r10, asr #19
+    mov     r1, r1, asr #3
+    pkhtb   r1, r1, r6, asr #19
+    uxtb16  r8, r11, ror #8
+    qadd16  r7, r7, r8
+    uxtb16  r8, r11
+    qadd16  r1, r1, r8
+    usat16  r7, #8, r7
+    usat16  r1, #8, r1
+    orr     r1, r1, r7, lsl #8
+    str     r9, [lr], r12
+    str     r1, [lr], r12
+    str     lr, [sp]
+    bne     vp8_dequant_dc_idct_loop2_v6
+
+; vpx_memset
+    sub     r0, r0, #32
+    add     sp, sp, #4
+
+    mov     r12, #0
+    str     r12, [r0]
+    str     r12, [r0, #4]
+    str     r12, [r0, #8]
+    str     r12, [r0, #12]
+    str     r12, [r0, #16]
+    str     r12, [r0, #20]
+    str     r12, [r0, #24]
+    str     r12, [r0, #28]
+
+    ldmia   sp!, {r4 - r11, pc}
+    ENDP    ; |vp8_dequant_dc_idct_add_v6|
+
+; Constant Pool
+cospi8sqrt2minus1 DCD 0x00004E7B
+sinpi8sqrt2       DCD 0x00008A8C
+c0x00040004       DCD 0x00040004
+
+    END

vp8/decoder/arm/armv6/dequant_idct_v6.asm

@@ -0,0 +1,196 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license and patent
+;  grant that can be found in the LICENSE file in the root of the source
+;  tree. All contributing project authors may be found in the AUTHORS
+;  file in the root of the source tree.
+;
+
+    EXPORT |vp8_dequant_idct_add_v6|
+
+    AREA |.text|, CODE, READONLY
+;void vp8_dequant_idct_v6(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride)
+; r0 = input
+; r1 = dq
+; r2 = pred
+; r3 = dest
+; sp + 36 = pitch  ; +4 = 40
+; sp + 40 = stride  ; +4 = 44
+
+
+|vp8_dequant_idct_add_v6| PROC
+    stmdb   sp!, {r4-r11, lr}
+
+    ldr     r4, [r0]                ;input
+    ldr     r5, [r1], #4            ;dq
+
+    sub     sp, sp, #4
+    str     r3, [sp]
+
+    mov     r12, #4
+
+vp8_dequant_add_loop
+    smulbb  r6, r4, r5
+    smultt  r7, r4, r5
+
+    ldr     r4, [r0, #4]            ;input
+    ldr     r5, [r1], #4            ;dq
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    smulbb  r6, r4, r5
+    smultt  r7, r4, r5
+
+    subs    r12, r12, #1
+
+    ldrne   r4, [r0, #4]
+    ldrne   r5, [r1], #4
+
+    strh    r6, [r0], #2
+    strh    r7, [r0], #2
+
+    bne     vp8_dequant_add_loop
+
+    sub     r0, r0, #32
+    mov     r1, r0
+
+; short_idct4x4llm_v6_dual
+    ldr     r3, cospi8sqrt2minus1
+    ldr     r4, sinpi8sqrt2
+    ldr     r6, [r0, #8]
+    mov     r5, #2
+vp8_dequant_idct_loop1_v6
+    ldr     r12, [r0, #24]
+    ldr     r14, [r0, #16]
+    smulwt  r9, r3, r6
+    smulwb  r7, r3, r6
+    smulwt  r10, r4, r6
+    smulwb  r8, r4, r6
+    pkhbt   r7, r7, r9, lsl #16
+    smulwt  r11, r3, r12
+    pkhbt   r8, r8, r10, lsl #16
+    uadd16  r6, r6, r7
+    smulwt  r7, r4, r12
+    smulwb  r9, r3, r12
+    smulwb  r10, r4, r12
+    subs    r5, r5, #1
+    pkhbt   r9, r9, r11, lsl #16
+    ldr     r11, [r0], #4
+    pkhbt   r10, r10, r7, lsl #16
+    uadd16  r7, r12, r9
+    usub16  r7, r8, r7
+    uadd16  r6, r6, r10
+    uadd16  r10, r11, r14
+    usub16  r8, r11, r14
+    uadd16  r9, r10, r6
+    usub16  r10, r10, r6
+    uadd16  r6, r8, r7
+    usub16  r7, r8, r7
+    str     r6, [r1, #8]
+    ldrne   r6, [r0, #8]
+    str     r7, [r1, #16]
+    str     r10, [r1, #24]
+    str     r9, [r1], #4
+    bne     vp8_dequant_idct_loop1_v6
+
+    mov     r5, #2
+    sub     r0, r1, #8
+vp8_dequant_idct_loop2_v6
+    ldr     r6, [r0], #4
+    ldr     r7, [r0], #4
+    ldr     r8, [r0], #4
+    ldr     r9, [r0], #4
+    smulwt  r1, r3, r6
+    smulwt  r12, r4, r6
+    smulwt  lr, r3, r8
+    smulwt  r10, r4, r8
+    pkhbt   r11, r8, r6, lsl #16
+    pkhbt   r1, lr, r1, lsl #16
+    pkhbt   r12, r10, r12, lsl #16
+    pkhtb   r6, r6, r8, asr #16
+    uadd16  r6, r1, r6
+    pkhbt   lr, r9, r7, lsl #16
+    uadd16  r10, r11, lr
+    usub16  lr, r11, lr
+    pkhtb   r8, r7, r9, asr #16
+    subs    r5, r5, #1
+    smulwt  r1, r3, r8
+    smulwb  r7, r3, r8
+    smulwt  r11, r4, r8
+    smulwb  r9, r4, r8
+    pkhbt   r1, r7, r1, lsl #16
+    uadd16  r8, r1, r8
+    pkhbt   r11, r9, r11, lsl #16
+    usub16  r1, r12, r8
+    uadd16  r8, r11, r6
+    ldr     r9, c0x00040004
+    ldr     r12, [sp, #40]
+    uadd16  r6, r10, r8
+    usub16  r7, r10, r8
+    uadd16  r7, r7, r9
+    uadd16  r6, r6, r9
+    uadd16  r10, r14, r1
+    usub16  r1, r14, r1
+    uadd16  r10, r10, r9
+    uadd16  r1, r1, r9
+    ldr     r11, [r2], r12
+    mov     r8, r7, asr #3
+    pkhtb   r9, r8, r10, asr #19
+    mov     r8, r1, asr #3
+    pkhtb   r8, r8, r6, asr #19
+    uxtb16  lr, r11, ror #8
+    qadd16  r9, r9, lr
+    uxtb16  lr, r11
+    qadd16  r8, r8, lr
+    usat16  r9, #8, r9
+    usat16  r8, #8, r8
+    orr     r9, r8, r9, lsl #8
+    ldr     r11, [r2], r12
+    ldr     lr, [sp]
+    ldr     r12, [sp, #44]
+    mov     r7, r7, lsl #16
+    mov     r1, r1, lsl #16
+    mov     r10, r10, lsl #16
+    mov     r6, r6, lsl #16
+    mov     r7, r7, asr #3
+    pkhtb   r7, r7, r10, asr #19
+    mov     r1, r1, asr #3
+    pkhtb   r1, r1, r6, asr #19
+    uxtb16  r8, r11, ror #8
+    qadd16  r7, r7, r8
+    uxtb16  r8, r11
+    qadd16  r1, r1, r8
+    usat16  r7, #8, r7
+    usat16  r1, #8, r1
+    orr     r1, r1, r7, lsl #8
+    str     r9, [lr], r12
+    str     r1, [lr], r12
+    str     lr, [sp]
+    bne     vp8_dequant_idct_loop2_v6
+
+; vpx_memset
+    sub     r0, r0, #32
+    add     sp, sp, #4
+
+    mov     r12, #0
+    str     r12, [r0]
+    str     r12, [r0, #4]
+    str     r12, [r0, #8]
+    str     r12, [r0, #12]
+    str     r12, [r0, #16]
+    str     r12, [r0, #20]
+    str     r12, [r0, #24]
+    str     r12, [r0, #28]
+
+    ldmia   sp!, {r4 - r11, pc}
+    ENDP    ; |vp8_dequant_idct_add_v6|
+
+; Constant Pool
+cospi8sqrt2minus1 DCD 0x00004E7B
+sinpi8sqrt2       DCD 0x00008A8C
+c0x00040004       DCD 0x00040004
+
+    END

vp8/decoder/arm/armv6/dequantdcidct_v6.asm

@@ -1,203 +0,0 @@
-;
-;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_dequant_dc_idct_v6|
-    ; ARM
-    ; REQUIRE8
-    ; PRESERVE8
-
-    AREA    |.text|, CODE, READONLY  ; name this block of code
-;void vp8_dequant_dc_idct_v6(short *input, short *dq, short *output, int pitch,int Dc)
-|vp8_dequant_dc_idct_v6| PROC
-    stmdb   sp!, {r4-r11, lr}
-
-    ldr     r6, [sp, #36]           ;load Dc
-
-    ldr     r4, [r0]                ;input
-    ldr     r5, [r1], #4            ;dq
-
-    sub     sp, sp, #4
-    str     r0, [sp]
-
-    smultt  r7, r4, r5
-
-    ldr     r4, [r0, #4]            ;input
-    ldr     r5, [r1], #4            ;dq
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    smulbb  r6, r4, r5
-    smultt  r7, r4, r5
-
-    ldr     r4, [r0, #4]            ;input
-    ldr     r5, [r1], #4            ;dq
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    mov     r12, #3
-
-dequant_dc_idct_loop
-    smulbb  r6, r4, r5
-    smultt  r7, r4, r5
-
-    ldr     r4, [r0, #4]            ;input
-    ldr     r5, [r1], #4            ;dq
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    smulbb  r6, r4, r5
-    smultt  r7, r4, r5
-
-    subs    r12, r12, #1
-
-    ldrne   r4, [r0, #4]
-    ldrne   r5, [r1], #4
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    bne     dequant_dc_idct_loop
-
-    sub     r0, r0, #32
-    mov     r1, r2
-    mov     r2, r3
-
-; short_idct4x4llm_v6_dual
-
-    mov r3, #0x00004E00 ;                   cos
-    orr r3, r3, #0x0000007B ; cospi8sqrt2minus1
-    mov r4, #0x00008A00 ;                       sin
-    orr r4, r4, #0x0000008C ; sinpi8sqrt2
-    mov r5, #0x2    ; i=2                           i
-loop1_dual_11
-    ldr r6, [r0, #(4*2)]    ; i5 | i4                               5|4
-    ldr r12, [r0, #(12*2)]  ; i13 | i12                                                     13|12
-    ldr r14, [r0, #(8*2)]   ; i9 | i8                                                               9|8
-
-    smulwt  r9, r3, r6  ; (ip[5] * cospi8sqrt2minus1) >> 16                                         5c
-    smulwb  r7, r3, r6  ; (ip[4] * cospi8sqrt2minus1) >> 16                                 4c
-    smulwt  r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16                                               5s
-    smulwb  r8, r4, r6  ; (ip[4] * sinpi8sqrt2) >> 16                                       4s
-    pkhbt   r7, r7, r9, lsl #16 ; 5c | 4c
-    smulwt  r11, r3, r12    ; (ip[13] * cospi8sqrt2minus1) >> 16                                                    13c
-    pkhbt   r8, r8, r10, lsl #16    ; 5s | 4s
-    uadd16  r6, r6, r7  ; 5c+5 | 4c+4
-    smulwt  r7, r4, r12 ; (ip[13] * sinpi8sqrt2) >> 16                                  13s
-    smulwb  r9, r3, r12 ; (ip[12] * cospi8sqrt2minus1) >> 16                                            12c
-    smulwb  r10, r4, r12    ; (ip[12] * sinpi8sqrt2) >> 16                                              12s
-    subs    r5, r5, #0x1    ; i--                           --
-    pkhbt   r9, r9, r11, lsl #16    ; 13c | 12c
-    ldr r11, [r0], #0x4 ; i1 | i0       ++                                          1|0
-    pkhbt   r10, r10, r7, lsl #16   ; 13s | 12s
-    uadd16  r7, r12, r9 ; 13c+13 | 12c+12
-    usub16  r7, r8, r7  ; c                                 c
-    uadd16  r6, r6, r10 ; d                             d
-    uadd16  r10, r11, r14   ; a                                             a
-    usub16  r8, r11, r14    ; b                                     b
-    uadd16  r9, r10, r6 ; a+d                                           a+d
-    usub16  r10, r10, r6    ; a-d                                               a-d
-    uadd16  r6, r8, r7  ; b+c                               b+c
-    usub16  r7, r8, r7  ; b-c                                   b-c
-    str r6, [r1, r2]    ; o5 | o4
-    add r6, r2, r2  ; pitch * 2                             p2
-    str r7, [r1, r6]    ; o9 | o8
-    add r6,  r6, r2 ; pitch * 3                             p3
-    str r10, [r1, r6]   ; o13 | o12
-    str r9, [r1], #0x4  ; o1 | o0           ++
-    bne loop1_dual_11   ;
-    mov r5, #0x2    ; i=2                           i
-    sub r0, r1, #8  ; reset input/output        i/o
-loop2_dual_22
-    ldr r6, [r0, r2]    ; i5 | i4                               5|4
-    ldr r1, [r0]    ; i1 | i0           1|0
-    ldr r12, [r0, #0x4] ; i3 | i2                                                       3|2
-    add r14, r2, #0x4   ; pitch + 2                                                             p+2
-    ldr r14, [r0, r14]  ; i7 | i6                                                               7|6
-    smulwt  r9, r3, r6  ; (ip[5] * cospi8sqrt2minus1) >> 16                                         5c
-    smulwt  r7, r3, r1  ; (ip[1] * cospi8sqrt2minus1) >> 16                                 1c
-    smulwt  r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16                                               5s
-    smulwt  r8, r4, r1  ; (ip[1] * sinpi8sqrt2) >> 16                                       1s
-    pkhbt   r11, r6, r1, lsl #16    ; i0 | i4                                                   0|4
-    pkhbt   r7, r9, r7, lsl #16 ; 1c | 5c
-    pkhbt   r8, r10, r8, lsl #16    ; 1s | 5s = temp1 <20>                                     tc1
-    pkhtb   r1, r1, r6, asr #16 ; i1 | i5           1|5
-    uadd16  r1, r7, r1  ; 1c+1 | 5c+5 = temp2 (d)           td2
-    pkhbt   r9, r14, r12, lsl #16   ; i2 | i6                                           2|6
-    uadd16  r10, r11, r9    ; a                                             a
-    usub16  r9, r11, r9 ; b                                         b
-    pkhtb   r6, r12, r14, asr #16   ; i3 | i7                               3|7
-    subs    r5, r5, #0x1    ; i--                           --
-    smulwt  r7, r3, r6  ; (ip[3] * cospi8sqrt2minus1) >> 16                                 3c
-    smulwt  r11, r4, r6 ; (ip[3] * sinpi8sqrt2) >> 16                                                   3s
-    smulwb  r12, r3, r6 ; (ip[7] * cospi8sqrt2minus1) >> 16                                                     7c
-    smulwb  r14, r4, r6 ; (ip[7] * sinpi8sqrt2) >> 16                                                               7s
-
-    pkhbt   r7, r12, r7, lsl #16    ; 3c | 7c
-    pkhbt   r11, r14, r11, lsl #16  ; 3s | 7s = temp1 (d)                                                   td1
-    uadd16  r6, r7, r6  ; 3c+3 | 7c+7 = temp2  (c)                              tc2
-    usub16  r12, r8, r6 ; c (o1 | o5)                                                       c
-    uadd16  r6, r11, r1 ; d (o3 | o7)                               d
-    uadd16  r7, r10, r6 ; a+d                                   a+d
-    mov r8, #0x4    ; set up 4's                                        4
-    orr r8, r8, #0x40000    ;                                       4|4
-    usub16  r6, r10, r6 ; a-d                               a-d
-    uadd16  r6, r6, r8  ; a-d+4                             3|7
-    uadd16  r7, r7, r8  ; a+d+4                                 0|4
-    uadd16  r10, r9, r12    ; b+c                                               b+c
-    usub16  r1, r9, r12 ; b-c           b-c
-    uadd16  r10, r10, r8    ; b+c+4                                             1|5
-    uadd16  r1, r1, r8  ; b-c+4         2|6
-    mov r8, r10, asr #19    ; o1 >> 3
-    strh    r8, [r0, #2]    ; o1
-    mov r8, r1, asr #19 ; o2 >> 3
-    strh    r8, [r0, #4]    ; o2
-    mov r8, r6, asr #19 ; o3 >> 3
-    strh    r8, [r0, #6]    ; o3
-    mov r8, r7, asr #19 ; o0 >> 3
-    strh    r8, [r0], r2    ; o0        +p
-    sxth    r10, r10    ;
-    mov r8, r10, asr #3 ; o5 >> 3
-    strh    r8, [r0, #2]    ; o5
-    sxth    r1, r1  ;
-    mov r8, r1, asr #3  ; o6 >> 3
-    strh    r8, [r0, #4]    ; o6
-    sxth    r6, r6  ;
-    mov r8, r6, asr #3  ; o7 >> 3
-    strh    r8, [r0, #6]    ; o7
-    sxth    r7, r7  ;
-    mov r8, r7, asr #3  ; o4 >> 3
-    strh    r8, [r0], r2    ; o4        +p
-;;;;;   subs    r5, r5, #0x1    ; i--                           --
-    bne loop2_dual_22   ;
-
-
-;vpx_memset
-    ldr     r0, [sp]
-    add     sp, sp, #4
-
-    mov     r12, #0
-    str     r12, [r0]
-    str     r12, [r0, #4]
-    str     r12, [r0, #8]
-    str     r12, [r0, #12]
-    str     r12, [r0, #16]
-    str     r12, [r0, #20]
-    str     r12, [r0, #24]
-    str     r12, [r0, #28]
-
-    ldmia   sp!, {r4 - r11, pc} ; replace vars, return                      restore
-
-    ENDP    ;|vp8_dequant_dc_idct_v68|
-
-    END

vp8/decoder/arm/armv6/dequantidct_v6.asm

@@ -1,184 +0,0 @@
-;
-;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_dequant_idct_v6|
-    ; ARM
-    ; REQUIRE8
-    ; PRESERVE8
-
-    AREA    |.text|, CODE, READONLY  ; name this block of code
-;void vp8_dequant_idct_v6(short *input, short *dq, short *output, int pitch)
-|vp8_dequant_idct_v6| PROC
-    stmdb   sp!, {r4-r11, lr}
-
-    ldr     r4, [r0]            ;input
-    ldr     r5, [r1], #4            ;dq
-
-    sub     sp, sp, #4
-    str     r0, [sp]
-
-    mov     r12, #4
-
-dequant_idct_loop
-    smulbb  r6, r4, r5
-    smultt  r7, r4, r5
-
-    ldr     r4, [r0, #4]            ;input
-    ldr     r5, [r1], #4        ;dq
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    smulbb  r6, r4, r5
-    smultt  r7, r4, r5
-
-    subs    r12, r12, #1
-
-    ldrne   r4, [r0, #4]
-    ldrne   r5, [r1], #4
-
-    strh    r6, [r0], #2
-    strh    r7, [r0], #2
-
-    bne     dequant_idct_loop
-
-    sub     r0, r0, #32
-    mov     r1, r2
-    mov     r2, r3
-
-; short_idct4x4llm_v6_dual
-
-    mov r3, #0x00004E00 ;                   cos
-    orr r3, r3, #0x0000007B ; cospi8sqrt2minus1
-    mov r4, #0x00008A00 ;                       sin
-    orr r4, r4, #0x0000008C ; sinpi8sqrt2
-    mov r5, #0x2    ; i=2                           i
-loop1_dual_1
-    ldr r6, [r0, #(4*2)]    ; i5 | i4                               5|4
-    ldr r12, [r0, #(12*2)]  ; i13 | i12                                                     13|12
-    ldr r14, [r0, #(8*2)]   ; i9 | i8                                                               9|8
-
-    smulwt  r9, r3, r6  ; (ip[5] * cospi8sqrt2minus1) >> 16                                         5c
-    smulwb  r7, r3, r6  ; (ip[4] * cospi8sqrt2minus1) >> 16                                 4c
-    smulwt  r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16                                               5s
-    smulwb  r8, r4, r6  ; (ip[4] * sinpi8sqrt2) >> 16                                       4s
-    pkhbt   r7, r7, r9, lsl #16 ; 5c | 4c
-    smulwt  r11, r3, r12    ; (ip[13] * cospi8sqrt2minus1) >> 16                                                    13c
-    pkhbt   r8, r8, r10, lsl #16    ; 5s | 4s
-    uadd16  r6, r6, r7  ; 5c+5 | 4c+4
-    smulwt  r7, r4, r12 ; (ip[13] * sinpi8sqrt2) >> 16                                  13s
-    smulwb  r9, r3, r12 ; (ip[12] * cospi8sqrt2minus1) >> 16                                            12c
-    smulwb  r10, r4, r12    ; (ip[12] * sinpi8sqrt2) >> 16                                              12s
-    subs    r5, r5, #0x1    ; i--                           --
-    pkhbt   r9, r9, r11, lsl #16    ; 13c | 12c
-    ldr r11, [r0], #0x4 ; i1 | i0       ++                                          1|0
-    pkhbt   r10, r10, r7, lsl #16   ; 13s | 12s
-    uadd16  r7, r12, r9 ; 13c+13 | 12c+12
-    usub16  r7, r8, r7  ; c                                 c
-    uadd16  r6, r6, r10 ; d                             d
-    uadd16  r10, r11, r14   ; a                                             a
-    usub16  r8, r11, r14    ; b                                     b
-    uadd16  r9, r10, r6 ; a+d                                           a+d
-    usub16  r10, r10, r6    ; a-d                                               a-d
-    uadd16  r6, r8, r7  ; b+c                               b+c
-    usub16  r7, r8, r7  ; b-c                                   b-c
-    str r6, [r1, r2]    ; o5 | o4
-    add r6, r2, r2  ; pitch * 2                             p2
-    str r7, [r1, r6]    ; o9 | o8
-    add r6,  r6, r2 ; pitch * 3                             p3
-    str r10, [r1, r6]   ; o13 | o12
-    str r9, [r1], #0x4  ; o1 | o0           ++
-    bne loop1_dual_1    ;
-    mov r5, #0x2    ; i=2                           i
-    sub r0, r1, #8  ; reset input/output        i/o
-loop2_dual_2
-    ldr r6, [r0, r2]    ; i5 | i4                               5|4
-    ldr r1, [r0]    ; i1 | i0           1|0
-    ldr r12, [r0, #0x4] ; i3 | i2                                                       3|2
-    add r14, r2, #0x4   ; pitch + 2                                                             p+2
-    ldr r14, [r0, r14]  ; i7 | i6                                                               7|6
-    smulwt  r9, r3, r6  ; (ip[5] * cospi8sqrt2minus1) >> 16                                         5c
-    smulwt  r7, r3, r1  ; (ip[1] * cospi8sqrt2minus1) >> 16                                 1c
-    smulwt  r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16                                               5s
-    smulwt  r8, r4, r1  ; (ip[1] * sinpi8sqrt2) >> 16                                       1s
-    pkhbt   r11, r6, r1, lsl #16    ; i0 | i4                                                   0|4
-    pkhbt   r7, r9, r7, lsl #16 ; 1c | 5c
-    pkhbt   r8, r10, r8, lsl #16    ; 1s | 5s = temp1 <20>                                     tc1
-    pkhtb   r1, r1, r6, asr #16 ; i1 | i5           1|5
-    uadd16  r1, r7, r1  ; 1c+1 | 5c+5 = temp2 (d)           td2
-    pkhbt   r9, r14, r12, lsl #16   ; i2 | i6                                           2|6
-    uadd16  r10, r11, r9    ; a                                             a
-    usub16  r9, r11, r9 ; b                                         b
-    pkhtb   r6, r12, r14, asr #16   ; i3 | i7                               3|7
-    subs    r5, r5, #0x1    ; i--                           --
-    smulwt  r7, r3, r6  ; (ip[3] * cospi8sqrt2minus1) >> 16                                 3c
-    smulwt  r11, r4, r6 ; (ip[3] * sinpi8sqrt2) >> 16                                                   3s
-    smulwb  r12, r3, r6 ; (ip[7] * cospi8sqrt2minus1) >> 16                                                     7c
-    smulwb  r14, r4, r6 ; (ip[7] * sinpi8sqrt2) >> 16                                                               7s
-
-    pkhbt   r7, r12, r7, lsl #16    ; 3c | 7c
-    pkhbt   r11, r14, r11, lsl #16  ; 3s | 7s = temp1 (d)                                                   td1
-    uadd16  r6, r7, r6  ; 3c+3 | 7c+7 = temp2  (c)                              tc2
-    usub16  r12, r8, r6 ; c (o1 | o5)                                                       c
-    uadd16  r6, r11, r1 ; d (o3 | o7)                               d
-    uadd16  r7, r10, r6 ; a+d                                   a+d
-    mov r8, #0x4    ; set up 4's                                        4
-    orr r8, r8, #0x40000    ;                                       4|4
-    usub16  r6, r10, r6 ; a-d                               a-d
-    uadd16  r6, r6, r8  ; a-d+4                             3|7
-    uadd16  r7, r7, r8  ; a+d+4                                 0|4
-    uadd16  r10, r9, r12    ; b+c                                               b+c
-    usub16  r1, r9, r12 ; b-c           b-c
-    uadd16  r10, r10, r8    ; b+c+4                                             1|5
-    uadd16  r1, r1, r8  ; b-c+4         2|6
-    mov r8, r10, asr #19    ; o1 >> 3
-    strh    r8, [r0, #2]    ; o1
-    mov r8, r1, asr #19 ; o2 >> 3
-    strh    r8, [r0, #4]    ; o2
-    mov r8, r6, asr #19 ; o3 >> 3
-    strh    r8, [r0, #6]    ; o3
-    mov r8, r7, asr #19 ; o0 >> 3
-    strh    r8, [r0], r2    ; o0        +p
-    sxth    r10, r10    ;
-    mov r8, r10, asr #3 ; o5 >> 3
-    strh    r8, [r0, #2]    ; o5
-    sxth    r1, r1  ;
-    mov r8, r1, asr #3  ; o6 >> 3
-    strh    r8, [r0, #4]    ; o6
-    sxth    r6, r6  ;
-    mov r8, r6, asr #3  ; o7 >> 3
-    strh    r8, [r0, #6]    ; o7
-    sxth    r7, r7  ;
-    mov r8, r7, asr #3  ; o4 >> 3
-    strh    r8, [r0], r2    ; o4        +p
-;;;;;   subs    r5, r5, #0x1    ; i--                           --
-    bne loop2_dual_2    ;
-            ;
-
-;vpx_memset
-    ldr     r0, [sp]
-    add     sp, sp, #4
-
-    mov     r12, #0
-    str     r12, [r0]
-    str     r12, [r0, #4]
-    str     r12, [r0, #8]
-    str     r12, [r0, #12]
-    str     r12, [r0, #16]
-    str     r12, [r0, #20]
-    str     r12, [r0, #24]
-    str     r12, [r0, #28]
-
-    ldmia   sp!, {r4 - r11, pc} ; replace vars, return                      restore
-
-    ENDP    ;|vp8_dequant_idct_v6|
-
-    END

vp8/decoder/arm/armv6/idct_blk_v6.c

@@ -0,0 +1,151 @@
+/*
+ *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "idct.h"
+#include "dequantize.h"
+
+void vp8_dequant_dc_idct_add_y_block_v6
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs, short *dc)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_dc_idct_add_v6 (q, dq, pre, dst, 16, stride, dc[0]);
+        else
+            vp8_dc_only_idct_add_v6 (dc[0], pre, dst, 16, stride);
+
+        if (eobs[1] > 1)
+            vp8_dequant_dc_idct_add_v6 (q+16, dq, pre+4, dst+4, 16, stride, dc[1]);
+        else
+            vp8_dc_only_idct_add_v6 (dc[1], pre+4, dst+4, 16, stride);
+
+        if (eobs[2] > 1)
+            vp8_dequant_dc_idct_add_v6 (q+32, dq, pre+8, dst+8, 16, stride, dc[2]);
+        else
+            vp8_dc_only_idct_add_v6 (dc[2], pre+8, dst+8, 16, stride);
+
+        if (eobs[3] > 1)
+            vp8_dequant_dc_idct_add_v6 (q+48, dq, pre+12, dst+12, 16, stride, dc[3]);
+        else
+            vp8_dc_only_idct_add_v6 (dc[3], pre+12, dst+12, 16, stride);
+
+        q    += 64;
+        dc   += 4;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_y_block_v6
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_v6 (q, dq, pre, dst, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[0]*dq[0], pre, dst, 16, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_v6 (q+16, dq, pre+4, dst+4, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[16]*dq[0], pre+4, dst+4, 16, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        if (eobs[2] > 1)
+            vp8_dequant_idct_add_v6 (q+32, dq, pre+8, dst+8, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[32]*dq[0], pre+8, dst+8, 16, stride);
+            ((int *)(q+32))[0] = 0;
+        }
+
+        if (eobs[3] > 1)
+            vp8_dequant_idct_add_v6 (q+48, dq, pre+12, dst+12, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[48]*dq[0], pre+12, dst+12, 16, stride);
+            ((int *)(q+48))[0] = 0;
+        }
+
+        q    += 64;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_v6
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_v6 (q, dq, pre, dstu, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[0]*dq[0], pre, dstu, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_v6 (q+16, dq, pre+4, dstu+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[16]*dq[0], pre+4, dstu+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstu += 4*stride;
+        eobs += 2;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_v6 (q, dq, pre, dstv, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[0]*dq[0], pre, dstv, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_v6 (q+16, dq, pre+4, dstv+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_v6 (q[16]*dq[0], pre+4, dstv+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstv += 4*stride;
+        eobs += 2;
+    }
+}

vp8/decoder/arm/dequantize_arm.h

@@ -14,32 +14,56 @@
 
 #if HAVE_ARMV6
 extern prototype_dequant_block(vp8_dequantize_b_v6);
-extern prototype_dequant_idct(vp8_dequant_idct_v6);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_v6);
-extern prototype_dequant_idct_dc(vp8_dequant_dc_idct_v6);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_v6);
+extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_v6);
+extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_v6);
+extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_v6);
 
 #undef  vp8_dequant_block
 #define vp8_dequant_block vp8_dequantize_b_v6
 
-#undef  vp8_dequant_idct
+#undef vp8_dequant_idct_add
-#define vp8_dequant_idct vp8_dequant_idct_v6
+#define vp8_dequant_idct_add vp8_dequant_idct_add_v6
 
-#undef  vp8_dequant_idct_dc
+#undef vp8_dequant_dc_idct_add
-#define vp8_dequant_idct_dc vp8_dequant_dc_idct_v6
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_v6
+
+#undef vp8_dequant_dc_idct_add_y_block
+#define vp8_dequant_dc_idct_add_y_block vp8_dequant_dc_idct_add_y_block_v6
+
+#undef vp8_dequant_idct_add_y_block
+#define vp8_dequant_idct_add_y_block vp8_dequant_idct_add_y_block_v6
+
+#undef vp8_dequant_idct_add_uv_block
+#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_v6
 #endif
 
 #if HAVE_ARMV7
 extern prototype_dequant_block(vp8_dequantize_b_neon);
-extern prototype_dequant_idct(vp8_dequant_idct_neon);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_neon);
-extern prototype_dequant_idct_dc(vp8_dequant_dc_idct_neon);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_neon);
+extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_neon);
+extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_neon);
+extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_neon);
 
 #undef  vp8_dequant_block
 #define vp8_dequant_block vp8_dequantize_b_neon
 
-#undef  vp8_dequant_idct
+#undef vp8_dequant_idct_add
-#define vp8_dequant_idct vp8_dequant_idct_neon
+#define vp8_dequant_idct_add vp8_dequant_idct_add_neon
 
-#undef  vp8_dequant_idct_dc
+#undef vp8_dequant_dc_idct_add
-#define vp8_dequant_idct_dc vp8_dequant_dc_idct_neon
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_neon
+
+#undef vp8_dequant_dc_idct_add_y_block
+#define vp8_dequant_dc_idct_add_y_block vp8_dequant_dc_idct_add_y_block_neon
+
+#undef vp8_dequant_idct_add_y_block
+#define vp8_dequant_idct_add_y_block vp8_dequant_idct_add_y_block_neon
+
+#undef vp8_dequant_idct_add_uv_block
+#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_neon
 #endif
 
 #endif

vp8/decoder/arm/detokenize.asm

@@ -0,0 +1,320 @@
+;
+;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license
+;  that can be found in the LICENSE file in the root of the source
+;  tree. An additional intellectual property rights grant can be found
+;  in the file PATENTS.  All contributing project authors may
+;  be found in the AUTHORS file in the root of the source tree.
+;
+
+
+    EXPORT  |vp8_decode_mb_tokens_v6|
+
+    AREA    |.text|, CODE, READONLY  ; name this block of code
+
+    INCLUDE vpx_asm_offsets.asm
+
+l_qcoeff    EQU     0
+l_i         EQU     4
+l_type      EQU     8
+l_stop      EQU     12
+l_c         EQU     16
+l_l_ptr     EQU     20
+l_a_ptr     EQU     24
+l_bc        EQU     28
+l_coef_ptr  EQU     32
+l_stacksize EQU     64
+
+
+;; constant offsets -- these should be created at build time
+c_block2above_offset         EQU 25
+c_entropy_nodes              EQU 11
+c_dct_eob_token              EQU 11
+
+|vp8_decode_mb_tokens_v6| PROC
+    stmdb       sp!, {r4 - r11, lr}
+    sub         sp, sp, #l_stacksize
+    mov         r7, r1                      ; type
+    mov         r9, r0                      ; detoken
+
+    ldr         r1, [r9, #detok_current_bc]
+    ldr         r0, [r9, #detok_qcoeff_start_ptr]
+    mov         r11, #0                     ; i
+    mov         r3, #16                     ; stop
+
+    cmp         r7, #1                      ; type ?= 1
+    addeq       r11, r11, #24               ; i = 24
+    addeq       r3, r3, #8                  ; stop = 24
+    addeq       r0, r0, #3, 24              ; qcoefptr += 24*16
+
+    str         r0, [sp, #l_qcoeff]
+    str         r11, [sp, #l_i]
+    str         r7, [sp, #l_type]
+    str         r3, [sp, #l_stop]
+    str         r1, [sp, #l_bc]
+
+    add         lr, r9, r7, lsl #2          ; detoken + type*4
+
+    ldr         r8, [r1, #bool_decoder_user_buffer]
+
+    ldr         r10, [lr, #detok_coef_probs]
+    ldr         r5, [r1, #bool_decoder_count]
+    ldr         r6, [r1, #bool_decoder_range]
+    ldr         r4, [r1, #bool_decoder_value]
+
+    str         r10, [sp, #l_coef_ptr]
+
+BLOCK_LOOP
+    ldr         r3, [r9, #detok_ptr_block2leftabove]
+    ldr         r1, [r9, #detok_L]
+    ldr         r2, [r9, #detok_A]
+    ldrb        r12, [r3, r11]!             ; block2left[i]
+    ldrb        r3, [r3, #c_block2above_offset]; block2above[i]
+
+    cmp         r7, #0                      ; c = !type
+    moveq       r7, #1
+    movne       r7, #0
+
+    ldrb        r0, [r1, r12]!              ; *(L += block2left[i])
+    ldrb        r3, [r2, r3]!               ; *(A += block2above[i])
+    mov         lr, #c_entropy_nodes        ; ENTROPY_NODES = 11
+
+; VP8_COMBINEENTROPYCONTETEXTS(t, *a, *l) => t = ((*a) != 0) + ((*l) !=0)
+    cmp         r0, #0                      ; *l ?= 0
+    movne       r0, #1
+    cmp         r3, #0                      ; *a ?= 0
+    addne       r0, r0, #1                  ; t
+
+    str         r1, [sp, #l_l_ptr]          ; save &l
+    str         r2, [sp, #l_a_ptr]          ; save &a
+    smlabb      r0, r0, lr, r10             ; Prob = coef_probs + (t * ENTROPY_NODES)
+    mov         r1, #0                      ; t = 0
+    str         r7, [sp, #l_c]
+
+    ;align 4
+COEFF_LOOP
+    ldr         r3, [r9, #detok_ptr_coef_bands_x]
+    ldr         lr, [r9, #detok_coef_tree_ptr]
+    ;STALL
+    ldrb        r3, [r3, r7]                ; coef_bands_x[c]
+    ;STALL
+    ;STALL
+    add         r0, r0, r3                  ; Prob += coef_bands_x[c]
+
+get_token_loop
+    ldrb        r2, [r0, +r1, asr #1]       ; Prob[t >> 1]
+    mov         r3, r6, lsl #8              ; range << 8
+    sub         r3, r3, #256                ; (range << 8) - (1 << 8)
+    mov         r10, #1                     ; 1
+
+    smlawb      r2, r3, r2, r10             ; split = 1 + (((range-1) * probability) >> 8)
+
+    ldrb        r12, [r8]                   ; load cx data byte in stall slot : r8 = bufptr
+    ;++
+
+    subs        r3, r4, r2, lsl #24         ; value-(split<<24): used later to calculate shift for NORMALIZE
+    addhs       r1, r1, #1                  ; t += 1
+    movhs       r4, r3                      ; value -= bigsplit (split << 24)
+    subhs       r2, r6, r2                  ; range -= split
+ ;   movlo       r6, r2                      ; range = split
+
+    ldrsb     r1, [lr, r1]                  ; t = onyx_coef_tree_ptr[t]
+
+; NORMALIZE
+    clz         r3, r2                      ; vp8dx_bitreader_norm[range] + 24
+    sub         r3, r3, #24                 ; vp8dx_bitreader_norm[range]
+    subs        r5, r5, r3                  ; count -= shift
+    mov         r6, r2, lsl r3              ; range <<= shift
+    mov         r4, r4, lsl r3              ; value <<= shift
+
+; if count <= 0, += BR_COUNT; value |= *bufptr++ << (BR_COUNT-count); BR_COUNT = 8, but need to upshift values by +16
+    addle         r5, r5, #8                ; count += 8
+    rsble         r3, r5, #24               ; 24 - count
+    addle         r8, r8, #1                ; bufptr++
+    orrle         r4, r4, r12, lsl r3       ; value |= *bufptr << shift + 16
+
+    cmp         r1, #0                      ; t ?= 0
+    bgt         get_token_loop              ; while (t > 0)
+
+    cmn         r1, #c_dct_eob_token        ; if(t == -DCT_EOB_TOKEN)
+    beq         END_OF_BLOCK                ; break
+
+    rsb         lr, r1, #0                  ; v = -t;
+
+    cmp         lr, #4                      ; if(v > FOUR_TOKEN)
+    ble         SKIP_EXTRABITS
+
+    ldr         r3, [r9, #detok_teb_base_ptr]
+    mov         r11, #1                     ; 1 in split = 1 + ... nope, v+= 1 << bits_count
+    add         r7, r3, lr, lsl #4          ; detok_teb_base_ptr + (v << 4)
+
+    ldrsh       lr, [r7, #tokenextrabits_min_val] ; v = teb_ptr->min_val
+    ldrsh       r0, [r7, #tokenextrabits_length] ; bits_count = teb_ptr->Length
+
+extrabits_loop
+    add         r3, r0, r7                  ; &teb_ptr->Probs[bits_count]
+
+    ldrb        r2, [r3, #4]                ; probability. why +4?
+    mov         r3, r6, lsl #8              ; range << 8
+    sub         r3, r3, #256                ; range << 8 + 1 << 8
+
+    smlawb      r2, r3, r2, r11             ; split = 1 +  (((range-1) * probability) >> 8)
+
+    ldrb        r12, [r8]                   ; *bufptr
+    ;++
+
+    subs        r10, r4, r2, lsl #24        ; value - (split<<24)
+    movhs       r4, r10                     ; value = value - (split << 24)
+    subhs       r2, r6, r2                  ; range = range - split
+    addhs       lr, lr, r11, lsl r0         ; v += ((UINT16)1<<bits_count)
+
+; NORMALIZE
+    clz         r3, r2                      ; shift - leading zeros in split
+    sub         r3, r3, #24                 ; don't count first 3 bytes
+    subs        r5, r5, r3                  ; count -= shift
+    mov         r6, r2, lsl r3              ; range = range << shift
+    mov         r4, r4, lsl r3              ; value <<= shift
+
+    addle       r5, r5, #8                  ; count += BR_COUNT
+    addle       r8, r8, #1                  ; bufptr++
+    rsble       r3, r5, #24                 ; BR_COUNT - count
+    orrle       r4, r4, r12, lsl r3         ; value |= *bufptr << (BR_COUNT - count)
+
+    subs        r0, r0, #1                  ; bits_count --
+    bpl         extrabits_loop
+
+
+SKIP_EXTRABITS
+    ldr         r11, [sp, #l_qcoeff]
+    ldr         r0, [sp, #l_coef_ptr]       ; Prob = coef_probs
+
+    cmp         r1, #0                      ; check for nonzero token - if (t)
+    beq         SKIP_EOB_CHECK              ; if t is zero, we will skip the eob table chec
+
+    add         r3, r6, #1                  ; range + 1
+    mov         r2, r3, lsr #1              ; split = (range + 1) >> 1
+
+    subs        r3, r4, r2, lsl #24         ; value - (split<<24)
+    movhs       r4, r3                      ; value -= (split << 24)
+    subhs       r2, r6, r2                  ; range -= split
+    mvnhs       r3, lr                      ; -v
+    addhs       lr, r3, #1                  ; v = (v ^ -1) + 1
+
+; NORMALIZE
+    clz         r3, r2                      ; leading 0s in split
+    sub         r3, r3, #24                 ; shift
+    subs        r5, r5, r3                  ; count -= shift
+    mov         r6, r2, lsl r3              ; range <<= shift
+    mov         r4, r4, lsl r3              ; value <<= shift
+    ldrleb      r2, [r8], #1                ; *(bufptr++)
+    addle       r5, r5, #8                  ; count += 8
+    rsble       r3, r5, #24                 ; BR_COUNT - count
+    orrle       r4, r4, r2, lsl r3          ; value |= *bufptr << (BR_COUNT - count)
+
+    add         r0, r0, #11                 ; Prob += ENTROPY_NODES (11)
+
+    cmn         r1, #1                      ; t < -ONE_TOKEN
+
+    addlt       r0, r0, #11                 ; Prob += ENTROPY_NODES (11)
+
+    mvn         r1, #1                      ; t = -1 ???? C is -2
+
+SKIP_EOB_CHECK
+    ldr         r7, [sp, #l_c]              ; c
+    ldr         r3, [r9, #detok_scan]
+    add         r1, r1, #2                  ; t+= 2
+    cmp         r7, #15                     ; c should will be one higher
+
+    ldr         r3, [r3, +r7, lsl #2]       ; scan[c] this needs pre-inc c value
+    add         r7, r7, #1                  ; c++
+    add         r3, r11, r3, lsl #1         ; qcoeff + scan[c]
+
+    str         r7, [sp, #l_c]              ; store c
+    strh        lr, [r3]                    ; qcoef_ptr[scan[c]] = v
+
+    blt         COEFF_LOOP
+
+    sub         r7, r7, #1                  ; if(t != -DCT_EOB_TOKEN) --c
+
+END_OF_BLOCK
+    ldr         r3, [sp, #l_type]           ; type
+    ldr         r10, [sp, #l_coef_ptr]      ; coef_ptr
+    ldr         r0, [sp, #l_qcoeff]         ; qcoeff
+    ldr         r11, [sp, #l_i]             ; i
+    ldr         r12, [sp, #l_stop]          ; stop
+
+    cmp         r3, #0                      ; type ?= 0
+    moveq       r1, #1
+    movne       r1, #0
+    add         r3, r11, r9                 ; detok + i
+
+    cmp         r7, r1                      ; c ?= !type
+    strb        r7, [r3, #detok_eob]        ; eob[i] = c
+
+    ldr         r7, [sp, #l_l_ptr]          ; l
+    ldr         r2, [sp, #l_a_ptr]          ; a
+    movne       r3, #1                      ; t
+    moveq       r3, #0
+
+    add         r0, r0, #32                 ; qcoeff += 32 (16 * 2?)
+    add         r11, r11, #1                ; i++
+    strb        r3, [r7]                    ; *l = t
+    strb        r3, [r2]                    ; *a = t
+    str         r0, [sp, #l_qcoeff]         ; qcoeff
+    str         r11, [sp, #l_i]             ; i
+
+    cmp         r11, r12                    ; i < stop
+    ldr         r7, [sp, #l_type]           ; type
+
+    blt         BLOCK_LOOP
+
+    cmp         r11, #25                    ; i ?= 25
+    bne         ln2_decode_mb_to
+
+    ldr         r12, [r9, #detok_qcoeff_start_ptr]
+    ldr         r10, [r9, #detok_coef_probs]
+    mov         r7, #0                      ; type/i = 0
+    mov         r3, #16                     ; stop = 16
+    str         r12, [sp, #l_qcoeff]        ; qcoeff_ptr = qcoeff_start_ptr
+    str         r7, [sp, #l_i]
+    str         r7, [sp, #l_type]
+    str         r3, [sp, #l_stop]
+
+    str         r10, [sp, #l_coef_ptr]      ; coef_probs = coef_probs[type=0]
+
+    b           BLOCK_LOOP
+
+ln2_decode_mb_to
+    cmp         r11, #16                    ; i ?= 16
+    bne         ln1_decode_mb_to
+
+    mov         r10, #detok_coef_probs
+    add         r10, r10, #2*4              ; coef_probs[type]
+    ldr         r10, [r9, r10]              ; detok + detok_coef_probs[type]
+
+    mov         r7, #2                      ; type = 2
+    mov         r3, #24                     ; stop = 24
+
+    str         r7, [sp, #l_type]
+    str         r3, [sp, #l_stop]
+
+    str         r10, [sp, #l_coef_ptr]      ; coef_probs = coef_probs[type]
+    b           BLOCK_LOOP
+
+ln1_decode_mb_to
+    ldr         r2, [sp, #l_bc]
+    mov         r0, #0
+    nop
+
+    str         r8, [r2, #bool_decoder_user_buffer]
+    str         r5, [r2, #bool_decoder_count]
+    str         r4, [r2, #bool_decoder_value]
+    str         r6, [r2, #bool_decoder_range]
+
+    add         sp, sp, #l_stacksize
+    ldmia       sp!, {r4 - r11, pc}
+
+    ENDP  ; |vp8_decode_mb_tokens_v6|
+
+    END

vp8/decoder/arm/detokenize_arm.h

@@ -7,13 +7,16 @@
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
-#ifndef BIT_OPS_H
-#define BIT_OPS_H
 
-/* Evaluates to a mask with n bits set */
-#define BITS_MASK(n) ((1<<(n))-1)
 
-/* Returns len bits, with the LSB at position bit */
+#ifndef DETOKENIZE_ARM_H
-#define BITS_GET(val, bit, len) (((val)>>(bit))&BITS_MASK(len))
+#define DETOKENIZE_ARM_H
+
+#if HAVE_ARMV6
+#if CONFIG_ARM_ASM_DETOK
+void vp8_init_detokenizer(VP8D_COMP *dx);
+void vp8_decode_mb_tokens_v6(DETOK *detoken, int type);
+#endif
+#endif
 
 #endif

vp8/decoder/arm/detokenizearm_sjl.c

@@ -1,731 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "type_aliases.h"
-#include "blockd.h"
-#include "onyxd_int.h"
-#include "vpx_mem/vpx_mem.h"
-#include "vpx_ports/mem.h"
-
-#define BR_COUNT 8
-#define BOOL_DATA UINT8
-
-#define OCB_X PREV_COEF_CONTEXTS * ENTROPY_NODES
-//ALIGN16 UINT16 onyx_coef_bands_x[16] = { 0, 1*OCB_X, 2*OCB_X, 3*OCB_X, 6*OCB_X, 4*OCB_X, 5*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 6*OCB_X, 7*OCB_X};
-DECLARE_ALIGNED(16, UINT8, vp8_coef_bands_x[16]) = { 0, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 6 * OCB_X, 4 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 7 * OCB_X};
-
-#define EOB_CONTEXT_NODE            0
-#define ZERO_CONTEXT_NODE           1
-#define ONE_CONTEXT_NODE            2
-#define LOW_VAL_CONTEXT_NODE        3
-#define TWO_CONTEXT_NODE            4
-#define THREE_CONTEXT_NODE          5
-#define HIGH_LOW_CONTEXT_NODE       6
-#define CAT_ONE_CONTEXT_NODE        7
-#define CAT_THREEFOUR_CONTEXT_NODE  8
-#define CAT_THREE_CONTEXT_NODE      9
-#define CAT_FIVE_CONTEXT_NODE       10
-
-
-
-
-DECLARE_ALIGNED(16, static const TOKENEXTRABITS, vp8d_token_extra_bits2[MAX_ENTROPY_TOKENS]) =
-{
-    {  0, -1, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },  //ZERO_TOKEN
-    {  1, 0, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },   //ONE_TOKEN
-    {  2, 0, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },   //TWO_TOKEN
-    {  3, 0, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },   //THREE_TOKEN
-    {  4, 0, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },   //FOUR_TOKEN
-    {  5, 0, { 159, 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },  //DCT_VAL_CATEGORY1
-    {  7, 1, { 145, 165, 0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } }, //DCT_VAL_CATEGORY2
-    { 11, 2, { 140, 148, 173, 0,  0,  0,  0,  0,  0,  0,  0,  0   } }, //DCT_VAL_CATEGORY3
-    { 19, 3, { 135, 140, 155, 176, 0,  0,  0,  0,  0,  0,  0,  0   } }, //DCT_VAL_CATEGORY4
-    { 35, 4, { 130, 134, 141, 157, 180, 0,  0,  0,  0,  0,  0,  0   } }, //DCT_VAL_CATEGORY5
-    { 67, 10, { 129, 130, 133, 140, 153, 177, 196, 230, 243, 254, 254, 0   } }, //DCT_VAL_CATEGORY6
-    {  0, -1, { 0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0   } },  // EOB TOKEN
-};
-
-/*
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-*/
-DECLARE_ALIGNED(16, const UINT8, vp8_block2context_leftabove[25*3]) =
-{
-    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, //end of vp8_block2context
-    0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 1, 1, 0, 0, 1, 1, 0, //end of vp8_block2left
-    0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 0, 1, 0, 1, 0, 1, 0 //end of vp8_block2above
-};
-
-/*
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-*/
-
-void vp8_reset_mb_tokens_context(MACROBLOCKD *x)
-{
-    ENTROPY_CONTEXT **const A = x->above_context;
-    ENTROPY_CONTEXT(* const L)[4] = x->left_context;
-
-    ENTROPY_CONTEXT *a;
-    ENTROPY_CONTEXT *l;
-    int i;
-
-    for (i = 0; i < 24; i++)
-    {
-
-        a = A[ vp8_block2context[i] ] + vp8_block2above[i];
-        l = L[ vp8_block2context[i] ] + vp8_block2left[i];
-
-        *a = *l = 0;
-    }
-
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
-    {
-        a = A[Y2CONTEXT] + vp8_block2above[24];
-        l = L[Y2CONTEXT] + vp8_block2left[24];
-        *a = *l = 0;
-    }
-
-
-}
-
-#define ONYXBLOCK2CONTEXT_OFFSET    0
-#define ONYXBLOCK2LEFT_OFFSET       25
-#define ONYXBLOCK2ABOVE_OFFSET 50
-
-DECLARE_ALIGNED(16, const static unsigned char, norm[128]) =
-{
-    0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-    2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-    1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
-};
-
-/*
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-*/
-void init_detokenizer(VP8D_COMP *dx)
-{
-    const VP8_COMMON *const oc = & dx->common;
-    MACROBLOCKD *x = & dx->mb;
-
-    dx->detoken.norm_ptr = (unsigned char *)norm;
-    dx->detoken.vp8_coef_tree_ptr = (vp8_tree_index *)vp8_coef_tree;
-    dx->detoken.ptr_onyxblock2context_leftabove = (UINT8 *)vp8_block2context_leftabove;
-    dx->detoken.ptr_onyx_coef_bands_x = vp8_coef_bands_x;
-    dx->detoken.scan = (int *)vp8_default_zig_zag1d;
-    dx->detoken.teb_base_ptr = (TOKENEXTRABITS *)vp8d_token_extra_bits2;
-
-    dx->detoken.qcoeff_start_ptr = &x->qcoeff[0];
-
-
-    dx->detoken.coef_probs[0] = (unsigned char *)(oc->fc.coef_probs [0] [ 0 ] [0]);
-    dx->detoken.coef_probs[1] = (unsigned char *)(oc->fc.coef_probs [1] [ 0 ] [0]);
-    dx->detoken.coef_probs[2] = (unsigned char *)(oc->fc.coef_probs [2] [ 0 ] [0]);
-    dx->detoken.coef_probs[3] = (unsigned char *)(oc->fc.coef_probs [3] [ 0 ] [0]);
-
-}
-
-/*
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-*/
-
-
-//shift = norm[range]; \
-//      shift = norm_ptr[range]; \
-
-#define NORMALIZE \
-    /*if(range < 0x80)*/                            \
-    { \
-        shift = detoken->norm_ptr[range]; \
-        range <<= shift; \
-        value <<= shift; \
-        count -= shift; \
-        if(count <= 0) \
-        { \
-            count += BR_COUNT ; \
-            value |= (*bufptr) << (BR_COUNT-count); \
-            bufptr++; \
-        } \
-    }
-#if 1
-#define DECODE_AND_APPLYSIGN(value_to_sign) \
-    split = (range + 1) >> 1; \
-    if ( (value >> 24) < split ) \
-    { \
-        range = split; \
-        v= value_to_sign; \
-    } \
-    else \
-    { \
-        range = range-split; \
-        value = value-(split<<24); \
-        v = -value_to_sign; \
-    } \
-    range +=range;                   \
-    value +=value;                   \
-    if (!--count) \
-    { \
-        count = BR_COUNT; \
-        value |= *bufptr; \
-        bufptr++; \
-    }
-
-#define DECODE_AND_BRANCH_IF_ZERO(probability,branch) \
-    { \
-        split = 1 +  ((( probability*(range-1) ) )>> 8); \
-        if ( (value >> 24) < split ) \
-        { \
-            range = split; \
-            NORMALIZE \
-            goto branch; \
-        } \
-        value -= (split<<24); \
-        range = range - split; \
-        NORMALIZE \
-    }
-
-#define DECODE_AND_LOOP_IF_ZERO(probability,branch) \
-    { \
-        split = 1 + ((( probability*(range-1) ) ) >> 8); \
-        if ( (value >> 24) < split ) \
-        { \
-            range = split; \
-            NORMALIZE \
-            Prob = coef_probs; \
-            ++c; \
-            Prob += vp8_coef_bands_x[c]; \
-            goto branch; \
-        } \
-        value -= (split<<24); \
-        range = range - split; \
-        NORMALIZE \
-    }
-
-#define DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val) \
-    DECODE_AND_APPLYSIGN(val) \
-    Prob = coef_probs + (ENTROPY_NODES*2); \
-    if(c < 15){\
-        qcoeff_ptr [ scan[c] ] = (INT16) v; \
-        ++c; \
-        goto DO_WHILE; }\
-    qcoeff_ptr [ scan[15] ] = (INT16) v; \
-    goto BLOCK_FINISHED;
-
-
-#define DECODE_EXTRABIT_AND_ADJUST_VAL(t,bits_count)\
-    split = 1 +  (((range-1) * vp8d_token_extra_bits2[t].Probs[bits_count]) >> 8); \
-    if(value >= (split<<24))\
-    {\
-        range = range-split;\
-        value = value-(split<<24);\
-        val += ((UINT16)1<<bits_count);\
-    }\
-    else\
-    {\
-        range = split;\
-    }\
-    NORMALIZE
-#endif
-
-#if 0
-int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
-{
-    ENTROPY_CONTEXT **const A = x->above_context;
-    ENTROPY_CONTEXT(* const L)[4] = x->left_context;
-    const VP8_COMMON *const oc = & dx->common;
-
-    BOOL_DECODER *bc = x->current_bc;
-
-    ENTROPY_CONTEXT *a;
-    ENTROPY_CONTEXT *l;
-    int i;
-
-    int eobtotal = 0;
-
-    register int count;
-
-    BOOL_DATA *bufptr;
-    register unsigned int range;
-    register unsigned int value;
-    const int *scan;
-    register unsigned int shift;
-    UINT32 split;
-    INT16 *qcoeff_ptr;
-
-    UINT8 *coef_probs;
-    int type;
-    int stop;
-    INT16 val, bits_count;
-    INT16 c;
-    INT16 t;
-    INT16 v;
-    vp8_prob *Prob;
-
-    //int *scan;
-    type = 3;
-    i = 0;
-    stop = 16;
-
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
-    {
-        i = 24;
-        stop = 24;
-        type = 1;
-        qcoeff_ptr = &x->qcoeff[24*16];
-        scan = vp8_default_zig_zag1d;
-        eobtotal -= 16;
-    }
-    else
-    {
-        scan = vp8_default_zig_zag1d;
-        qcoeff_ptr = &x->qcoeff[0];
-    }
-
-    count   = bc->count;
-    range   = bc->range;
-    value   = bc->value;
-    bufptr  = &bc->buffer[bc->pos];
-
-
-    coef_probs = (unsigned char *)(oc->fc.coef_probs [type] [ 0 ] [0]);
-
-BLOCK_LOOP:
-    a = A[ vp8_block2context[i] ] + vp8_block2above[i];
-    l = L[ vp8_block2context[i] ] + vp8_block2left[i];
-    c = (INT16)(!type);
-
-    VP8_COMBINEENTROPYCONTEXTS(t, *a, *l);
-    Prob = coef_probs;
-    Prob += t * ENTROPY_NODES;
-
-DO_WHILE:
-    Prob += vp8_coef_bands_x[c];
-    DECODE_AND_BRANCH_IF_ZERO(Prob[EOB_CONTEXT_NODE], BLOCK_FINISHED);
-
-CHECK_0_:
-    DECODE_AND_LOOP_IF_ZERO(Prob[ZERO_CONTEXT_NODE], CHECK_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[ONE_CONTEXT_NODE], ONE_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[LOW_VAL_CONTEXT_NODE], LOW_VAL_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[HIGH_LOW_CONTEXT_NODE], HIGH_LOW_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[CAT_THREEFOUR_CONTEXT_NODE], CAT_THREEFOUR_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[CAT_FIVE_CONTEXT_NODE], CAT_FIVE_CONTEXT_NODE_0_);
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY6].min_val;
-    bits_count = vp8d_token_extra_bits2[DCT_VAL_CATEGORY6].Length;
-
-    do
-    {
-        DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY6, bits_count);
-        bits_count -- ;
-    }
-    while (bits_count >= 0);
-
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_FIVE_CONTEXT_NODE_0_:
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY5].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 4);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 3);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_THREEFOUR_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(Prob[CAT_THREE_CONTEXT_NODE], CAT_THREE_CONTEXT_NODE_0_);
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY4].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 3);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_THREE_CONTEXT_NODE_0_:
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY3].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-HIGH_LOW_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(Prob[CAT_ONE_CONTEXT_NODE], CAT_ONE_CONTEXT_NODE_0_);
-
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY2].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY2, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY2, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_ONE_CONTEXT_NODE_0_:
-    val = vp8d_token_extra_bits2[DCT_VAL_CATEGORY1].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY1, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-LOW_VAL_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(Prob[TWO_CONTEXT_NODE], TWO_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(Prob[THREE_CONTEXT_NODE], THREE_CONTEXT_NODE_0_);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(4);
-
-THREE_CONTEXT_NODE_0_:
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(3);
-
-TWO_CONTEXT_NODE_0_:
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(2);
-
-ONE_CONTEXT_NODE_0_:
-    DECODE_AND_APPLYSIGN(1);
-    Prob = coef_probs + ENTROPY_NODES;
-
-    if (c < 15)
-    {
-        qcoeff_ptr [ scan[c] ] = (INT16) v;
-        ++c;
-        goto DO_WHILE;
-    }
-
-    qcoeff_ptr [ scan[15] ] = (INT16) v;
-BLOCK_FINISHED:
-    t = ((x->Block[i].eob = c) != !type);   // any nonzero data?
-    eobtotal += x->Block[i].eob;
-    *a = *l = t;
-    qcoeff_ptr += 16;
-
-    i++;
-
-    if (i < stop)
-        goto BLOCK_LOOP;
-
-    if (i == 25)
-    {
-        scan = vp8_default_zig_zag1d;//x->scan_order1d;
-        type = 0;
-        i = 0;
-        stop = 16;
-        coef_probs = (unsigned char *)(oc->fc.coef_probs [type] [ 0 ] [0]);
-        qcoeff_ptr = &x->qcoeff[0];
-        goto BLOCK_LOOP;
-    }
-
-    if (i == 16)
-    {
-        type = 2;
-        coef_probs = (unsigned char *)(oc->fc.coef_probs [type] [ 0 ] [0]);
-        stop = 24;
-        goto BLOCK_LOOP;
-    }
-
-    bc->count = count;
-    bc->value = value;
-    bc->range = range;
-    bc->pos  = bufptr - bc->buffer;
-    return eobtotal;
-
-}
-//#endif
-#else
-/*
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-*/
-
-#if 0
-//uses relative offsets
-
-const vp8_tree_index vp8_coef_tree_x[ 22] =   /* corresponding _CONTEXT_NODEs */
-{
-    -DCT_EOB_TOKEN, 1,                             /* 0 = EOB */
-    -ZERO_TOKEN, 1,                               /* 1 = ZERO */
-    -ONE_TOKEN, 1,                               /* 2 = ONE */
-    2, 5,                                       /* 3 = LOW_VAL */
-    -TWO_TOKEN, 1,                         /* 4 = TWO */
-    -THREE_TOKEN, -FOUR_TOKEN,                /* 5 = THREE */
-    2, 3,                                  /* 6 = HIGH_LOW */
-    -DCT_VAL_CATEGORY1, -DCT_VAL_CATEGORY2,   /* 7 = CAT_ONE */
-    2, 3,                                 /* 8 = CAT_THREEFOUR */
-    -DCT_VAL_CATEGORY3, -DCT_VAL_CATEGORY4,  /* 9 = CAT_THREE */
-    -DCT_VAL_CATEGORY5, -DCT_VAL_CATEGORY6   /* 10 = CAT_FIVE */
-};
-#endif
-
-#define _SCALEDOWN 8 //16 //8
-
-int vp8_decode_mb_tokens_v5(DETOK *detoken, int type);
-
-int vp8_decode_mb_tokens_v5_c(DETOK *detoken, int type)
-{
-    BOOL_DECODER *bc = detoken->current_bc;
-
-    ENTROPY_CONTEXT *a;
-    ENTROPY_CONTEXT *l;
-    int i;
-
-    register int count;
-
-    BOOL_DATA *bufptr;
-    register unsigned int range;
-    register unsigned int value;
-    register unsigned int shift;
-    UINT32 split;
-    INT16 *qcoeff_ptr;
-
-    UINT8 *coef_probs;
-//  int type;
-    int stop;
-    INT16 c;
-    INT16 t;
-    INT16 v;
-    vp8_prob *Prob;
-
-
-
-//  type = 3;
-    i = 0;
-    stop = 16;
-    qcoeff_ptr = detoken->qcoeff_start_ptr;
-
-//  if( detoken->mode != B_PRED && detoken->mode != SPLITMV)
-    if (type == 1)
-    {
-        i += 24;
-        stop += 8; //24;
-//      type = 1;
-        qcoeff_ptr += 24 * 16;
-//      eobtotal-=16;
-    }
-
-    count   = bc->count;
-    range   = bc->range;
-    value   = bc->value;
-    bufptr  = &bc->buffer[bc->pos];
-
-
-    coef_probs = detoken->coef_probs[type]; //(unsigned char *)( oc->fc.coef_probs [type] [ 0 ] [0]);
-
-BLOCK_LOOP:
-    a = detoken->A[ detoken->ptr_onyxblock2context_leftabove[i] ];
-    l = detoken->L[ detoken->ptr_onyxblock2context_leftabove[i] ];
-    c = !type;
-    a += detoken->ptr_onyxblock2context_leftabove[i + ONYXBLOCK2ABOVE_OFFSET];
-    l += detoken->ptr_onyxblock2context_leftabove[i + ONYXBLOCK2LEFT_OFFSET];
-
-    //#define ONYX_COMBINEENTROPYCONTEXTS( Dest, A, B) \
-    //Dest = ((A)!=0) + ((B)!=0);
-
-    VP8_COMBINEENTROPYCONTEXTS(t, *a, *l);
-
-    Prob = coef_probs;
-    Prob += t * ENTROPY_NODES;
-    t = 0;
-
-    do
-    {
-
-        {
-//                  onyx_tree_index * onyx_coef_tree_ptr = onyx_coef_tree_x;
-
-            Prob += detoken->ptr_onyx_coef_bands_x[c];
-
-        GET_TOKEN_START:
-
-            do
-            {
-                split = 1 + (((range - 1) * (Prob[t>>1])) >> 8);
-
-                if (value >> 24 >= split)
-                {
-                    range = range - split;
-                    value = value - (split << 24);
-                    t += 1;
-
-                    //used to eliminate else branch
-                    split = range;
-                }
-
-                range = split;
-
-                t = detoken->vp8_coef_tree_ptr[ t ];
-
-                NORMALIZE
-
-            }
-            while (t  > 0) ;
-        }
-    GET_TOKEN_STOP:
-
-        if (t == -DCT_EOB_TOKEN)
-        {
-            break;
-        }
-
-        v = -t;
-
-        if (v > FOUR_TOKEN)
-        {
-            INT16 bits_count;
-            TOKENEXTRABITS *teb_ptr;
-
-//                      teb_ptr = &onyxd_token_extra_bits2[t];
-//                  teb_ptr = &onyxd_token_extra_bits2[v];
-            teb_ptr = &detoken->teb_base_ptr[v];
-
-
-            v = teb_ptr->min_val;
-            bits_count = teb_ptr->Length;
-
-            do
-            {
-                split = 1 + (((range - 1) * teb_ptr->Probs[bits_count]) >> _SCALEDOWN);
-
-                if ((value >> 24) >= split)
-                {
-                    range = range - split;
-                    value = value - (split << 24);
-                    v += ((UINT16)1 << bits_count);
-
-                    //used to eliminate else branch
-                    split = range;
-                }
-
-                range = split;
-
-                NORMALIZE
-
-                bits_count -- ;
-            }
-            while (bits_count >= 0);
-        }
-
-        Prob = coef_probs;
-
-        if (t)
-        {
-            split = 1 + (((range - 1) * vp8_prob_half) >> 8);
-
-            if ((value >> 24) >= split)
-            {
-                range = range - split;
-                value = value - (split << 24);
-                v = (v ^ -1) + 1;           /* negate w/out conditionals */
-
-                //used to eliminate else branch
-                split = range;
-            }
-
-            range = split;
-
-            NORMALIZE
-            Prob += ENTROPY_NODES;
-
-            if (t < -ONE_TOKEN)
-                Prob += ENTROPY_NODES;
-
-            t = -2;
-        }
-
-        //if t is zero, we will skip the eob table check
-        t += 2;
-        qcoeff_ptr [detoken->scan [c] ] = (INT16) v;
-
-    }
-    while (++c < 16);
-
-    if (t != -DCT_EOB_TOKEN)
-    {
-        --c;
-    }
-
-    t = ((detoken->eob[i] = c) != !type);   // any nonzero data?
-//  eobtotal += detoken->eob[i];
-    *a = *l = t;
-    qcoeff_ptr += 16;
-
-    i++;
-
-    if (i < stop)
-        goto BLOCK_LOOP;
-
-    if (i == 25)
-    {
-        type = 0;
-        i = 0;
-        stop = 16;
-//      coef_probs = (unsigned char *)(oc->fc.coef_probs [type] [ 0 ] [0]);
-        coef_probs = detoken->coef_probs[type]; //(unsigned char *)( oc->fc.coef_probs [type] [ 0 ] [0]);
-        qcoeff_ptr = detoken->qcoeff_start_ptr;
-        goto BLOCK_LOOP;
-    }
-
-    if (i == 16)
-    {
-        type = 2;
-//      coef_probs =(unsigned char *)( oc->fc.coef_probs [type] [ 0 ] [0]);
-        coef_probs = detoken->coef_probs[type]; //(unsigned char *)( oc->fc.coef_probs [type] [ 0 ] [0]);
-        stop = 24;
-        goto BLOCK_LOOP;
-    }
-
-    bc->count = count;
-    bc->value = value;
-    bc->range = range;
-    bc->pos  = bufptr - bc->buffer;
-    return 0;
-}
-//#if 0
-int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
-{
-//  const ONYX_COMMON * const oc = & dx->common;
-    int eobtotal = 0;
-    int i, type;
-    /*
-        dx->detoken.norm_ptr = norm;
-        dx->detoken.onyx_coef_tree_ptr = onyx_coef_tree;
-        dx->detoken.ptr_onyxblock2context_leftabove = ONYXBLOCK2CONTEXT_LEFTABOVE;
-        dx->detoken.ptr_onyx_coef_bands_x = onyx_coef_bands_x;
-        dx->detoken.scan = default_zig_zag1d;
-        dx->detoken.teb_base_ptr = onyxd_token_extra_bits2;
-
-        dx->detoken.qcoeff_start_ptr = &x->qcoeff[0];
-
-        dx->detoken.A = x->above_context;
-        dx->detoken.L = x->left_context;
-
-        dx->detoken.coef_probs[0] = (unsigned char *)( oc->fc.coef_probs [0] [ 0 ] [0]);
-        dx->detoken.coef_probs[1] = (unsigned char *)( oc->fc.coef_probs [1] [ 0 ] [0]);
-        dx->detoken.coef_probs[2] = (unsigned char *)( oc->fc.coef_probs [2] [ 0 ] [0]);
-        dx->detoken.coef_probs[3] = (unsigned char *)( oc->fc.coef_probs [3] [ 0 ] [0]);
-    */
-
-    dx->detoken.current_bc = x->current_bc;
-    dx->detoken.A = x->above_context;
-    dx->detoken.L = x->left_context;
-
-    type = 3;
-
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
-    {
-        type = 1;
-        eobtotal -= 16;
-    }
-
-    vp8_decode_mb_tokens_v5(&dx->detoken, type);
-
-    for (i = 0; i < 25; i++)
-    {
-        x->Block[i].eob = dx->detoken.eob[i];
-        eobtotal += dx->detoken.eob[i];
-    }
-
-    return eobtotal;
-}
-#endif

vp8/decoder/arm/detokenizearm_v6.asm

@@ -1,365 +0,0 @@
-;
-;  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_decode_mb_tokens_v5|
-
-    AREA    |.text|, CODE, READONLY  ; name this block of code
-
-    INCLUDE vpx_asm_offsets.asm
-
-l_qcoeff    EQU     0
-l_i         EQU     4
-l_type      EQU     8
-l_stop      EQU     12
-l_c         EQU     16
-l_l_ptr      EQU     20
-l_a_ptr      EQU     24
-l_bc        EQU     28
-l_coef_ptr   EQU     32
-l_stacksize EQU     64
-
-
-;; constant offsets -- these should be created at build time
-c_onyxblock2left_offset      EQU 25
-c_onyxblock2above_offset     EQU 50
-c_entropy_nodes              EQU 11
-c_dct_eob_token              EQU 11
-
-|vp8_decode_mb_tokens_v5| PROC
-    stmdb       sp!, {r4 - r11, lr}
-    sub         sp, sp, #l_stacksize
-    mov         r7, r1
-    mov         r9, r0                      ;DETOK *detoken
-
-    ldr         r1, [r9, #detok_current_bc]
-    ldr         r0, [r9, #detok_qcoeff_start_ptr]
-    mov         r11, #0
-    mov         r3, #0x10
-
-    cmp         r7, #1
-    addeq       r11, r11, #24
-    addeq       r3, r3, #8
-    addeq       r0, r0, #3, 24
-
-    str         r0, [sp, #l_qcoeff]
-    str         r11, [sp, #l_i]
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-    str         r1, [sp, #l_bc]
-
-    add         lr, r9, r7, lsl #2
-
-    ldr         r2, [r1, #bool_decoder_buffer]
-    ldr         r3, [r1, #bool_decoder_pos]
-
-    ldr         r10, [lr, #detok_coef_probs]
-    ldr         r5, [r1, #bool_decoder_count]
-    ldr         r6, [r1, #bool_decoder_range]
-    ldr         r4, [r1, #bool_decoder_value]
-    add         r8, r2, r3
-
-    str         r10, [sp, #l_coef_ptr]
-
-
-    ;align 4
-BLOCK_LOOP
-    ldr         r3, [r9, #detok_ptr_onyxblock2context_leftabove]
-    ldr         r2, [r9, #DETOK_A]
-    ldr         r1, [r9, #DETOK_L]
-    ldrb        r12, [r3, +r11]                                 ; detoken->ptr_onyxblock2context_leftabove[i]
-
-    cmp         r7, #0                                          ; check type
-    moveq       r7, #1
-    movne       r7, #0
-
-    ldr         r0, [r2, +r12, lsl #2]                          ; a
-    add         r1, r1, r12, lsl #4
-    add         r3, r3, r11
-
-    ldrb        r2, [r3, #c_onyxblock2above_offset]
-    ldrb        r3, [r3, #c_onyxblock2left_offset]
-    mov         lr, #c_entropy_nodes
-;;  ;++
-
-    ldr         r2, [r0, +r2, lsl #2]!
-    add         r3, r1, r3, lsl #2
-    str         r3, [sp, #l_l_ptr]
-    ldr         r3, [r3]
-
-    cmp         r2, #0
-    movne       r2, #1
-    cmp         r3, #0
-    addne       r2, r2, #1
-
-    str         r0, [sp, #l_a_ptr]
-    smlabb      r0, r2, lr, r10
-    mov         r1, #0                                          ; t = 0
-    str         r7, [sp, #l_c]
-
-    ;align 4
-COEFF_LOOP
-    ldr         r3, [r9, #detok_ptr_onyx_coef_bands_x]
-    ldr         lr, [r9, #detok_onyx_coef_tree_ptr]
-
-;;the following two lines are used if onyx_coef_bands_x is UINT16
-;;  add         r3, r3, r7, lsl #1
-;;  ldrh        r3, [r3]
-
-;;the following line is used if onyx_coef_bands_x is UINT8
-    ldrb        r3, [r7, +r3]
-
-
-;;  ;++
-;;  pld         [r8]
-    ;++
-    add         r0, r0, r3
-
-    ;align 4
-get_token_loop
-    ldrb        r2, [r0, +r1, asr #1]
-    mov         r3, r6, lsl #8
-    sub         r3, r3, #256                    ;split = 1 +  (((range-1) * probability) >> 8)
-    mov         r10, #1
-
-    smlawb      r2, r3, r2, r10
-    ldrb        r12, [r8]                       ;load cx data byte in stall slot
-    ;++
-
-    subs        r3, r4, r2, lsl #24             ;x = value-(split<<24)
-    addhs       r1, r1, #1                      ;t += 1
-    movhs       r4, r3                          ;update value
-    subhs       r2, r6, r2                      ;range = range - split
-    movlo       r6, r2
-
-;;; ldrsbhs     r1, [r1, +lr]
-    ldrsb     r1, [r1, +lr]
-
-
-;; use branch for short pipelines ???
-;;  cmp         r2, #0x80
-;;  bcs         |$LN22@decode_mb_to|
-
-    clz         r3, r2
-    sub         r3, r3, #24
-    subs        r5, r5, r3
-    mov         r6, r2, lsl r3
-    mov         r4, r4, lsl r3
-
-;; use branch for short pipelines ???
-;;  bgt         |$LN22@decode_mb_to|
-
-    addle         r5, r5, #8
-    rsble         r3, r5, #8
-    addle         r8, r8, #1
-    orrle         r4, r4, r12, lsl r3
-
-;;|$LN22@decode_mb_to|
-
-    cmp         r1, #0
-    bgt         get_token_loop
-
-    cmn         r1, #c_dct_eob_token             ;if(t == -DCT_EOB_TOKEN)
-    beq         END_OF_BLOCK
-
-    rsb         lr, r1, #0                      ;v = -t;
-
-    cmp         lr, #4                          ;if(v > FOUR_TOKEN)
-    ble         SKIP_EXTRABITS
-
-    ldr         r3, [r9, #detok_teb_base_ptr]
-    mov         r11, #1
-    add         r7, r3, lr, lsl #4
-
-    ldrsh       lr, [r7, #tokenextrabits_min_val];v = teb_ptr->min_val
-    ldrsh       r0, [r7, #tokenextrabits_length];bits_count = teb_ptr->Length
-
-extrabits_loop
-    add         r3, r0, r7
-
-    ldrb        r2, [r3, #4]
-    mov         r3, r6, lsl #8
-    sub         r3, r3, #256                    ;split = 1 +  (((range-1) * probability) >> 8)
-    mov         r10, #1
-
-    smlawb      r2, r3, r2, r10
-    ldrb        r12, [r8]
-    ;++
-
-    subs        r10, r4, r2, lsl #24            ;x = value-(split<<24)
-    movhs       r4, r10                         ;update value
-    subhs       r2, r6, r2                      ;range = range - split
-    addhs       lr, lr, r11, lsl r0             ;v += ((UINT16)1<<bits_count)
-    movlo       r6, r2                          ;range = split
-
-
-;; use branch for short pipelines ???
-;;  cmp         r2, #0x80
-;;  bcs         |$LN10@decode_mb_to|
-
-    clz         r3, r2
-    sub         r3, r3, #24
-    subs        r5, r5, r3
-    mov         r6, r2, lsl r3                  ;range
-    mov         r4, r4, lsl r3                  ;value
-
-    addle       r5, r5, #8
-    addle       r8, r8, #1
-    rsble       r3, r5, #8
-    orrle       r4, r4, r12, lsl r3
-
-;;|$LN10@decode_mb_to|
-    subs         r0, r0, #1
-    bpl         extrabits_loop
-
-
-SKIP_EXTRABITS
-    ldr         r11, [sp, #l_qcoeff]
-    ldr         r0, [sp, #l_coef_ptr]
-
-    cmp         r1, #0                          ;check for nonzero token
-    beq         SKIP_EOB_CHECK              ;if t is zero, we will skip the eob table chec
-
-    sub         r3, r6, #1                      ;range - 1
-    ;++
-    mov         r3, r3, lsl #7                  ; *= onyx_prob_half  (128)
-    ;++
-    mov         r3, r3, lsr #8
-    add         r2, r3, #1                      ;split
-
-    subs        r3, r4, r2, lsl #24             ;x = value-(split<<24)
-    movhs       r4, r3                          ;update value
-    subhs       r2, r6, r2                      ;range = range - split
-    mvnhs       r3, lr
-    addhs       lr, r3, #1                      ;v = (v ^ -1) + 1
-    movlo       r6, r2                          ;range = split
-
-;; use branch for short pipelines ???
-;;  cmp         r2, #0x80
-;;  bcs         |$LN6@decode_mb_to|
-
-    clz         r3, r2
-    sub         r3, r3, #24
-    subs        r5, r5, r3
-    mov         r6, r2, lsl r3
-    mov         r4, r4, lsl r3
-    ldrleb      r2, [r8], #1
-    addle       r5, r5, #8
-    rsble       r3, r5, #8
-    orrle       r4, r4, r2, lsl r3
-
-;;|$LN6@decode_mb_to|
-    add         r0, r0, #0xB
-
-    cmn         r1, #1
-
-    addlt       r0, r0, #0xB
-
-    mvn         r1, #1
-
-SKIP_EOB_CHECK
-    ldr         r7, [sp, #l_c]
-    ldr         r3, [r9, #detok_scan]
-    add         r1, r1, #2
-    cmp         r7, #(0x10 - 1)                     ;assume one less for now.... increment below
-
-    ldr         r3, [r3, +r7, lsl #2]
-    add         r7, r7, #1
-    add         r3, r11, r3, lsl #1
-
-    str         r7, [sp, #l_c]
-    strh        lr, [r3]
-
-    blt         COEFF_LOOP
-
-    sub         r7, r7, #1                          ;if(t != -DCT_EOB_TOKEN) --c
-
-END_OF_BLOCK
-    ldr         r3, [sp, #l_type]
-    ldr         r10, [sp, #l_coef_ptr]
-    ldr         r0, [sp, #l_qcoeff]
-    ldr         r11, [sp, #l_i]
-    ldr         r12, [sp, #l_stop]
-
-    cmp         r3, #0
-    moveq       r1, #1
-    movne       r1, #0
-    add         r3, r11, r9
-
-    cmp         r7, r1
-    strb        r7, [r3, #detok_eob]
-
-    ldr         r7, [sp, #l_l_ptr]
-    ldr         r2, [sp, #l_a_ptr]
-    movne       r3, #1
-    moveq       r3, #0
-
-    add         r0, r0, #0x20
-    add         r11, r11, #1
-    str         r3, [r7]
-    str         r3, [r2]
-    str         r0, [sp, #l_qcoeff]
-    str         r11, [sp, #l_i]
-
-    cmp         r11, r12                            ;i >= stop ?
-    ldr         r7, [sp, #l_type]
-    mov         lr, #0xB
-
-    blt         BLOCK_LOOP
-
-    cmp         r11, #0x19
-    bne         ln2_decode_mb_to
-
-    ldr         r12, [r9, #detok_qcoeff_start_ptr]
-    ldr         r10, [r9, #detok_coef_probs]
-    mov         r7, #0
-    mov         r3, #0x10
-    str         r12, [sp, #l_qcoeff]
-    str         r7, [sp, #l_i]
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-
-    str         r10, [sp, #l_coef_ptr]
-
-    b           BLOCK_LOOP
-
-ln2_decode_mb_to
-    cmp         r11, #0x10
-    bne         ln1_decode_mb_to
-
-    ldr         r10, [r9, #0x30]
-
-    mov         r7, #2
-    mov         r3, #0x18
-
-    str         r7, [sp, #l_type]
-    str         r3, [sp, #l_stop]
-
-    str         r10, [sp, #l_coef_ptr]
-    b           BLOCK_LOOP
-
-ln1_decode_mb_to
-    ldr         r2, [sp, #l_bc]
-    mov         r0, #0
-    nop
-
-    ldr         r3, [r2, #bool_decoder_buffer]
-    str         r5, [r2, #bool_decoder_count]
-    str         r4, [r2, #bool_decoder_value]
-    sub         r3, r8, r3
-    str         r3, [r2, #bool_decoder_pos]
-    str         r6, [r2, #bool_decoder_range]
-
-    add         sp, sp, #l_stacksize
-    ldmia       sp!, {r4 - r11, pc}
-
-    ENDP  ; |vp8_decode_mb_tokens_v5|
-
-    END

vp8/decoder/arm/dsystemdependent.c

@@ -23,8 +23,6 @@ void vp8_dmachine_specific_config(VP8D_COMP *pbi)
     pbi->mb.rtcd         = &pbi->common.rtcd;
 #if HAVE_ARMV7
     pbi->dequant.block   = vp8_dequantize_b_neon;
-    pbi->dequant.idct    = vp8_dequant_idct_neon;
-    pbi->dequant.idct_dc = vp8_dequant_dc_idct_neon;
     pbi->dboolhuff.start = vp8dx_start_decode_c;
     pbi->dboolhuff.fill  = vp8dx_bool_decoder_fill_c;
     pbi->dboolhuff.debool = vp8dx_decode_bool_c;
@@ -32,8 +30,6 @@ void vp8_dmachine_specific_config(VP8D_COMP *pbi)
 
 #elif HAVE_ARMV6
     pbi->dequant.block   = vp8_dequantize_b_v6;
-    pbi->dequant.idct    = vp8_dequant_idct_v6;
-    pbi->dequant.idct_dc = vp8_dequant_dc_idct_v6;
     pbi->dboolhuff.start = vp8dx_start_decode_c;
     pbi->dboolhuff.fill  = vp8dx_bool_decoder_fill_c;
     pbi->dboolhuff.debool = vp8dx_decode_bool_c;

vp8/decoder/arm/neon/dequant_dc_idct_neon.asm

@@ -9,31 +9,43 @@
 ;
 
 
-    EXPORT  |vp8_dequant_dc_idct_neon|
+    EXPORT  |vp8_dequant_dc_idct_add_neon|
     ARM
     REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp8_dequant_dc_idct_c(short *input, short *dq, short *output, int pitch, int Dc);
+;void vp8_dequant_dc_idct_add_neon(short *input, short *dq, unsigned char *pred,
+;                                  unsigned char *dest, int pitch, int stride,
+;                                  int Dc);
 ; r0    short *input,
 ; r1    short *dq,
-; r2    short *output,
+; r2    unsigned char *pred
-; r3    int pitch,
+; r3    unsigned char *dest
-; (stack)   int Dc
+; sp    int pitch
-|vp8_dequant_dc_idct_neon| PROC
+; sp+4  int stride
+; sp+8  int Dc
+|vp8_dequant_dc_idct_add_neon| PROC
     vld1.16         {q3, q4}, [r0]
     vld1.16         {q5, q6}, [r1]
 
-    ldr             r1, [sp]                ;load Dc from stack
+    ldr             r1, [sp, #8]            ;load Dc from stack
 
-    ldr             r12, _dcidct_coeff_
+    ldr             r12, _CONSTANTS_
 
     vmul.i16        q1, q3, q5              ;input for short_idct4x4llm_neon
     vmul.i16        q2, q4, q6
 
     vmov.16         d2[0], r1
 
+    ldr             r1, [sp]                ; pitch
+    vld1.32         {d14[0]}, [r2], r1
+    vld1.32         {d14[1]}, [r2], r1
+    vld1.32         {d15[0]}, [r2], r1
+    vld1.32         {d15[1]}, [r2]
+
+    ldr             r1, [sp, #4]            ; stride
+
 ;|short_idct4x4llm_neon| PROC
     vld1.16         {d0}, [r12]
     vswp            d3, d4                  ;q2(vp[4] vp[12])
@@ -47,14 +59,9 @@
     vshr.s16        q3, q3, #1
     vshr.s16        q4, q4, #1
 
-    vqadd.s16       q3, q3, q2              ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+    vqadd.s16       q3, q3, q2
     vqadd.s16       q4, q4, q2
 
-    ;d6 - c1:temp1
-    ;d7 - d1:temp2
-    ;d8 - d1:temp1
-    ;d9 - c1:temp2
-
     vqsub.s16       d10, d6, d9             ;c1
     vqadd.s16       d11, d7, d8             ;d1
 
@@ -83,7 +90,7 @@
     vshr.s16        q3, q3, #1
     vshr.s16        q4, q4, #1
 
-    vqadd.s16       q3, q3, q2              ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+    vqadd.s16       q3, q3, q2
     vqadd.s16       q4, q4, q2
 
     vqsub.s16       d10, d6, d9             ;c1
@@ -101,34 +108,29 @@
     vrshr.s16       d4, d4, #3
     vrshr.s16       d5, d5, #3
 
-    add             r1, r2, r3
-    add             r12, r1, r3
-    add             r0, r12, r3
-
     vtrn.32         d2, d4
     vtrn.32         d3, d5
     vtrn.16         d2, d3
     vtrn.16         d4, d5
 
-    vst1.16         {d2}, [r2]
+    vaddw.u8        q1, q1, d14
-    vst1.16         {d3}, [r1]
+    vaddw.u8        q2, q2, d15
-    vst1.16         {d4}, [r12]
+
-    vst1.16         {d5}, [r0]
+    vqmovun.s16     d0, q1
+    vqmovun.s16     d1, q2
+
+    vst1.32         {d0[0]}, [r3], r1
+    vst1.32         {d0[1]}, [r3], r1
+    vst1.32         {d1[0]}, [r3], r1
+    vst1.32         {d1[1]}, [r3]
 
     bx             lr
 
-    ENDP
+    ENDP           ; |vp8_dequant_dc_idct_add_neon|
 
-;-----------------
+; Constant Pool
-    AREA    dcidct4x4_dat, DATA, READWRITE          ;read/write by default
+_CONSTANTS_       DCD cospi8sqrt2minus1
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
+cospi8sqrt2minus1 DCD 0x4e7b4e7b
-;One word each is reserved. Label filter_coeff can be used to access the data.
+sinpi8sqrt2       DCD 0x8a8c8a8c
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-_dcidct_coeff_
-    DCD     dcidct_coeff
-dcidct_coeff
-    DCD     0x4e7b4e7b, 0x8a8c8a8c
-
-;20091, 20091, 35468, 35468
 
     END

vp8/decoder/arm/neon/dequant_idct_neon.asm

@@ -9,22 +9,33 @@
 ;
 
 
-    EXPORT  |vp8_dequant_idct_neon|
+    EXPORT  |vp8_dequant_idct_add_neon|
     ARM
     REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp8_dequant_idct_c(short *input, short *dq, short *output, int pitch);
+;void vp8_dequant_idct_neon(short *input, short *dq, unsigned char *pred,
+;                           unsigned char *dest, int pitch, int stride)
 ; r0    short *input,
 ; r1    short *dq,
-; r2    short *output,
+; r2    unsigned char *pred
-; r3    int pitch,
+; r3    unsigned char *dest
-|vp8_dequant_idct_neon| PROC
+; sp    int pitch
+; sp+4  int stride
+
+|vp8_dequant_idct_add_neon| PROC
     vld1.16         {q3, q4}, [r0]
     vld1.16         {q5, q6}, [r1]
+    ldr             r1, [sp]                ; pitch
+    vld1.32         {d14[0]}, [r2], r1
+    vld1.32         {d14[1]}, [r2], r1
+    vld1.32         {d15[0]}, [r2], r1
+    vld1.32         {d15[1]}, [r2]
 
-    ldr             r12, _didct_coeff_
+    ldr             r1, [sp, #4]            ; stride
+
+    ldr             r12, _CONSTANTS_
 
     vmul.i16        q1, q3, q5              ;input for short_idct4x4llm_neon
     vmul.i16        q2, q4, q6
@@ -42,14 +53,9 @@
     vshr.s16        q3, q3, #1
     vshr.s16        q4, q4, #1
 
-    vqadd.s16       q3, q3, q2              ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+    vqadd.s16       q3, q3, q2
     vqadd.s16       q4, q4, q2
 
-    ;d6 - c1:temp1
-    ;d7 - d1:temp2
-    ;d8 - d1:temp1
-    ;d9 - c1:temp2
-
     vqsub.s16       d10, d6, d9             ;c1
     vqadd.s16       d11, d7, d8             ;d1
 
@@ -78,7 +84,7 @@
     vshr.s16        q3, q3, #1
     vshr.s16        q4, q4, #1
 
-    vqadd.s16       q3, q3, q2              ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+    vqadd.s16       q3, q3, q2
     vqadd.s16       q4, q4, q2
 
     vqsub.s16       d10, d6, d9             ;c1
@@ -96,34 +102,29 @@
     vrshr.s16       d4, d4, #3
     vrshr.s16       d5, d5, #3
 
-    add             r1, r2, r3
-    add             r12, r1, r3
-    add             r0, r12, r3
-
     vtrn.32         d2, d4
     vtrn.32         d3, d5
     vtrn.16         d2, d3
     vtrn.16         d4, d5
 
-    vst1.16         {d2}, [r2]
+    vaddw.u8        q1, q1, d14
-    vst1.16         {d3}, [r1]
+    vaddw.u8        q2, q2, d15
-    vst1.16         {d4}, [r12]
+
-    vst1.16         {d5}, [r0]
+    vqmovun.s16     d0, q1
+    vqmovun.s16     d1, q2
+
+    vst1.32         {d0[0]}, [r3], r1
+    vst1.32         {d0[1]}, [r3], r1
+    vst1.32         {d1[0]}, [r3], r1
+    vst1.32         {d1[1]}, [r3]
 
     bx             lr
 
-    ENDP
+    ENDP           ; |vp8_dequant_idct_add_neon|
 
-;-----------------
+; Constant Pool
-    AREA    didct4x4_dat, DATA, READWRITE           ;read/write by default
+_CONSTANTS_       DCD cospi8sqrt2minus1
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
+cospi8sqrt2minus1 DCD 0x4e7b4e7b
-;One word each is reserved. Label filter_coeff can be used to access the data.
+sinpi8sqrt2       DCD 0x8a8c8a8c
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-_didct_coeff_
-    DCD     didct_coeff
-didct_coeff
-    DCD     0x4e7b4e7b, 0x8a8c8a8c
-
-;20091, 20091, 35468, 35468
 
     END

vp8/decoder/arm/neon/idct_blk_neon.c

@@ -0,0 +1,151 @@
+/*
+ *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "idct.h"
+#include "dequantize.h"
+
+void vp8_dequant_dc_idct_add_y_block_neon
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs, short *dc)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_dc_idct_add_neon (q, dq, pre, dst, 16, stride, dc[0]);
+        else
+            vp8_dc_only_idct_add_neon (dc[0], pre, dst, 16, stride);
+
+        if (eobs[1] > 1)
+            vp8_dequant_dc_idct_add_neon (q+16, dq, pre+4, dst+4, 16, stride, dc[1]);
+        else
+            vp8_dc_only_idct_add_neon (dc[1], pre+4, dst+4, 16, stride);
+
+        if (eobs[2] > 1)
+            vp8_dequant_dc_idct_add_neon (q+32, dq, pre+8, dst+8, 16, stride, dc[2]);
+        else
+            vp8_dc_only_idct_add_neon (dc[2], pre+8, dst+8, 16, stride);
+
+        if (eobs[3] > 1)
+            vp8_dequant_dc_idct_add_neon (q+48, dq, pre+12, dst+12, 16, stride, dc[3]);
+        else
+            vp8_dc_only_idct_add_neon (dc[3], pre+12, dst+12, 16, stride);
+
+        q    += 64;
+        dc   += 4;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_y_block_neon
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_neon (q, dq, pre, dst, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[0]*dq[0], pre, dst, 16, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_neon (q+16, dq, pre+4, dst+4, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[16]*dq[0], pre+4, dst+4, 16, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        if (eobs[2] > 1)
+            vp8_dequant_idct_add_neon (q+32, dq, pre+8, dst+8, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[32]*dq[0], pre+8, dst+8, 16, stride);
+            ((int *)(q+32))[0] = 0;
+        }
+
+        if (eobs[3] > 1)
+            vp8_dequant_idct_add_neon (q+48, dq, pre+12, dst+12, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[48]*dq[0], pre+12, dst+12, 16, stride);
+            ((int *)(q+48))[0] = 0;
+        }
+
+        q    += 64;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_neon
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_neon (q, dq, pre, dstu, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[0]*dq[0], pre, dstu, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_neon (q+16, dq, pre+4, dstu+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[16]*dq[0], pre+4, dstu+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstu += 4*stride;
+        eobs += 2;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_neon (q, dq, pre, dstv, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[0]*dq[0], pre, dstv, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_neon (q+16, dq, pre+4, dstv+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_neon (q[16]*dq[0], pre+4, dstv+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstv += 4*stride;
+        eobs += 2;
+    }
+}

vp8/decoder/dboolhuff.c

@@ -13,7 +13,7 @@
 #include "vpx_ports/mem.h"
 #include "vpx_mem/vpx_mem.h"
 
-DECLARE_ALIGNED(16, const unsigned int, vp8dx_bitreader_norm[256]) =
+DECLARE_ALIGNED(16, const unsigned char, vp8dx_bitreader_norm[256]) =
 {
     0, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
     2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,

vp8/decoder/dboolhuff.h

@@ -95,7 +95,7 @@ typedef struct vp8_dboolhuff_rtcd_vtable {
 #define IF_RTCD(x) NULL
 //#endif
 
-DECLARE_ALIGNED(16, extern const unsigned int, vp8dx_bitreader_norm[256]);
+DECLARE_ALIGNED(16, extern const unsigned char, vp8dx_bitreader_norm[256]);
 
 /* wrapper functions to hide RTCD. static means inline means hopefully no
  * penalty

vp8/decoder/decodemv.c

@@ -226,13 +226,14 @@ void vp8_decode_mode_mvs(VP8D_COMP *pbi)
                         int mv_contz;
 
                         while (j != L[++k])
-                            if (k >= 16)
+                        {
 #if CONFIG_DEBUG
+                            if (k >= 16)
+                            {
                                 assert(0);
-
+                            }
-#else
-                                ;
 #endif
+                        }
 
                         mv_contz = vp8_mv_cont(&(vp8_left_bmi(mi, k)->mv.as_mv), &(vp8_above_bmi(mi, k, mis)->mv.as_mv));
 

vp8/decoder/decoderthreading.h

@@ -18,6 +18,7 @@
 
 extern void vp8_mtdecode_mb_rows(VP8D_COMP *pbi,
                                  MACROBLOCKD *xd);
+extern void vp8_mt_loop_filter_frame(VP8D_COMP *pbi);
 extern void vp8_stop_lfthread(VP8D_COMP *pbi);
 extern void vp8_start_lfthread(VP8D_COMP *pbi);
 extern void vp8_decoder_remove_threads(VP8D_COMP *pbi);

vp8/decoder/decodframe.c

@@ -21,7 +21,7 @@
 #include "alloccommon.h"
 #include "entropymode.h"
 #include "quant_common.h"
-#include "segmentation_common.h"
+
 #include "setupintrarecon.h"
 #include "demode.h"
 #include "decodemv.h"
@@ -113,7 +113,7 @@ static void mb_init_dequantizer(VP8D_COMP *pbi, MACROBLOCKD *xd)
 // to dst buffer, we can write the result directly to dst buffer. This eliminates unnecessary copy.
 static void skip_recon_mb(VP8D_COMP *pbi, MACROBLOCKD *xd)
 {
-    if (xd->frame_type == KEY_FRAME  ||  xd->mbmi.ref_frame == INTRA_FRAME)
+    if (xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
     {
 
         vp8_build_intra_predictors_mbuv_s(xd);
@@ -126,7 +126,6 @@ static void skip_recon_mb(VP8D_COMP *pbi, MACROBLOCKD *xd)
     }
 }
 
-
 static void clamp_mv_to_umv_border(MV *mv, const MACROBLOCKD *xd)
 {
     /* If the MV points so far into the UMV border that no visible pixels
@@ -165,7 +164,7 @@ static void clamp_uvmv_to_umv_border(MV *mv, const MACROBLOCKD *xd)
 
 static void clamp_mvs(MACROBLOCKD *xd)
 {
-    if (xd->mbmi.mode == SPLITMV)
+    if (xd->mode_info_context->mbmi.mode == SPLITMV)
     {
         int i;
 
@@ -176,48 +175,69 @@ static void clamp_mvs(MACROBLOCKD *xd)
     }
     else
     {
-        clamp_mv_to_umv_border(&xd->mbmi.mv.as_mv, xd);
+        clamp_mv_to_umv_border(&xd->mode_info_context->mbmi.mv.as_mv, xd);
         clamp_uvmv_to_umv_border(&xd->block[16].bmi.mv.as_mv, xd);
     }
 
 }
 
-static void reconstruct_mb(VP8D_COMP *pbi, MACROBLOCKD *xd)
+void vp8_decode_macroblock(VP8D_COMP *pbi, MACROBLOCKD *xd)
 {
-    if (xd->frame_type == KEY_FRAME  ||  xd->mbmi.ref_frame == INTRA_FRAME)
+    int eobtotal = 0;
-    {
+    int i, do_clamp = xd->mode_info_context->mbmi.need_to_clamp_mvs;
-        vp8_build_intra_predictors_mbuv(xd);
 
-        if (xd->mbmi.mode != B_PRED)
+    if (xd->mode_info_context->mbmi.mb_skip_coeff)
     {
-            vp8_build_intra_predictors_mby_ptr(xd);
+        vp8_reset_mb_tokens_context(xd);
-            vp8_recon16x16mb(RTCD_VTABLE(recon), xd);
     }
     else
     {
-            vp8_recon_intra4x4mb(RTCD_VTABLE(recon), xd);
+        eobtotal = vp8_decode_mb_tokens(pbi, xd);
+    }
+
+    /* Perform temporary clamping of the MV to be used for prediction */
+    if (do_clamp)
+    {
+        clamp_mvs(xd);
+    }
+
+    xd->mode_info_context->mbmi.dc_diff = 1;
+
+    if (xd->mode_info_context->mbmi.mode != B_PRED && xd->mode_info_context->mbmi.mode != SPLITMV && eobtotal == 0)
+    {
+        xd->mode_info_context->mbmi.dc_diff = 0;
+        skip_recon_mb(pbi, xd);
+        return;
+    }
+
+    if (xd->segmentation_enabled)
+        mb_init_dequantizer(pbi, xd);
+
+    // do prediction
+    if (xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
+    {
+        vp8_build_intra_predictors_mbuv(xd);
+
+        if (xd->mode_info_context->mbmi.mode != B_PRED)
+        {
+            vp8_build_intra_predictors_mby_ptr(xd);
+        } else {
+            vp8_intra_prediction_down_copy(xd);
         }
     }
     else
     {
         vp8_build_inter_predictors_mb(xd);
-        vp8_recon16x16mb(RTCD_VTABLE(recon), xd);
-    }
     }
 
-
+    // dequantization and idct
-static void de_quantand_idct(VP8D_COMP *pbi, MACROBLOCKD *xd)
+    if (xd->mode_info_context->mbmi.mode != B_PRED && xd->mode_info_context->mbmi.mode != SPLITMV)
     {
-    int i;
         BLOCKD *b = &xd->block[24];
-
-
-    if (xd->mbmi.mode != B_PRED && xd->mbmi.mode != SPLITMV)
-    {
         DEQUANT_INVOKE(&pbi->dequant, block)(b);
 
         // do 2nd order transform on the dc block
-        if (b->eob > 1)
+        if (xd->eobs[24] > 1)
         {
             IDCT_INVOKE(RTCD_VTABLE(idct), iwalsh16)(&b->dqcoeff[0], b->diff);
             ((int *)b->qcoeff)[0] = 0;
@@ -235,115 +255,47 @@ static void de_quantand_idct(VP8D_COMP *pbi, MACROBLOCKD *xd)
             ((int *)b->qcoeff)[0] = 0;
         }
 
-
+        DEQUANT_INVOKE (&pbi->dequant, dc_idct_add_y_block)
+                        (xd->qcoeff, &xd->block[0].dequant[0][0],
+                         xd->predictor, xd->dst.y_buffer,
+                         xd->dst.y_stride, xd->eobs, xd->block[24].diff);
+    }
+    else if ((xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) && xd->mode_info_context->mbmi.mode == B_PRED)
+    {
         for (i = 0; i < 16; i++)
         {
 
-            b = &xd->block[i];
+            BLOCKD *b = &xd->block[i];
+            vp8_predict_intra4x4(b, b->bmi.mode, b->predictor);
 
-            if (b->eob > 1)
+            if (xd->eobs[i] > 1)
             {
-                DEQUANT_INVOKE(&pbi->dequant, idct_dc)(b->qcoeff, &b->dequant[0][0], b->diff, 32, xd->block[24].diff[i]);
+                DEQUANT_INVOKE(&pbi->dequant, idct_add)
+                    (b->qcoeff, &b->dequant[0][0],  b->predictor,
+                    *(b->base_dst) + b->dst, 16, b->dst_stride);
             }
             else
             {
-                IDCT_INVOKE(RTCD_VTABLE(idct), idct1_scalar)(xd->block[24].diff[i], b->diff, 32);
+                IDCT_INVOKE(RTCD_VTABLE(idct), idct1_scalar_add)
-            }
+                    (b->qcoeff[0] * b->dequant[0][0], b->predictor,
-        }
+                    *(b->base_dst) + b->dst, 16, b->dst_stride);
-
-        for (i = 16; i < 24; i++)
-        {
-            b = &xd->block[i];
-
-            if (b->eob > 1)
-            {
-                DEQUANT_INVOKE(&pbi->dequant, idct)(b->qcoeff, &b->dequant[0][0], b->diff, 16);
-            }
-            else
-            {
-                IDCT_INVOKE(RTCD_VTABLE(idct), idct1_scalar)(b->qcoeff[0] * b->dequant[0][0], b->diff, 16);
                 ((int *)b->qcoeff)[0] = 0;
             }
         }
+
     }
     else
     {
-        for (i = 0; i < 24; i++)
+        DEQUANT_INVOKE (&pbi->dequant, idct_add_y_block)
-        {
+                        (xd->qcoeff, &xd->block[0].dequant[0][0],
-
+                         xd->predictor, xd->dst.y_buffer,
-            b = &xd->block[i];
+                         xd->dst.y_stride, xd->eobs);
-
-            if (b->eob > 1)
-            {
-                DEQUANT_INVOKE(&pbi->dequant, idct)(b->qcoeff, &b->dequant[0][0], b->diff, (32 - (i & 16)));
-            }
-            else
-            {
-                IDCT_INVOKE(RTCD_VTABLE(idct), idct1_scalar)(b->qcoeff[0] * b->dequant[0][0], b->diff, (32 - (i & 16)));
-                ((int *)b->qcoeff)[0] = 0;
-            }
-        }
-    }
     }
 
-void vp8_decode_macroblock(VP8D_COMP *pbi, MACROBLOCKD *xd)
+    DEQUANT_INVOKE (&pbi->dequant, idct_add_uv_block)
-{
+                    (xd->qcoeff+16*16, &xd->block[16].dequant[0][0],
-    int eobtotal = 0;
+                     xd->predictor+16*16, xd->dst.u_buffer, xd->dst.v_buffer,
-    MV  orig_mvs[24];
+                     xd->dst.uv_stride, xd->eobs+16);
-    int i, do_clamp = xd->mbmi.need_to_clamp_mvs;
-
-    if (xd->mbmi.mb_skip_coeff)
-    {
-        vp8_reset_mb_tokens_context(xd);
-    }
-    else
-    {
-        eobtotal = vp8_decode_mb_tokens(pbi, xd);
-    }
-
-    /* Perform temporary clamping of the MV to be used for prediction */
-    if (do_clamp)
-    {
-        if (xd->mbmi.mode == SPLITMV)
-            for (i=0; i<24; i++)
-                orig_mvs[i] = xd->block[i].bmi.mv.as_mv;
-        else
-        {
-            orig_mvs[0] = xd->mbmi.mv.as_mv;
-            orig_mvs[1] = xd->block[16].bmi.mv.as_mv;
-        }
-        clamp_mvs(xd);
-    }
-
-    xd->mode_info_context->mbmi.dc_diff = 1;
-
-    if (xd->mbmi.mode != B_PRED && xd->mbmi.mode != SPLITMV && eobtotal == 0)
-    {
-        xd->mode_info_context->mbmi.dc_diff = 0;
-        skip_recon_mb(pbi, xd);
-    }
-    else
-    {
-        if (xd->segmentation_enabled)
-            mb_init_dequantizer(pbi, xd);
-
-        de_quantand_idct(pbi, xd);
-        reconstruct_mb(pbi, xd);
-    }
-
-
-    /* Restore the original MV so as not to affect the entropy context. */
-    if (do_clamp)
-    {
-        if (xd->mbmi.mode == SPLITMV)
-            for (i=0; i<24; i++)
-                xd->block[i].bmi.mv.as_mv = orig_mvs[i];
-        else
-        {
-            xd->mbmi.mv.as_mv = orig_mvs[0];
-            xd->block[16].bmi.mv.as_mv = orig_mvs[1];
-        }
-    }
 }
 
 static int get_delta_q(vp8_reader *bc, int prev, int *q_update)
@@ -381,18 +333,17 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
     int i;
     int recon_yoffset, recon_uvoffset;
     int mb_col;
-    int recon_y_stride = pc->last_frame.y_stride;
+    int ref_fb_idx = pc->lst_fb_idx;
-    int recon_uv_stride = pc->last_frame.uv_stride;
+    int dst_fb_idx = pc->new_fb_idx;
+    int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
+    int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;
 
-    vpx_memset(pc->left_context, 0, sizeof(pc->left_context));
+    vpx_memset(&pc->left_context, 0, sizeof(pc->left_context));
     recon_yoffset = mb_row * recon_y_stride * 16;
     recon_uvoffset = mb_row * recon_uv_stride * 8;
     // reset above block coeffs
 
-    xd->above_context[Y1CONTEXT] = pc->above_context[Y1CONTEXT];
+    xd->above_context = pc->above_context;
-    xd->above_context[UCONTEXT ] = pc->above_context[UCONTEXT];
-    xd->above_context[VCONTEXT ] = pc->above_context[VCONTEXT];
-    xd->above_context[Y2CONTEXT] = pc->above_context[Y2CONTEXT];
     xd->up_available = (mb_row != 0);
 
     xd->mb_to_top_edge = -((mb_row * 16)) << 3;
@@ -400,12 +351,8 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
 
     for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
     {
-        // Take a copy of the mode and Mv information for this macroblock into the xd->mbmi
-        // the partition_bmi array is unused in the decoder, so don't copy it.
-        vpx_memcpy(&xd->mbmi, &xd->mode_info_context->mbmi,
-                   sizeof(MB_MODE_INFO) - sizeof(xd->mbmi.partition_bmi));
 
-        if (xd->mbmi.mode == SPLITMV || xd->mbmi.mode == B_PRED)
+        if (xd->mode_info_context->mbmi.mode == SPLITMV || xd->mode_info_context->mbmi.mode == B_PRED)
         {
             for (i = 0; i < 16; i++)
             {
@@ -419,33 +366,23 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
         xd->mb_to_left_edge = -((mb_col * 16) << 3);
         xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
 
-        xd->dst.y_buffer = pc->new_frame.y_buffer + recon_yoffset;
+        xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
-        xd->dst.u_buffer = pc->new_frame.u_buffer + recon_uvoffset;
+        xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
-        xd->dst.v_buffer = pc->new_frame.v_buffer + recon_uvoffset;
+        xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
 
         xd->left_available = (mb_col != 0);
 
         // Select the appropriate reference frame for this MB
-        if (xd->mbmi.ref_frame == LAST_FRAME)
+        if (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)
-        {
+            ref_fb_idx = pc->lst_fb_idx;
-            xd->pre.y_buffer = pc->last_frame.y_buffer + recon_yoffset;
+        else if (xd->mode_info_context->mbmi.ref_frame == GOLDEN_FRAME)
-            xd->pre.u_buffer = pc->last_frame.u_buffer + recon_uvoffset;
+            ref_fb_idx = pc->gld_fb_idx;
-            xd->pre.v_buffer = pc->last_frame.v_buffer + recon_uvoffset;
-        }
-        else if (xd->mbmi.ref_frame == GOLDEN_FRAME)
-        {
-            // Golden frame reconstruction buffer
-            xd->pre.y_buffer = pc->golden_frame.y_buffer + recon_yoffset;
-            xd->pre.u_buffer = pc->golden_frame.u_buffer + recon_uvoffset;
-            xd->pre.v_buffer = pc->golden_frame.v_buffer + recon_uvoffset;
-        }
         else
-        {
+            ref_fb_idx = pc->alt_fb_idx;
-            // Alternate reference frame reconstruction buffer
+
-            xd->pre.y_buffer = pc->alt_ref_frame.y_buffer + recon_yoffset;
+        xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
-            xd->pre.u_buffer = pc->alt_ref_frame.u_buffer + recon_uvoffset;
+        xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
-            xd->pre.v_buffer = pc->alt_ref_frame.v_buffer + recon_uvoffset;
+        xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
-        }
 
         vp8_build_uvmvs(xd, pc->full_pixel);
 
@@ -463,19 +400,14 @@ void vp8_decode_mb_row(VP8D_COMP *pbi,
 
         ++xd->mode_info_context;  /* next mb */
 
-        xd->gf_active_ptr++;      // GF useage flag for next MB
+        xd->above_context++;
-
-        xd->above_context[Y1CONTEXT] += 4;
-        xd->above_context[UCONTEXT ] += 2;
-        xd->above_context[VCONTEXT ] += 2;
-        xd->above_context[Y2CONTEXT] ++;
 
         pbi->current_mb_col_main = mb_col;
     }
 
     // adjust to the next row of mbs
     vp8_extend_mb_row(
-        &pc->new_frame,
+        &pc->yv12_fb[dst_fb_idx],
         xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
     );
 
@@ -623,10 +555,10 @@ static void init_frame(VP8D_COMP *pbi)
         }
     }
 
-    xd->left_context = pc->left_context;
+    xd->left_context = &pc->left_context;
     xd->mode_info_context = pc->mi;
     xd->frame_type = pc->frame_type;
-    xd->mbmi.mode = DC_PRED;
+    xd->mode_info_context->mbmi.mode = DC_PRED;
     xd->mode_info_stride = pc->mode_info_stride;
 }
 
@@ -890,11 +822,11 @@ int vp8_decode_frame(VP8D_COMP *pbi)
                     }
     }
 
-    vpx_memcpy(&xd->pre, &pc->last_frame, sizeof(YV12_BUFFER_CONFIG));
+    vpx_memcpy(&xd->pre, &pc->yv12_fb[pc->lst_fb_idx], sizeof(YV12_BUFFER_CONFIG));
-    vpx_memcpy(&xd->dst, &pc->new_frame, sizeof(YV12_BUFFER_CONFIG));
+    vpx_memcpy(&xd->dst, &pc->yv12_fb[pc->new_fb_idx], sizeof(YV12_BUFFER_CONFIG));
 
     // set up frame new frame for intra coded blocks
-    vp8_setup_intra_recon(&pc->new_frame);
+    vp8_setup_intra_recon(&pc->yv12_fb[pc->new_fb_idx]);
 
     vp8_setup_block_dptrs(xd);
 
@@ -911,14 +843,7 @@ int vp8_decode_frame(VP8D_COMP *pbi)
     else
         vp8_decode_mode_mvs(pbi);
 
-    // reset since these guys are used as iterators
+    vpx_memset(pc->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES) * pc->mb_cols);
-    vpx_memset(pc->above_context[Y1CONTEXT], 0, sizeof(ENTROPY_CONTEXT) * pc->mb_cols * 4);
-    vpx_memset(pc->above_context[UCONTEXT ], 0, sizeof(ENTROPY_CONTEXT) * pc->mb_cols * 2);
-    vpx_memset(pc->above_context[VCONTEXT ], 0, sizeof(ENTROPY_CONTEXT) * pc->mb_cols * 2);
-    vpx_memset(pc->above_context[Y2CONTEXT], 0, sizeof(ENTROPY_CONTEXT) * pc->mb_cols);
-
-    xd->gf_active_ptr = (signed char *)pc->gf_active_flags;     // Point to base of GF active flags data structure
-
 
     vpx_memcpy(&xd->block[0].bmi, &xd->mode_info_context->bmi[0], sizeof(B_MODE_INFO));
 

vp8/decoder/dequantize.c

@@ -32,8 +32,12 @@ void vp8_dequantize_b_c(BLOCKD *d)
     }
 }
 
-void vp8_dequant_idct_c(short *input, short *dq, short *output, int pitch)
+void vp8_dequant_idct_add_c(short *input, short *dq, unsigned char *pred,
+                            unsigned char *dest, int pitch, int stride)
 {
+    short output[16];
+    short *diff_ptr = output;
+    int r, c;
     int i;
 
     for (i = 0; i < 16; i++)
@@ -41,13 +45,40 @@ void vp8_dequant_idct_c(short *input, short *dq, short *output, int pitch)
         input[i] = dq[i] * input[i];
     }
 
-    vp8_short_idct4x4llm_c(input, output, pitch);
+    // the idct halves ( >> 1) the pitch
+    vp8_short_idct4x4llm_c(input, output, 4 << 1);
+
     vpx_memset(input, 0, 32);
+
+    for (r = 0; r < 4; r++)
+    {
+        for (c = 0; c < 4; c++)
+        {
+            int a = diff_ptr[c] + pred[c];
+
+            if (a < 0)
+                a = 0;
+
+            if (a > 255)
+                a = 255;
+
+            dest[c] = (unsigned char) a;
         }
 
-void vp8_dequant_dc_idct_c(short *input, short *dq, short *output, int pitch, int Dc)
+        dest += stride;
+        diff_ptr += 4;
+        pred += pitch;
+    }
+}
+
+void vp8_dequant_dc_idct_add_c(short *input, short *dq, unsigned char *pred,
+                               unsigned char *dest, int pitch, int stride,
+                               int Dc)
 {
     int i;
+    short output[16];
+    short *diff_ptr = output;
+    int r, c;
 
     input[0] = (short)Dc;
 
@@ -56,6 +87,28 @@ void vp8_dequant_dc_idct_c(short *input, short *dq, short *output, int pitch, in
         input[i] = dq[i] * input[i];
     }
 
-    vp8_short_idct4x4llm_c(input, output, pitch);
+    // the idct halves ( >> 1) the pitch
+    vp8_short_idct4x4llm_c(input, output, 4 << 1);
+
     vpx_memset(input, 0, 32);
+
+    for (r = 0; r < 4; r++)
+    {
+        for (c = 0; c < 4; c++)
+        {
+            int a = diff_ptr[c] + pred[c];
+
+            if (a < 0)
+                a = 0;
+
+            if (a > 255)
+                a = 255;
+
+            dest[c] = (unsigned char) a;
+        }
+
+        dest += stride;
+        diff_ptr += 4;
+        pred += pitch;
+    }
 }

vp8/decoder/dequantize.h

@@ -16,11 +16,31 @@
 #define prototype_dequant_block(sym) \
     void sym(BLOCKD *x)
 
-#define prototype_dequant_idct(sym) \
+#define prototype_dequant_idct_add(sym) \
-    void sym(short *input, short *dq, short *output, int pitch)
+    void sym(short *input, short *dq, \
+             unsigned char *pred, unsigned char *output, \
+             int pitch, int stride)
 
-#define prototype_dequant_idct_dc(sym) \
+#define prototype_dequant_dc_idct_add(sym) \
-    void sym(short *input, short *dq, short *output, int pitch, int dc)
+    void sym(short *input, short *dq, \
+             unsigned char *pred, unsigned char *output, \
+             int pitch, int stride, \
+             int dc)
+
+#define prototype_dequant_dc_idct_add_y_block(sym) \
+    void sym(short *q, short *dq, \
+             unsigned char *pre, unsigned char *dst, \
+             int stride, char *eobs, short *dc)
+
+#define prototype_dequant_idct_add_y_block(sym) \
+    void sym(short *q, short *dq, \
+             unsigned char *pre, unsigned char *dst, \
+             int stride, char *eobs)
+
+#define prototype_dequant_idct_add_uv_block(sym) \
+    void sym(short *q, short *dq, \
+             unsigned char *pre, unsigned char *dst_u, \
+             unsigned char *dst_v, int stride, char *eobs)
 
 #if ARCH_X86 || ARCH_X86_64
 #include "x86/dequantize_x86.h"
@@ -35,25 +55,52 @@
 #endif
 extern prototype_dequant_block(vp8_dequant_block);
 
-#ifndef vp8_dequant_idct
+#ifndef vp8_dequant_idct_add
-#define vp8_dequant_idct vp8_dequant_idct_c
+#define vp8_dequant_idct_add vp8_dequant_idct_add_c
 #endif
-extern prototype_dequant_idct(vp8_dequant_idct);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add);
 
-#ifndef vp8_dequant_idct_dc
+#ifndef vp8_dequant_dc_idct_add
-#define vp8_dequant_idct_dc vp8_dequant_dc_idct_c
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_c
 #endif
-extern prototype_dequant_idct_dc(vp8_dequant_idct_dc);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add);
+
+#ifndef vp8_dequant_dc_idct_add_y_block
+#define vp8_dequant_dc_idct_add_y_block vp8_dequant_dc_idct_add_y_block_c
+#endif
+extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block);
+
+#ifndef vp8_dequant_idct_add_y_block
+#define vp8_dequant_idct_add_y_block vp8_dequant_idct_add_y_block_c
+#endif
+extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block);
+
+#ifndef vp8_dequant_idct_add_uv_block
+#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_c
+#endif
+extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block);
 
 
 typedef prototype_dequant_block((*vp8_dequant_block_fn_t));
-typedef prototype_dequant_idct((*vp8_dequant_idct_fn_t));
+
-typedef prototype_dequant_idct_dc((*vp8_dequant_idct_dc_fn_t));
+typedef prototype_dequant_idct_add((*vp8_dequant_idct_add_fn_t));
+
+typedef prototype_dequant_dc_idct_add((*vp8_dequant_dc_idct_add_fn_t));
+
+typedef prototype_dequant_dc_idct_add_y_block((*vp8_dequant_dc_idct_add_y_block_fn_t));
+
+typedef prototype_dequant_idct_add_y_block((*vp8_dequant_idct_add_y_block_fn_t));
+
+typedef prototype_dequant_idct_add_uv_block((*vp8_dequant_idct_add_uv_block_fn_t));
+
 typedef struct
 {
     vp8_dequant_block_fn_t               block;
-    vp8_dequant_idct_fn_t     idct;
+    vp8_dequant_idct_add_fn_t            idct_add;
-    vp8_dequant_idct_dc_fn_t  idct_dc;
+    vp8_dequant_dc_idct_add_fn_t         dc_idct_add;
+    vp8_dequant_dc_idct_add_y_block_fn_t dc_idct_add_y_block;
+    vp8_dequant_idct_add_y_block_fn_t    idct_add_y_block;
+    vp8_dequant_idct_add_uv_block_fn_t   idct_add_uv_block;
 } vp8_dequant_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT

vp8/decoder/detokenize.c

@@ -14,11 +14,12 @@
 #include "onyxd_int.h"
 #include "vpx_mem/vpx_mem.h"
 #include "vpx_ports/mem.h"
+#include "detokenize.h"
 
 #define BOOL_DATA UINT8
 
 #define OCB_X PREV_COEF_CONTEXTS * ENTROPY_NODES
-DECLARE_ALIGNED(16, UINT16, vp8_coef_bands_x[16]) = { 0, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 6 * OCB_X, 4 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 7 * OCB_X};
+DECLARE_ALIGNED(16, UINT8, vp8_coef_bands_x[16]) = { 0, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 6 * OCB_X, 4 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 7 * OCB_X};
 #define EOB_CONTEXT_NODE            0
 #define ZERO_CONTEXT_NODE           1
 #define ONE_CONTEXT_NODE            2
@@ -60,50 +61,50 @@ DECLARE_ALIGNED(16, static const TOKENEXTRABITS, vp8d_token_extra_bits2[MAX_ENTR
 
 void vp8_reset_mb_tokens_context(MACROBLOCKD *x)
 {
-    ENTROPY_CONTEXT **const A = x->above_context;
-    ENTROPY_CONTEXT(* const L)[4] = x->left_context;
-
-    ENTROPY_CONTEXT *a;
-    ENTROPY_CONTEXT *l;
-
-    /* Clear entropy contexts for Y blocks */
-    a = A[Y1CONTEXT];
-    l = L[Y1CONTEXT];
-    *a = 0;
-    *(a+1) = 0;
-    *(a+2) = 0;
-    *(a+3) = 0;
-    *l = 0;
-    *(l+1) = 0;
-    *(l+2) = 0;
-    *(l+3) = 0;
-
-    /* Clear entropy contexts for U blocks */
-    a = A[UCONTEXT];
-    l = L[UCONTEXT];
-    *a = 0;
-    *(a+1) = 0;
-    *l = 0;
-    *(l+1) = 0;
-
-    /* Clear entropy contexts for V blocks */
-    a = A[VCONTEXT];
-    l = L[VCONTEXT];
-    *a = 0;
-    *(a+1) = 0;
-    *l = 0;
-    *(l+1) = 0;
-
     /* Clear entropy contexts for Y2 blocks */
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != B_PRED && x->mode_info_context->mbmi.mode != SPLITMV)
     {
-        a = A[Y2CONTEXT];
+        vpx_memset(x->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
-        l = L[Y2CONTEXT];
+        vpx_memset(x->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES));
-        *a = 0;
+    }
-        *l = 0;
+    else
+    {
+        vpx_memset(x->above_context, 0, sizeof(ENTROPY_CONTEXT_PLANES)-1);
+        vpx_memset(x->left_context, 0, sizeof(ENTROPY_CONTEXT_PLANES)-1);
     }
 }
-DECLARE_ALIGNED(16, extern const unsigned int, vp8dx_bitreader_norm[256]);
+
+#if CONFIG_ARM_ASM_DETOK
+// mashup of vp8_block2left and vp8_block2above so we only need one pointer
+// for the assembly version.
+DECLARE_ALIGNED(16, const UINT8, vp8_block2leftabove[25*2]) =
+{
+    //vp8_block2left
+    0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
+    //vp8_block2above
+    0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 4, 5, 4, 5, 6, 7, 6, 7, 8
+};
+
+void vp8_init_detokenizer(VP8D_COMP *dx)
+{
+    const VP8_COMMON *const oc = & dx->common;
+    MACROBLOCKD *x = & dx->mb;
+
+    dx->detoken.vp8_coef_tree_ptr = vp8_coef_tree;
+    dx->detoken.ptr_block2leftabove = vp8_block2leftabove;
+    dx->detoken.ptr_coef_bands_x = vp8_coef_bands_x;
+    dx->detoken.scan = vp8_default_zig_zag1d;
+    dx->detoken.teb_base_ptr = vp8d_token_extra_bits2;
+    dx->detoken.qcoeff_start_ptr = &x->qcoeff[0];
+
+    dx->detoken.coef_probs[0] = (oc->fc.coef_probs [0] [ 0 ] [0]);
+    dx->detoken.coef_probs[1] = (oc->fc.coef_probs [1] [ 0 ] [0]);
+    dx->detoken.coef_probs[2] = (oc->fc.coef_probs [2] [ 0 ] [0]);
+    dx->detoken.coef_probs[3] = (oc->fc.coef_probs [3] [ 0 ] [0]);
+}
+#endif
+
+DECLARE_ALIGNED(16, extern const unsigned char, vp8dx_bitreader_norm[256]);
 #define FILL \
     if(count < 0) \
         VP8DX_BOOL_DECODER_FILL(count, value, bufptr, bufend);
@@ -200,14 +201,45 @@ DECLARE_ALIGNED(16, extern const unsigned int, vp8dx_bitreader_norm[256]);
     }\
     NORMALIZE
 
+#if CONFIG_ARM_ASM_DETOK
 int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
 {
-    ENTROPY_CONTEXT **const A = x->above_context;
+    int eobtotal = 0;
-    ENTROPY_CONTEXT(* const L)[4] = x->left_context;
+    int i, type;
+
+    dx->detoken.current_bc = x->current_bc;
+    dx->detoken.A = x->above_context;
+    dx->detoken.L = x->left_context;
+
+    type = 3;
+
+    if (x->mode_info_context->mbmi.mode != B_PRED && x->mode_info_context->mbmi.mode != SPLITMV)
+    {
+        type = 1;
+        eobtotal -= 16;
+    }
+
+    vp8_decode_mb_tokens_v6(&dx->detoken, type);
+
+    for (i = 0; i < 25; i++)
+    {
+        x->eobs[i] = dx->detoken.eob[i];
+        eobtotal += dx->detoken.eob[i];
+    }
+
+    return eobtotal;
+}
+#else
+int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
+{
+    ENTROPY_CONTEXT *A = (ENTROPY_CONTEXT *)x->above_context;
+    ENTROPY_CONTEXT *L = (ENTROPY_CONTEXT *)x->left_context;
     const VP8_COMMON *const oc = & dx->common;
 
     BOOL_DECODER *bc = x->current_bc;
 
+    char *eobs = x->eobs;
+
     ENTROPY_CONTEXT *a;
     ENTROPY_CONTEXT *l;
     int i;
@@ -231,29 +263,24 @@ int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
     int stop;
     INT16 val, bits_count;
     INT16 c;
-    INT16 t;
     INT16 v;
     const vp8_prob *Prob;
 
-    //int *scan;
     type = 3;
     i = 0;
     stop = 16;
 
-    if (x->mbmi.mode != B_PRED && x->mbmi.mode != SPLITMV)
+    scan = vp8_default_zig_zag1d;
+    qcoeff_ptr = &x->qcoeff[0];
+
+    if (x->mode_info_context->mbmi.mode != B_PRED && x->mode_info_context->mbmi.mode != SPLITMV)
     {
         i = 24;
         stop = 24;
         type = 1;
-        qcoeff_ptr = &x->qcoeff[24*16];
+        qcoeff_ptr += 24*16;
-        scan = vp8_default_zig_zag1d;
         eobtotal -= 16;
     }
-    else
-    {
-        scan = vp8_default_zig_zag1d;
-        qcoeff_ptr = &x->qcoeff[0];
-    }
 
     bufend  = bc->user_buffer_end;
     bufptr  = bc->user_buffer;
@@ -265,13 +292,15 @@ int vp8_decode_mb_tokens(VP8D_COMP *dx, MACROBLOCKD *x)
     coef_probs = oc->fc.coef_probs [type] [ 0 ] [0];
 
 BLOCK_LOOP:
-    a = A[ vp8_block2context[i] ] + vp8_block2above[i];
+    a = A + vp8_block2above[i];
-    l = L[ vp8_block2context[i] ] + vp8_block2left[i];
+    l = L + vp8_block2left[i];
+
     c = (INT16)(!type);
 
-    VP8_COMBINEENTROPYCONTEXTS(t, *a, *l);
+//    Dest = ((A)!=0) + ((B)!=0);
+    VP8_COMBINEENTROPYCONTEXTS(v, *a, *l);
     Prob = coef_probs;
-    Prob += t * ENTROPY_NODES;
+    Prob += v * ENTROPY_NODES;
 
 DO_WHILE:
     Prob += vp8_coef_bands_x[c];
@@ -358,9 +387,8 @@ ONE_CONTEXT_NODE_0_:
 
     qcoeff_ptr [ scan[15] ] = (INT16) v;
 BLOCK_FINISHED:
-    t = ((x->block[i].eob = c) != !type);   // any nonzero data?
+    *a = *l = ((eobs[i] = c) != !type);   // any nonzero data?
-    eobtotal += x->block[i].eob;
+    eobtotal += c;
-    *a = *l = t;
     qcoeff_ptr += 16;
 
     i++;
@@ -370,12 +398,11 @@ BLOCK_FINISHED:
 
     if (i == 25)
     {
-        scan = vp8_default_zig_zag1d;//x->scan_order1d;
         type = 0;
         i = 0;
         stop = 16;
         coef_probs = oc->fc.coef_probs [type] [ 0 ] [0];
-        qcoeff_ptr = &x->qcoeff[0];
+        qcoeff_ptr -= (24*16 + 16);
         goto BLOCK_LOOP;
     }
 
@@ -395,3 +422,4 @@ BLOCK_FINISHED:
     return eobtotal;
 
 }
+#endif //!CONFIG_ASM_DETOK

vp8/decoder/detokenize.h

@@ -9,12 +9,16 @@
  */
 
 
-#ifndef detokenize_h
+#ifndef DETOKENIZE_H
-#define detokenize_h 1
+#define DETOKENIZE_H
 
 #include "onyxd_int.h"
 
+#if ARCH_ARM
+#include "arm/detokenize_arm.h"
+#endif
+
 void vp8_reset_mb_tokens_context(MACROBLOCKD *x);
 int vp8_decode_mb_tokens(VP8D_COMP *, MACROBLOCKD *);
 
-#endif /* detokenize_h */
+#endif /* DETOKENIZE_H */

vp8/decoder/generic/dsystemdependent.c

@@ -21,8 +21,11 @@ void vp8_dmachine_specific_config(VP8D_COMP *pbi)
 #if CONFIG_RUNTIME_CPU_DETECT
     pbi->mb.rtcd                     = &pbi->common.rtcd;
     pbi->dequant.block               = vp8_dequantize_b_c;
-    pbi->dequant.idct    = vp8_dequant_idct_c;
+    pbi->dequant.idct_add            = vp8_dequant_idct_add_c;
-    pbi->dequant.idct_dc = vp8_dequant_dc_idct_c;
+    pbi->dequant.dc_idct_add         = vp8_dequant_dc_idct_add_c;
+    pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_c;
+    pbi->dequant.idct_add_y_block    = vp8_dequant_idct_add_y_block_c;
+    pbi->dequant.idct_add_uv_block   = vp8_dequant_idct_add_uv_block_c;
     pbi->dboolhuff.start             = vp8dx_start_decode_c;
     pbi->dboolhuff.fill              = vp8dx_bool_decoder_fill_c;
 #if 0 //For use with RTCD, when implemented

vp8/decoder/idct_blk.c

@@ -0,0 +1,116 @@
+/*
+ *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "idct.h"
+#include "dequantize.h"
+
+void vp8_dequant_dc_idct_add_y_block_c
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs, short *dc)
+{
+    int i, j;
+
+    for (i = 0; i < 4; i++)
+    {
+        for (j = 0; j < 4; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_dc_idct_add_c (q, dq, pre, dst, 16, stride, dc[0]);
+            else
+                vp8_dc_only_idct_add_c (dc[0], pre, dst, 16, stride);
+
+            q   += 16;
+            pre += 4;
+            dst += 4;
+            dc  ++;
+        }
+
+        pre += 64 - 16;
+        dst += 4*stride - 16;
+    }
+}
+
+void vp8_dequant_idct_add_y_block_c
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs)
+{
+    int i, j;
+
+    for (i = 0; i < 4; i++)
+    {
+        for (j = 0; j < 4; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_c (q, dq, pre, dst, 16, stride);
+            else
+            {
+                vp8_dc_only_idct_add_c (q[0]*dq[0], pre, dst, 16, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q   += 16;
+            pre += 4;
+            dst += 4;
+        }
+
+        pre += 64 - 16;
+        dst += 4*stride - 16;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_c
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    int i, j;
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 2; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_c (q, dq, pre, dstu, 8, stride);
+            else
+            {
+                vp8_dc_only_idct_add_c (q[0]*dq[0], pre, dstu, 8, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q    += 16;
+            pre  += 4;
+            dstu += 4;
+        }
+
+        pre  += 32 - 8;
+        dstu += 4*stride - 8;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 2; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_c (q, dq, pre, dstv, 8, stride);
+            else
+            {
+                vp8_dc_only_idct_add_c (q[0]*dq[0], pre, dstv, 8, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q    += 16;
+            pre  += 4;
+            dstv += 4;
+        }
+
+        pre  += 32 - 8;
+        dstv += 4*stride - 8;
+    }
+}

vp8/decoder/onyxd_if.c

@@ -24,18 +24,16 @@
 #include "threading.h"
 #include "decoderthreading.h"
 #include <stdio.h>
-#include "segmentation_common.h"
+
 #include "quant_common.h"
 #include "vpx_scale/vpxscale.h"
 #include "systemdependent.h"
 #include "vpx_ports/vpx_timer.h"
-
+#include "detokenize.h"
 
 extern void vp8_init_loop_filter(VP8_COMMON *cm);
-
 extern void vp8cx_init_de_quantizer(VP8D_COMP *pbi);
 
-// DEBUG code
 #if CONFIG_DEBUG
 void vp8_recon_write_yuv_frame(unsigned char *name, YV12_BUFFER_CONFIG *s)
 {
@@ -129,6 +127,9 @@ VP8D_PTR vp8dx_create_decompressor(VP8D_CONFIG *oxcf)
         cm->last_sharpness_level = cm->sharpness_level;
     }
 
+#if CONFIG_ARM_ASM_DETOK
+    vp8_init_detokenizer(pbi);
+#endif
     pbi->common.error.setjmp = 0;
     return (VP8D_PTR) pbi;
 }
@@ -180,38 +181,38 @@ int vp8dx_get_reference(VP8D_PTR ptr, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_C
 {
     VP8D_COMP *pbi = (VP8D_COMP *) ptr;
     VP8_COMMON *cm = &pbi->common;
+    int ref_fb_idx;
 
     if (ref_frame_flag == VP8_LAST_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->last_frame, sd);
+        ref_fb_idx = cm->lst_fb_idx;
-
     else if (ref_frame_flag == VP8_GOLD_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->golden_frame, sd);
+        ref_fb_idx = cm->gld_fb_idx;
-
     else if (ref_frame_flag == VP8_ALT_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->alt_ref_frame, sd);
+        ref_fb_idx = cm->alt_fb_idx;
-
     else
         return -1;
 
+    vp8_yv12_copy_frame_ptr(&cm->yv12_fb[ref_fb_idx], sd);
+
     return 0;
 }
 int vp8dx_set_reference(VP8D_PTR ptr, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd)
 {
     VP8D_COMP *pbi = (VP8D_COMP *) ptr;
     VP8_COMMON *cm = &pbi->common;
+    int ref_fb_idx;
 
     if (ref_frame_flag == VP8_LAST_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->last_frame);
+        ref_fb_idx = cm->lst_fb_idx;
-
     else if (ref_frame_flag == VP8_GOLD_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->golden_frame);
+        ref_fb_idx = cm->gld_fb_idx;
-
     else if (ref_frame_flag == VP8_ALT_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->alt_ref_frame);
+        ref_fb_idx = cm->alt_fb_idx;
-
     else
         return -1;
 
+    vp8_yv12_copy_frame_ptr(sd, &cm->yv12_fb[ref_fb_idx]);
+
     return 0;
 }
 
@@ -221,12 +222,95 @@ extern void vp8_push_neon(INT64 *store);
 extern void vp8_pop_neon(INT64 *store);
 static INT64 dx_store_reg[8];
 #endif
+
+static int get_free_fb (VP8_COMMON *cm)
+{
+    int i;
+    for (i = 0; i < NUM_YV12_BUFFERS; i++)
+        if (cm->fb_idx_ref_cnt[i] == 0)
+            break;
+
+    cm->fb_idx_ref_cnt[i] = 1;
+    return i;
+}
+
+static void ref_cnt_fb (int *buf, int *idx, int new_idx)
+{
+    if (buf[*idx] > 0)
+        buf[*idx]--;
+
+    *idx = new_idx;
+
+    buf[new_idx]++;
+}
+
+// If any buffer copy / swapping is signalled it should be done here.
+static int swap_frame_buffers (VP8_COMMON *cm)
+{
+    int fb_to_update_with, err = 0;
+
+    if (cm->refresh_last_frame)
+        fb_to_update_with = cm->lst_fb_idx;
+    else
+        fb_to_update_with = cm->new_fb_idx;
+
+    // The alternate reference frame or golden frame can be updated
+    //  using the new, last, or golden/alt ref frame.  If it
+    //  is updated using the newly decoded frame it is a refresh.
+    //  An update using the last or golden/alt ref frame is a copy.
+    if (cm->copy_buffer_to_arf)
+    {
+        int new_fb = 0;
+
+        if (cm->copy_buffer_to_arf == 1)
+            new_fb = fb_to_update_with;
+        else if (cm->copy_buffer_to_arf == 2)
+            new_fb = cm->gld_fb_idx;
+        else
+            err = -1;
+
+        ref_cnt_fb (cm->fb_idx_ref_cnt, &cm->alt_fb_idx, new_fb);
+    }
+
+    if (cm->copy_buffer_to_gf)
+    {
+        int new_fb = 0;
+
+        if (cm->copy_buffer_to_gf == 1)
+            new_fb = fb_to_update_with;
+        else if (cm->copy_buffer_to_gf == 2)
+            new_fb = cm->alt_fb_idx;
+        else
+            err = -1;
+
+        ref_cnt_fb (cm->fb_idx_ref_cnt, &cm->gld_fb_idx, new_fb);
+    }
+
+    if (cm->refresh_golden_frame)
+        ref_cnt_fb (cm->fb_idx_ref_cnt, &cm->gld_fb_idx, cm->new_fb_idx);
+
+    if (cm->refresh_alt_ref_frame)
+        ref_cnt_fb (cm->fb_idx_ref_cnt, &cm->alt_fb_idx, cm->new_fb_idx);
+
+    if (cm->refresh_last_frame)
+    {
+        ref_cnt_fb (cm->fb_idx_ref_cnt, &cm->lst_fb_idx, cm->new_fb_idx);
+
+        cm->frame_to_show = &cm->yv12_fb[cm->lst_fb_idx];
+    }
+    else
+        cm->frame_to_show = &cm->yv12_fb[cm->new_fb_idx];
+
+    cm->fb_idx_ref_cnt[cm->new_fb_idx]--;
+
+    return err;
+}
+
 int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsigned char *source, INT64 time_stamp)
 {
     VP8D_COMP *pbi = (VP8D_COMP *) ptr;
     VP8_COMMON *cm = &pbi->common;
     int retcode = 0;
-
     struct vpx_usec_timer timer;
 
 //  if(pbi->ready_for_new_data == 0)
@@ -257,6 +341,8 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
     pbi->Source = source;
     pbi->source_sz = size;
 
+    cm->new_fb_idx = get_free_fb (cm);
+
     retcode = vp8_decode_frame(pbi);
 
     if (retcode < 0)
@@ -269,23 +355,17 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
         return retcode;
     }
 
-    // Update the GF useage maps.
-    vp8_update_gf_useage_maps(cm, &pbi->mb);
-
     if (pbi->b_multithreaded_lf && pbi->common.filter_level != 0)
         vp8_stop_lfthread(pbi);
 
-    if (cm->refresh_last_frame)
+    if (swap_frame_buffers (cm))
     {
-        vp8_swap_yv12_buffer(&cm->last_frame, &cm->new_frame);
+        pbi->common.error.error_code = VPX_CODEC_ERROR;
-
+        pbi->common.error.setjmp = 0;
-        cm->frame_to_show = &cm->last_frame;
+        return -1;
-    }
-    else
-    {
-        cm->frame_to_show = &cm->new_frame;
     }
 
+/*
     if (!pbi->b_multithreaded_lf)
     {
         struct vpx_usec_timer lpftimer;
@@ -293,16 +373,45 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
         // Apply the loop filter if appropriate.
 
         if (cm->filter_level > 0)
-        {
             vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
-            cm->last_frame_type = cm->frame_type;
-            cm->last_filter_type = cm->filter_type;
-            cm->last_sharpness_level = cm->sharpness_level;
-
-        }
 
         vpx_usec_timer_mark(&lpftimer);
         pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
+    }else{
+      struct vpx_usec_timer lpftimer;
+      vpx_usec_timer_start(&lpftimer);
+      // Apply the loop filter if appropriate.
+
+      if (cm->filter_level > 0)
+          vp8_mt_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
+
+      vpx_usec_timer_mark(&lpftimer);
+      pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
+    }
+    if (cm->filter_level > 0) {
+        cm->last_frame_type = cm->frame_type;
+        cm->last_filter_type = cm->filter_type;
+        cm->last_sharpness_level = cm->sharpness_level;
+    }
+*/
+
+    if(pbi->common.filter_level)
+    {
+        struct vpx_usec_timer lpftimer;
+        vpx_usec_timer_start(&lpftimer);
+        // Apply the loop filter if appropriate.
+
+        if (pbi->b_multithreaded_lf && cm->multi_token_partition != ONE_PARTITION)
+            vp8_mt_loop_filter_frame(pbi);   //cm, &pbi->mb, cm->filter_level);
+        else
+            vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
+
+        vpx_usec_timer_mark(&lpftimer);
+        pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
+
+        cm->last_frame_type = cm->frame_type;
+        cm->last_filter_type = cm->filter_type;
+        cm->last_sharpness_level = cm->sharpness_level;
     }
 
     vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show);
@@ -314,49 +423,6 @@ int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, const unsign
         write_dx_frame_to_file(cm->frame_to_show, cm->current_video_frame);
 #endif
 
-    // If any buffer copy / swaping is signalled it should be done here.
-    if (cm->copy_buffer_to_arf)
-    {
-        if (cm->copy_buffer_to_arf == 1)
-        {
-            if (cm->refresh_last_frame)
-                vp8_yv12_copy_frame_ptr(&cm->new_frame, &cm->alt_ref_frame);
-            else
-                vp8_yv12_copy_frame_ptr(&cm->last_frame, &cm->alt_ref_frame);
-        }
-        else if (cm->copy_buffer_to_arf == 2)
-            vp8_yv12_copy_frame_ptr(&cm->golden_frame, &cm->alt_ref_frame);
-    }
-
-    if (cm->copy_buffer_to_gf)
-    {
-        if (cm->copy_buffer_to_gf == 1)
-        {
-            if (cm->refresh_last_frame)
-                vp8_yv12_copy_frame_ptr(&cm->new_frame, &cm->golden_frame);
-            else
-                vp8_yv12_copy_frame_ptr(&cm->last_frame, &cm->golden_frame);
-        }
-        else if (cm->copy_buffer_to_gf == 2)
-            vp8_yv12_copy_frame_ptr(&cm->alt_ref_frame, &cm->golden_frame);
-    }
-
-    // Should the golden or alternate reference frame be refreshed?
-    if (cm->refresh_golden_frame || cm->refresh_alt_ref_frame)
-    {
-        if (cm->refresh_golden_frame)
-            vp8_yv12_copy_frame_ptr(cm->frame_to_show, &cm->golden_frame);
-
-        if (cm->refresh_alt_ref_frame)
-            vp8_yv12_copy_frame_ptr(cm->frame_to_show, &cm->alt_ref_frame);
-
-        //vpx_log("Decoder: recovery frame received \n");
-
-        // Update data structures that monitors GF useage
-        vpx_memset(cm->gf_active_flags, 1, (cm->mb_rows * cm->mb_cols));
-        cm->gf_active_count = cm->mb_rows * cm->mb_cols;
-    }
-
     vp8_clear_system_state();
 
     vpx_usec_timer_mark(&timer);

vp8/decoder/onyxd_if_sjl.c

@@ -1,399 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "onyxc_int.h"
-#include "postproc.h"
-#include "onyxd.h"
-#include "onyxd_int.h"
-#include "vpx_mem/vpx_mem.h"
-#include "alloccommon.h"
-#include "vpx_scale/yv12extend.h"
-#include "loopfilter.h"
-#include "swapyv12buffer.h"
-#include "g_common.h"
-#include "threading.h"
-#include "decoderthreading.h"
-#include <stdio.h>
-#include "segmentation_common.h"
-#include "quant_common.h"
-#include "vpx_scale/vpxscale.h"
-#include "systemdependent.h"
-#include "vpx_ports/vpx_timer.h"
-
-
-#ifndef VPX_NO_GLOBALS
-static int init_ct = 0;
-#else
-# include "vpx_global_handling.h"
-# define init_ct ((int)vpxglobalm(onyxd,init_ct))
-#endif
-
-extern void vp8_init_loop_filter(VP8_COMMON *cm);
-
-extern void vp8cx_init_de_quantizer(VP8D_COMP *pbi);
-extern void init_detokenizer(VP8D_COMP *dx);
-
-// DEBUG code
-void vp8_recon_write_yuv_frame(unsigned char *name, YV12_BUFFER_CONFIG *s)
-{
-    FILE *yuv_file = fopen((char *)name, "ab");
-    unsigned char *src = s->y_buffer;
-    int h = s->y_height;
-
-    do
-    {
-        fwrite(src, s->y_width, 1,  yuv_file);
-        src += s->y_stride;
-    }
-    while (--h);
-
-    src = s->u_buffer;
-    h = s->uv_height;
-
-    do
-    {
-        fwrite(src, s->uv_width, 1,  yuv_file);
-        src += s->uv_stride;
-    }
-    while (--h);
-
-    src = s->v_buffer;
-    h = s->uv_height;
-
-    do
-    {
-        fwrite(src, s->uv_width, 1, yuv_file);
-        src += s->uv_stride;
-    }
-    while (--h);
-
-    fclose(yuv_file);
-}
-
-void vp8dx_initialize()
-{
-    if (!init_ct++)
-    {
-        vp8_initialize_common();
-        vp8_scale_machine_specific_config();
-    }
-}
-
-void vp8dx_shutdown()
-{
-    if (!--init_ct)
-    {
-        vp8_shutdown_common();
-    }
-}
-
-
-VP8D_PTR vp8dx_create_decompressor(VP8D_CONFIG *oxcf)
-{
-    VP8D_COMP *pbi = vpx_memalign(32, sizeof(VP8D_COMP));
-
-    if (!pbi)
-        return NULL;
-
-    vpx_memset(pbi, 0, sizeof(VP8D_COMP));
-
-    vp8dx_initialize();
-
-    vp8_create_common(&pbi->common);
-    vp8_dmachine_specific_config(pbi);
-
-    pbi->common.current_video_frame = 0;
-    pbi->ready_for_new_data = 1;
-
-    pbi->CPUFreq = 0; //vp8_get_processor_freq();
-    pbi->max_threads = oxcf->max_threads;
-    vp8_decoder_create_threads(pbi);
-
-    //vp8cx_init_de_quantizer() is first called here. Add check in frame_init_dequantizer() to avoid
-    // unnecessary calling of vp8cx_init_de_quantizer() for every frame.
-    vp8cx_init_de_quantizer(pbi);
-
-    {
-        VP8_COMMON *cm = &pbi->common;
-
-        vp8_init_loop_filter(cm);
-        cm->last_frame_type = KEY_FRAME;
-        cm->last_filter_type = cm->filter_type;
-        cm->last_sharpness_level = cm->sharpness_level;
-    }
-
-    init_detokenizer(pbi);
-
-    return (VP8D_PTR) pbi;
-}
-void vp8dx_remove_decompressor(VP8D_PTR ptr)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) ptr;
-
-    if (!pbi)
-        return;
-
-    vp8_decoder_remove_threads(pbi);
-    vp8_remove_common(&pbi->common);
-    vpx_free(pbi);
-    vp8dx_shutdown();
-
-}
-
-void vp8dx_set_setting(VP8D_PTR comp, VP8D_SETTING oxst, int x)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) comp;
-
-    (void) pbi;
-    (void) x;
-
-    switch (oxst)
-    {
-    case VP8D_OK:
-        break;
-    }
-}
-
-int vp8dx_get_setting(VP8D_PTR comp, VP8D_SETTING oxst)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) comp;
-
-    (void) pbi;
-
-    switch (oxst)
-    {
-    case VP8D_OK:
-        break;
-    }
-
-    return -1;
-}
-
-int vp8dx_get_reference(VP8D_PTR ptr, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) ptr;
-    VP8_COMMON *cm = &pbi->common;
-
-    if (ref_frame_flag == VP8_LAST_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->last_frame, sd);
-
-    else if (ref_frame_flag == VP8_GOLD_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->golden_frame, sd);
-
-    else if (ref_frame_flag == VP8_ALT_FLAG)
-        vp8_yv12_copy_frame_ptr(&cm->alt_ref_frame, sd);
-
-    else
-        return -1;
-
-    return 0;
-}
-int vp8dx_set_reference(VP8D_PTR ptr, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) ptr;
-    VP8_COMMON *cm = &pbi->common;
-
-    if (ref_frame_flag == VP8_LAST_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->last_frame);
-
-    else if (ref_frame_flag == VP8_GOLD_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->golden_frame);
-
-    else if (ref_frame_flag == VP8_ALT_FLAG)
-        vp8_yv12_copy_frame_ptr(sd, &cm->alt_ref_frame);
-
-    else
-        return -1;
-
-    return 0;
-}
-int vp8dx_receive_compressed_data(VP8D_PTR ptr, unsigned long size, char *source, INT64 time_stamp)
-{
-    VP8D_COMP *pbi = (VP8D_COMP *) ptr;
-    VP8_COMMON *cm = &pbi->common;
-    int retcode = 0;
-
-    struct vpx_usec_timer timer;
-    (void) size;
-
-//  if(pbi->ready_for_new_data == 0)
-//      return -1;
-
-    vpx_usec_timer_start(&timer);
-
-    if (ptr == 0)
-    {
-        return -1;
-    }
-
-    //cm->current_video_frame++;
-    pbi->Source = source;
-
-    retcode = vp8_decode_frame(pbi);
-
-    if (retcode < 0)
-        return retcode;
-
-    // Update the GF useage maps.
-    vp8_update_gf_useage_maps(cm, &pbi->mb);
-
-    if (pbi->b_multithreaded)
-        vp8_stop_lfthread(pbi);
-
-    if (cm->refresh_last_frame)
-    {
-        vp8_swap_yv12_buffer(&cm->last_frame, &cm->new_frame);
-
-        cm->frame_to_show = &cm->last_frame;
-    }
-    else
-    {
-        cm->frame_to_show = &cm->new_frame;
-    }
-
-    if (!pbi->b_multithreaded)
-    {
-        struct vpx_usec_timer lpftimer;
-        vpx_usec_timer_start(&lpftimer);
-        // Apply the loop filter if appropriate.
-
-        if (cm->filter_level > 0)
-        {
-            vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
-            cm->last_frame_type = cm->frame_type;
-            cm->last_filter_type = cm->filter_type;
-            cm->last_sharpness_level = cm->sharpness_level;
-
-        }
-
-        vpx_usec_timer_mark(&lpftimer);
-        pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
-    }
-
-    vp8_yv12_extend_frame_borders_ptr(cm->frame_to_show);
-
-#if 0
-    // DEBUG code
-    //vp8_recon_write_yuv_frame("recon.yuv", cm->frame_to_show);
-    if (cm->current_video_frame <= 5)
-        write_dx_frame_to_file(cm->frame_to_show, cm->current_video_frame);
-#endif
-
-    // If any buffer copy / swaping is signalled it should be done here.
-    if (cm->copy_buffer_to_arf)
-    {
-        if (cm->copy_buffer_to_arf == 1)
-        {
-            if (cm->refresh_last_frame)
-                vp8_yv12_copy_frame_ptr(&cm->new_frame, &cm->alt_ref_frame);
-            else
-                vp8_yv12_copy_frame_ptr(&cm->last_frame, &cm->alt_ref_frame);
-        }
-        else if (cm->copy_buffer_to_arf == 2)
-            vp8_yv12_copy_frame_ptr(&cm->golden_frame, &cm->alt_ref_frame);
-    }
-
-    if (cm->copy_buffer_to_gf)
-    {
-        if (cm->copy_buffer_to_gf == 1)
-        {
-            if (cm->refresh_last_frame)
-                vp8_yv12_copy_frame_ptr(&cm->new_frame, &cm->golden_frame);
-            else
-                vp8_yv12_copy_frame_ptr(&cm->last_frame, &cm->golden_frame);
-        }
-        else if (cm->copy_buffer_to_gf == 2)
-            vp8_yv12_copy_frame_ptr(&cm->alt_ref_frame, &cm->golden_frame);
-    }
-
-    // Should the golden or alternate reference frame be refreshed?
-    if (cm->refresh_golden_frame || cm->refresh_alt_ref_frame)
-    {
-        if (cm->refresh_golden_frame)
-            vp8_yv12_copy_frame_ptr(cm->frame_to_show, &cm->golden_frame);
-
-        if (cm->refresh_alt_ref_frame)
-            vp8_yv12_copy_frame_ptr(cm->frame_to_show, &cm->alt_ref_frame);
-
-        //vpx_log("Decoder: recovery frame received \n");
-
-        // Update data structures that monitors GF useage
-        vpx_memset(cm->gf_active_flags, 1, (cm->mb_rows * cm->mb_cols));
-        cm->gf_active_count = cm->mb_rows * cm->mb_cols;
-    }
-
-    vp8_clear_system_state();
-
-    vpx_usec_timer_mark(&timer);
-    pbi->decode_microseconds = vpx_usec_timer_elapsed(&timer);
-
-    pbi->time_decoding += pbi->decode_microseconds;
-
-//  vp8_print_modes_and_motion_vectors( cm->mi, cm->mb_rows,cm->mb_cols, cm->current_video_frame);
-
-    cm->current_video_frame++;
-    pbi->ready_for_new_data = 0;
-    pbi->last_time_stamp = time_stamp;
-
-    {
-        int i;
-        INT64 earliest_time = pbi->dr[0].time_stamp;
-        INT64 latest_time = pbi->dr[0].time_stamp;
-        INT64 time_diff = 0;
-        int bytes = 0;
-
-        pbi->dr[pbi->common.current_video_frame&0xf].size = pbi->bc.pos + pbi->bc2.pos + 4;;
-        pbi->dr[pbi->common.current_video_frame&0xf].time_stamp = time_stamp;
-
-        for (i = 0; i < 16; i++)
-        {
-
-            bytes += pbi->dr[i].size;
-
-            if (pbi->dr[i].time_stamp < earliest_time)
-                earliest_time = pbi->dr[i].time_stamp;
-
-            if (pbi->dr[i].time_stamp > latest_time)
-                latest_time = pbi->dr[i].time_stamp;
-        }
-
-        time_diff = latest_time - earliest_time;
-
-        if (time_diff > 0)
-        {
-            pbi->common.bitrate = 80000.00 * bytes / time_diff  ;
-            pbi->common.framerate = 160000000.00 / time_diff ;
-        }
-
-    }
-    return retcode;
-}
-int vp8dx_get_raw_frame(VP8D_PTR ptr, YV12_BUFFER_CONFIG *sd, INT64 *time_stamp, INT64 *time_end_stamp, int deblock_level,  int noise_level, int flags)
-{
-    int ret = -1;
-    VP8D_COMP *pbi = (VP8D_COMP *) ptr;
-
-    if (pbi->ready_for_new_data == 1)
-        return ret;
-
-    // ie no raw frame to show!!!
-    if (pbi->common.show_frame == 0)
-        return ret;
-
-    pbi->ready_for_new_data = 1;
-    *time_stamp = pbi->last_time_stamp;
-    *time_end_stamp = 0;
-
-    sd->clrtype = pbi->common.clr_type;
-    ret = vp8_post_proc_frame(&pbi->common, sd, deblock_level, noise_level, flags);
-    vp8_clear_system_state();
-    return ret;
-}

vp8/decoder/onyxd_int.h

@@ -48,21 +48,20 @@ typedef struct
 
 typedef struct
 {
-    int *scan;
+    int const *scan;
-    UINT8 *ptr_onyxblock2context_leftabove;
+    UINT8 const *ptr_block2leftabove;
-    vp8_tree_index *vp8_coef_tree_ptr;  //onyx_coef_tree_ptr; ???
+    vp8_tree_index const *vp8_coef_tree_ptr;
-    TOKENEXTRABITS *teb_base_ptr;
+    TOKENEXTRABITS const *teb_base_ptr;
     unsigned char *norm_ptr;
-//  UINT16 *ptr_onyx_coef_bands_x;
+    UINT8 *ptr_coef_bands_x;
-    UINT8 *ptr_onyx_coef_bands_x;
 
-    ENTROPY_CONTEXT   **A;
+    ENTROPY_CONTEXT_PLANES *A;
-    ENTROPY_CONTEXT(*L)[4];
+    ENTROPY_CONTEXT_PLANES *L;
 
     INT16 *qcoeff_start_ptr;
     BOOL_DECODER *current_bc;
 
-    UINT8 *coef_probs[4];
+    vp8_prob const *coef_probs[4];
 
     UINT8 eob[25];
 
@@ -95,20 +94,22 @@ typedef struct VP8Decompressor
     int current_mb_col_main;
     int decoding_thread_count;
     int allocated_decoding_thread_count;
+    int *current_mb_col;                  //Each row remembers its already decoded column.
+    int mt_baseline_filter_level[MAX_MB_SEGMENTS];
 
     // variable for threading
     DECLARE_ALIGNED(16, MACROBLOCKD, lpfmb);
 #if CONFIG_MULTITHREAD
-    pthread_t           h_thread_lpf;         // thread for postprocessing
+    //pthread_t           h_thread_lpf;         // thread for postprocessing
-    sem_t               h_event_lpf;          // Event for post_proc completed
+    sem_t               h_event_end_lpf;          // Event for post_proc completed
-    sem_t               h_event_start_lpf;
+    sem_t               *h_event_start_lpf;
 #endif
     MB_ROW_DEC           *mb_row_di;
     DECODETHREAD_DATA   *de_thread_data;
 #if CONFIG_MULTITHREAD
     pthread_t           *h_decoding_thread;
-    sem_t               *h_event_mbrdecoding;
+    sem_t               *h_event_start_decoding;
-    sem_t               h_event_main;
+    sem_t               h_event_end_decoding;
     // end of threading data
 #endif
     vp8_reader *mbc;

vp8/decoder/threading.c

@@ -12,6 +12,9 @@
 #ifndef WIN32
 # include <unistd.h>
 #endif
+#ifdef __APPLE__
+#include <mach/mach_init.h>
+#endif
 #include "onyxd_int.h"
 #include "vpx_mem/vpx_mem.h"
 #include "threading.h"
@@ -20,6 +23,8 @@
 #include "extend.h"
 #include "vpx_ports/vpx_timer.h"
 
+#define MAX_ROWS 256
+
 extern void vp8_decode_mb_row(VP8D_COMP *pbi,
                               VP8_COMMON *pc,
                               int mb_row,
@@ -28,11 +33,10 @@ extern void vp8_decode_mb_row(VP8D_COMP *pbi,
 extern void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel);
 extern void vp8_decode_macroblock(VP8D_COMP *pbi, MACROBLOCKD *xd);
 
+void vp8_thread_loop_filter(VP8D_COMP *pbi, MB_ROW_DEC *mbrd, int ithread);
+
 void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC *mbrd, int count)
 {
-
-
-
 #if CONFIG_MULTITHREAD
     VP8_COMMON *const pc = & pbi->common;
     int i, j;
@@ -43,13 +47,10 @@ void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC
 #if CONFIG_RUNTIME_CPU_DETECT
         mbd->rtcd = xd->rtcd;
 #endif
-
-
         mbd->subpixel_predict        = xd->subpixel_predict;
         mbd->subpixel_predict8x4     = xd->subpixel_predict8x4;
         mbd->subpixel_predict8x8     = xd->subpixel_predict8x8;
         mbd->subpixel_predict16x16   = xd->subpixel_predict16x16;
-        mbd->gf_active_ptr            = xd->gf_active_ptr;
 
         mbd->mode_info        = pc->mi - 1;
         mbd->mode_info_context = pc->mi   + pc->mode_info_stride * (i + 1);
@@ -59,11 +60,8 @@ void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC
         mbd->frames_since_golden      = pc->frames_since_golden;
         mbd->frames_till_alt_ref_frame  = pc->frames_till_alt_ref_frame;
 
-        mbd->pre = pc->last_frame;
+        mbd->pre = pc->yv12_fb[pc->lst_fb_idx];
-        mbd->dst = pc->new_frame;
+        mbd->dst = pc->yv12_fb[pc->new_fb_idx];
-
-
-
 
         vp8_setup_block_dptrs(mbd);
         vp8_build_block_doffsets(mbd);
@@ -71,9 +69,6 @@ void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC
         mbd->mb_segement_abs_delta     = xd->mb_segement_abs_delta;
         vpx_memcpy(mbd->segment_feature_data, xd->segment_feature_data, sizeof(xd->segment_feature_data));
 
-        mbd->mbmi.mode = DC_PRED;
-        mbd->mbmi.uv_mode = DC_PRED;
-
         mbd->current_bc = &pbi->bc2;
 
         for (j = 0; j < 25; j++)
@@ -82,6 +77,8 @@ void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC
         }
     }
 
+    for (i=0; i< pc->mb_rows; i++)
+        pbi->current_mb_col[i]=-1;
 #else
     (void) pbi;
     (void) xd;
@@ -90,6 +87,69 @@ void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC
 #endif
 }
 
+void vp8_setup_loop_filter_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC *mbrd, int count)
+{
+#if CONFIG_MULTITHREAD
+    VP8_COMMON *const pc = & pbi->common;
+    int i, j;
+
+    for (i = 0; i < count; i++)
+    {
+        MACROBLOCKD *mbd = &mbrd[i].mbd;
+//#if CONFIG_RUNTIME_CPU_DETECT
+//        mbd->rtcd = xd->rtcd;
+//#endif
+
+        //mbd->subpixel_predict        = xd->subpixel_predict;
+        //mbd->subpixel_predict8x4     = xd->subpixel_predict8x4;
+        //mbd->subpixel_predict8x8     = xd->subpixel_predict8x8;
+        //mbd->subpixel_predict16x16   = xd->subpixel_predict16x16;
+
+        mbd->mode_info        = pc->mi - 1;
+        mbd->mode_info_context = pc->mi   + pc->mode_info_stride * (i + 1);
+        mbd->mode_info_stride  = pc->mode_info_stride;
+
+        //mbd->frame_type = pc->frame_type;
+        //mbd->frames_since_golden      = pc->frames_since_golden;
+        //mbd->frames_till_alt_ref_frame  = pc->frames_till_alt_ref_frame;
+
+        //mbd->pre = pc->yv12_fb[pc->lst_fb_idx];
+        //mbd->dst = pc->yv12_fb[pc->new_fb_idx];
+
+        //vp8_setup_block_dptrs(mbd);
+        //vp8_build_block_doffsets(mbd);
+        mbd->segmentation_enabled    = xd->segmentation_enabled;  //
+        mbd->mb_segement_abs_delta     = xd->mb_segement_abs_delta;  //
+        vpx_memcpy(mbd->segment_feature_data, xd->segment_feature_data, sizeof(xd->segment_feature_data));   //
+
+        //signed char ref_lf_deltas[MAX_REF_LF_DELTAS];
+        vpx_memcpy(mbd->ref_lf_deltas, xd->ref_lf_deltas, sizeof(xd->ref_lf_deltas));
+        //signed char mode_lf_deltas[MAX_MODE_LF_DELTAS];
+        vpx_memcpy(mbd->mode_lf_deltas, xd->mode_lf_deltas, sizeof(xd->mode_lf_deltas));
+        //unsigned char mode_ref_lf_delta_enabled;
+        //unsigned char mode_ref_lf_delta_update;
+        mbd->mode_ref_lf_delta_enabled    = xd->mode_ref_lf_delta_enabled;
+        mbd->mode_ref_lf_delta_update    = xd->mode_ref_lf_delta_update;
+
+        //mbd->mbmi.mode = DC_PRED;
+        //mbd->mbmi.uv_mode = DC_PRED;
+        //mbd->current_bc = &pbi->bc2;
+
+        //for (j = 0; j < 25; j++)
+        //{
+        //    mbd->block[j].dequant = xd->block[j].dequant;
+        //}
+    }
+
+    for (i=0; i< pc->mb_rows; i++)
+        pbi->current_mb_col[i]=-1;
+#else
+    (void) pbi;
+    (void) xd;
+    (void) mbrd;
+    (void) count;
+#endif
+}
 
 THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
 {
@@ -97,48 +157,51 @@ THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
     int ithread = ((DECODETHREAD_DATA *)p_data)->ithread;
     VP8D_COMP *pbi = (VP8D_COMP *)(((DECODETHREAD_DATA *)p_data)->ptr1);
     MB_ROW_DEC *mbrd = (MB_ROW_DEC *)(((DECODETHREAD_DATA *)p_data)->ptr2);
-    ENTROPY_CONTEXT mb_row_left_context[4][4];
+    ENTROPY_CONTEXT_PLANES mb_row_left_context;
 
     while (1)
     {
+        int current_filter_level = 0;
+
         if (pbi->b_multithreaded_rd == 0)
             break;
 
-        //if(WaitForSingleObject(pbi->h_event_mbrdecoding[ithread], INFINITE) == WAIT_OBJECT_0)
+        //if(WaitForSingleObject(pbi->h_event_start_decoding[ithread], INFINITE) == WAIT_OBJECT_0)
-        if (sem_wait(&pbi->h_event_mbrdecoding[ithread]) == 0)
+        if (sem_wait(&pbi->h_event_start_decoding[ithread]) == 0)
         {
             if (pbi->b_multithreaded_rd == 0)
                 break;
             else
             {
                 VP8_COMMON *pc = &pbi->common;
-                int mb_row       = mbrd->mb_row;
                 MACROBLOCKD *xd = &mbrd->mbd;
 
-                //printf("ithread:%d mb_row %d\n", ithread, mb_row);
+                int mb_row;
+                int num_part = 1 << pbi->common.multi_token_partition;
+                volatile int *last_row_current_mb_col;
+
+                for (mb_row = ithread+1; mb_row < pc->mb_rows; mb_row += (pbi->decoding_thread_count + 1))
+                {
                     int i;
                     int recon_yoffset, recon_uvoffset;
                     int mb_col;
-                int recon_y_stride = pc->last_frame.y_stride;
+                    int ref_fb_idx = pc->lst_fb_idx;
-                int recon_uv_stride = pc->last_frame.uv_stride;
+                    int dst_fb_idx = pc->new_fb_idx;
+                    int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
+                    int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;
 
-                volatile int *last_row_current_mb_col;
+                    pbi->mb_row_di[ithread].mb_row = mb_row;
+                    pbi->mb_row_di[ithread].mbd.current_bc =  &pbi->mbc[mb_row%num_part];
 
-                if (ithread > 0)
+                    last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];
-                    last_row_current_mb_col = &pbi->mb_row_di[ithread-1].current_mb_col;
-                else
-                    last_row_current_mb_col = &pbi->current_mb_col_main;
 
                     recon_yoffset = mb_row * recon_y_stride * 16;
                     recon_uvoffset = mb_row * recon_uv_stride * 8;
                     // reset above block coeffs
 
-                xd->above_context[Y1CONTEXT] = pc->above_context[Y1CONTEXT];
+                    xd->above_context = pc->above_context;
-                xd->above_context[UCONTEXT ] = pc->above_context[UCONTEXT];
+                    xd->left_context = &mb_row_left_context;
-                xd->above_context[VCONTEXT ] = pc->above_context[VCONTEXT];
+                    vpx_memset(&mb_row_left_context, 0, sizeof(mb_row_left_context));
-                xd->above_context[Y2CONTEXT] = pc->above_context[Y2CONTEXT];
-                xd->left_context = mb_row_left_context;
-                vpx_memset(mb_row_left_context, 0, sizeof(mb_row_left_context));
                     xd->up_available = (mb_row != 0);
 
                     xd->mb_to_top_edge = -((mb_row * 16)) << 3;
@@ -146,19 +209,16 @@ THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
 
                     for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
                     {
-
+                        if ((mb_col & 7) == 0)
-                    while (mb_col > (*last_row_current_mb_col - 1) && *last_row_current_mb_col != pc->mb_cols - 1)
+                        {
+                            while (mb_col > (*last_row_current_mb_col - 8) && *last_row_current_mb_col != pc->mb_cols - 1)
                             {
                                 x86_pause_hint();
                                 thread_sleep(0);
                             }
+                        }
 
-                    // Take a copy of the mode and Mv information for this macroblock into the xd->mbmi
+                        if (xd->mode_info_context->mbmi.mode == SPLITMV || xd->mode_info_context->mbmi.mode == B_PRED)
-                    // the partition_bmi array is unused in the decoder, so don't copy it.
-                    vpx_memcpy(&xd->mbmi, &xd->mode_info_context->mbmi,
-                               sizeof(MB_MODE_INFO) - sizeof(xd->mbmi.partition_bmi));
-
-                    if (xd->mbmi.mode == SPLITMV || xd->mbmi.mode == B_PRED)
                         {
                             for (i = 0; i < 16; i++)
                             {
@@ -172,57 +232,42 @@ THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
                         xd->mb_to_left_edge = -((mb_col * 16) << 3);
                         xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
 
-                    xd->dst.y_buffer = pc->new_frame.y_buffer + recon_yoffset;
+                        xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
-                    xd->dst.u_buffer = pc->new_frame.u_buffer + recon_uvoffset;
+                        xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
-                    xd->dst.v_buffer = pc->new_frame.v_buffer + recon_uvoffset;
+                        xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
 
                         xd->left_available = (mb_col != 0);
 
                         // Select the appropriate reference frame for this MB
-                    if (xd->mbmi.ref_frame == LAST_FRAME)
+                        if (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)
-                    {
+                            ref_fb_idx = pc->lst_fb_idx;
-                        xd->pre.y_buffer = pc->last_frame.y_buffer + recon_yoffset;
+                        else if (xd->mode_info_context->mbmi.ref_frame == GOLDEN_FRAME)
-                        xd->pre.u_buffer = pc->last_frame.u_buffer + recon_uvoffset;
+                            ref_fb_idx = pc->gld_fb_idx;
-                        xd->pre.v_buffer = pc->last_frame.v_buffer + recon_uvoffset;
-                    }
-                    else if (xd->mbmi.ref_frame == GOLDEN_FRAME)
-                    {
-                        // Golden frame reconstruction buffer
-                        xd->pre.y_buffer = pc->golden_frame.y_buffer + recon_yoffset;
-                        xd->pre.u_buffer = pc->golden_frame.u_buffer + recon_uvoffset;
-                        xd->pre.v_buffer = pc->golden_frame.v_buffer + recon_uvoffset;
-                    }
                         else
-                    {
+                            ref_fb_idx = pc->alt_fb_idx;
-                        // Alternate reference frame reconstruction buffer
+
-                        xd->pre.y_buffer = pc->alt_ref_frame.y_buffer + recon_yoffset;
+                        xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
-                        xd->pre.u_buffer = pc->alt_ref_frame.u_buffer + recon_uvoffset;
+                        xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
-                        xd->pre.v_buffer = pc->alt_ref_frame.v_buffer + recon_uvoffset;
+                        xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
-                    }
 
                         vp8_build_uvmvs(xd, pc->full_pixel);
 
                         vp8_decode_macroblock(pbi, xd);
 
-
                         recon_yoffset += 16;
                         recon_uvoffset += 8;
 
                         ++xd->mode_info_context;  /* next mb */
 
-                    xd->gf_active_ptr++;      // GF useage flag for next MB
+                        xd->above_context++;
-
-                    xd->above_context[Y1CONTEXT] += 4;
-                    xd->above_context[UCONTEXT ] += 2;
-                    xd->above_context[VCONTEXT ] += 2;
-                    xd->above_context[Y2CONTEXT] ++;
-                    pbi->mb_row_di[ithread].current_mb_col = mb_col;
 
+                        //pbi->mb_row_di[ithread].current_mb_col = mb_col;
+                        pbi->current_mb_col[mb_row] = mb_col;
                     }
 
                     // adjust to the next row of mbs
                     vp8_extend_mb_row(
-                    &pc->new_frame,
+                    &pc->yv12_fb[dst_fb_idx],
                     xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
                     );
 
@@ -231,23 +276,27 @@ THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
                     // since we have multithread
                     xd->mode_info_context += xd->mode_info_stride * pbi->decoding_thread_count;
 
-                //memcpy(&pbi->lpfmb, &pbi->mb, sizeof(pbi->mb));
-                if ((mb_row & 1) == 1)
-                {
                     pbi->last_mb_row_decoded = mb_row;
-                    //printf("S%d", pbi->last_mb_row_decoded);
+
+                }
+            }
         }
 
-                if (ithread == (pbi->decoding_thread_count - 1) || mb_row == pc->mb_rows - 1)
+        // If |pbi->common.filter_level| is 0 the value can change in-between
+        // the sem_post and the check to call vp8_thread_loop_filter.
+        current_filter_level = pbi->common.filter_level;
+
+        //  add this to each frame
+        if ((mbrd->mb_row == pbi->common.mb_rows-1) || ((mbrd->mb_row == pbi->common.mb_rows-2) && (pbi->common.mb_rows % (pbi->decoding_thread_count+1))==1))
         {
-                    //SetEvent(pbi->h_event_main);
+            //SetEvent(pbi->h_event_end_decoding);
-                    sem_post(&pbi->h_event_main);
+            sem_post(&pbi->h_event_end_decoding);
-
-                }
-            }
-        }
         }
 
+        if ((pbi->b_multithreaded_lf) && (current_filter_level))
+            vp8_thread_loop_filter(pbi, mbrd, ithread);
+
+    }
 #else
     (void) p_data;
 #endif
@@ -255,93 +304,60 @@ THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
     return 0 ;
 }
 
-THREAD_FUNCTION vp8_thread_loop_filter(void *p_data)
+
+void vp8_thread_loop_filter(VP8D_COMP *pbi, MB_ROW_DEC *mbrd, int ithread)
 {
 #if CONFIG_MULTITHREAD
-    VP8D_COMP *pbi = (VP8D_COMP *)p_data;
 
-    while (1)
+        if (sem_wait(&pbi->h_event_start_lpf[ithread]) == 0)
         {
-        if (pbi->b_multithreaded_lf == 0)
+           // if (pbi->b_multithreaded_lf == 0) // we're shutting down      ????
-            break;
+           //     break;
-
+           // else
-        //printf("before waiting for start_lpf\n");
-
-        //if(WaitForSingleObject(pbi->h_event_start_lpf, INFINITE) == WAIT_OBJECT_0)
-        if (sem_wait(&pbi->h_event_start_lpf) == 0)
             {
-            if (pbi->b_multithreaded_lf == 0) // we're shutting down
-                break;
-            else
-            {
-
                 VP8_COMMON *cm  = &pbi->common;
-                MACROBLOCKD *mbd = &pbi->lpfmb;
+                MACROBLOCKD *mbd = &mbrd->mbd;
                 int default_filt_lvl = pbi->common.filter_level;
 
-                YV12_BUFFER_CONFIG *post = &cm->new_frame;
+                YV12_BUFFER_CONFIG *post = cm->frame_to_show;
                 loop_filter_info *lfi = cm->lf_info;
+                //int frame_type = cm->frame_type;
 
                 int mb_row;
                 int mb_col;
 
-
-                int baseline_filter_level[MAX_MB_SEGMENTS];
                 int filter_level;
                 int alt_flt_enabled = mbd->segmentation_enabled;
 
                 int i;
                 unsigned char *y_ptr, *u_ptr, *v_ptr;
 
-                volatile int *last_mb_row_decoded = &pbi->last_mb_row_decoded;
+                volatile int *last_row_current_mb_col;
-
-                //MODE_INFO * this_mb_mode_info = cm->mi;
-                mbd->mode_info_context = cm->mi;          // Point at base of Mb MODE_INFO list
-
-                // Note the baseline filter values for each segment
-                if (alt_flt_enabled)
-                {
-                    for (i = 0; i < MAX_MB_SEGMENTS; i++)
-                    {
-                        if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-                            baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-                        else
-                        {
-                            baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-                            baseline_filter_level[i] = (baseline_filter_level[i] >= 0) ? ((baseline_filter_level[i] <= MAX_LOOP_FILTER) ? baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  // Clamp to valid range
-                        }
-                    }
-                }
-                else
-                {
-                    for (i = 0; i < MAX_MB_SEGMENTS; i++)
-                        baseline_filter_level[i] = default_filt_lvl;
-                }
-
-                // Initialize the loop filter for this frame.
-                vp8_init_loop_filter(cm);
 
                 // Set up the buffer pointers
-                y_ptr = post->y_buffer;
+                y_ptr = post->y_buffer + post->y_stride  * 16 * (ithread +1);
-                u_ptr = post->u_buffer;
+                u_ptr = post->u_buffer + post->uv_stride *  8 * (ithread +1);
-                v_ptr = post->v_buffer;
+                v_ptr = post->v_buffer + post->uv_stride *  8 * (ithread +1);
 
                 // vp8_filter each macro block
-                for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
+                for (mb_row = ithread+1; mb_row < cm->mb_rows; mb_row+= (pbi->decoding_thread_count + 1))
                 {
+                    last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];
 
-                    while (mb_row >= *last_mb_row_decoded)
-                    {
-                        x86_pause_hint();
-                        thread_sleep(0);
-                    }
-
-                    //printf("R%d", mb_row);
                     for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
                     {
                         int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
 
-                        filter_level = baseline_filter_level[Segment];
+                        if ((mb_col & 7) == 0)
+                        {
+                            while (mb_col > (*last_row_current_mb_col-8) && *last_row_current_mb_col != cm->mb_cols - 1)
+                            {
+                                x86_pause_hint();
+                                thread_sleep(0);
+                            }
+                        }
+
+                        filter_level = pbi->mt_baseline_filter_level[Segment];
 
                         // Apply any context driven MB level adjustment
                         vp8_adjust_mb_lf_value(mbd, &filter_level);
@@ -367,29 +383,28 @@ THREAD_FUNCTION vp8_thread_loop_filter(void *p_data)
                         v_ptr += 8;
 
                         mbd->mode_info_context++;     // step to next MB
-
+                        pbi->current_mb_col[mb_row] = mb_col;
                     }
 
-                    y_ptr += post->y_stride  * 16 - post->y_width;
-                    u_ptr += post->uv_stride *  8 - post->uv_width;
-                    v_ptr += post->uv_stride *  8 - post->uv_width;
-
                     mbd->mode_info_context++;         // Skip border mb
-                }
 
-                //printf("R%d\n", mb_row);
+                    y_ptr += post->y_stride  * 16 * (pbi->decoding_thread_count + 1) - post->y_width;
-                // When done, signal main thread that ME is finished
+                    u_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
-                //SetEvent(pbi->h_event_lpf);
+                    v_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
-                sem_post(&pbi->h_event_lpf);
-            }
 
+                    mbd->mode_info_context += pbi->decoding_thread_count * mbd->mode_info_stride;         // Skip border mb
+                }
             }
         }
 
+        //  add this to each frame
+        if ((mbrd->mb_row == pbi->common.mb_rows-1) || ((mbrd->mb_row == pbi->common.mb_rows-2) && (pbi->common.mb_rows % (pbi->decoding_thread_count+1))==1))
+        {
+          sem_post(&pbi->h_event_end_lpf);
+        }
 #else
-    (void) p_data;
+    (void) pbi;
 #endif
-    return 0;
 }
 
 void vp8_decoder_create_threads(VP8D_COMP *pbi)
@@ -401,39 +416,38 @@ void vp8_decoder_create_threads(VP8D_COMP *pbi)
     pbi->b_multithreaded_rd = 0;
     pbi->b_multithreaded_lf = 0;
     pbi->allocated_decoding_thread_count = 0;
-    core_count = (pbi->max_threads > 16) ? 16 : pbi->max_threads; //vp8_get_proc_core_count();
+    core_count = (pbi->max_threads > 16) ? 16 : pbi->max_threads;
-    if (core_count > 1)
-    {
-        sem_init(&pbi->h_event_lpf, 0, 0);
-        sem_init(&pbi->h_event_start_lpf, 0, 0);
-        pbi->b_multithreaded_lf = 1;
-        pthread_create(&pbi->h_thread_lpf, 0, vp8_thread_loop_filter, (pbi));
-    }
 
     if (core_count > 1)
     {
         pbi->b_multithreaded_rd = 1;
+        pbi->b_multithreaded_lf = 1;  // this can be merged with pbi->b_multithreaded_rd ?
         pbi->decoding_thread_count = core_count -1;
 
         CHECK_MEM_ERROR(pbi->h_decoding_thread, vpx_malloc(sizeof(pthread_t) * pbi->decoding_thread_count));
-        CHECK_MEM_ERROR(pbi->h_event_mbrdecoding, vpx_malloc(sizeof(sem_t) * pbi->decoding_thread_count));
+        CHECK_MEM_ERROR(pbi->h_event_start_decoding, vpx_malloc(sizeof(sem_t) * pbi->decoding_thread_count));
         CHECK_MEM_ERROR(pbi->mb_row_di, vpx_memalign(32, sizeof(MB_ROW_DEC) * pbi->decoding_thread_count));
         vpx_memset(pbi->mb_row_di, 0, sizeof(MB_ROW_DEC) * pbi->decoding_thread_count);
         CHECK_MEM_ERROR(pbi->de_thread_data, vpx_malloc(sizeof(DECODETHREAD_DATA) * pbi->decoding_thread_count));
 
+        CHECK_MEM_ERROR(pbi->current_mb_col, vpx_malloc(sizeof(int) * MAX_ROWS));  // pc->mb_rows));
+        CHECK_MEM_ERROR(pbi->h_event_start_lpf, vpx_malloc(sizeof(sem_t) * pbi->decoding_thread_count));
+
         for (ithread = 0; ithread < pbi->decoding_thread_count; ithread++)
         {
-            sem_init(&pbi->h_event_mbrdecoding[ithread], 0, 0);
+            sem_init(&pbi->h_event_start_decoding[ithread], 0, 0);
+            sem_init(&pbi->h_event_start_lpf[ithread], 0, 0);
 
             pbi->de_thread_data[ithread].ithread  = ithread;
             pbi->de_thread_data[ithread].ptr1     = (void *)pbi;
             pbi->de_thread_data[ithread].ptr2     = (void *) &pbi->mb_row_di[ithread];
 
             pthread_create(&pbi->h_decoding_thread[ithread], 0, vp8_thread_decoding_proc, (&pbi->de_thread_data[ithread]));
-
         }
 
-        sem_init(&pbi->h_event_main, 0, 0);
+        sem_init(&pbi->h_event_end_decoding, 0, 0);
+        sem_init(&pbi->h_event_end_lpf, 0, 0);
+
         pbi->allocated_decoding_thread_count = pbi->decoding_thread_count;
     }
 
@@ -448,39 +462,35 @@ void vp8_decoder_remove_threads(VP8D_COMP *pbi)
 
     if (pbi->b_multithreaded_lf)
     {
+        int i;
         pbi->b_multithreaded_lf = 0;
-        sem_post(&pbi->h_event_start_lpf);
+
-        pthread_join(pbi->h_thread_lpf, 0);
+        for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
-        sem_destroy(&pbi->h_event_start_lpf);
+            sem_destroy(&pbi->h_event_start_lpf[i]);
+
+        sem_destroy(&pbi->h_event_end_lpf);
     }
 
     //shutdown MB Decoding thread;
     if (pbi->b_multithreaded_rd)
-    {
-        pbi->b_multithreaded_rd = 0;
-        // allow all threads to exit
     {
         int i;
 
+        pbi->b_multithreaded_rd = 0;
+
+        // allow all threads to exit
         for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
         {
-
+            sem_post(&pbi->h_event_start_decoding[i]);
-                sem_post(&pbi->h_event_mbrdecoding[i]);
             pthread_join(pbi->h_decoding_thread[i], NULL);
         }
-        }
-        {
 
-            int i;
         for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
         {
-                sem_destroy(&pbi->h_event_mbrdecoding[i]);
+            sem_destroy(&pbi->h_event_start_decoding[i]);
         }
 
-
+        sem_destroy(&pbi->h_event_end_decoding);
-        }
-
-        sem_destroy(&pbi->h_event_main);
 
         if (pbi->h_decoding_thread)
         {
@@ -488,10 +498,16 @@ void vp8_decoder_remove_threads(VP8D_COMP *pbi)
             pbi->h_decoding_thread = NULL;
         }
 
-        if (pbi->h_event_mbrdecoding)
+        if (pbi->h_event_start_decoding)
         {
-            vpx_free(pbi->h_event_mbrdecoding);
+            vpx_free(pbi->h_event_start_decoding);
-            pbi->h_event_mbrdecoding = NULL;
+            pbi->h_event_start_decoding = NULL;
+        }
+
+        if (pbi->h_event_start_lpf)
+        {
+            vpx_free(pbi->h_event_start_lpf);
+            pbi->h_event_start_lpf = NULL;
         }
 
         if (pbi->mb_row_di)
@@ -505,8 +521,13 @@ void vp8_decoder_remove_threads(VP8D_COMP *pbi)
             vpx_free(pbi->de_thread_data);
             pbi->de_thread_data = NULL;
         }
-    }
 
+        if (pbi->current_mb_col)
+        {
+            vpx_free(pbi->current_mb_col);
+            pbi->current_mb_col = NULL ;
+        }
+    }
 #else
     (void) pbi;
 #endif
@@ -516,9 +537,12 @@ void vp8_decoder_remove_threads(VP8D_COMP *pbi)
 void vp8_start_lfthread(VP8D_COMP *pbi)
 {
 #if CONFIG_MULTITHREAD
+  /*
     memcpy(&pbi->lpfmb, &pbi->mb, sizeof(pbi->mb));
     pbi->last_mb_row_decoded = 0;
     sem_post(&pbi->h_event_start_lpf);
+    */
+    (void) pbi;
 #else
     (void) pbi;
 #endif
@@ -527,14 +551,17 @@ void vp8_start_lfthread(VP8D_COMP *pbi)
 void vp8_stop_lfthread(VP8D_COMP *pbi)
 {
 #if CONFIG_MULTITHREAD
+  /*
     struct vpx_usec_timer timer;
 
     vpx_usec_timer_start(&timer);
 
-    sem_wait(&pbi->h_event_lpf);
+    sem_wait(&pbi->h_event_end_lpf);
 
     vpx_usec_timer_mark(&timer);
     pbi->time_loop_filtering += vpx_usec_timer_elapsed(&timer);
+    */
+    (void) pbi;
 #else
     (void) pbi;
 #endif
@@ -550,49 +577,246 @@ void vp8_mtdecode_mb_rows(VP8D_COMP *pbi,
 
     int ibc = 0;
     int num_part = 1 << pbi->common.multi_token_partition;
+    int i;
+    volatile int *last_row_current_mb_col = NULL;
 
     vp8_setup_decoding_thread_data(pbi, xd, pbi->mb_row_di, pbi->decoding_thread_count);
 
+    for (i = 0; i < pbi->decoding_thread_count; i++)
+        sem_post(&pbi->h_event_start_decoding[i]);
+
     for (mb_row = 0; mb_row < pc->mb_rows; mb_row += (pbi->decoding_thread_count + 1))
     {
         int i;
-        pbi->current_mb_col_main = -1;
 
-        xd->current_bc = &pbi->mbc[ibc];
+        xd->current_bc = &pbi->mbc[mb_row%num_part];
-        ibc++ ;
 
-        if (ibc == num_part)
+        //vp8_decode_mb_row(pbi, pc, mb_row, xd);
-            ibc = 0;
-
-        for (i = 0; i < pbi->decoding_thread_count; i++)
         {
-            if ((mb_row + i + 1) >= pc->mb_rows)
+            int i;
-                break;
+            int recon_yoffset, recon_uvoffset;
+            int mb_col;
+            int ref_fb_idx = pc->lst_fb_idx;
+            int dst_fb_idx = pc->new_fb_idx;
+            int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
+            int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;
 
-            pbi->mb_row_di[i].mb_row = mb_row + i + 1;
+           // volatile int *last_row_current_mb_col = NULL;
-            pbi->mb_row_di[i].mbd.current_bc =  &pbi->mbc[ibc];
+            if (mb_row > 0)
-            ibc++;
+                last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];
 
-            if (ibc == num_part)
+            vpx_memset(&pc->left_context, 0, sizeof(pc->left_context));
-                ibc = 0;
+            recon_yoffset = mb_row * recon_y_stride * 16;
+            recon_uvoffset = mb_row * recon_uv_stride * 8;
+            // reset above block coeffs
 
-            pbi->mb_row_di[i].current_mb_col = -1;
+            xd->above_context = pc->above_context;
-            sem_post(&pbi->h_event_mbrdecoding[i]);
+            xd->up_available = (mb_row != 0);
-        }
 
-        vp8_decode_mb_row(pbi, pc, mb_row, xd);
+            xd->mb_to_top_edge = -((mb_row * 16)) << 3;
+            xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;
 
-        xd->mode_info_context += xd->mode_info_stride * pbi->decoding_thread_count;
+            for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
-
-        if (mb_row < pc->mb_rows - 1)
             {
-            sem_wait(&pbi->h_event_main);
+                if ( mb_row > 0 && (mb_col & 7) == 0){
+                    while (mb_col > (*last_row_current_mb_col - 8) && *last_row_current_mb_col != pc->mb_cols - 1)
+                    {
+                        x86_pause_hint();
+                        thread_sleep(0);
                     }
                 }
 
+                if (xd->mode_info_context->mbmi.mode == SPLITMV || xd->mode_info_context->mbmi.mode == B_PRED)
+                {
+                    for (i = 0; i < 16; i++)
+                    {
+                        BLOCKD *d = &xd->block[i];
+                        vpx_memcpy(&d->bmi, &xd->mode_info_context->bmi[i], sizeof(B_MODE_INFO));
+                    }
+                }
+
+                // Distance of Mb to the various image edges.
+                // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
+                xd->mb_to_left_edge = -((mb_col * 16) << 3);
+                xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
+
+                xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
+                xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
+                xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
+
+                xd->left_available = (mb_col != 0);
+
+                // Select the appropriate reference frame for this MB
+                if (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME)
+                    ref_fb_idx = pc->lst_fb_idx;
+                else if (xd->mode_info_context->mbmi.ref_frame == GOLDEN_FRAME)
+                    ref_fb_idx = pc->gld_fb_idx;
+                else
+                    ref_fb_idx = pc->alt_fb_idx;
+
+                xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
+                xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
+                xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
+
+                vp8_build_uvmvs(xd, pc->full_pixel);
+
+                vp8_decode_macroblock(pbi, xd);
+
+                recon_yoffset += 16;
+                recon_uvoffset += 8;
+
+                ++xd->mode_info_context;  /* next mb */
+
+                xd->above_context++;
+
+                //pbi->current_mb_col_main = mb_col;
+                pbi->current_mb_col[mb_row] = mb_col;
+            }
+
+            // adjust to the next row of mbs
+            vp8_extend_mb_row(
+                &pc->yv12_fb[dst_fb_idx],
+                xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
+            );
+
+            ++xd->mode_info_context;      /* skip prediction column */
+
             pbi->last_mb_row_decoded = mb_row;
+        }
+        xd->mode_info_context += xd->mode_info_stride * pbi->decoding_thread_count;
+    }
+
+    sem_wait(&pbi->h_event_end_decoding);   // add back for each frame
 #else
     (void) pbi;
     (void) xd;
 #endif
 }
+
+
+void vp8_mt_loop_filter_frame( VP8D_COMP *pbi)
+{
+#if CONFIG_MULTITHREAD
+    VP8_COMMON *cm  = &pbi->common;
+    MACROBLOCKD *mbd = &pbi->mb;
+    int default_filt_lvl = pbi->common.filter_level;
+
+    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
+    loop_filter_info *lfi = cm->lf_info;
+    int frame_type = cm->frame_type;
+
+    int mb_row;
+    int mb_col;
+
+    int filter_level;
+    int alt_flt_enabled = mbd->segmentation_enabled;
+
+    int i;
+    unsigned char *y_ptr, *u_ptr, *v_ptr;
+
+    volatile int *last_row_current_mb_col=NULL;
+
+    vp8_setup_loop_filter_thread_data(pbi, mbd, pbi->mb_row_di, pbi->decoding_thread_count);
+
+    mbd->mode_info_context = cm->mi;          // Point at base of Mb MODE_INFO list
+
+    // Note the baseline filter values for each segment
+    if (alt_flt_enabled)
+    {
+        for (i = 0; i < MAX_MB_SEGMENTS; i++)
+        {
+            // Abs value
+            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
+                pbi->mt_baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+            // Delta Value
+            else
+            {
+                pbi->mt_baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+                pbi->mt_baseline_filter_level[i] = (pbi->mt_baseline_filter_level[i] >= 0) ? ((pbi->mt_baseline_filter_level[i] <= MAX_LOOP_FILTER) ? pbi->mt_baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  // Clamp to valid range
+            }
+        }
+    }
+    else
+    {
+        for (i = 0; i < MAX_MB_SEGMENTS; i++)
+            pbi->mt_baseline_filter_level[i] = default_filt_lvl;
+    }
+
+    // Initialize the loop filter for this frame.
+    if ((cm->last_filter_type != cm->filter_type) || (cm->last_sharpness_level != cm->sharpness_level))
+        vp8_init_loop_filter(cm);
+    else if (frame_type != cm->last_frame_type)
+        vp8_frame_init_loop_filter(lfi, frame_type);
+
+    for (i = 0; i < pbi->decoding_thread_count; i++)
+        sem_post(&pbi->h_event_start_lpf[i]);
+        // sem_post(&pbi->h_event_start_lpf);
+
+    // Set up the buffer pointers
+    y_ptr = post->y_buffer;
+    u_ptr = post->u_buffer;
+    v_ptr = post->v_buffer;
+
+    // vp8_filter each macro block
+    for (mb_row = 0; mb_row < cm->mb_rows; mb_row+= (pbi->decoding_thread_count + 1))
+    {
+        if (mb_row > 0)
+            last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];
+
+        for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+        {
+            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+
+            if ( mb_row > 0 && (mb_col & 7) == 0){
+            // if ( mb_row > 0 ){
+                while (mb_col > (*last_row_current_mb_col-8) && *last_row_current_mb_col != cm->mb_cols - 1)
+                {
+                    x86_pause_hint();
+                    thread_sleep(0);
+                }
+            }
+
+            filter_level = pbi->mt_baseline_filter_level[Segment];
+
+            // Distance of Mb to the various image edges.
+            // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
+            // Apply any context driven MB level adjustment
+            vp8_adjust_mb_lf_value(mbd, &filter_level);
+
+            if (filter_level)
+            {
+                if (mb_col > 0)
+                    cm->lf_mbv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+
+                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                    cm->lf_bv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+
+                // don't apply across umv border
+                if (mb_row > 0)
+                    cm->lf_mbh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+
+                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                    cm->lf_bh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
+            }
+
+            y_ptr += 16;
+            u_ptr += 8;
+            v_ptr += 8;
+
+            mbd->mode_info_context++;     // step to next MB
+            pbi->current_mb_col[mb_row] = mb_col;
+        }
+        mbd->mode_info_context++;         // Skip border mb
+
+        //update for multi-thread
+        y_ptr += post->y_stride  * 16 * (pbi->decoding_thread_count + 1) - post->y_width;
+        u_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
+        v_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
+        mbd->mode_info_context += pbi->decoding_thread_count * mbd->mode_info_stride;
+    }
+
+    sem_wait(&pbi->h_event_end_lpf);
+#else
+    (void) pbi;
+#endif
+}

vp8/decoder/x86/dequantize_mmx.asm

@@ -50,12 +50,12 @@ sym(vp8_dequantize_b_impl_mmx):
     ret
 
 
-;void dequant_idct_mmx(short *input, short *dq, short *output, int pitch)
+;void dequant_idct_add_mmx(short *input, short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride)
-global sym(vp8_dequant_idct_mmx)
+global sym(vp8_dequant_idct_add_mmx)
-sym(vp8_dequant_idct_mmx):
+sym(vp8_dequant_idct_add_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 6
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -77,7 +77,8 @@ sym(vp8_dequant_idct_mmx):
         movq        mm3,    [rax+24]
         pmullw      mm3,    [rdx+24]
 
-        mov         rdx,    arg(2) ;output
+        mov         rdx,    arg(3) ;dest
+        mov         rsi,    arg(2) ;pred
         pxor        mm7,    mm7
 
 
@@ -88,7 +89,8 @@ sym(vp8_dequant_idct_mmx):
         movq        [rax+24],mm7
 
 
-        movsxd      rax,            dword ptr arg(3) ;pitch
+        movsxd      rax,            dword ptr arg(4) ;pitch
+        movsxd      rdi,            dword ptr arg(5) ;stride
 
         psubw       mm0,            mm2             ; b1= 0-2
         paddw       mm2,            mm2             ;
@@ -207,13 +209,34 @@ sym(vp8_dequant_idct_mmx):
         punpckldq   mm2,            mm4             ; 32 22 12 02
         punpckhdq   mm5,            mm4             ; 33 23 13 03
 
-        movq        [rdx],          mm0
+        pxor        mm7,            mm7
 
-        movq        [rdx+rax],      mm1
+        movd        mm4,            [rsi]
-        movq        [rdx+rax*2],    mm2
+        punpcklbw   mm4,            mm7
+        paddsw      mm0,            mm4
+        packuswb    mm0,            mm7
+        movd        [rdx],          mm0
 
-        add         rdx,            rax
+        movd        mm4,            [rsi+rax]
-        movq        [rdx+rax*2],    mm5
+        punpcklbw   mm4,            mm7
+        paddsw      mm1,            mm4
+        packuswb    mm1,            mm7
+        movd        [rdx+rdi],      mm1
+
+        movd        mm4,            [rsi+2*rax]
+        punpcklbw   mm4,            mm7
+        paddsw      mm2,            mm4
+        packuswb    mm2,            mm7
+        movd        [rdx+rdi*2],    mm2
+
+        add         rdx,            rdi
+        add         rsi,            rax
+
+        movd        mm4,            [rsi+2*rax]
+        punpcklbw   mm4,            mm7
+        paddsw      mm5,            mm4
+        packuswb    mm5,            mm7
+        movd        [rdx+rdi*2],    mm5
 
     ; begin epilog
     pop rdi
@@ -224,12 +247,12 @@ sym(vp8_dequant_idct_mmx):
     ret
 
 
-;void dequant_dc_idct_mmx(short *input, short *dq, short *output, int pitch, int Dc)
+;void dequant_dc_idct_add_mmx(short *input, short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride, int Dc)
-global sym(vp8_dequant_dc_idct_mmx)
+global sym(vp8_dequant_dc_idct_add_mmx)
-sym(vp8_dequant_dc_idct_mmx):
+sym(vp8_dequant_dc_idct_add_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
+    SHADOW_ARGS_TO_STACK 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -238,8 +261,6 @@ sym(vp8_dequant_dc_idct_mmx):
         mov         rax,    arg(0) ;input
         mov         rdx,    arg(1) ;dq
 
-        movsxd      rcx,    dword ptr arg(4) ;Dc
-
         movq        mm0,    [rax   ]
         pmullw      mm0,    [rdx]
 
@@ -252,7 +273,8 @@ sym(vp8_dequant_dc_idct_mmx):
         movq        mm3,    [rax+24]
         pmullw      mm3,    [rdx+24]
 
-        mov         rdx,    arg(2) ;output
+        mov         rdx,    arg(3) ;dest
+        mov         rsi,    arg(2) ;pred
         pxor        mm7,    mm7
 
 
@@ -262,8 +284,15 @@ sym(vp8_dequant_dc_idct_mmx):
         movq        [rax+16],mm7
         movq        [rax+24],mm7
 
-        pinsrw      mm0,    rcx,  0
+        ; move lower word of Dc to lower word of mm0
-        movsxd      rax,            dword ptr arg(3) ;pitch
+        psrlq       mm0,    16
+        movzx       rcx,    word ptr arg(6) ;Dc
+        psllq       mm0,    16
+        movd        mm7,    rcx
+        por         mm0,    mm7
+
+        movsxd      rax,            dword ptr arg(4) ;pitch
+        movsxd      rdi,            dword ptr arg(5) ;stride
 
         psubw       mm0,            mm2             ; b1= 0-2
         paddw       mm2,            mm2             ;
@@ -382,13 +411,34 @@ sym(vp8_dequant_dc_idct_mmx):
         punpckldq   mm2,            mm4             ; 32 22 12 02
         punpckhdq   mm5,            mm4             ; 33 23 13 03
 
-        movq        [rdx],          mm0
+        pxor        mm7,            mm7
 
-        movq        [rdx+rax],      mm1
+        movd        mm4,            [rsi]
-        movq        [rdx+rax*2],    mm2
+        punpcklbw   mm4,            mm7
+        paddsw      mm0,            mm4
+        packuswb    mm0,            mm7
+        movd        [rdx],          mm0
 
-        add         rdx,            rax
+        movd        mm4,            [rsi+rax]
-        movq        [rdx+rax*2],    mm5
+        punpcklbw   mm4,            mm7
+        paddsw      mm1,            mm4
+        packuswb    mm1,            mm7
+        movd        [rdx+rdi],      mm1
+
+        movd        mm4,            [rsi+2*rax]
+        punpcklbw   mm4,            mm7
+        paddsw      mm2,            mm4
+        packuswb    mm2,            mm7
+        movd        [rdx+rdi*2],    mm2
+
+        add         rdx,            rdi
+        add         rsi,            rax
+
+        movd        mm4,            [rsi+2*rax]
+        punpcklbw   mm4,            mm7
+        paddsw      mm5,            mm4
+        packuswb    mm5,            mm7
+        movd        [rdx+rdi*2],    mm5
 
     ; begin epilog
     pop rdi

vp8/decoder/x86/dequantize_x86.h

@@ -21,19 +21,48 @@
  */
 #if HAVE_MMX
 extern prototype_dequant_block(vp8_dequantize_b_mmx);
-extern prototype_dequant_idct(vp8_dequant_idct_mmx);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_mmx);
-extern prototype_dequant_idct_dc(vp8_dequant_dc_idct_mmx);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_mmx);
-
+extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_mmx);
+extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_mmx);
+extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_mmx);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_dequant_block
 #define vp8_dequant_block vp8_dequantize_b_mmx
 
-#undef  vp8_dequant_idct
+#undef  vp8_dequant_idct_add
-#define vp8_dequant_idct vp8_dequant_idct_mmx
+#define vp8_dequant_idct_add vp8_dequant_idct_add_mmx
 
-#undef  vp8_dequant_idct_dc
+#undef  vp8_dequant_dc_idct_add
-#define vp8_dequant_idct_dc vp8_dequant_dc_idct_mmx
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_mmx
+
+#undef vp8_dequant_dc_idct_add_y_block
+#define vp8_dequant_dc_idct_add_y_block vp8_dequant_dc_idct_add_y_block_mmx
+
+#undef vp8_dequant_idct_add_y_block
+#define vp8_dequant_idct_add_y_block vp8_dequant_idct_add_y_block_mmx
+
+#undef vp8_dequant_idct_add_uv_block
+#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_mmx
+
+#endif
+#endif
+
+#if HAVE_SSE2
+extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_sse2);
+extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_sse2);
+extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_sse2);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+#undef vp8_dequant_dc_idct_add_y_block
+#define vp8_dequant_dc_idct_add_y_block vp8_dequant_dc_idct_add_y_block_sse2
+
+#undef vp8_dequant_idct_add_y_block
+#define vp8_dequant_idct_add_y_block vp8_dequant_idct_add_y_block_sse2
+
+#undef vp8_dequant_idct_add_uv_block
+#define vp8_dequant_idct_add_uv_block vp8_dequant_idct_add_uv_block_sse2
 
 #endif
 #endif

vp8/decoder/x86/idct_blk_mmx.c

@@ -0,0 +1,151 @@
+/*
+ *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "idct.h"
+#include "dequantize.h"
+
+void vp8_dequant_dc_idct_add_y_block_mmx
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs, short *dc)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_dc_idct_add_mmx (q, dq, pre, dst, 16, stride, dc[0]);
+        else
+            vp8_dc_only_idct_add_mmx (dc[0], pre, dst, 16, stride);
+
+        if (eobs[1] > 1)
+            vp8_dequant_dc_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride, dc[1]);
+        else
+            vp8_dc_only_idct_add_mmx (dc[1], pre+4, dst+4, 16, stride);
+
+        if (eobs[2] > 1)
+            vp8_dequant_dc_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride, dc[2]);
+        else
+            vp8_dc_only_idct_add_mmx (dc[2], pre+8, dst+8, 16, stride);
+
+        if (eobs[3] > 1)
+            vp8_dequant_dc_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride, dc[3]);
+        else
+            vp8_dc_only_idct_add_mmx (dc[3], pre+12, dst+12, 16, stride);
+
+        q    += 64;
+        dc   += 4;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_y_block_mmx
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_mmx (q, dq, pre, dst, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dst, 16, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dst+4, 16, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        if (eobs[2] > 1)
+            vp8_dequant_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[32]*dq[0], pre+8, dst+8, 16, stride);
+            ((int *)(q+32))[0] = 0;
+        }
+
+        if (eobs[3] > 1)
+            vp8_dequant_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[48]*dq[0], pre+12, dst+12, 16, stride);
+            ((int *)(q+48))[0] = 0;
+        }
+
+        q    += 64;
+        pre  += 64;
+        dst  += 4*stride;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_mmx
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_mmx (q, dq, pre, dstu, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstu, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dstu+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstu+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstu += 4*stride;
+        eobs += 2;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        if (eobs[0] > 1)
+            vp8_dequant_idct_add_mmx (q, dq, pre, dstv, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstv, 8, stride);
+            ((int *)q)[0] = 0;
+        }
+
+        if (eobs[1] > 1)
+            vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dstv+4, 8, stride);
+        else
+        {
+            vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstv+4, 8, stride);
+            ((int *)(q+16))[0] = 0;
+        }
+
+        q    += 32;
+        pre  += 32;
+        dstv += 4*stride;
+        eobs += 2;
+    }
+}

vp8/decoder/x86/idct_blk_sse2.c

@@ -0,0 +1,114 @@
+/*
+ *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_ports/config.h"
+#include "idct.h"
+#include "dequantize.h"
+
+void idct_dequant_dc_0_2x_sse2
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int dst_stride, short *dc);
+void idct_dequant_dc_full_2x_sse2
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int dst_stride, short *dc);
+
+void idct_dequant_0_2x_sse2
+            (short *q, short *dq ,unsigned char *pre,
+             unsigned char *dst, int dst_stride, int blk_stride);
+void idct_dequant_full_2x_sse2
+            (short *q, short *dq ,unsigned char *pre,
+             unsigned char *dst, int dst_stride, int blk_stride);
+
+void vp8_dequant_dc_idct_add_y_block_sse2
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs, short *dc)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (((short *)(eobs))[0] & 0xfefe)
+            idct_dequant_dc_full_2x_sse2 (q, dq, pre, dst, stride, dc);
+        else
+            idct_dequant_dc_0_2x_sse2 (q, dq, pre, dst, stride, dc);
+
+        if (((short *)(eobs))[1] & 0xfefe)
+            idct_dequant_dc_full_2x_sse2 (q+32, dq, pre+8, dst+8, stride, dc+2);
+        else
+            idct_dequant_dc_0_2x_sse2 (q+32, dq, pre+8, dst+8, stride, dc+2);
+
+        q    += 64;
+        dc   += 4;
+        pre  += 64;
+        dst  += stride*4;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_y_block_sse2
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dst, int stride, char *eobs)
+{
+    int i;
+
+    for (i = 0; i < 4; i++)
+    {
+        if (((short *)(eobs))[0] & 0xfefe)
+            idct_dequant_full_2x_sse2 (q, dq, pre, dst, stride, 16);
+        else
+            idct_dequant_0_2x_sse2 (q, dq, pre, dst, stride, 16);
+
+        if (((short *)(eobs))[1] & 0xfefe)
+            idct_dequant_full_2x_sse2 (q+32, dq, pre+8, dst+8, stride, 16);
+        else
+            idct_dequant_0_2x_sse2 (q+32, dq, pre+8, dst+8, stride, 16);
+
+        q    += 64;
+        pre  += 64;
+        dst  += stride*4;
+        eobs += 4;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_sse2
+            (short *q, short *dq, unsigned char *pre,
+             unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    if (((short *)(eobs))[0] & 0xfefe)
+        idct_dequant_full_2x_sse2 (q, dq, pre, dstu, stride, 8);
+    else
+        idct_dequant_0_2x_sse2 (q, dq, pre, dstu, stride, 8);
+
+    q    += 32;
+    pre  += 32;
+    dstu += stride*4;
+
+    if (((short *)(eobs))[1] & 0xfefe)
+        idct_dequant_full_2x_sse2 (q, dq, pre, dstu, stride, 8);
+    else
+        idct_dequant_0_2x_sse2 (q, dq, pre, dstu, stride, 8);
+
+    q    += 32;
+    pre  += 32;
+
+    if (((short *)(eobs))[2] & 0xfefe)
+        idct_dequant_full_2x_sse2 (q, dq, pre, dstv, stride, 8);
+    else
+        idct_dequant_0_2x_sse2 (q, dq, pre, dstv, stride, 8);
+
+    q    += 32;
+    pre  += 32;
+    dstv += stride*4;
+
+    if (((short *)(eobs))[3] & 0xfefe)
+        idct_dequant_full_2x_sse2 (q, dq, pre, dstv, stride, 8);
+    else
+        idct_dequant_0_2x_sse2 (q, dq, pre, dstv, stride, 8);
+}

vp8/decoder/x86/x86_dsystemdependent.c

@@ -39,14 +39,24 @@ void vp8_arch_x86_decode_init(VP8D_COMP *pbi)
 #if CONFIG_RUNTIME_CPU_DETECT
     /* Override default functions with fastest ones for this CPU. */
 #if HAVE_MMX
-
     if (flags & HAS_MMX)
     {
         pbi->dequant.block               = vp8_dequantize_b_mmx;
-        pbi->dequant.idct    = vp8_dequant_idct_mmx;
+        pbi->dequant.idct_add            = vp8_dequant_idct_add_mmx;
-        pbi->dequant.idct_dc = vp8_dequant_dc_idct_mmx;
+        pbi->dequant.dc_idct_add         = vp8_dequant_dc_idct_add_mmx;
+        pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_mmx;
+        pbi->dequant.idct_add_y_block    = vp8_dequant_idct_add_y_block_mmx;
+        pbi->dequant.idct_add_uv_block   = vp8_dequant_idct_add_uv_block_mmx;
     }
+#endif
+#if HAVE_SSE2
+    if (flags & HAS_SSE2)
+    {
+        pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_sse2;
+        pbi->dequant.idct_add_y_block    = vp8_dequant_idct_add_y_block_sse2;
+        pbi->dequant.idct_add_uv_block   = vp8_dequant_idct_add_uv_block_sse2;
+    }
+#endif
 
-#endif
 #endif
 }

vp8/dixie/bool_decoder.h

@@ -1,144 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license and
- *  patent grant that can be found in the LICENSE file in the root of
- *  the source tree. All contributing project authors may be found in
- *  the AUTHORS file in the root of the source tree.
- */
-
-
-#ifndef BOOL_DECODER_H
-#define BOOL_DECODER_H
-#include <stddef.h>
-
-struct bool_decoder
-{
-    const unsigned char *input;      /* next compressed data byte */
-    size_t               input_len;  /* length of the input buffer */
-    unsigned int         range;      /* identical to encoder's range */
-    unsigned int         value;      /* contains at least 8 significant
-                                      * bits */
-    int                  bit_count;  /* # of bits shifted out of value,
-                                      * max 7 */
-};
-
-
-static void
-init_bool_decoder(struct bool_decoder *d,
-                  const unsigned char *start_partition,
-                  size_t               sz)
-{
-    if (sz >= 2)
-    {
-        d->value = (start_partition[0] << 8) /* first 2 input bytes */
-                   | start_partition[1];
-        d->input = start_partition + 2;      /* ptr to next byte */
-        d->input_len = sz - 2;
-    }
-    else
-    {
-        d->value = 0;
-        d->input = NULL;
-        d->input_len = 0;
-    }
-
-    d->range = 255;    /* initial range is full */
-    d->bit_count = 0;  /* have not yet shifted out any bits */
-}
-
-
-static int bool_get(struct bool_decoder *d, int probability)
-{
-    /* range and split are identical to the corresponding values
-       used by the encoder when this bool was written */
-
-    unsigned int  split = 1 + (((d->range - 1) * probability) >> 8);
-    unsigned int  SPLIT = split << 8;
-    int           retval;           /* will be 0 or 1 */
-
-    if (d->value >= SPLIT)    /* encoded a one */
-    {
-        retval = 1;
-        d->range -= split;  /* reduce range */
-        d->value -= SPLIT;  /* subtract off left endpoint of interval */
-    }
-    else                  /* encoded a zero */
-    {
-        retval = 0;
-        d->range = split; /* reduce range, no change in left endpoint */
-    }
-
-    while (d->range < 128)    /* shift out irrelevant value bits */
-    {
-        d->value <<= 1;
-        d->range <<= 1;
-
-        if (++d->bit_count == 8)    /* shift in new bits 8 at a time */
-        {
-            d->bit_count = 0;
-
-            if (d->input_len)
-            {
-                d->value |= *d->input++;
-                d->input_len--;
-            }
-        }
-    }
-
-    return retval;
-}
-
-
-static int bool_get_bit(struct bool_decoder *br)
-{
-    return bool_get(br, 128);
-}
-
-
-static int bool_get_uint(struct bool_decoder *br, int bits)
-{
-    int z = 0;
-    int bit;
-
-    for (bit = bits - 1; bit >= 0; bit--)
-    {
-        z |= (bool_get_bit(br) << bit);
-    }
-
-    return z;
-}
-
-
-static int bool_get_int(struct bool_decoder *br, int bits)
-{
-    int z = 0;
-    int bit;
-
-    for (bit = bits - 1; bit >= 0; bit--)
-    {
-        z |= (bool_get_bit(br) << bit);
-    }
-
-    return bool_get_bit(br) ? -z : z;
-}
-
-
-static int bool_maybe_get_int(struct bool_decoder *br, int bits)
-{
-    return bool_get_bit(br) ? bool_get_int(br, bits) : 0;
-}
-
-
-static int
-bool_read_tree(struct bool_decoder *bool,
-               const int           *t,
-               const unsigned char *p)
-{
-    int i = 0;
-
-    while ((i = t[ i + bool_get(bool, p[i>>1])]) > 0) ;
-
-    return -i;
-}
-#endif

vp8/dixie/dequant_data.h

@@ -1,38 +0,0 @@
-static const int dc_q_lookup[128] =
-{
-    4,    5,    6,    7,    8,    9,    10,   10,
-    11,   12,   13,   14,   15,   16,   17,   17,
-    18,   19,   20,   20,   21,   21,   22,   22,
-    23,   23,   24,   25,   25,   26,   27,   28,
-    29,   30,   31,   32,   33,   34,   35,   36,
-    37,   37,   38,   39,   40,   41,   42,   43,
-    44,   45,   46,   46,   47,   48,   49,   50,
-    51,   52,   53,   54,   55,   56,   57,   58,
-    59,   60,   61,   62,   63,   64,   65,   66,
-    67,   68,   69,   70,   71,   72,   73,   74,
-    75,   76,   76,   77,   78,   79,   80,   81,
-    82,   83,   84,   85,   86,   87,   88,   89,
-    91,   93,   95,   96,   98,   100,  101,  102,
-    104,  106,  108,  110,  112,  114,  116,  118,
-    122,  124,  126,  128,  130,  132,  134,  136,
-    138,  140,  143,  145,  148,  151,  154,  157
-};
-static const int ac_q_lookup[128] =
-{
-    4,    5,    6,    7,    8,    9,    10,   11,
-    12,   13,   14,   15,   16,   17,   18,   19,
-    20,   21,   22,   23,   24,   25,   26,   27,
-    28,   29,   30,   31,   32,   33,   34,   35,
-    36,   37,   38,   39,   40,   41,   42,   43,
-    44,   45,   46,   47,   48,   49,   50,   51,
-    52,   53,   54,   55,   56,   57,   58,   60,
-    62,   64,   66,   68,   70,   72,   74,   76,
-    78,   80,   82,   84,   86,   88,   90,   92,
-    94,   96,   98,   100,  102,  104,  106,  108,
-    110,  112,  114,  116,  119,  122,  125,  128,
-    131,  134,  137,  140,  143,  146,  149,  152,
-    155,  158,  161,  164,  167,  170,  173,  177,
-    181,  185,  189,  193,  197,  201,  205,  209,
-    213,  217,  221,  225,  229,  234,  239,  245,
-    249,  254,  259,  264,  269,  274,  279,  284
-};

vp8/dixie/dixie.c

@@ -1,560 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include "vpx/internal/vpx_codec_internal.h"
-#include "bit_ops.h"
-#include "dixie.h"
-#include "vp8_prob_data.h"
-#include "dequant_data.h"
-#include "modemv.h"
-#include "tokens.h"
-#include "predict.h"
-#include "dixie_loopfilter.h"
-#include <string.h>
-#include <assert.h>
-
-enum
-{
-    FRAME_HEADER_SZ = 3,
-    KEYFRAME_HEADER_SZ = 7
-};
-
-
-#define ARRAY_COPY(a,b) {\
-    assert(sizeof(a)==sizeof(b));memcpy(a,b,sizeof(a));}
-static void
-decode_entropy_header(struct vp8_decoder_ctx    *ctx,
-                      struct bool_decoder       *bool,
-                      struct vp8_entropy_hdr    *hdr)
-{
-    int i, j, k, l;
-
-    /* Read coefficient probability updates */
-    for (i = 0; i < BLOCK_TYPES; i++)
-        for (j = 0; j < COEF_BANDS; j++)
-            for (k = 0; k < PREV_COEF_CONTEXTS; k++)
-                for (l = 0; l < ENTROPY_NODES; l++)
-                    if (bool_get(bool,
-                                 k_coeff_entropy_update_probs
-                                     [i][j][k][l]))
-                        hdr->coeff_probs[i][j][k][l] =
-                            bool_get_uint(bool, 8);
-
-    /* Read coefficient skip mode probability */
-    hdr->coeff_skip_enabled = bool_get_bit(bool);
-
-    if (hdr->coeff_skip_enabled)
-        hdr->coeff_skip_prob = bool_get_uint(bool, 8);
-
-    /* Parse interframe probability updates */
-    if (!ctx->frame_hdr.is_keyframe)
-    {
-        hdr->prob_inter = bool_get_uint(bool, 8);
-        hdr->prob_last  = bool_get_uint(bool, 8);
-        hdr->prob_gf    = bool_get_uint(bool, 8);
-
-        if (bool_get_bit(bool))
-            for (i = 0; i < 4; i++)
-                hdr->y_mode_probs[i] = bool_get_uint(bool, 8);
-
-        if (bool_get_bit(bool))
-            for (i = 0; i < 3; i++)
-                hdr->uv_mode_probs[i] = bool_get_uint(bool, 8);
-
-        for (i = 0; i < 2; i++)
-            for (j = 0; j < MV_PROB_CNT; j++)
-                if (bool_get(bool, k_mv_entropy_update_probs[i][j]))
-                {
-                    int x = bool_get_uint(bool, 7);
-                    hdr->mv_probs[i][j] = x ? x << 1 : 1;
-                }
-    }
-}
-
-
-static void
-decode_reference_header(struct vp8_decoder_ctx    *ctx,
-                        struct bool_decoder       *bool,
-                        struct vp8_reference_hdr  *hdr)
-{
-    unsigned int key = ctx->frame_hdr.is_keyframe;
-
-    hdr->refresh_gf    = key ? 1 : bool_get_bit(bool);
-    hdr->refresh_arf   = key ? 1 : bool_get_bit(bool);
-    hdr->copy_gf       = key ? 0 : !hdr->refresh_gf
-                         ? bool_get_uint(bool, 2) : 0;
-    hdr->copy_arf      = key ? 0 : !hdr->refresh_arf
-                         ? bool_get_uint(bool, 2) : 0;
-    hdr->sign_bias[GOLDEN_FRAME] = key ? 0 : bool_get_bit(bool);
-    hdr->sign_bias[ALTREF_FRAME] = key ? 0 : bool_get_bit(bool);
-    hdr->refresh_entropy = bool_get_bit(bool);
-    hdr->refresh_last  = key ? 1 : bool_get_bit(bool);
-}
-
-
-static void
-decode_quantizer_header(struct vp8_decoder_ctx    *ctx,
-                        struct bool_decoder       *bool,
-                        struct vp8_quant_hdr      *hdr)
-{
-    int update;
-    int last_q = hdr->q_index;
-
-    hdr->q_index = bool_get_uint(bool, 7);
-    update = last_q != hdr->q_index;
-    update |= (hdr->y1_dc_delta_q = bool_maybe_get_int(bool, 4));
-    update |= (hdr->y2_dc_delta_q = bool_maybe_get_int(bool, 4));
-    update |= (hdr->y2_ac_delta_q = bool_maybe_get_int(bool, 4));
-    update |= (hdr->uv_dc_delta_q = bool_maybe_get_int(bool, 4));
-    update |= (hdr->uv_ac_delta_q = bool_maybe_get_int(bool, 4));
-    hdr->delta_update = update;
-}
-
-
-static void
-decode_and_init_token_partitions(struct vp8_decoder_ctx    *ctx,
-                                 struct bool_decoder       *bool,
-                                 const unsigned char       *data,
-                                 unsigned int               sz,
-                                 struct vp8_token_hdr      *hdr)
-{
-    int i;
-
-    hdr->partitions = 1 << bool_get_uint(bool, 2);
-
-    if (sz < 3 *(hdr->partitions - 1))
-        vpx_internal_error(&ctx->error, VPX_CODEC_CORRUPT_FRAME,
-                           "Truncated packet found parsing partition"
-                           " lengths.");
-
-    sz -= 3 * (hdr->partitions - 1);
-
-    for (i = 0; i < hdr->partitions; i++)
-    {
-        if (i < hdr->partitions - 1)
-        {
-            hdr->partition_sz[i] = (data[2] << 16)
-                                   | (data[1] << 8) | data[0];
-            data += 3;
-        }
-        else
-            hdr->partition_sz[i] = sz;
-
-        if (sz < hdr->partition_sz[i])
-            vpx_internal_error(&ctx->error, VPX_CODEC_CORRUPT_FRAME,
-                               "Truncated partition %d", i);
-
-        sz -= hdr->partition_sz[i];
-    }
-
-
-    for (i = 0; i < ctx->token_hdr.partitions; i++)
-    {
-        init_bool_decoder(&ctx->tokens[i].bool, data,
-                          ctx->token_hdr.partition_sz[i]);
-        data += ctx->token_hdr.partition_sz[i];
-    }
-}
-
-
-static void
-decode_loopfilter_header(struct vp8_decoder_ctx    *ctx,
-                         struct bool_decoder       *bool,
-                         struct vp8_loopfilter_hdr *hdr)
-{
-    if (ctx->frame_hdr.is_keyframe)
-        memset(hdr, 0, sizeof(*hdr));
-
-    hdr->use_simple    = bool_get_bit(bool);
-    hdr->level         = bool_get_uint(bool, 6);
-    hdr->sharpness     = bool_get_uint(bool, 3);
-    hdr->delta_enabled = bool_get_bit(bool);
-
-    if (hdr->delta_enabled && bool_get_bit(bool))
-    {
-        int i;
-
-        for (i = 0; i < BLOCK_CONTEXTS; i++)
-            hdr->ref_delta[i] = bool_maybe_get_int(bool, 6);
-
-        for (i = 0; i < BLOCK_CONTEXTS; i++)
-            hdr->mode_delta[i] = bool_maybe_get_int(bool, 6);
-    }
-}
-
-
-static void
-decode_segmentation_header(struct vp8_decoder_ctx *ctx,
-                           struct bool_decoder    *bool,
-                           struct vp8_segment_hdr *hdr)
-{
-    if (ctx->frame_hdr.is_keyframe)
-        memset(hdr, 0, sizeof(*hdr));
-
-    hdr->enabled = bool_get_bit(bool);
-
-    if (hdr->enabled)
-    {
-        int i;
-
-        hdr->update_map = bool_get_bit(bool);
-        hdr->update_data = bool_get_bit(bool);
-
-        if (hdr->update_data)
-        {
-            hdr->abs = bool_get_bit(bool);
-
-            for (i = 0; i < MAX_MB_SEGMENTS; i++)
-                hdr->quant_idx[i] = bool_maybe_get_int(bool, 7);
-
-            for (i = 0; i < MAX_MB_SEGMENTS; i++)
-                hdr->lf_level[i] = bool_maybe_get_int(bool, 6);
-        }
-
-        if (hdr->update_map)
-        {
-            for (i = 0; i < MB_FEATURE_TREE_PROBS; i++)
-                hdr->tree_probs[i] = bool_get_bit(bool)
-                                     ? bool_get_uint(bool, 8)
-                                     : 255;
-        }
-    }
-    else
-    {
-        hdr->update_map = 0;
-        hdr->update_data = 0;
-    }
-}
-
-
-static void
-dequant_global_init(struct dequant_factors dqf[MAX_MB_SEGMENTS])
-{
-    int i;
-
-    for (i = 0; i < MAX_MB_SEGMENTS; i++)
-        dqf[i].quant_idx = -1;
-}
-
-
-static int
-clamp_q(int q)
-{
-    if (q < 0) return 0;
-    else if (q > 127) return 127;
-
-    return q;
-}
-
-
-static int
-dc_q(int q)
-{
-    return dc_q_lookup[clamp_q(q)];
-}
-
-
-static int
-ac_q(int q)
-{
-    return ac_q_lookup[clamp_q(q)];
-}
-
-
-static void
-dequant_init(struct dequant_factors        factors[MAX_MB_SEGMENTS],
-             const struct vp8_segment_hdr *seg,
-             const struct vp8_quant_hdr   *quant_hdr)
-{
-    int i, q;
-    struct dequant_factors *dqf = factors;
-
-    for (i = 0; i < (seg->enabled ? MAX_MB_SEGMENTS : 1); i++)
-    {
-        q = quant_hdr->q_index;
-
-        if (seg->enabled)
-            q = (!seg->abs) ? q + seg->quant_idx[i] : seg->quant_idx[i];
-
-        if (dqf->quant_idx != q || quant_hdr->delta_update)
-        {
-            dqf->factor[TOKEN_BLOCK_Y1][0] =
-                dc_q(q + quant_hdr->y1_dc_delta_q);
-            dqf->factor[TOKEN_BLOCK_Y1][1] =
-                ac_q(q);
-            dqf->factor[TOKEN_BLOCK_UV][0] =
-                dc_q(q + quant_hdr->uv_dc_delta_q);
-            dqf->factor[TOKEN_BLOCK_UV][1] =
-                ac_q(q + quant_hdr->uv_ac_delta_q);
-            dqf->factor[TOKEN_BLOCK_Y2][0] =
-                dc_q(q + quant_hdr->y2_dc_delta_q) * 2;
-            dqf->factor[TOKEN_BLOCK_Y2][1] =
-                ac_q(q + quant_hdr->y2_ac_delta_q) * 155 / 100;
-
-            if (dqf->factor[TOKEN_BLOCK_Y2][1] < 8)
-                dqf->factor[TOKEN_BLOCK_Y2][1] = 8;
-
-            if (dqf->factor[TOKEN_BLOCK_UV][0] > 132)
-                dqf->factor[TOKEN_BLOCK_UV][0] = 132;
-
-            dqf->quant_idx = q;
-        }
-
-        dqf++;
-    }
-}
-
-
-static void
-decode_frame(struct vp8_decoder_ctx *ctx,
-             const unsigned char    *data,
-             unsigned int            sz)
-{
-    vpx_codec_err_t  res;
-    struct bool_decoder  bool;
-    int                  i, row, partition;
-
-    ctx->saved_entropy_valid = 0;
-
-    if ((res = vp8_parse_frame_header(data, sz, &ctx->frame_hdr)))
-        vpx_internal_error(&ctx->error, res,
-                           "Failed to parse frame header");
-
-    if (ctx->frame_hdr.is_experimental)
-        vpx_internal_error(&ctx->error, VPX_CODEC_UNSUP_BITSTREAM,
-                           "Experimental bitstreams not supported.");
-
-    data += FRAME_HEADER_SZ;
-    sz -= FRAME_HEADER_SZ;
-
-    if (ctx->frame_hdr.is_keyframe)
-    {
-        data += KEYFRAME_HEADER_SZ;
-        sz -= KEYFRAME_HEADER_SZ;
-        ctx->mb_cols = (ctx->frame_hdr.kf.w + 15) / 16;
-        ctx->mb_rows = (ctx->frame_hdr.kf.h + 15) / 16;
-    }
-
-    /* Start the bitreader for the header/entropy partition */
-    init_bool_decoder(&bool, data, ctx->frame_hdr.part0_sz);
-
-    /* Skip the colorspace and clamping bits */
-    if (ctx->frame_hdr.is_keyframe)
-        if (bool_get_uint(&bool, 2))
-            vpx_internal_error(&ctx->error, VPX_CODEC_UNSUP_BITSTREAM,
-                               "Reserved bits not supported.");
-
-    decode_segmentation_header(ctx, &bool, &ctx->segment_hdr);
-    decode_loopfilter_header(ctx, &bool, &ctx->loopfilter_hdr);
-    decode_and_init_token_partitions(ctx,
-                                     &bool,
-                                     data + ctx->frame_hdr.part0_sz,
-                                     sz - ctx->frame_hdr.part0_sz,
-                                     &ctx->token_hdr);
-    decode_quantizer_header(ctx, &bool, &ctx->quant_hdr);
-    decode_reference_header(ctx, &bool, &ctx->reference_hdr);
-
-    /* Set keyframe entropy defaults. These get updated on keyframes
-     * regardless of the refresh_entropy setting.
-     */
-    if (ctx->frame_hdr.is_keyframe)
-    {
-        ARRAY_COPY(ctx->entropy_hdr.coeff_probs,
-                   k_default_coeff_probs);
-        ARRAY_COPY(ctx->entropy_hdr.mv_probs,
-                   k_default_mv_probs);
-        ARRAY_COPY(ctx->entropy_hdr.y_mode_probs,
-                   k_default_y_mode_probs);
-        ARRAY_COPY(ctx->entropy_hdr.uv_mode_probs,
-                   k_default_uv_mode_probs);
-    }
-
-    if (!ctx->reference_hdr.refresh_entropy)
-    {
-        ctx->saved_entropy = ctx->entropy_hdr;
-        ctx->saved_entropy_valid = 1;
-    }
-
-    decode_entropy_header(ctx, &bool, &ctx->entropy_hdr);
-
-    vp8_dixie_modemv_init(ctx);
-    vp8_dixie_tokens_init(ctx);
-    vp8_dixie_predict_init(ctx);
-    dequant_init(ctx->dequant_factors, &ctx->segment_hdr,
-                 &ctx->quant_hdr);
-
-    for (row = 0, partition = 0; row < ctx->mb_rows; row++)
-    {
-        vp8_dixie_modemv_process_row(ctx, &bool, row, 0, ctx->mb_cols);
-        vp8_dixie_tokens_process_row(ctx, partition, row, 0,
-                                     ctx->mb_cols);
-        vp8_dixie_predict_process_row(ctx, row, 0, ctx->mb_cols);
-
-        if (ctx->loopfilter_hdr.level && row)
-            vp8_dixie_loopfilter_process_row(ctx, row - 1, 0,
-                                             ctx->mb_cols);
-
-        if (++partition == ctx->token_hdr.partitions)
-            partition = 0;
-    }
-
-    if (ctx->loopfilter_hdr.level)
-        vp8_dixie_loopfilter_process_row(ctx, row - 1, 0, ctx->mb_cols);
-
-    ctx->frame_cnt++;
-
-    if (!ctx->reference_hdr.refresh_entropy)
-    {
-        ctx->entropy_hdr = ctx->saved_entropy;
-        ctx->saved_entropy_valid = 0;
-    }
-
-    /* Handle reference frame updates */
-    if (ctx->reference_hdr.copy_arf == 1)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[ALTREF_FRAME]);
-        ctx->ref_frames[ALTREF_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[LAST_FRAME]);
-    }
-    else if (ctx->reference_hdr.copy_arf == 2)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[ALTREF_FRAME]);
-        ctx->ref_frames[ALTREF_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[GOLDEN_FRAME]);
-    }
-
-    if (ctx->reference_hdr.copy_gf == 1)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[GOLDEN_FRAME]);
-        ctx->ref_frames[GOLDEN_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[LAST_FRAME]);
-    }
-    else if (ctx->reference_hdr.copy_gf == 2)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[GOLDEN_FRAME]);
-        ctx->ref_frames[GOLDEN_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[ALTREF_FRAME]);
-    }
-
-    if (ctx->reference_hdr.refresh_gf)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[GOLDEN_FRAME]);
-        ctx->ref_frames[GOLDEN_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[CURRENT_FRAME]);
-    }
-
-    if (ctx->reference_hdr.refresh_arf)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[ALTREF_FRAME]);
-        ctx->ref_frames[ALTREF_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[CURRENT_FRAME]);
-    }
-
-    if (ctx->reference_hdr.refresh_last)
-    {
-        vp8_dixie_release_ref_frame(ctx->ref_frames[LAST_FRAME]);
-        ctx->ref_frames[LAST_FRAME] =
-            vp8_dixie_ref_frame(ctx->ref_frames[CURRENT_FRAME]);
-    }
-
-}
-
-
-void
-vp8_dixie_decode_init(struct vp8_decoder_ctx *ctx)
-{
-    dequant_global_init(ctx->dequant_factors);
-}
-
-
-#define CHECK_FOR_UPDATE(lval,rval,update_flag) do {\
-        unsigned int old = lval; \
-        update_flag |= (old != (lval = rval)); \
-    } while(0)
-
-vpx_codec_err_t
-vp8_parse_frame_header(const unsigned char   *data,
-                       unsigned int           sz,
-                       struct vp8_frame_hdr  *hdr)
-{
-    unsigned long raw;
-
-    if (sz < 10)
-        return VPX_CODEC_CORRUPT_FRAME;
-
-    /* The frame header is defined as a three byte little endian
-     * value
-     */
-    raw = data[0] | (data[1] << 8) | (data[2] << 16);
-    hdr->is_keyframe     = !BITS_GET(raw, 0, 1);
-    hdr->version         = BITS_GET(raw, 1, 2);
-    hdr->is_experimental = BITS_GET(raw, 3, 1);
-    hdr->is_shown        = BITS_GET(raw, 4, 1);
-    hdr->part0_sz        = BITS_GET(raw, 5, 19);
-
-    if (sz <= hdr->part0_sz + (hdr->is_keyframe ? 10 : 3))
-        return VPX_CODEC_CORRUPT_FRAME;
-
-    hdr->frame_size_updated = 0;
-
-    if (hdr->is_keyframe)
-    {
-        unsigned int update = 0;
-
-        /* Keyframe header consists of a three byte sync code followed
-         * by the width and height and associated scaling factors.
-         */
-        if (data[3] != 0x9d || data[4] != 0x01 || data[5] != 0x2a)
-            return VPX_CODEC_UNSUP_BITSTREAM;
-
-        raw = data[6] | (data[7] << 8)
-              | (data[8] << 16) | (data[9] << 24);
-        CHECK_FOR_UPDATE(hdr->kf.w,       BITS_GET(raw,  0, 14),
-                         update);
-        CHECK_FOR_UPDATE(hdr->kf.scale_w, BITS_GET(raw, 14,  2),
-                         update);
-        CHECK_FOR_UPDATE(hdr->kf.h,       BITS_GET(raw, 16, 14),
-                         update);
-        CHECK_FOR_UPDATE(hdr->kf.scale_h, BITS_GET(raw, 30,  2),
-                         update);
-
-        hdr->frame_size_updated = update;
-
-        if (!hdr->kf.w || !hdr->kf.h)
-            return VPX_CODEC_UNSUP_BITSTREAM;
-    }
-
-    return VPX_CODEC_OK;
-}
-
-
-vpx_codec_err_t
-vp8_dixie_decode_frame(struct vp8_decoder_ctx *ctx,
-                       const unsigned char    *data,
-                       unsigned int            sz)
-{
-    volatile struct vp8_decoder_ctx *ctx_ = ctx;
-
-    ctx->error.error_code = VPX_CODEC_OK;
-    ctx->error.has_detail = 0;
-
-    if (!setjmp(ctx->error.jmp))
-        decode_frame(ctx, data, sz);
-
-    return ctx_->error.error_code;
-}
-
-
-void
-vp8_dixie_decode_destroy(struct vp8_decoder_ctx *ctx)
-{
-    vp8_dixie_predict_destroy(ctx);
-    vp8_dixie_tokens_destroy(ctx);
-    vp8_dixie_modemv_destroy(ctx);
-}

vp8/dixie/dixie.h

@@ -1,308 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef DIXIE_H
-#define DIXIE_H
-#include "vpx/internal/vpx_codec_internal.h"
-#include "bool_decoder.h"
-
-struct vp8_frame_hdr
-{
-    unsigned int is_keyframe;      /* Frame is a keyframe */
-    unsigned int is_experimental;  /* Frame is a keyframe */
-    unsigned int version;          /* Bitstream version */
-    unsigned int is_shown;         /* Frame is to be displayed. */
-    unsigned int part0_sz;         /* Partition 0 length, in bytes */
-
-    struct vp8_kf_hdr
-    {
-        unsigned int w;        /* Width */
-        unsigned int h;        /* Height */
-        unsigned int scale_w;  /* Scaling factor, Width */
-        unsigned int scale_h;  /* Scaling factor, Height */
-    } kf;
-
-    unsigned int frame_size_updated; /* Flag to indicate a resolution
-                                      * update.
-                                      */
-};
-
-
-enum
-{
-    MB_FEATURE_TREE_PROBS = 3,
-    MAX_MB_SEGMENTS = 4
-};
-
-
-struct vp8_segment_hdr
-{
-    unsigned int         enabled;
-    unsigned int         update_data;
-    unsigned int         update_map;
-    unsigned int         abs;    /* 0=deltas, 1=absolute values */
-    unsigned int         tree_probs[MB_FEATURE_TREE_PROBS];
-    int                  lf_level[MAX_MB_SEGMENTS];
-    int                  quant_idx[MAX_MB_SEGMENTS];
-};
-
-
-enum
-{
-    BLOCK_CONTEXTS = 4
-};
-
-
-struct vp8_loopfilter_hdr
-{
-    unsigned int         use_simple;
-    unsigned int         level;
-    unsigned int         sharpness;
-    unsigned int         delta_enabled;
-    int                  ref_delta[BLOCK_CONTEXTS];
-    int                  mode_delta[BLOCK_CONTEXTS];
-};
-
-
-enum
-{
-    MAX_PARTITIONS = 8
-};
-
-struct vp8_token_hdr
-{
-    unsigned int        partitions;
-    unsigned int        partition_sz[MAX_PARTITIONS];
-};
-
-
-struct vp8_quant_hdr
-{
-    unsigned int       q_index;
-    int                delta_update;
-    int                y1_dc_delta_q;
-    int                y2_dc_delta_q;
-    int                y2_ac_delta_q;
-    int                uv_dc_delta_q;
-    int                uv_ac_delta_q;
-};
-
-
-struct vp8_reference_hdr
-{
-    unsigned int refresh_last;
-    unsigned int refresh_gf;
-    unsigned int refresh_arf;
-    unsigned int copy_gf;
-    unsigned int copy_arf;
-    unsigned int sign_bias[4];
-    unsigned int refresh_entropy;
-};
-
-
-enum
-{
-    BLOCK_TYPES        = 4,
-    PREV_COEF_CONTEXTS = 3,
-    COEF_BANDS         = 8,
-    ENTROPY_NODES      = 11,
-};
-typedef unsigned char coeff_probs_table_t[BLOCK_TYPES][COEF_BANDS]
-[PREV_COEF_CONTEXTS]
-[ENTROPY_NODES];
-
-
-enum
-{
-    MV_PROB_CNT = 2 + 8 - 1 + 10 /* from entropymv.h */
-};
-typedef unsigned char mv_component_probs_t[MV_PROB_CNT];
-
-
-struct vp8_entropy_hdr
-{
-    coeff_probs_table_t   coeff_probs;
-    mv_component_probs_t  mv_probs[2];
-    unsigned int          coeff_skip_enabled;
-    unsigned char         coeff_skip_prob;
-    unsigned char         y_mode_probs[4];
-    unsigned char         uv_mode_probs[3];
-    unsigned char         prob_inter;
-    unsigned char         prob_last;
-    unsigned char         prob_gf;
-};
-
-
-enum reference_frame
-{
-    CURRENT_FRAME,
-    LAST_FRAME,
-    GOLDEN_FRAME,
-    ALTREF_FRAME,
-    NUM_REF_FRAMES
-};
-
-
-enum prediction_mode
-{
-    /* 16x16 intra modes */
-    DC_PRED, V_PRED, H_PRED, TM_PRED, B_PRED,
-
-    /* 16x16 inter modes */
-    NEARESTMV, NEARMV, ZEROMV, NEWMV, SPLITMV,
-
-    MB_MODE_COUNT,
-
-    /* 4x4 intra modes */
-    B_DC_PRED = 0, B_TM_PRED, B_VE_PRED, B_HE_PRED, B_LD_PRED,
-    B_RD_PRED, B_VR_PRED, B_VL_PRED, B_HD_PRED, B_HU_PRED,
-
-    /* 4x4 inter modes */
-    LEFT4X4, ABOVE4X4, ZERO4X4, NEW4X4,
-
-    B_MODE_COUNT
-};
-
-
-enum splitmv_partitioning
-{
-    SPLITMV_16X8,
-    SPLITMV_8X16,
-    SPLITMV_8X8,
-    SPLITMV_4X4
-};
-
-
-typedef short filter_t[6];
-
-
-typedef union mv
-{
-    struct
-    {
-        int16_t x, y;
-    }  d;
-    uint32_t               raw;
-} mv_t;
-
-
-struct mb_base_info
-{
-    unsigned char y_mode     : 4;
-    unsigned char uv_mode    : 4;
-    unsigned char segment_id : 2;
-    unsigned char ref_frame  : 2;
-    unsigned char skip_coeff : 1;
-    unsigned char need_mc_border : 1;
-    enum splitmv_partitioning  partitioning : 2;
-    union mv      mv;
-    unsigned int  eob_mask;
-};
-
-
-struct mb_info
-{
-    struct mb_base_info base;
-    union
-    {
-        union mv              mvs[16];
-        enum prediction_mode  modes[16];
-    } split;
-};
-
-
-/* A "token entropy context" has 4 Y values, 2 U, 2 V, and 1 Y2 */
-typedef int token_entropy_ctx_t[4 + 2 + 2 + 1];
-
-struct token_decoder
-{
-    struct bool_decoder  bool;
-    token_entropy_ctx_t  left_token_entropy_ctx;
-    short               *coeffs;
-};
-
-enum token_block_type
-{
-    TOKEN_BLOCK_Y1,
-    TOKEN_BLOCK_UV,
-    TOKEN_BLOCK_Y2,
-    TOKEN_BLOCK_TYPES,
-};
-
-struct dequant_factors
-{
-    int   quant_idx;
-    short factor[TOKEN_BLOCK_TYPES][2]; /* [ Y1, UV, Y2 ] [ DC, AC ] */
-};
-
-
-struct ref_cnt_img
-{
-    vpx_image_t  img;
-    unsigned int ref_cnt;
-};
-
-
-struct vp8_decoder_ctx
-{
-    struct vpx_internal_error_info  error;
-    unsigned int                    frame_cnt;
-
-    struct vp8_frame_hdr            frame_hdr;
-    struct vp8_segment_hdr          segment_hdr;
-    struct vp8_loopfilter_hdr       loopfilter_hdr;
-    struct vp8_token_hdr            token_hdr;
-    struct vp8_quant_hdr            quant_hdr;
-    struct vp8_reference_hdr        reference_hdr;
-    struct vp8_entropy_hdr          entropy_hdr;
-
-    struct vp8_entropy_hdr          saved_entropy;
-    unsigned int                    saved_entropy_valid;
-
-    unsigned int                    mb_rows;
-    unsigned int                    mb_cols;
-    struct mb_info                 *mb_info_storage;
-    struct mb_info                **mb_info_rows_storage;
-    struct mb_info                **mb_info_rows;
-
-    token_entropy_ctx_t            *above_token_entropy_ctx;
-    struct token_decoder            tokens[MAX_PARTITIONS];
-    struct dequant_factors          dequant_factors[MAX_MB_SEGMENTS];
-
-    struct ref_cnt_img              frame_strg[NUM_REF_FRAMES];
-    struct ref_cnt_img             *ref_frames[NUM_REF_FRAMES];
-    ptrdiff_t                       ref_frame_offsets[4];
-
-    const filter_t                 *subpixel_filters;
-};
-
-
-void
-vp8_dixie_decode_init(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_decode_destroy(struct vp8_decoder_ctx *ctx);
-
-
-vpx_codec_err_t
-vp8_parse_frame_header(const unsigned char   *data,
-                       unsigned int           sz,
-                       struct vp8_frame_hdr  *hdr);
-
-
-vpx_codec_err_t
-vp8_dixie_decode_frame(struct vp8_decoder_ctx *ctx,
-                       const unsigned char    *data,
-                       unsigned int            sz);
-
-
-#define CLAMP_255(x) ((x)<0?0:((x)>255?255:(x)))
-
-#endif

vp8/dixie/dixie_loopfilter.c

@@ -1,530 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include "dixie.h"
-#include "dixie_loopfilter.h"
-
-#define ABS(x) ((x) >= 0 ? (x) : -(x))
-
-#define p3 pixels[-4*stride]
-#define p2 pixels[-3*stride]
-#define p1 pixels[-2*stride]
-#define p0 pixels[-1*stride]
-#define q0 pixels[ 0*stride]
-#define q1 pixels[ 1*stride]
-#define q2 pixels[ 2*stride]
-#define q3 pixels[ 3*stride]
-
-#define static
-static int
-saturate_int8(int x)
-{
-    if (x < -128)
-        return -128;
-
-    if (x > 127)
-        return 127;
-
-    return x;
-}
-
-
-static int
-saturate_uint8(int x)
-{
-    if (x < 0)
-        return 0;
-
-    if (x > 255)
-        return 255;
-
-    return x;
-}
-
-
-static int
-high_edge_variance(unsigned char *pixels,
-                   int            stride,
-                   int            hev_threshold)
-{
-    return ABS(p1 - p0) > hev_threshold || ABS(q1 - q0) > hev_threshold;
-}
-
-
-static int
-simple_threshold(unsigned char *pixels,
-                 int            stride,
-                 int            filter_limit)
-{
-    return (ABS(p0 - q0) * 2 + (ABS(p1 - q1) >> 1)) <= filter_limit;
-}
-
-
-static int
-normal_threshold(unsigned char *pixels,
-                 int            stride,
-                 int            edge_limit,
-                 int            interior_limit)
-{
-    int E = edge_limit;
-    int I = interior_limit;
-
-    return simple_threshold(pixels, stride, 2 * E + I)
-           && ABS(p3 - p2) <= I && ABS(p2 - p1) <= I
-           && ABS(p1 - p0) <= I && ABS(q3 - q2) <= I
-           && ABS(q2 - q1) <= I && ABS(q1 - q0) <= I;
-}
-
-
-static void
-filter_common(unsigned char *pixels,
-              int            stride,
-              int            use_outer_taps)
-{
-    int a, f1, f2;
-
-    a = 3 * (q0 - p0);
-
-    if (use_outer_taps)
-        a += saturate_int8(p1 - q1);
-
-    a = saturate_int8(a);
-
-    f1 = ((a + 4 > 127) ? 127 : a + 4) >> 3;
-    f2 = ((a + 3 > 127) ? 127 : a + 3) >> 3;
-
-    p0 = saturate_uint8(p0 + f2);
-    q0 = saturate_uint8(q0 - f1);
-
-    if (!use_outer_taps)
-    {
-        /* This handles the case of subblock_filter()
-         * (from the bitstream guide.
-         */
-        a = (f1 + 1) >> 1;
-        p1 = saturate_uint8(p1 + a);
-        q1 = saturate_uint8(q1 - a);
-    }
-}
-
-
-static void
-filter_mb_edge(unsigned char *pixels,
-               int            stride)
-{
-    int w, a;
-
-    w = saturate_int8(saturate_int8(p1 - q1) + 3 * (q0 - p0));
-
-    a = (27 * w + 63) >> 7;
-    p0 = saturate_uint8(p0 + a);
-    q0 = saturate_uint8(q0 - a);
-
-    a = (18 * w + 63) >> 7;
-    p1 = saturate_uint8(p1 + a);
-    q1 = saturate_uint8(q1 - a);
-
-    a = (9 * w + 63) >> 7;
-    p2 = saturate_uint8(p2 + a);
-    q2 = saturate_uint8(q2 - a);
-
-}
-
-
-static void
-filter_mb_v_edge(unsigned char *src,
-                 int            stride,
-                 int            edge_limit,
-                 int            interior_limit,
-                 int            hev_threshold,
-                 int            size)
-{
-    int i;
-
-    for (i = 0; i < 8 * size; i++)
-    {
-        if (normal_threshold(src, 1, edge_limit, interior_limit))
-        {
-            if (high_edge_variance(src, 1, hev_threshold))
-                filter_common(src, 1, 1);
-            else
-                filter_mb_edge(src, 1);
-        }
-
-        src += stride;
-    }
-}
-
-
-static void
-filter_subblock_v_edge(unsigned char *src,
-                       int            stride,
-                       int            edge_limit,
-                       int            interior_limit,
-                       int            hev_threshold,
-                       int            size)
-{
-    int i;
-
-    for (i = 0; i < 8 * size; i++)
-    {
-        if (normal_threshold(src, 1, edge_limit, interior_limit))
-            filter_common(src, 1,
-                          high_edge_variance(src, 1, hev_threshold));
-
-        src += stride;
-    }
-}
-
-
-static void
-filter_mb_h_edge(unsigned char *src,
-                 int            stride,
-                 int            edge_limit,
-                 int            interior_limit,
-                 int            hev_threshold,
-                 int            size)
-{
-    int i;
-
-    for (i = 0; i < 8 * size; i++)
-    {
-        if (normal_threshold(src, stride, edge_limit, interior_limit))
-        {
-            if (high_edge_variance(src, stride, hev_threshold))
-                filter_common(src, stride, 1);
-            else
-                filter_mb_edge(src, stride);
-        }
-
-        src += 1;
-    }
-}
-
-
-static void
-filter_subblock_h_edge(unsigned char *src,
-                       int            stride,
-                       int            edge_limit,
-                       int            interior_limit,
-                       int            hev_threshold,
-                       int            size)
-{
-    int i;
-
-    for (i = 0; i < 8 * size; i++)
-    {
-        if (normal_threshold(src, stride, edge_limit, interior_limit))
-            filter_common(src, stride,
-                          high_edge_variance(src, stride,
-                                             hev_threshold));
-
-        src += 1;
-    }
-}
-
-
-static void
-filter_v_edge_simple(unsigned char *src,
-                     int            stride,
-                     int            filter_limit)
-{
-    int i;
-
-    for (i = 0; i < 16; i++)
-    {
-        if (simple_threshold(src, 1, filter_limit))
-            filter_common(src, 1, 1);
-
-        src += stride;
-    }
-}
-
-
-static void
-filter_h_edge_simple(unsigned char *src,
-                     int            stride,
-                     int            filter_limit)
-{
-    int i;
-
-    for (i = 0; i < 16; i++)
-    {
-        if (simple_threshold(src, stride, filter_limit))
-            filter_common(src, stride, 1);
-
-        src += 1;
-    }
-}
-
-
-static void
-calculate_filter_parameters(struct vp8_decoder_ctx *ctx,
-                            struct mb_info         *mbi,
-                            int                    *edge_limit_,
-                            int                    *interior_limit_,
-                            int                    *hev_threshold_)
-{
-    int filter_level, interior_limit, hev_threshold;
-
-    /* Reference code/spec seems to conflate filter_level and
-     * edge_limit
-     */
-
-    filter_level = ctx->loopfilter_hdr.level;
-
-    if (ctx->segment_hdr.enabled)
-    {
-        if (!ctx->segment_hdr.abs)
-            filter_level +=
-                ctx->segment_hdr.lf_level[mbi->base.segment_id];
-        else
-            filter_level =
-                ctx->segment_hdr.lf_level[mbi->base.segment_id];
-    }
-
-    if (ctx->loopfilter_hdr.delta_enabled)
-    {
-        filter_level +=
-            ctx->loopfilter_hdr.ref_delta[mbi->base.ref_frame];
-
-        if (mbi->base.ref_frame == CURRENT_FRAME)
-        {
-            if (mbi->base.y_mode == B_PRED)
-                filter_level += ctx->loopfilter_hdr.mode_delta[0];
-        }
-        else if (mbi->base.y_mode == ZEROMV)
-            filter_level += ctx->loopfilter_hdr.mode_delta[1];
-        else if (mbi->base.y_mode == SPLITMV)
-            filter_level += ctx->loopfilter_hdr.mode_delta[3];
-        else
-            filter_level += ctx->loopfilter_hdr.mode_delta[2];
-    }
-
-    if (filter_level > 63)
-        filter_level = 63;
-    else if (filter_level < 0)
-        filter_level = 0;
-
-    interior_limit = filter_level;
-
-    if (ctx->loopfilter_hdr.sharpness)
-    {
-        interior_limit >>= ctx->loopfilter_hdr.sharpness > 4 ? 2 : 1;
-
-        if (interior_limit > 9 - ctx->loopfilter_hdr.sharpness)
-            interior_limit = 9 - ctx->loopfilter_hdr.sharpness;
-    }
-
-    if (interior_limit < 1)
-        interior_limit = 1;
-
-    hev_threshold = (filter_level >= 15);
-
-    if (filter_level >= 40)
-        hev_threshold++;
-
-    if (filter_level >= 20 && !ctx->frame_hdr.is_keyframe)
-        hev_threshold++;
-
-    *edge_limit_ = filter_level;
-    *interior_limit_ = interior_limit;
-    *hev_threshold_ = hev_threshold;
-}
-
-
-static void
-filter_row_normal(struct vp8_decoder_ctx *ctx,
-                  unsigned int            row,
-                  unsigned int            start_col,
-                  unsigned int            num_cols)
-{
-    unsigned char  *y, *u, *v;
-    int             stride, uv_stride;
-    struct mb_info *mbi;
-    unsigned int    col;
-
-    /* Adjust pointers based on row, start_col */
-    stride    = ctx->ref_frames[CURRENT_FRAME]->img.stride[PLANE_Y];
-    uv_stride = ctx->ref_frames[CURRENT_FRAME]->img.stride[PLANE_U];
-    y = ctx->ref_frames[CURRENT_FRAME]->img.planes[PLANE_Y];
-    u = ctx->ref_frames[CURRENT_FRAME]->img.planes[PLANE_U];
-    v = ctx->ref_frames[CURRENT_FRAME]->img.planes[PLANE_V];
-    y += (stride * row + start_col) * 16;
-    u += (uv_stride * row + start_col) * 8;
-    v += (uv_stride * row + start_col) * 8;
-    mbi = ctx->mb_info_rows[row] + start_col;
-
-    for (col = start_col; col < start_col + num_cols; col++)
-    {
-        int edge_limit, interior_limit, hev_threshold;
-
-        /* TODO: only need to recalculate every MB if segmentation is
-         * enabled.
-         */
-        calculate_filter_parameters(ctx, mbi, &edge_limit,
-                                    &interior_limit, &hev_threshold);
-
-        if (edge_limit)
-        {
-            if (col)
-            {
-                filter_mb_v_edge(y, stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 2);
-                filter_mb_v_edge(u, uv_stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 1);
-                filter_mb_v_edge(v, uv_stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 1);
-            }
-
-            /* NOTE: This conditional is actually dependent on the
-             * number of coefficients decoded, not the skip flag as
-             * coded in the bitstream. The tokens task is expected to
-             * set 31 if there is *any* non-zero data.
-             */
-            if (mbi->base.eob_mask
-                || mbi->base.y_mode == SPLITMV
-                || mbi->base.y_mode == B_PRED)
-            {
-                filter_subblock_v_edge(y + 4, stride, edge_limit,
-                                       interior_limit, hev_threshold,
-                                       2);
-                filter_subblock_v_edge(y + 8, stride, edge_limit,
-                                       interior_limit, hev_threshold,
-                                       2);
-                filter_subblock_v_edge(y + 12, stride, edge_limit,
-                                       interior_limit, hev_threshold,
-                                       2);
-                filter_subblock_v_edge(u + 4, uv_stride, edge_limit,
-                                       interior_limit, hev_threshold,
-                                       1);
-                filter_subblock_v_edge(v + 4, uv_stride, edge_limit,
-                                       interior_limit, hev_threshold,
-                                       1);
-            }
-
-            if (row)
-            {
-                filter_mb_h_edge(y, stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 2);
-                filter_mb_h_edge(u, uv_stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 1);
-                filter_mb_h_edge(v, uv_stride, edge_limit + 2,
-                                 interior_limit, hev_threshold, 1);
-            }
-
-            if (mbi->base.eob_mask
-                || mbi->base.y_mode == SPLITMV
-                || mbi->base.y_mode == B_PRED)
-            {
-                filter_subblock_h_edge(y + 4 * stride, stride,
-                                       edge_limit, interior_limit,
-                                       hev_threshold, 2);
-                filter_subblock_h_edge(y + 8 * stride, stride,
-                                       edge_limit, interior_limit,
-                                       hev_threshold, 2);
-                filter_subblock_h_edge(y + 12 * stride, stride,
-                                       edge_limit, interior_limit,
-                                       hev_threshold, 2);
-                filter_subblock_h_edge(u + 4 * uv_stride, uv_stride,
-                                       edge_limit, interior_limit,
-                                       hev_threshold, 1);
-                filter_subblock_h_edge(v + 4 * uv_stride, uv_stride,
-                                       edge_limit, interior_limit,
-                                       hev_threshold, 1);
-            }
-        }
-
-        y += 16;
-        u += 8;
-        v += 8;
-        mbi++;
-    }
-}
-
-
-static void
-filter_row_simple(struct vp8_decoder_ctx *ctx,
-                  unsigned int            row,
-                  unsigned int            start_col,
-                  unsigned int            num_cols)
-{
-    unsigned char  *y;
-    int             stride;
-    struct mb_info *mbi;
-    unsigned int    col;
-
-    /* Adjust pointers based on row, start_col */
-    stride    = ctx->ref_frames[CURRENT_FRAME]->img.stride[PLANE_Y];
-    y = ctx->ref_frames[CURRENT_FRAME]->img.planes[PLANE_Y];
-    y += (stride * row + start_col) * 16;
-    mbi = ctx->mb_info_rows[row] + start_col;
-
-    for (col = start_col; col < start_col + num_cols; col++)
-    {
-        int edge_limit, interior_limit, hev_threshold;
-
-        /* TODO: only need to recalculate every MB if segmentation is
-         * enabled.
-         */
-        calculate_filter_parameters(ctx, mbi, &edge_limit,
-                                    &interior_limit, &hev_threshold);
-
-        if (edge_limit)
-        {
-
-            /* NOTE: This conditional is actually dependent on the
-             * number of coefficients decoded, not the skip flag as
-             * coded in the bitstream. The tokens task is expected to
-             * set 31 if there is *any* non-zero data.
-             */
-            int filter_subblocks = (mbi->base.eob_mask
-                                    || mbi->base.y_mode == SPLITMV
-                                    || mbi->base.y_mode == B_PRED);
-            int mb_limit = (edge_limit + 2) * 2 + interior_limit;
-            int b_limit = edge_limit * 2 + interior_limit;
-
-            if (col)
-                filter_v_edge_simple(y, stride, mb_limit);
-
-            if (filter_subblocks)
-            {
-                filter_v_edge_simple(y + 4, stride, b_limit);
-                filter_v_edge_simple(y + 8, stride, b_limit);
-                filter_v_edge_simple(y + 12, stride, b_limit);
-            }
-
-            if (row)
-                filter_h_edge_simple(y, stride, mb_limit);
-
-            if (filter_subblocks)
-            {
-                filter_h_edge_simple(y + 4 * stride, stride, b_limit);
-                filter_h_edge_simple(y + 8 * stride, stride, b_limit);
-                filter_h_edge_simple(y + 12 * stride, stride, b_limit);
-            }
-        }
-
-        y += 16;
-        mbi++;
-    }
-}
-
-
-void
-vp8_dixie_loopfilter_process_row(struct vp8_decoder_ctx *ctx,
-                                 unsigned int            row,
-                                 unsigned int            start_col,
-                                 unsigned int            num_cols)
-{
-    if (ctx->loopfilter_hdr.use_simple)
-        filter_row_simple(ctx, row, start_col, num_cols);
-    else
-        filter_row_normal(ctx, row, start_col, num_cols);
-}

vp8/dixie/dixie_loopfilter.h

@@ -1,19 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef DIXIE_LOOPFILTER_H
-#define DIXIE_LOOPFILTER_H
-
-void
-vp8_dixie_loopfilter_process_row(struct vp8_decoder_ctx *ctx,
-                                 unsigned int            row,
-                                 unsigned int            start_col,
-                                 unsigned int            num_cols);
-
-#endif

vp8/dixie/idct_add.c

@@ -1,142 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include "dixie.h"
-#include "idct_add.h"
-#include <assert.h>
-
-void
-vp8_dixie_walsh(const short *input, short *output)
-{
-    int i;
-    int a1, b1, c1, d1;
-    int a2, b2, c2, d2;
-    const short *ip = input;
-    short *op = output;
-
-    for (i = 0; i < 4; i++)
-    {
-        a1 = ip[0] + ip[12];
-        b1 = ip[4] + ip[8];
-        c1 = ip[4] - ip[8];
-        d1 = ip[0] - ip[12];
-
-        op[0] = a1 + b1;
-        op[4] = c1 + d1;
-        op[8] = a1 - b1;
-        op[12] = d1 - c1;
-        ip++;
-        op++;
-    }
-
-    ip = output;
-    op = output;
-
-    for (i = 0; i < 4; i++)
-    {
-        a1 = ip[0] + ip[3];
-        b1 = ip[1] + ip[2];
-        c1 = ip[1] - ip[2];
-        d1 = ip[0] - ip[3];
-
-        a2 = a1 + b1;
-        b2 = c1 + d1;
-        c2 = a1 - b1;
-        d2 = d1 - c1;
-
-        op[0] = (a2 + 3) >> 3;
-        op[1] = (b2 + 3) >> 3;
-        op[2] = (c2 + 3) >> 3;
-        op[3] = (d2 + 3) >> 3;
-
-        ip += 4;
-        op += 4;
-    }
-}
-
-
-#define cospi8sqrt2minus1 20091
-#define sinpi8sqrt2       35468
-#define rounding          0
-static void
-idct_columns(const short *input, short *output)
-{
-    int i;
-    int a1, b1, c1, d1;
-
-    const short *ip = input;
-    short *op = output;
-    int temp1, temp2;
-    int shortpitch = 4;
-
-    for (i = 0; i < 4; i++)
-    {
-        a1 = ip[0] + ip[8];
-        b1 = ip[0] - ip[8];
-
-        temp1 = (ip[4] * sinpi8sqrt2 + rounding) >> 16;
-        temp2 = ip[12] +
-            ((ip[12] * cospi8sqrt2minus1 + rounding) >> 16);
-        c1 = temp1 - temp2;
-
-        temp1 = ip[4] +
-            ((ip[4] * cospi8sqrt2minus1 + rounding) >> 16);
-        temp2 = (ip[12] * sinpi8sqrt2 + rounding) >> 16;
-        d1 = temp1 + temp2;
-
-        op[shortpitch*0] = a1 + d1;
-        op[shortpitch*3] = a1 - d1;
-
-        op[shortpitch*1] = b1 + c1;
-        op[shortpitch*2] = b1 - c1;
-
-        ip++;
-        op++;
-    }
-}
-
-
-void
-vp8_dixie_idct_add(unsigned char        *recon,
-                   const unsigned char  *predict,
-                   int                   stride,
-                   const short          *coeffs)
-{
-    int i;
-    int a1, b1, c1, d1, temp1, temp2;
-    short tmp[16];
-
-    idct_columns(coeffs, tmp);
-    coeffs = tmp;
-
-    for (i = 0; i < 4; i++)
-    {
-        a1 = coeffs[0] + coeffs[2];
-        b1 = coeffs[0] - coeffs[2];
-
-        temp1 = (coeffs[1] * sinpi8sqrt2 + rounding) >> 16;
-        temp2 = coeffs[3] +
-            ((coeffs[3] * cospi8sqrt2minus1 + rounding) >> 16);
-        c1 = temp1 - temp2;
-
-        temp1 = coeffs[1] +
-            ((coeffs[1] * cospi8sqrt2minus1 + rounding) >> 16);
-        temp2 = (coeffs[3] * sinpi8sqrt2 + rounding) >> 16;
-        d1 = temp1 + temp2;
-
-        recon[0] = CLAMP_255(predict[0] + ((a1 + d1 + 4) >> 3));
-        recon[3] = CLAMP_255(predict[3] + ((a1 - d1 + 4) >> 3));
-        recon[1] = CLAMP_255(predict[1] + ((b1 + c1 + 4) >> 3));
-        recon[2] = CLAMP_255(predict[2] + ((b1 - c1 + 4) >> 3));
-
-        coeffs += 4;
-        recon += stride;
-        predict += stride;
-    }
-}

vp8/dixie/idct_add.h

@@ -1,35 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef IDCT_ADD_H
-#define IDCT_ADD_H
-
-void
-vp8_dixie_idct_add_init(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_idct_add(unsigned char        *recon,
-                   const unsigned char  *predict,
-                   int                   stride,
-                   const short          *coeffs);
-
-
-void
-vp8_dixie_walsh(const short *in, short *out);
-
-
-void
-vp8_dixie_idct_add_process_row(struct vp8_decoder_ctx *ctx,
-                               short                  *coeffs,
-                               unsigned int            row,
-                               unsigned int            start_col,
-                               unsigned int            num_cols);
-
-#endif

vp8/dixie/modemv.c

@@ -1,686 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include "dixie.h"
-#include "modemv_data.h"
-#include <stdlib.h>
-#include <assert.h>
-
-
-struct mv_clamp_rect
-{
-    int to_left, to_right, to_top, to_bottom;
-};
-
-
-static union mv
-        clamp_mv(union mv raw, const struct mv_clamp_rect *bounds)
-{
-    union mv newmv;
-
-    newmv.d.x = (raw.d.x < bounds->to_left)
-                ? bounds->to_left : raw.d.x;
-    newmv.d.x = (raw.d.x > bounds->to_right)
-                ? bounds->to_right : newmv.d.x;
-    newmv.d.y = (raw.d.y < bounds->to_top)
-                ? bounds->to_top : raw.d.y;
-    newmv.d.y = (raw.d.y > bounds->to_bottom)
-                ? bounds->to_bottom : newmv.d.y;
-    return newmv;
-}
-
-
-static int
-read_segment_id(struct bool_decoder *bool, struct vp8_segment_hdr *seg)
-{
-    return bool_get(bool, seg->tree_probs[0])
-           ? 2 + bool_get(bool, seg->tree_probs[2])
-           : bool_get(bool, seg->tree_probs[1]);
-}
-
-
-static enum prediction_mode
-above_block_mode(const struct mb_info *this,
-                 const struct mb_info *above,
-                 unsigned int b)
-{
-    if (b < 4)
-    {
-        switch (above->base.y_mode)
-        {
-        case DC_PRED:
-            return B_DC_PRED;
-        case V_PRED:
-            return B_VE_PRED;
-        case H_PRED:
-            return B_HE_PRED;
-        case TM_PRED:
-            return B_TM_PRED;
-        case B_PRED:
-            return above->split.modes[b+12];
-        default:
-            assert(0);
-        }
-    }
-
-    return this->split.modes[b-4];
-}
-
-
-static enum prediction_mode
-left_block_mode(const struct mb_info *this,
-                const struct mb_info *left,
-                unsigned int b)
-{
-    if (!(b & 3))
-    {
-        switch (left->base.y_mode)
-        {
-        case DC_PRED:
-            return B_DC_PRED;
-        case V_PRED:
-            return B_VE_PRED;
-        case H_PRED:
-            return B_HE_PRED;
-        case TM_PRED:
-            return B_TM_PRED;
-        case B_PRED:
-            return left->split.modes[b+3];
-        default:
-            assert(0);
-        }
-    }
-
-    return this->split.modes[b-1];
-}
-
-
-static void
-decode_kf_mb_mode(struct mb_info      *this,
-                  struct mb_info      *left,
-                  struct mb_info      *above,
-                  struct bool_decoder *bool)
-{
-    int y_mode, uv_mode;
-
-    y_mode = bool_read_tree(bool, kf_y_mode_tree, kf_y_mode_probs);
-
-    if (y_mode == B_PRED)
-    {
-        unsigned int i;
-
-        for (i = 0; i < 16; i++)
-        {
-            enum prediction_mode a = above_block_mode(this, above, i);
-            enum prediction_mode l = left_block_mode(this, left, i);
-            enum prediction_mode b;
-
-            b = bool_read_tree(bool, b_mode_tree,
-                               kf_b_mode_probs[a][l]);
-            this->split.modes[i] = b;
-        }
-    }
-
-    uv_mode = bool_read_tree(bool, uv_mode_tree, kf_uv_mode_probs);
-
-    this->base.y_mode = y_mode;
-    this->base.uv_mode = uv_mode;
-    this->base.mv.raw = 0;
-    this->base.ref_frame = 0;
-}
-
-
-static void
-decode_intra_mb_mode(struct mb_info         *this,
-                     struct vp8_entropy_hdr *hdr,
-                     struct bool_decoder    *bool)
-{
-    /* Like decode_kf_mb_mode, but with probabilities transmitted in the
-     * bitstream and no context on the above/left block mode.
-     */
-    int y_mode, uv_mode;
-
-    y_mode = bool_read_tree(bool, y_mode_tree, hdr->y_mode_probs);
-
-    if (y_mode == B_PRED)
-    {
-        unsigned int i;
-
-        for (i = 0; i < 16; i++)
-        {
-            enum prediction_mode b;
-
-            b = bool_read_tree(bool, b_mode_tree, default_b_mode_probs);
-            this->split.modes[i] = b;
-        }
-    }
-
-    uv_mode = bool_read_tree(bool, uv_mode_tree, hdr->uv_mode_probs);
-
-    this->base.y_mode = y_mode;
-    this->base.uv_mode = uv_mode;
-    this->base.mv.raw = 0;
-    this->base.ref_frame = CURRENT_FRAME;
-}
-
-
-static int
-read_mv_component(struct bool_decoder *bool,
-                  const unsigned char  mvc[MV_PROB_CNT])
-{
-    enum {IS_SHORT, SIGN, SHORT, BITS = SHORT + 8 - 1, LONG_WIDTH = 10};
-    int x = 0;
-
-    if (bool_get(bool, mvc[IS_SHORT])) /* Large */
-    {
-        int i = 0;
-
-        for (i = 0; i < 3; i++)
-            x += bool_get(bool, mvc[BITS + i]) << i;
-
-        /* Skip bit 3, which is sometimes implicit */
-        for (i = LONG_WIDTH - 1; i > 3; i--)
-            x += bool_get(bool, mvc[BITS + i]) << i;
-
-        if (!(x & 0xFFF0)  ||  bool_get(bool, mvc[BITS + 3]))
-            x += 8;
-    }
-    else   /* small */
-        x = bool_read_tree(bool, small_mv_tree, mvc + SHORT);
-
-    if (x && bool_get(bool, mvc[SIGN]))
-        x = -x;
-
-    return x << 1;
-}
-
-
-static mv_t
-above_block_mv(const struct mb_info *this,
-               const struct mb_info *above,
-               unsigned int          b)
-{
-    if (b < 4)
-    {
-        if (above->base.y_mode == SPLITMV)
-            return above->split.mvs[b+12];
-
-        return above->base.mv;
-    }
-
-    return this->split.mvs[b-4];
-}
-
-
-static mv_t
-left_block_mv(const struct mb_info *this,
-              const struct mb_info *left,
-              unsigned int          b)
-{
-    if (!(b & 3))
-    {
-        if (left->base.y_mode == SPLITMV)
-            return left->split.mvs[b+3];
-
-        return left->base.mv;
-    }
-
-    return this->split.mvs[b-1];
-}
-
-
-static enum prediction_mode
-submv_ref(struct bool_decoder *bool, union mv l, union mv a)
-{
-    enum subblock_mv_ref
-    {
-        SUBMVREF_NORMAL,
-        SUBMVREF_LEFT_ZED,
-        SUBMVREF_ABOVE_ZED,
-        SUBMVREF_LEFT_ABOVE_SAME,
-        SUBMVREF_LEFT_ABOVE_ZED
-    };
-
-    int lez = !(l.raw);
-    int aez = !(a.raw);
-    int lea = l.raw == a.raw;
-    enum subblock_mv_ref ctx = SUBMVREF_NORMAL;
-
-    if (lea && lez)
-        ctx = SUBMVREF_LEFT_ABOVE_ZED;
-    else if (lea)
-        ctx = SUBMVREF_LEFT_ABOVE_SAME;
-    else if (aez)
-        ctx = SUBMVREF_ABOVE_ZED;
-    else if (lez)
-        ctx = SUBMVREF_LEFT_ZED;
-
-    return bool_read_tree(bool, submv_ref_tree, submv_ref_probs2[ctx]);
-}
-
-
-static void
-read_mv(struct bool_decoder  *bool,
-        union mv             *mv,
-        mv_component_probs_t  mvc[2])
-{
-    mv->d.y = read_mv_component(bool, mvc[0]);
-    mv->d.x = read_mv_component(bool, mvc[1]);
-}
-
-
-static void
-mv_bias(const struct mb_info *mb,
-        const unsigned int   sign_bias[3],
-        enum reference_frame ref_frame,
-        union mv             *mv)
-{
-    if (sign_bias[mb->base.ref_frame] ^ sign_bias[ref_frame])
-    {
-        mv->d.x *= -1;
-        mv->d.y *= -1;
-    }
-}
-
-
-enum near_mv_v
-{
-    CNT_BEST = 0,
-    CNT_ZEROZERO = 0,
-    CNT_NEAREST,
-    CNT_NEAR,
-    CNT_SPLITMV
-};
-
-
-static void
-find_near_mvs(const struct mb_info   *this,
-              const struct mb_info   *left,
-              const struct mb_info   *above,
-              const unsigned int      sign_bias[3],
-              union  mv               near_mvs[4],
-              int                     cnt[4])
-{
-    const struct mb_info *aboveleft = above - 1;
-    union  mv             *mv = near_mvs;
-    int                   *cntx = cnt;
-
-    /* Zero accumulators */
-    mv[0].raw = mv[1].raw = mv[2].raw = 0;
-    cnt[0] = cnt[1] = cnt[2] = cnt[3] = 0;
-
-    /* Process above */
-    if (above->base.ref_frame != CURRENT_FRAME)
-    {
-        if (above->base.mv.raw)
-        {
-            (++mv)->raw = above->base.mv.raw;
-            mv_bias(above, sign_bias, this->base.ref_frame, mv);
-            ++cntx;
-        }
-
-        *cntx += 2;
-    }
-
-    /* Process left */
-    if (left->base.ref_frame != CURRENT_FRAME)
-    {
-        if (left->base.mv.raw)
-        {
-            union mv this_mv;
-
-            this_mv.raw = left->base.mv.raw;
-            mv_bias(left, sign_bias, this->base.ref_frame, &this_mv);
-
-            if (this_mv.raw != mv->raw)
-            {
-                (++mv)->raw = this_mv.raw;
-                ++cntx;
-            }
-
-            *cntx += 2;
-        }
-        else
-            cnt[CNT_ZEROZERO] += 2;
-    }
-
-    /* Process above left */
-    if (aboveleft->base.ref_frame != CURRENT_FRAME)
-    {
-        if (aboveleft->base.mv.raw)
-        {
-            union mv this_mv;
-
-            this_mv.raw = aboveleft->base.mv.raw;
-            mv_bias(aboveleft, sign_bias, this->base.ref_frame,
-                    &this_mv);
-
-            if (this_mv.raw != mv->raw)
-            {
-                (++mv)->raw = this_mv.raw;
-                ++cntx;
-            }
-
-            *cntx += 1;
-        }
-        else
-            cnt[CNT_ZEROZERO] += 1;
-    }
-
-    /* If we have three distinct MV's ... */
-    if (cnt[CNT_SPLITMV])
-    {
-        /* See if above-left MV can be merged with NEAREST */
-        if (mv->raw == near_mvs[CNT_NEAREST].raw)
-            cnt[CNT_NEAREST] += 1;
-    }
-
-    cnt[CNT_SPLITMV] = ((above->base.y_mode == SPLITMV)
-                        + (left->base.y_mode == SPLITMV)) * 2
-                       + (aboveleft->base.y_mode == SPLITMV);
-
-    /* Swap near and nearest if necessary */
-    if (cnt[CNT_NEAR] > cnt[CNT_NEAREST])
-    {
-        int tmp;
-        tmp = cnt[CNT_NEAREST];
-        cnt[CNT_NEAREST] = cnt[CNT_NEAR];
-        cnt[CNT_NEAR] = tmp;
-        tmp = near_mvs[CNT_NEAREST].raw;
-        near_mvs[CNT_NEAREST].raw = near_mvs[CNT_NEAR].raw;
-        near_mvs[CNT_NEAR].raw = tmp;
-    }
-
-    /* Use near_mvs[CNT_BEST] to store the "best" MV. Note that this
-     * storage shares the same address as near_mvs[CNT_ZEROZERO].
-     */
-    if (cnt[CNT_NEAREST] >= cnt[CNT_BEST])
-        near_mvs[CNT_BEST] = near_mvs[CNT_NEAREST];
-}
-
-
-static void
-decode_split_mv(struct mb_info         *this,
-                const struct mb_info   *left,
-                const struct mb_info   *above,
-                struct vp8_entropy_hdr *hdr,
-                union  mv              *best_mv,
-                struct bool_decoder    *bool)
-{
-    const int *partition;
-    int        j, k, mask, partition_id;
-
-    partition_id = bool_read_tree(bool, split_mv_tree, split_mv_probs);
-    partition = mv_partitions[partition_id];
-    this->base.partitioning = partition_id;
-
-    for (j = 0, mask = 0; mask < 65535; j++)
-    {
-        union mv mv, left_mv, above_mv;
-        enum prediction_mode subblock_mode;
-
-        /* Find the first subblock in this partition. */
-        for (k = 0; j != partition[k]; k++);
-
-        /* Decode the next MV */
-        left_mv = left_block_mv(this, left, k);
-        above_mv = above_block_mv(this, above, k);
-        subblock_mode = submv_ref(bool, left_mv,  above_mv);
-
-        switch (subblock_mode)
-        {
-        case LEFT4X4:
-            mv = left_mv;
-            break;
-        case ABOVE4X4:
-            mv = above_mv;
-            break;
-        case ZERO4X4:
-            mv.raw = 0;
-            break;
-        case NEW4X4:
-            read_mv(bool, &mv, hdr->mv_probs);
-            mv.d.x += best_mv->d.x;
-            mv.d.y += best_mv->d.y;
-            break;
-        default:
-            assert(0);
-        }
-
-        /* Fill the MV's for this partition */
-        for (; k < 16; k++)
-            if (j == partition[k])
-            {
-                this->split.mvs[k] = mv;
-                mask |= 1 << k;
-            }
-    }
-}
-
-
-static int
-need_mc_border(union mv mv, int l, int t, int b_w, int w, int h)
-{
-    int b, r;
-
-    /* Get distance to edge for top-left pixel */
-    l += (mv.d.x >> 3);
-    t += (mv.d.y >> 3);
-
-    /* Get distance to edge for bottom-right pixel */
-    r = w - (l + b_w);
-    b = h - (t + b_w);
-
-    return (l >> 1 < 2 || r >> 1 < 3 || t >> 1 < 2 || b >> 1 < 3);
-}
-
-static void
-decode_mvs(struct vp8_decoder_ctx       *ctx,
-           struct mb_info               *this,
-           const struct mb_info         *left,
-           const struct mb_info         *above,
-           const struct mv_clamp_rect   *bounds,
-           struct bool_decoder          *bool)
-{
-    struct vp8_entropy_hdr *hdr = &ctx->entropy_hdr;
-    union mv          near_mvs[4];
-    union mv          clamped_best_mv;
-    int               mv_cnts[4];
-    unsigned char     probs[4];
-    enum {BEST, NEAREST, NEAR};
-    int x, y, w, h, b;
-
-    this->base.ref_frame = bool_get(bool, hdr->prob_last)
-                           ? 2 + bool_get(bool, hdr->prob_gf)
-                           : 1;
-
-    find_near_mvs(this, this - 1, above, ctx->reference_hdr.sign_bias,
-                  near_mvs, mv_cnts);
-    probs[0] = mv_counts_to_probs[mv_cnts[0]][0];
-    probs[1] = mv_counts_to_probs[mv_cnts[1]][1];
-    probs[2] = mv_counts_to_probs[mv_cnts[2]][2];
-    probs[3] = mv_counts_to_probs[mv_cnts[3]][3];
-
-    this->base.y_mode = bool_read_tree(bool, mv_ref_tree, probs);
-    this->base.uv_mode = this->base.y_mode;
-
-    this->base.need_mc_border = 0;
-    x = (-bounds->to_left - 128) >> 3;
-    y = (-bounds->to_top - 128) >> 3;
-    w = ctx->mb_cols * 16;
-    h = ctx->mb_rows * 16;
-
-    switch (this->base.y_mode)
-    {
-    case NEARESTMV:
-        this->base.mv = clamp_mv(near_mvs[NEAREST], bounds);
-        break;
-    case NEARMV:
-        this->base.mv = clamp_mv(near_mvs[NEAR], bounds);
-        break;
-    case ZEROMV:
-        this->base.mv.raw = 0;
-        return; //skip need_mc_border check
-    case NEWMV:
-        clamped_best_mv = clamp_mv(near_mvs[BEST], bounds);
-        read_mv(bool, &this->base.mv, hdr->mv_probs);
-        this->base.mv.d.x += clamped_best_mv.d.x;
-        this->base.mv.d.y += clamped_best_mv.d.y;
-        break;
-    case SPLITMV:
-    {
-        union mv          chroma_mv[4] = {{{0}}};
-
-        clamped_best_mv = clamp_mv(near_mvs[BEST], bounds);
-        decode_split_mv(this, left, above, hdr, &clamped_best_mv, bool);
-        this->base.mv = this->split.mvs[15];
-
-        for (b = 0; b < 16; b++)
-        {
-            chroma_mv[(b>>1&1) + (b>>2&2)].d.x +=
-                this->split.mvs[b].d.x;
-            chroma_mv[(b>>1&1) + (b>>2&2)].d.y +=
-                this->split.mvs[b].d.y;
-
-            if (need_mc_border(this->split.mvs[b],
-            x + (b & 3) * 4, y + (b & ~3), 4, w, h))
-            {
-                this->base.need_mc_border = 1;
-                break;
-            }
-        }
-
-        for (b = 0; b < 4; b++)
-        {
-            chroma_mv[b].d.x += 4 + 8 * (chroma_mv[b].d.x >> 31);
-            chroma_mv[b].d.y += 4 + 8 * (chroma_mv[b].d.y >> 31);
-            chroma_mv[b].d.x /= 4;
-            chroma_mv[b].d.y /= 4;
-
-            //note we're passing in non-subsampled coordinates
-            if (need_mc_border(chroma_mv[b],
-            x + (b & 1) * 8, y + (b >> 1) * 8, 16, w, h))
-            {
-                this->base.need_mc_border = 1;
-                break;
-            }
-        }
-
-        return; //skip need_mc_border check
-    }
-    default:
-        assert(0);
-    }
-
-    if (need_mc_border(this->base.mv, x, y, 16, w, h))
-        this->base.need_mc_border = 1;
-}
-
-
-void
-vp8_dixie_modemv_process_row(struct vp8_decoder_ctx *ctx,
-struct bool_decoder    *bool,
-int                     row,
-int                     start_col,
-int                     num_cols)
-{
-    struct mb_info       *above, *this;
-    unsigned int          col;
-    struct mv_clamp_rect  bounds;
-
-    this = ctx->mb_info_rows[row] + start_col;
-    above = ctx->mb_info_rows[row - 1] + start_col;
-
-    /* Calculate the eighth-pel MV bounds using a 1 MB border. */
-    bounds.to_left   = -((start_col + 1) << 7);
-    bounds.to_right  = (ctx->mb_cols - start_col) << 7;
-    bounds.to_top    = -((row + 1) << 7);
-    bounds.to_bottom = (ctx->mb_rows - row) << 7;
-
-    for (col = start_col; col < start_col + num_cols; col++)
-    {
-        if (ctx->segment_hdr.update_map)
-            this->base.segment_id = read_segment_id(bool,
-            &ctx->segment_hdr);
-
-        if (ctx->entropy_hdr.coeff_skip_enabled)
-            this->base.skip_coeff = bool_get(bool,
-            ctx->entropy_hdr.coeff_skip_prob);
-
-        if (ctx->frame_hdr.is_keyframe)
-        {
-            if (!ctx->segment_hdr.update_map)
-                this->base.segment_id = 0;
-
-            decode_kf_mb_mode(this, this - 1, above, bool);
-        }
-        else
-        {
-            if (bool_get(bool, ctx->entropy_hdr.prob_inter))
-                decode_mvs(ctx, this, this - 1, above, &bounds, bool);
-            else
-                decode_intra_mb_mode(this, &ctx->entropy_hdr, bool);
-
-            bounds.to_left -= 16 << 3;
-            bounds.to_right -= 16 << 3;
-        }
-
-        /* Advance to next mb */
-        this++;
-        above++;
-    }
-}
-
-
-void
-vp8_dixie_modemv_init(struct vp8_decoder_ctx *ctx)
-{
-    unsigned int    mbi_w, mbi_h, i;
-    struct mb_info *mbi;
-
-    mbi_w = ctx->mb_cols + 1; /* For left border col */
-    mbi_h = ctx->mb_rows + 1; /* For above border row */
-
-    if (ctx->frame_hdr.frame_size_updated)
-    {
-        free(ctx->mb_info_storage);
-        ctx->mb_info_storage = NULL;
-        free(ctx->mb_info_rows_storage);
-        ctx->mb_info_rows_storage = NULL;
-    }
-
-    if (!ctx->mb_info_storage)
-        ctx->mb_info_storage = calloc(mbi_w * mbi_h,
-        sizeof(*ctx->mb_info_storage));
-
-    if (!ctx->mb_info_rows_storage)
-        ctx->mb_info_rows_storage = calloc(mbi_h,
-        sizeof(*ctx->mb_info_rows_storage));
-
-    /* Set up row pointers */
-    mbi = ctx->mb_info_storage + 1;
-
-    for (i = 0; i < mbi_h; i++)
-    {
-        ctx->mb_info_rows_storage[i] = mbi;
-        mbi += mbi_w;
-    }
-
-    ctx->mb_info_rows = ctx->mb_info_rows_storage + 1;
-}
-
-
-void
-vp8_dixie_modemv_destroy(struct vp8_decoder_ctx *ctx)
-{
-    free(ctx->mb_info_storage);
-    ctx->mb_info_storage = NULL;
-    free(ctx->mb_info_rows_storage);
-    ctx->mb_info_rows_storage = NULL;
-}

vp8/dixie/modemv.h

@@ -1,28 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef MODEMV_H
-#define MODEMV_H
-
-void
-vp8_dixie_modemv_init(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_modemv_destroy(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_modemv_process_row(struct vp8_decoder_ctx *ctx,
-                             struct bool_decoder    *bool,
-                             int                     row,
-                             int                     start_col,
-                             int                     num_cols);
-
-#endif

vp8/dixie/modemv_data.h

@@ -1,216 +0,0 @@
-static const unsigned char kf_y_mode_probs[]  = { 145, 156, 163, 128};
-static const unsigned char kf_uv_mode_probs[] = { 142, 114, 183};
-static const unsigned char kf_b_mode_probs[10][10][9] =
-{
-  { /* above mode 0 */
-    { /* left mode 0 */ 231, 120,  48,  89, 115, 113, 120, 152, 112},
-    { /* left mode 1 */ 152, 179,  64, 126, 170, 118,  46,  70,  95},
-    { /* left mode 2 */ 175,  69, 143,  80,  85,  82,  72, 155, 103},
-    { /* left mode 3 */  56,  58,  10, 171, 218, 189,  17,  13, 152},
-    { /* left mode 4 */ 144,  71,  10,  38, 171, 213, 144,  34,  26},
-    { /* left mode 5 */ 114,  26,  17, 163,  44, 195,  21,  10, 173},
-    { /* left mode 6 */ 121,  24,  80, 195,  26,  62,  44,  64,  85},
-    { /* left mode 7 */ 170,  46,  55,  19, 136, 160,  33, 206,  71},
-    { /* left mode 8 */  63,  20, 8, 114, 114, 208,  12,   9, 226},
-    { /* left mode 9 */  81,  40,  11,  96, 182,  84,  29,  16,  36}
-  },
-  { /* above mode 1 */
-    { /* left mode 0 */ 134, 183,  89, 137,  98, 101, 106, 165, 148},
-    { /* left mode 1 */  72, 187, 100, 130, 157, 111,  32,  75,  80},
-    { /* left mode 2 */  66, 102, 167,  99,  74,  62,  40, 234, 128},
-    { /* left mode 3 */  41,  53, 9, 178, 241, 141,  26,   8, 107},
-    { /* left mode 4 */ 104,  79,  12,  27, 217, 255,  87,  17,   7},
-    { /* left mode 5 */  74,  43,  26, 146,  73, 166,  49,  23, 157},
-    { /* left mode 6 */  65,  38, 105, 160,  51,  52,  31, 115, 128},
-    { /* left mode 7 */  87,  68,  71,  44, 114,  51,  15, 186,  23},
-    { /* left mode 8 */  47,  41,  14, 110, 182, 183,  21,  17, 194},
-    { /* left mode 9 */  66,  45,  25, 102, 197, 189,  23,  18,  22}
-  },
-  { /* above mode 2 */
-    { /* left mode 0 */  88,  88, 147, 150,  42,  46,  45, 196, 205},
-    { /* left mode 1 */  43,  97, 183, 117,  85,  38,  35, 179,  61},
-    { /* left mode 2 */  39,  53, 200,  87,  26,  21,  43, 232, 171},
-    { /* left mode 3 */  56,  34,  51, 104, 114, 102,  29,  93,  77},
-    { /* left mode 4 */ 107,  54,  32,  26,  51,   1,  81,  43,  31},
-    { /* left mode 5 */  39,  28,  85, 171,  58, 165,  90,  98,  64},
-    { /* left mode 6 */  34,  22, 116, 206,  23,  34,  43, 166,  73},
-    { /* left mode 7 */  68,  25, 106,  22,  64, 171,  36, 225, 114},
-    { /* left mode 8 */  34,  19,  21, 102, 132, 188,  16,  76, 124},
-    { /* left mode 9 */  62,  18,  78,  95,  85,  57,  50,  48,  51}
-  },
-  { /* above mode 3 */
-    { /* left mode 0 */ 193, 101,  35, 159, 215, 111,  89,  46, 111},
-    { /* left mode 1 */  60, 148,  31, 172, 219, 228,  21,  18, 111},
-    { /* left mode 2 */ 112, 113,  77,  85, 179, 255,  38, 120, 114},
-    { /* left mode 3 */  40,  42, 1, 196, 245, 209,  10,  25, 109},
-    { /* left mode 4 */ 100,  80, 8,  43, 154,   1,  51,  26,  71},
-    { /* left mode 5 */  88,  43,  29, 140, 166, 213,  37,  43, 154},
-    { /* left mode 6 */  61,  63,  30, 155,  67,  45,  68,   1, 209},
-    { /* left mode 7 */ 142,  78,  78,  16, 255, 128,  34, 197, 171},
-    { /* left mode 8 */  41,  40, 5, 102, 211, 183, 4,   1, 221},
-    { /* left mode 9 */  51,  50,  17, 168, 209, 192,  23,  25,  82}
-  },
-  { /* above mode 4 */
-    { /* left mode 0 */ 125,  98,  42,  88, 104,  85, 117, 175,  82},
-    { /* left mode 1 */  95,  84,  53,  89, 128, 100, 113, 101,  45},
-    { /* left mode 2 */  75,  79, 123,  47,  51, 128,  81, 171,   1},
-    { /* left mode 3 */  57,  17, 5,  71, 102,  57,  53,  41,  49},
-    { /* left mode 4 */ 115,  21, 2,  10, 102, 255, 166,  23,   6},
-    { /* left mode 5 */  38,  33,  13, 121,  57,  73,  26,   1,  85},
-    { /* left mode 6 */  41,  10,  67, 138,  77, 110,  90,  47, 114},
-    { /* left mode 7 */ 101,  29,  16,  10,  85, 128, 101, 196,  26},
-    { /* left mode 8 */  57,  18,  10, 102, 102, 213,  34,  20,  43},
-    { /* left mode 9 */ 117,  20,  15,  36, 163, 128,  68,   1,  26}
-  },
-  { /* above mode 5 */
-    { /* left mode 0 */ 138,  31,  36, 171,  27, 166,  38,  44, 229},
-    { /* left mode 1 */  67,  87,  58, 169,  82, 115,  26,  59, 179},
-    { /* left mode 2 */  63,  59,  90, 180,  59, 166,  93,  73, 154},
-    { /* left mode 3 */  40,  40,  21, 116, 143, 209,  34,  39, 175},
-    { /* left mode 4 */  57,  46,  22,  24, 128,   1,  54,  17,  37},
-    { /* left mode 5 */  47,  15,  16, 183,  34, 223,  49,  45, 183},
-    { /* left mode 6 */  46,  17,  33, 183,   6,  98,  15,  32, 183},
-    { /* left mode 7 */  65,  32,  73, 115,  28, 128,  23, 128, 205},
-    { /* left mode 8 */  40,   3, 9, 115,  51, 192,  18,   6, 223},
-    { /* left mode 9 */  87,  37, 9, 115,  59,  77,  64,  21,  47}
-  },
-  { /* above mode 6 */
-    { /* left mode 0 */ 104,  55,  44, 218,   9,  54,  53, 130, 226},
-    { /* left mode 1 */  64,  90,  70, 205,  40,  41,  23,  26,  57},
-    { /* left mode 2 */  54,  57, 112, 184,   5,  41,  38, 166, 213},
-    { /* left mode 3 */  30,  34,  26, 133, 152, 116,  10,  32, 134},
-    { /* left mode 4 */  75,  32,  12,  51, 192, 255, 160,  43,  51},
-    { /* left mode 5 */  39,  19,  53, 221,  26, 114,  32,  73, 255},
-    { /* left mode 6 */  31,   9,  65, 234,   2,  15, 1, 118,  73},
-    { /* left mode 7 */  88,  31,  35,  67, 102,  85,  55, 186,  85},
-    { /* left mode 8 */  56,  21,  23, 111,  59, 205,  45,  37, 192},
-    { /* left mode 9 */  55,  38,  70, 124,  73, 102, 1,  34,  98}
-  },
-  { /* above mode 7 */
-    { /* left mode 0 */ 102,  61,  71,  37,  34,  53,  31, 243, 192},
-    { /* left mode 1 */  69,  60,  71,  38,  73, 119,  28, 222,  37},
-    { /* left mode 2 */  68,  45, 128,  34,   1,  47,  11, 245, 171},
-    { /* left mode 3 */  62,  17,  19,  70, 146,  85,  55,  62,  70},
-    { /* left mode 4 */  75,  15, 9,   9,  64, 255, 184, 119,  16},
-    { /* left mode 5 */  37,  43,  37, 154, 100, 163,  85, 160,   1},
-    { /* left mode 6 */  63,   9,  92, 136,  28,  64,  32, 201,  85},
-    { /* left mode 7 */  86,   6,  28,   5,  64, 255,  25, 248,   1},
-    { /* left mode 8 */  56,   8,  17, 132, 137, 255,  55, 116, 128},
-    { /* left mode 9 */  58,  15,  20,  82, 135,  57,  26, 121,  40}
-  },
-  { /* above mode 8 */
-    { /* left mode 0 */ 164,  50,  31, 137, 154, 133,  25,  35, 218},
-    { /* left mode 1 */  51, 103,  44, 131, 131, 123,  31,   6, 158},
-    { /* left mode 2 */  86,  40,  64, 135, 148, 224,  45, 183, 128},
-    { /* left mode 3 */  22,  26,  17, 131, 240, 154,  14,   1, 209},
-    { /* left mode 4 */  83,  12,  13,  54, 192, 255,  68,  47,  28},
-    { /* left mode 5 */  45,  16,  21,  91,  64, 222, 7,   1, 197},
-    { /* left mode 6 */  56,  21,  39, 155,  60, 138,  23, 102, 213},
-    { /* left mode 7 */  85,  26,  85,  85, 128, 128,  32, 146, 171},
-    { /* left mode 8 */  18,  11, 7,  63, 144, 171, 4,   4, 246},
-    { /* left mode 9 */  35,  27,  10, 146, 174, 171,  12,  26, 128}
-  },
-  { /* above mode 9 */
-    { /* left mode 0 */ 190,  80,  35,  99, 180,  80, 126,  54,  45},
-    { /* left mode 1 */  85, 126,  47,  87, 176,  51,  41,  20,  32},
-    { /* left mode 2 */ 101,  75, 128, 139, 118, 146, 116, 128,  85},
-    { /* left mode 3 */  56,  41,  15, 176, 236,  85,  37,   9,  62},
-    { /* left mode 4 */ 146,  36,  19,  30, 171, 255,  97,  27,  20},
-    { /* left mode 5 */  71,  30,  17, 119, 118, 255,  17,  18, 138},
-    { /* left mode 6 */ 101,  38,  60, 138,  55,  70,  43,  26, 142},
-    { /* left mode 7 */ 138,  45,  61,  62, 219,   1,  81, 188,  64},
-    { /* left mode 8 */  32,  41,  20, 117, 151, 142,  20,  21, 163},
-    { /* left mode 9 */ 112,  19,  12,  61, 195, 128,  48,   4,  24}
-  }
-};
-static const int kf_y_mode_tree[] =
-{
-  -B_PRED, 2,
-  4, 6,
-  -DC_PRED, -V_PRED,
-  -H_PRED, -TM_PRED
-};
-static const int y_mode_tree[] =
-{
-  -DC_PRED, 2,
-  4, 6,
-  -V_PRED, -H_PRED,
-  -TM_PRED, -B_PRED
-};
-static const int uv_mode_tree[6] =
-{
-  -DC_PRED, 2,
-  -V_PRED, 4,
-  -H_PRED, -TM_PRED
-};
-static const int b_mode_tree[18] =
-{
-  -B_DC_PRED, 2,                 /* 0 = DC_NODE */
-  -B_TM_PRED, 4,                /* 1 = TM_NODE */
-  -B_VE_PRED, 6,               /* 2 = VE_NODE */
-  8, 12,                  /* 3 = COM_NODE */
-  -B_HE_PRED, 10,              /* 4 = HE_NODE */
-  -B_RD_PRED, -B_VR_PRED,         /* 5 = RD_NODE */
-  -B_LD_PRED, 14,              /* 6 = LD_NODE */
-  -B_VL_PRED, 16,            /* 7 = VL_NODE */
-  -B_HD_PRED, -B_HU_PRED         /* 8 = HD_NODE */
-};
-static const int small_mv_tree[14] =
-{
-  2, 8,
-  4, 6,
-  -0, -1,
-  -2, -3,
-  10, 12,
-  -4, -5,
-  -6, -7
-};
-static const int mv_ref_tree[8] =
-{
-  -ZEROMV, 2,
-  -NEARESTMV, 4,
-  -NEARMV, 6,
-  -NEWMV, -SPLITMV
-};
-static const int submv_ref_tree[6] =
-{
-  -LEFT4X4, 2,
-  -ABOVE4X4, 4,
-  -ZERO4X4, -NEW4X4
-};
-static const int split_mv_tree[6] =
-{
-  -3, 2,
-  -2, 4,
-  -0, -1
-};
-static const unsigned char default_b_mode_probs[] =
-{ 120,  90,  79, 133,  87,  85,  80, 111, 151};
-static const unsigned char mv_counts_to_probs[6][4] =
-{
-  { 7,   1,   1, 143 },
-  {  14,  18,  14, 107 },
-  { 135,  64,  57,  68 },
-  {  60,  56, 128,  65 },
-  { 159, 134, 128,  34 },
-  { 234, 188, 128,  28 }
-
-};
-static const unsigned char split_mv_probs[3] =
-{ 110, 111, 150};
-static const unsigned char submv_ref_probs2[5][3] =
-{
-  { 147, 136, 18 },
-  { 106, 145,  1 },
-  { 179, 121,  1 },
-  { 223,   1, 34 },
-  { 208,   1,  1 }
-};
-
-const static int mv_partitions[4][16] =
-{
-  {0, 0, 0, 0, 0, 0, 0, 0, 1, 1,  1,  1,  1,  1,  1,  1 },
-  {0, 0, 1, 1, 0, 0, 1, 1, 0, 0,  1,  1,  0,  0,  1,  1 },
-  {0, 0, 1, 1, 0, 0, 1, 1, 2, 2,  3,  3,  2,  2,  3,  3 },
-  {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }
-};

vp8/dixie/predict.h

@@ -1,36 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef PREDICT_H
-#define PREDICT_H
-
-void
-vp8_dixie_predict_init(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_predict_destroy(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_predict_process_row(struct vp8_decoder_ctx *ctx,
-                              unsigned int            row,
-                              unsigned int            start_col,
-                              unsigned int            num_cols);
-
-void
-vp8_dixie_release_ref_frame(struct ref_cnt_img *rcimg);
-
-struct ref_cnt_img *
-vp8_dixie_ref_frame(struct ref_cnt_img *rcimg);
-
-struct ref_cnt_img *
-vp8_dixie_find_free_ref_frame(struct ref_cnt_img *frames);
-
-#endif

vp8/dixie/tokens.c

@@ -1,443 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#include "vpx/internal/vpx_codec_internal.h"
-#include "dixie.h"
-#include "tokens.h"
-#include <stdlib.h>
-#include <string.h>
-#include <malloc.h>
-
-
-enum
-{
-    EOB_CONTEXT_NODE,
-    ZERO_CONTEXT_NODE,
-    ONE_CONTEXT_NODE,
-    LOW_VAL_CONTEXT_NODE,
-    TWO_CONTEXT_NODE,
-    THREE_CONTEXT_NODE,
-    HIGH_LOW_CONTEXT_NODE,
-    CAT_ONE_CONTEXT_NODE,
-    CAT_THREEFOUR_CONTEXT_NODE,
-    CAT_THREE_CONTEXT_NODE,
-    CAT_FIVE_CONTEXT_NODE
-};
-enum
-{
-    ZERO_TOKEN,
-    ONE_TOKEN,
-    TWO_TOKEN,
-    THREE_TOKEN,
-    FOUR_TOKEN,
-    DCT_VAL_CATEGORY1,
-    DCT_VAL_CATEGORY2,
-    DCT_VAL_CATEGORY3,
-    DCT_VAL_CATEGORY4,
-    DCT_VAL_CATEGORY5,
-    DCT_VAL_CATEGORY6,
-    DCT_EOB_TOKEN,
-    MAX_ENTROPY_TOKENS
-};
-struct extrabits
-{
-    short         min_val;
-    short         length;
-    unsigned char probs[12];
-};
-static const unsigned int left_context_index[25] =
-{
-    0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3,
-    4, 4, 5, 5, 6, 6, 7, 7, 8
-};
-static const unsigned int above_context_index[25] =
-{
-    0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3,
-    4, 5, 4, 5, 6, 7, 6, 7, 8
-};
-#define X(n) ((n) * PREV_COEF_CONTEXTS * ENTROPY_NODES)
-static const unsigned int bands_x[16] =
-{
-    X(0), X(1), X(2), X(3), X(6), X(4), X(5), X(6),
-    X(6), X(6), X(6), X(6), X(6), X(6), X(6), X(7)
-};
-#undef X
-static const struct extrabits extrabits[MAX_ENTROPY_TOKENS] =
-{
-    {  0, -1, {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //ZERO_TOKEN
-    {  1, 0,  {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //ONE_TOKEN
-    {  2, 0,  {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //TWO_TOKEN
-    {  3, 0,  {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //THREE_TOKEN
-    {  4, 0,  {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //FOUR_TOKEN
-    {  5, 0,  { 159,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //DCT_VAL_CATEGORY1
-    {  7, 1,  { 145, 165,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //DCT_VAL_CATEGORY2
-    { 11, 2,  { 140, 148, 173,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //DCT_VAL_CATEGORY3
-    { 19, 3,  { 135, 140, 155, 176,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, //DCT_VAL_CATEGORY4
-    { 35, 4,  { 130, 134, 141, 157, 180,   0,
-                  0,   0,   0,   0,   0,   0   } }, //DCT_VAL_CATEGORY5
-    { 67, 10, { 129, 130, 133, 140, 153, 177,
-                196, 230, 243, 254, 254,   0   } }, //DCT_VAL_CATEGORY6
-    {  0, -1, {   0,   0,   0,   0,   0,   0,
-                  0,   0,   0,   0,   0,   0   } }, // EOB TOKEN
-};
-static const unsigned int zigzag[16] =
-{
-    0,  1,  4,  8,  5,  2,  3,  6,  9, 12, 13, 10,  7, 11, 14, 15
-};
-
-#define DECODE_AND_APPLYSIGN(value_to_sign) \
-    v = (bool_get_bit(bool) ? -value_to_sign \
-                            : value_to_sign) * dqf[!!c];
-
-#define DECODE_AND_BRANCH_IF_ZERO(probability,branch) \
-    if (!bool_get(bool, probability)) goto branch;
-
-#define DECODE_AND_LOOP_IF_ZERO(probability,branch) \
-    if (!bool_get(bool, probability)) \
-    { \
-        prob = type_probs; \
-        if(c<15) {\
-            ++c; \
-            prob += bands_x[c]; \
-            goto branch; \
-        }\
-        else \
-            goto BLOCK_FINISHED; /*for malformed input */\
-    }
-
-#define DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val) \
-    DECODE_AND_APPLYSIGN(val) \
-    prob = type_probs + (ENTROPY_NODES*2); \
-    if(c < 15){\
-        b_tokens[zigzag[c]] = v; \
-        ++c; \
-        goto DO_WHILE; }\
-    b_tokens[zigzag[15]] = v; \
-    goto BLOCK_FINISHED;
-
-
-#define DECODE_EXTRABIT_AND_ADJUST_VAL(t,bits_count)\
-    val += bool_get(bool, extrabits[t].probs[bits_count]) << bits_count;
-
-
-static int
-decode_mb_tokens(struct bool_decoder  *bool,
-                 token_entropy_ctx_t   left,
-                 token_entropy_ctx_t   above,
-                 short                *tokens,
-                 enum prediction_mode  mode,
-                 coeff_probs_table_t   probs,
-                 short                 factor[TOKEN_BLOCK_TYPES][2])
-{
-    int            i, stop, type;
-    int            c, t, v;
-    int            val, bits_count;
-    int            eob_mask;
-    short         *b_tokens;   /* tokens for this block */
-    unsigned char *type_probs; /* probabilities for this block type */
-    unsigned char *prob;
-    short         *dqf;
-
-    eob_mask = 0;
-
-    if (mode != B_PRED && mode != SPLITMV)
-    {
-        i = 24;
-        stop = 24;
-        type = 1;
-        b_tokens = tokens + 24 * 16;
-        dqf = factor[TOKEN_BLOCK_Y2];
-    }
-    else
-    {
-        i = 0;
-        stop = 16;
-        type = 3;
-        b_tokens = tokens;
-        dqf = factor[TOKEN_BLOCK_Y1];
-    }
-
-    /* Save a pointer to the coefficient probs for the current type.
-     * Need to repeat this whenever type changes.
-     */
-    type_probs = probs[type][0][0];
-
-BLOCK_LOOP:
-    t = left[left_context_index[i]] + above[above_context_index[i]];
-    c = !type; /* all blocks start at 0 except type 0, which starts
-                * at 1. */
-
-    prob = type_probs;
-    prob += t * ENTROPY_NODES;
-
-DO_WHILE:
-    prob += bands_x[c];
-    DECODE_AND_BRANCH_IF_ZERO(prob[EOB_CONTEXT_NODE], BLOCK_FINISHED);
-
-CHECK_0_:
-    DECODE_AND_LOOP_IF_ZERO(prob[ZERO_CONTEXT_NODE], CHECK_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[ONE_CONTEXT_NODE],
-                              ONE_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[LOW_VAL_CONTEXT_NODE],
-                              LOW_VAL_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[HIGH_LOW_CONTEXT_NODE],
-                              HIGH_LOW_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[CAT_THREEFOUR_CONTEXT_NODE],
-                              CAT_THREEFOUR_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[CAT_FIVE_CONTEXT_NODE],
-                              CAT_FIVE_CONTEXT_NODE_0_);
-    val = extrabits[DCT_VAL_CATEGORY6].min_val;
-    bits_count = extrabits[DCT_VAL_CATEGORY6].length;
-
-    do
-    {
-        DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY6, bits_count);
-        bits_count -- ;
-    }
-    while (bits_count >= 0);
-
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_FIVE_CONTEXT_NODE_0_:
-    val = extrabits[DCT_VAL_CATEGORY5].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 4);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 3);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY5, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_THREEFOUR_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(prob[CAT_THREE_CONTEXT_NODE],
-                              CAT_THREE_CONTEXT_NODE_0_);
-    val = extrabits[DCT_VAL_CATEGORY4].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 3);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY4, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_THREE_CONTEXT_NODE_0_:
-    val = extrabits[DCT_VAL_CATEGORY3].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 2);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY3, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-HIGH_LOW_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(prob[CAT_ONE_CONTEXT_NODE],
-                              CAT_ONE_CONTEXT_NODE_0_);
-
-    val = extrabits[DCT_VAL_CATEGORY2].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY2, 1);
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY2, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-CAT_ONE_CONTEXT_NODE_0_:
-    val = extrabits[DCT_VAL_CATEGORY1].min_val;
-    DECODE_EXTRABIT_AND_ADJUST_VAL(DCT_VAL_CATEGORY1, 0);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val);
-
-LOW_VAL_CONTEXT_NODE_0_:
-    DECODE_AND_BRANCH_IF_ZERO(prob[TWO_CONTEXT_NODE],
-                              TWO_CONTEXT_NODE_0_);
-    DECODE_AND_BRANCH_IF_ZERO(prob[THREE_CONTEXT_NODE],
-                              THREE_CONTEXT_NODE_0_);
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(4);
-
-THREE_CONTEXT_NODE_0_:
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(3);
-
-TWO_CONTEXT_NODE_0_:
-    DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(2);
-
-ONE_CONTEXT_NODE_0_:
-    DECODE_AND_APPLYSIGN(1);
-    prob = type_probs + ENTROPY_NODES;
-
-    if (c < 15)
-    {
-        b_tokens[zigzag[c]] = v;
-        ++c;
-        goto DO_WHILE;
-    }
-
-    b_tokens[zigzag[15]] = v;
-BLOCK_FINISHED:
-    eob_mask |= (c > 1) << i;
-    t = (c != !type);   // any nonzero data?
-    eob_mask |= t << 31;
-
-    left[left_context_index[i]] = above[above_context_index[i]] = t;
-    b_tokens += 16;
-
-    i++;
-
-    if (i < stop)
-        goto BLOCK_LOOP;
-
-    if (i == 25)
-    {
-        type = 0;
-        i = 0;
-        stop = 16;
-        type_probs = probs[type][0][0];
-        b_tokens = tokens;
-        dqf = factor[TOKEN_BLOCK_Y1];
-        goto BLOCK_LOOP;
-    }
-
-    if (i == 16)
-    {
-        type = 2;
-        type_probs = probs[type][0][0];
-        stop = 24;
-        dqf = factor[TOKEN_BLOCK_UV];
-        goto BLOCK_LOOP;
-    }
-
-    return eob_mask;
-}
-
-
-static void
-reset_row_context(token_entropy_ctx_t *left)
-{
-    memset(left, 0, sizeof(*left));
-}
-
-
-static void
-reset_above_context(token_entropy_ctx_t *above, unsigned int cols)
-{
-    memset(above, 0, cols * sizeof(*above));
-}
-
-
-static void
-reset_mb_context(token_entropy_ctx_t  *left,
-                 token_entropy_ctx_t  *above,
-                 enum prediction_mode  mode)
-{
-    /* Reset the macroblock context on the left and right. We have to
-     * preserve the context of the second order block if this mode
-     * would not have updated it.
-     */
-    memset(left, 0, sizeof((*left)[0]) * 8);
-    memset(above, 0, sizeof((*above)[0]) * 8);
-
-    if (mode != B_PRED && mode != SPLITMV)
-    {
-        (*left)[8] = 0;
-        (*above)[8] = 0;
-    }
-}
-
-
-void
-vp8_dixie_tokens_process_row(struct vp8_decoder_ctx *ctx,
-                             unsigned int            partition,
-                             unsigned int            row,
-                             unsigned int            start_col,
-                             unsigned int            num_cols)
-{
-    struct token_decoder *tokens = &ctx->tokens[partition];
-    short                *coeffs = tokens->coeffs + 25 * 16 * start_col;
-    unsigned int          col;
-    token_entropy_ctx_t  *above = ctx->above_token_entropy_ctx
-                                  + start_col;
-    token_entropy_ctx_t  *left = &tokens->left_token_entropy_ctx;
-    struct mb_info       *mbi = ctx->mb_info_rows[row] + start_col;
-
-    if (row == 0)
-        reset_above_context(above, num_cols);
-
-    if (start_col == 0)
-        reset_row_context(left);
-
-    for (col = start_col; col < start_col + num_cols; col++)
-    {
-        memset(coeffs, 0, 25 * 16 * sizeof(short));
-
-        if (mbi->base.skip_coeff)
-        {
-            reset_mb_context(left, above, mbi->base.y_mode);
-            mbi->base.eob_mask = 0;
-        }
-        else
-        {
-            struct dequant_factors *dqf;
-
-            dqf = ctx->dequant_factors  + mbi->base.segment_id;
-            mbi->base.eob_mask =
-                decode_mb_tokens(&tokens->bool,
-                                 *left, *above,
-                                 coeffs,
-                                 mbi->base.y_mode,
-                                 ctx->entropy_hdr.coeff_probs,
-                                 dqf->factor);
-        }
-
-        above++;
-        mbi++;
-        coeffs += 25 * 16;
-    }
-}
-
-
-void
-vp8_dixie_tokens_init(struct vp8_decoder_ctx *ctx)
-{
-    unsigned int  partitions = ctx->token_hdr.partitions;
-
-    if (ctx->frame_hdr.frame_size_updated)
-    {
-        unsigned int i;
-        unsigned int coeff_row_sz =
-            ctx->mb_cols * 25 * 16 * sizeof(short);
-
-        for (i = 0; i < partitions; i++)
-        {
-            free(ctx->tokens[i].coeffs);
-            ctx->tokens[i].coeffs = memalign(16, coeff_row_sz);
-
-            if (!ctx->tokens[i].coeffs)
-                vpx_internal_error(&ctx->error, VPX_CODEC_MEM_ERROR,
-                                   NULL);
-        }
-
-        free(ctx->above_token_entropy_ctx);
-        ctx->above_token_entropy_ctx =
-            calloc(ctx->mb_cols, sizeof(*ctx->above_token_entropy_ctx));
-
-        if (!ctx->above_token_entropy_ctx)
-            vpx_internal_error(&ctx->error, VPX_CODEC_MEM_ERROR, NULL);
-    }
-}
-
-
-void
-vp8_dixie_tokens_destroy(struct vp8_decoder_ctx *ctx)
-{
-    int i;
-
-    for (i = 0; i < MAX_PARTITIONS; i++)
-        free(ctx->tokens[i].coeffs);
-
-    free(ctx->above_token_entropy_ctx);
-}

vp8/dixie/tokens.h

@@ -1,28 +0,0 @@
-/*
- *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-#ifndef TOKENS_H
-#define TOKENS_H
-
-void
-vp8_dixie_tokens_init(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_tokens_destroy(struct vp8_decoder_ctx *ctx);
-
-
-void
-vp8_dixie_tokens_process_row(struct vp8_decoder_ctx *ctx,
-                             unsigned int            partition,
-                             unsigned int            row,
-                             unsigned int            start_col,
-                             unsigned int            num_cols);
-
-#endif