Update CHANGELOG for v0.9.7-p1

Change-Id: I5490a9cad2d6752832b6bf4ec1835c06a45eeb9b

.mailmap

@@ -2,3 +2,4 @@ Adrian Grange <agrange@google.com>
 Johann Koenig <johannkoenig@google.com>
 Tero Rintaluoma <teror@google.com> <tero.rintaluoma@on2.com>
 Tom Finegan <tomfinegan@google.com>
+Ralph Giles <giles@xiph.org> <giles@entropywave.com>

AUTHORS

@@ -4,8 +4,11 @@
 Aaron Watry <awatry@gmail.com>
 Adrian Grange <agrange@google.com>
 Alex Converse <alex.converse@gmail.com>
+Alexis Ballier <aballier@gentoo.org>
+Alok Ahuja <waveletcoeff@gmail.com>
 Andoni Morales Alastruey <ylatuya@gmail.com>
 Andres Mejia <mcitadel@gmail.com>
+Aron Rosenberg <arosenberg@logitech.com>
 Attila Nagy <attilanagy@google.com>
 Fabio Pedretti <fabio.ped@libero.it>
 Frank Galligan <fgalligan@google.com>
@@ -22,20 +25,29 @@ Jeff Muizelaar <jmuizelaar@mozilla.com>
 Jim Bankoski <jimbankoski@google.com>
 Johann Koenig <johannkoenig@google.com>
 John Koleszar <jkoleszar@google.com>
+Joshua Bleecher Snyder <josh@treelinelabs.com>
 Justin Clift <justin@salasaga.org>
 Justin Lebar <justin.lebar@gmail.com>
+Lou Quillio <louquillio@google.com>
 Luca Barbato <lu_zero@gentoo.org>
 Makoto Kato <makoto.kt@gmail.com>
 Martin Ettl <ettl.martin78@googlemail.com>
 Michael Kohler <michaelkohler@live.com>
+Mike Hommey <mhommey@mozilla.com>
 Mikhal Shemer <mikhal@google.com>
 Pascal Massimino <pascal.massimino@gmail.com>
 Patrik Westin <patrik.westin@gmail.com>
 Paul Wilkins <paulwilkins@google.com>
 Pavol Rusnak <stick@gk2.sk>
 Philip Jägenstedt <philipj@opera.com>
+Rafael Ávila de Espíndola <rafael.espindola@gmail.com>
+Ralph Giles <giles@xiph.org>
+Ronald S. Bultje <rbultje@google.com>
 Scott LaVarnway <slavarnway@google.com>
+Stefan Holmer <holmer@google.com>
+Taekhyun Kim <takim@nvidia.com>
 Tero Rintaluoma <teror@google.com>
+Thijs Vermeir <thijsvermeir@gmail.com>
 Timothy B. Terriberry <tterribe@xiph.org>
 Tom Finegan <tomfinegan@google.com>
 Yaowu Xu <yaowu@google.com>

CHANGELOG

@@ -1,3 +1,115 @@
+2011-08-15 v0.9.7-p1 "Cayuga" patch 1
+  This is an incremental bugfix release against Cayuga. All users of that
+  release are strongly encouraged to upgrade.
+
+    - Fix potential OOB reads (cdae03a)
+
+          An unbounded out of bounds read was discovered when the
+          decoder was requested to perform error concealment (new in
+          Cayuga) given a frame with corrupt partition sizes.
+
+          A bounded out of bounds read was discovered affecting all
+          versions of libvpx. Given an multipartition input frame that
+          is truncated between the mode/mv partition and the first
+          residiual paritition (in the block of partition offsets), up
+          to 3 extra bytes could have been read from the source buffer.
+          The code will not take any action regardless of the contents
+          of these undefined bytes, as the truncated buffer is detected
+          immediately following the read based on the calculated
+          starting position of the coefficient partition.
+
+    - Fix potential error concealment crash when the very first frame
+      is missing or corrupt (a609be5)
+
+    - Fix significant artifacts in error concealment (a4c2211, 99d870a)
+
+    - Revert 1-pass CBR rate control changes (e961317)
+      Further testing showed this change produced undesirable visual
+      artifacts, rolling back for now.
+
+
+2011-08-02 v0.9.7 "Cayuga"
+  Our third named release, focused on a faster, higher quality, encoder.
+
+  - Upgrading:
+    This release is backwards compatible with Aylesbury (v0.9.5) and
+    Bali (v0.9.6). Users of older releases should refer to the Upgrading
+    notes in this document for that release.
+
+  - Enhancements:
+          Stereo 3D format support for vpxenc
+          Runtime detection of available processor cores.
+          Allow specifying --end-usage by enum name
+          vpxdec: test for frame corruption
+          vpxenc: add quantizer histogram display
+          vpxenc: add rate histogram display
+          Set VPX_FRAME_IS_DROPPABLE
+          update configure for ios sdk 4.3
+          Avoid text relocations in ARM vp8 decoder
+          Generate a vpx.pc file for pkg-config.
+          New ways of passing encoded data between encoder and decoder.
+
+  - Speed:
+      This release includes across-the-board speed improvements to the
+      encoder. On x86, these measure at approximately 11.5% in Best mode,
+      21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
+      On ARM Cortex A9 with Neon extensions, real-time encoding of video
+      telephony content is 35% faster than Bali on single core and 48%
+      faster on multi-core. On the NVidia Tegra2 platform, real time
+      encoding is 40% faster than Bali.
+
+      Decoder speed was not a priority for this release, but improved
+      approximately 8.4% on x86.
+
+          Reduce motion vector search on alt-ref frame.
+          Encoder loopfilter running in its own thread
+          Reworked loopfilter to precalculate more parameters
+          SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
+          Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
+          Removed redundant checks
+          Reduced structure sizes
+          utilize preload in ARMv6 MC/LPF/Copy routines
+          ARM optimized quantization, dfct, variance, subtract
+          Increase chrow row alignment to 16 bytes.
+          disable trellis optimization for first pass
+          Write SSSE3 sub-pixel filter function
+          Improve SSE2 half-pixel filter funtions
+          Add vp8_sub_pixel_variance16x8_ssse3 function
+          Reduce unnecessary distortion computation
+          Use diamond search to replace full search
+          Preload reference area in sub-pixel motion search (real-time mode)
+
+  - Quality:
+      This release focused primarily on one-pass use cases, including
+      video conferencing. Low latency data rate control was significantly
+      improved, improving streamability over bandwidth constrained links.
+      Added support for error concealment, allowing frames to maintain
+      visual quality in the presence of substantial packet loss.
+
+          Add rc_max_intra_bitrate_pct control
+          Limit size of initial keyframe in one-pass.
+          Improve framerate adaptation
+          Improved 1-pass CBR rate control
+          Improved KF insertion after fades to still.
+          Improved key frame detection.
+          Improved activity masking (lower PSNR impact for same SSIM boost)
+          Improved interaction between GF and ARFs
+          Adding error-concealment to the decoder.
+          Adding support for independent partitions
+          Adjusted rate-distortion constants
+
+
+  - Bug Fixes:
+          Removed firstpass motion map
+          Fix parallel make install
+          Fix multithreaded encoding for 1 MB wide frame
+          Fixed iwalsh_neon build problems with RVDS4.1
+          Fix semaphore emulation, spin-wait intrinsics on Windows
+          Fix build with xcode4 and simplify GLOBAL.
+          Mark ARM asm objects as allowing a non-executable stack.
+          Fix vpxenc encoding incorrect webm file header on big endian
+
+
 2011-03-07 v0.9.6 "Bali"
   Our second named release, focused on a faster, higher quality, encoder.
 

build/make/Makefile

@@ -98,11 +98,11 @@ install::
 $(BUILD_PFX)%.c.d: %.c
 	$(if $(quiet),@echo "    [DEP] $@")
 	$(qexec)mkdir -p $(dir $@)
-	$(qexec)$(CC) $(CFLAGS) -M $< | $(fmt_deps) > $@
+	$(qexec)$(CC) $(INTERNAL_CFLAGS) $(CFLAGS) -M $< | $(fmt_deps) > $@
 
 $(BUILD_PFX)%.c.o: %.c
 	$(if $(quiet),@echo "    [CC] $@")
-	$(qexec)$(CC) $(CFLAGS) -c -o $@ $<
+	$(qexec)$(CC) $(INTERNAL_CFLAGS) $(CFLAGS) -c -o $@ $<
 
 $(BUILD_PFX)%.asm.d: %.asm
 	$(if $(quiet),@echo "    [DEP] $@")
@@ -124,6 +124,12 @@ $(BUILD_PFX)%.s.o: %.s
 	$(if $(quiet),@echo "    [AS] $@")
 	$(qexec)$(AS) $(ASFLAGS) -o $@ $<
 
+.PRECIOUS: %.c.S
+%.c.S: CFLAGS += -DINLINE_ASM
+$(BUILD_PFX)%.c.S: %.c
+	$(if $(quiet),@echo "    [GEN] $@")
+	$(qexec)$(CC) -S $(CFLAGS) -o $@ $<
+
 .PRECIOUS: %.asm.s
 $(BUILD_PFX)%.asm.s: %.asm
 	$(if $(quiet),@echo "    [ASM CONVERSION] $@")
@@ -152,8 +158,8 @@ endif
 # Rule to extract assembly constants from C sources
 #
 obj_int_extract: build/make/obj_int_extract.c
-	$(if $(quiet),echo "    [HOSTCC] $@")
-	$(qexec)$(HOSTCC) -I. -o $@ $<
+	$(if $(quiet),@echo "    [HOSTCC] $@")
+	$(qexec)$(HOSTCC) -I. -I$(SRC_PATH_BARE) -o $@ $<
 CLEAN-OBJS += obj_int_extract
 
 #
@@ -188,7 +194,7 @@ define linker_template
 $(1): $(filter-out -%,$(2))
 $(1):
 	$(if $(quiet),@echo    "    [LD] $$@")
-	$(qexec)$$(LD) $$(strip $$(LDFLAGS) -o $$@ $(2) $(3) $$(extralibs))
+	$(qexec)$$(LD) $$(strip $$(INTERNAL_LDFLAGS) $$(LDFLAGS) -o $$@ $(2) $(3) $$(extralibs))
 endef
 # make-3.80 has a bug with expanding large input strings to the eval function,
 # which was triggered in some cases by the following component of
@@ -255,7 +261,7 @@ ifeq ($(filter clean,$(MAKECMDGOALS)),)
 endif
 
 #
-# Configuration dependant rules
+# Configuration dependent rules
 #
 $(call pairmap,install_map_templates,$(INSTALL_MAPS))
 
@@ -330,12 +336,10 @@ ifneq ($(call enabled,DIST-SRCS),)
     DIST-SRCS-$(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-$(CONFIG_MSVS)  += build/x86-msvs/obj_int_extract.bat
     DIST-SRCS-$(CONFIG_RVCT) += build/make/armlink_adapter.sh
-    #
-    # This isn't really ARCH_ARM dependent, it's dependant on whether we're
-    # using assembly code or not (CONFIG_OPTIMIZATIONS maybe). Just use
-    # this for now.
-    DIST-SRCS-$(ARCH_ARM)    += build/make/obj_int_extract.c
+    # Include obj_int_extract if we use offsets from asm_*_offsets
+    DIST-SRCS-$(ARCH_ARM)$(ARCH_X86)$(ARCH_X86_64)    += build/make/obj_int_extract.c
     DIST-SRCS-$(ARCH_ARM)    += build/make/ads2gas.pl
     DIST-SRCS-yes            += $(target:-$(TOOLCHAIN)=).mk
 endif

build/make/ads2gas.pl

@@ -21,8 +21,14 @@ print "@ This file was created from a .asm file\n";
 print "@  using the ads2gas.pl script.\n";
 print "\t.equ DO1STROUNDING, 0\n";
 
+# Stack of procedure names.
+@proc_stack = ();
+
 while (<STDIN>)
 {
+    # Load and store alignment
+    s/@/,:/g;
+
     # Comment character
     s/;/@/g;
 
@@ -79,7 +85,10 @@ while (<STDIN>)
     s/CODE([0-9][0-9])/.code $1/;
 
     # No AREA required
-    s/^\s*AREA.*$/.text/;
+    # But ALIGNs in AREA must be obeyed
+    s/^\s*AREA.*ALIGN=([0-9])$/.text\n.p2align $1/;
+    # If no ALIGN, strip the AREA and align to 4 bytes
+    s/^\s*AREA.*$/.text\n.p2align 2/;
 
     # DCD to .word
     # This one is for incoming symbols
@@ -114,8 +123,8 @@ while (<STDIN>)
     # put the colon at the end of the line in the macro
     s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
 
-    # Strip ALIGN
-    s/\sALIGN/@ ALIGN/g;
+    # ALIGN directive
+    s/ALIGN/.balign/g;
 
     # Strip ARM
     s/\sARM/@ ARM/g;
@@ -127,9 +136,23 @@ while (<STDIN>)
     # Strip PRESERVE8
     s/\sPRESERVE8/@ PRESERVE8/g;
 
-    # Strip PROC and ENDPROC
-    s/\sPROC/@/g;
-    s/\sENDP/@/g;
+    # Use PROC and ENDP to give the symbols a .size directive.
+    # This makes them show up properly in debugging tools like gdb and valgrind.
+    if (/\bPROC\b/)
+    {
+        my $proc;
+        /^_([\.0-9A-Z_a-z]\w+)\b/;
+        $proc = $1;
+        push(@proc_stack, $proc) if ($proc);
+        s/\bPROC\b/@ $&/;
+    }
+    if (/\bENDP\b/)
+    {
+        my $proc;
+        s/\bENDP\b/@ $&/;
+        $proc = pop(@proc_stack);
+        $_ = "\t.size $proc, .-$proc".$_ if ($proc);
+    }
 
     # EQU directive
     s/(.*)EQU(.*)/.equ $1, $2/;
@@ -148,3 +171,6 @@ while (<STDIN>)
     next if /^\s*END\s*$/;
     print;
 }
+
+# Mark that this object doesn't need an executable stack.
+printf ("\t.section\t.note.GNU-stack,\"\",\%\%progbits\n");

build/make/ads2gas_apple.pl

@@ -41,6 +41,9 @@ sub trim($)
 
 while (<STDIN>)
 {
+    # Load and store alignment
+    s/@/,:/g;
+
     # Comment character
     s/;/@/g;
 
@@ -97,7 +100,10 @@ while (<STDIN>)
     s/CODE([0-9][0-9])/.code $1/;
 
     # No AREA required
-    s/^\s*AREA.*$/.text/;
+    # But ALIGNs in AREA must be obeyed
+    s/^\s*AREA.*ALIGN=([0-9])$/.text\n.p2align $1/;
+    # If no ALIGN, strip the AREA and align to 4 bytes
+    s/^\s*AREA.*$/.text\n.p2align 2/;
 
     # DCD to .word
     # This one is for incoming symbols
@@ -137,8 +143,8 @@ while (<STDIN>)
     # put the colon at the end of the line in the macro
     s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
 
-    # Strip ALIGN
-    s/\sALIGN/@ ALIGN/g;
+    # ALIGN directive
+    s/ALIGN/.balign/g;
 
     # Strip ARM
     s/\sARM/@ ARM/g;

build/make/configure.sh

@@ -83,7 +83,7 @@ Build options:
   ${toggle_werror}            treat warnings as errors, if possible
                               (not available with all compilers)
   ${toggle_optimizations}     turn on/off compiler optimization flags
-  ${toggle_pic}               turn on/off Position Independant Code
+  ${toggle_pic}               turn on/off Position Independent Code
   ${toggle_ccache}            turn on/off compiler cache
   ${toggle_debug}             enable/disable debug mode
   ${toggle_gprof}             enable/disable gprof profiling instrumentation
@@ -412,11 +412,14 @@ EOF
 write_common_target_config_h() {
     cat > ${TMP_H} << EOF
 /* This file automatically generated by configure. Do not edit! */
+#ifndef VPX_CONFIG_H
+#define VPX_CONFIG_H
 #define RESTRICT    ${RESTRICT}
 EOF
     print_config_h ARCH   "${TMP_H}" ${ARCH_LIST}
     print_config_h HAVE   "${TMP_H}" ${HAVE_LIST}
     print_config_h CONFIG "${TMP_H}" ${CONFIG_LIST}
+    echo "#endif /* VPX_CONFIG_H */" >> ${TMP_H}
     mkdir -p `dirname "$1"`
     cmp "$1" ${TMP_H} >/dev/null 2>&1 || mv ${TMP_H} "$1"
 }
@@ -626,7 +629,7 @@ process_common_toolchain() {
     case ${toolchain} in
         sparc-solaris-*)
             add_extralibs -lposix4
-            add_cflags "-DMUST_BE_ALIGNED"
+            disable fast_unaligned
             ;;
         *-solaris-*)
             add_extralibs -lposix4
@@ -639,8 +642,8 @@ process_common_toolchain() {
     # on arm, isa versions are supersets
     enabled armv7a && soft_enable armv7 ### DEBUG
     enabled armv7 && soft_enable armv6
-    enabled armv6 && soft_enable armv5te
-    enabled armv6 && soft_enable fast_unaligned
+    enabled armv7 || enabled armv6 && soft_enable armv5te
+    enabled armv7 || enabled armv6 && soft_enable fast_unaligned
     enabled iwmmxt2 && soft_enable iwmmxt
     enabled iwmmxt && soft_enable armv5te
 
@@ -689,7 +692,7 @@ process_common_toolchain() {
             if enabled armv7
                 then
                     check_add_cflags --cpu=Cortex-A8 --fpu=softvfp+vfpv3
-                    check_add_asflags --cpu=Cortex-A8 --fpu=none
+                    check_add_asflags --cpu=Cortex-A8 --fpu=softvfp+vfpv3
                 else
                     check_add_cflags --cpu=${tgt_isa##armv}
                     check_add_asflags --cpu=${tgt_isa##armv}
@@ -729,19 +732,18 @@ process_common_toolchain() {
             add_cflags -arch ${tgt_isa}
             add_ldflags -arch_only ${tgt_isa}
 
-            add_cflags  "-isysroot /Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS4.2.sdk"
+            add_cflags  "-isysroot ${SDK_PATH}/SDKs/iPhoneOS4.3.sdk"
 
             # This should be overridable
-            alt_libc=${SDK_PATH}/SDKs/iPhoneOS4.2.sdk
+            alt_libc=${SDK_PATH}/SDKs/iPhoneOS4.3.sdk
 
             # Add the paths for the alternate libc
-#            for d in usr/include usr/include/gcc/darwin/4.0/; do
-            for d in usr/include usr/include/gcc/darwin/4.0/ usr/lib/gcc/arm-apple-darwin10/4.2.1/include/; do
+            for d in usr/include usr/include/gcc/darwin/4.2/ usr/lib/gcc/arm-apple-darwin10/4.2.1/include/; do
                 try_dir="${alt_libc}/${d}"
                 [ -d "${try_dir}" ] && add_cflags -I"${try_dir}"
             done
 
-            for d in lib usr/lib; do
+            for d in lib usr/lib usr/lib/system; do
                 try_dir="${alt_libc}/${d}"
                 [ -d "${try_dir}" ] && add_ldflags -L"${try_dir}"
             done
@@ -752,41 +754,24 @@ process_common_toolchain() {
         linux*)
             enable linux
             if enabled rvct; then
-                # Compiling with RVCT requires an alternate libc (glibc) when
-                # targetting linux.
-                disabled builtin_libc \
-                    || die "Must supply --libc when targetting *-linux-rvct"
+                # Check if we have CodeSourcery GCC in PATH. Needed for
+                # libraries
+                hash arm-none-linux-gnueabi-gcc 2>&- || \
+                  die "Couldn't find CodeSourcery GCC from PATH"
 
-                # Set up compiler
-                add_cflags --library_interface=aeabi_glibc
-                add_cflags --no_hide_all
-                add_cflags --dwarf2
+                # Use armcc as a linker to enable translation of
+                # some gcc specific options such as -lm and -lpthread.
+                LD="armcc --translate_gcc"
 
-                # Set up linker
-                add_ldflags --sysv --no_startup --no_ref_cpp_init
-                add_ldflags --entry=_start
-                add_ldflags --keep '"*(.init)"' --keep '"*(.fini)"'
-                add_ldflags --keep '"*(.init_array)"' --keep '"*(.fini_array)"'
-                add_ldflags --dynamiclinker=/lib/ld-linux.so.3
-                add_extralibs libc.so.6 -lc_nonshared crt1.o crti.o crtn.o
+                # create configuration file (uses path to CodeSourcery GCC)
+                armcc --arm_linux_configure --arm_linux_config_file=arm_linux.cfg
 
-                # Add the paths for the alternate libc
-                for d in usr/include; do
-                    try_dir="${alt_libc}/${d}"
-                    [ -d "${try_dir}" ] && add_cflags -J"${try_dir}"
-                done
-                add_cflags -J"${RVCT31INC}"
-                for d in lib usr/lib; do
-                    try_dir="${alt_libc}/${d}"
-                    [ -d "${try_dir}" ] && add_ldflags -L"${try_dir}"
-                done
-
-
-                # glibc has some struct members named __align, which is a
-                # storage modifier in RVCT. If we need to use this modifier,
-                # we'll have to #undef it in our code. Note that this must
-                # happen AFTER all libc inclues.
-                add_cflags -D__align=x_align_x
+                add_cflags --arm_linux_paths --arm_linux_config_file=arm_linux.cfg
+                add_asflags --no_hide_all --apcs=/interwork
+                add_ldflags --arm_linux_paths --arm_linux_config_file=arm_linux.cfg
+                enabled pic && add_cflags --apcs=/fpic
+                enabled pic && add_asflags --apcs=/fpic
+                enabled shared && add_cflags --shared
             fi
         ;;
 
@@ -885,6 +870,8 @@ process_common_toolchain() {
                 link_with_cc=gcc
                 tune_cflags="-march="
             setup_gnu_toolchain
+                #for 32 bit x86 builds, -O3 did not turn on this flag
+                enabled optimizations && check_add_cflags -fomit-frame-pointer
                 ;;
         esac
 
@@ -952,15 +939,23 @@ process_common_toolchain() {
     enabled gcov &&
         check_add_cflags -fprofile-arcs -ftest-coverage &&
         check_add_ldflags -fprofile-arcs -ftest-coverage
+
     if enabled optimizations; then
-        enabled rvct && check_add_cflags -Otime
-        enabled small && check_add_cflags -O2 || check_add_cflags -O3
+        if enabled rvct; then
+            enabled small && check_add_cflags -Ospace || check_add_cflags -Otime
+        else
+            enabled small && check_add_cflags -O2 ||  check_add_cflags -O3
+        fi
     fi
 
-    # Position Independant Code (PIC) support, for building relocatable
+    # Position Independent Code (PIC) support, for building relocatable
     # shared objects
     enabled gcc && enabled pic && check_add_cflags -fPIC
 
+    # Work around longjmp interception on glibc >= 2.11, to improve binary
+    # compatibility. See http://code.google.com/p/webm/issues/detail?id=166
+    enabled linux && check_add_cflags -D_FORTIFY_SOURCE=0
+
     # Check for strip utility variant
     ${STRIP} -V 2>/dev/null | grep GNU >/dev/null && enable gnu_strip
 
@@ -979,6 +974,9 @@ EOF
         esac
     fi
 
+    # for sysconf(3) and friends.
+    check_header unistd.h
+
     # glibc needs these
     if enabled linux; then
         add_cflags -D_LARGEFILE_SOURCE

build/make/gen_msvs_proj.sh

@@ -33,6 +33,7 @@ Options:
     --proj-guid=GUID            GUID to use for the project
     --module-def=filename       File containing export definitions (for DLLs)
     --ver=version               Version (7,8,9) of visual studio to generate for
+    --src-path-bare=dir         Path to root of source tree
     -Ipath/to/include           Additional include directories
     -DFLAG[=value]              Preprocessor macros to define
     -Lpath/to/lib               Additional library search paths
@@ -191,6 +192,8 @@ for opt in "$@"; do
         ;;
         --lib) proj_kind="lib"
         ;;
+        --src-path-bare=*) src_path_bare="$optval"
+        ;;
         --static-crt) use_static_runtime=true
         ;;
         --ver=*)
@@ -335,6 +338,35 @@ generate_vcproj() {
         case "$target" in
             x86*)
                 case "$name" in
+                    obj_int_extract)
+                        tag Tool \
+                            Name="VCCLCompilerTool" \
+                            Optimization="0" \
+                            AdditionalIncludeDirectories="$incs" \
+                            PreprocessorDefinitions="WIN32;DEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE" \
+                            RuntimeLibrary="$debug_runtime" \
+                            WarningLevel="3" \
+                            Detect64BitPortabilityProblems="true" \
+                            DebugInformationFormat="1" \
+                    ;;
+                    vpx)
+                        tag Tool \
+                            Name="VCPreBuildEventTool" \
+                            CommandLine="call obj_int_extract.bat $src_path_bare" \
+
+                        tag Tool \
+                            Name="VCCLCompilerTool" \
+                            Optimization="0" \
+                            AdditionalIncludeDirectories="$incs" \
+                            PreprocessorDefinitions="WIN32;_DEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
+                            RuntimeLibrary="$debug_runtime" \
+                            UsePrecompiledHeader="0" \
+                            WarningLevel="3" \
+                            DebugInformationFormat="1" \
+                            Detect64BitPortabilityProblems="true" \
+
+                        $uses_asm && tag Tool Name="YASM"  IncludePaths="$incs" Debug="true"
+                    ;;
                     *)
                         tag Tool \
                             Name="VCCLCompilerTool" \
@@ -347,7 +379,7 @@ generate_vcproj() {
                             DebugInformationFormat="1" \
                             Detect64BitPortabilityProblems="true" \
 
-                        $uses_asm && tag Tool Name="YASM"  IncludePaths="$incs" Debug="1"
+                        $uses_asm && tag Tool Name="YASM"  IncludePaths="$incs" Debug="true"
                     ;;
                 esac
             ;;
@@ -358,6 +390,12 @@ generate_vcproj() {
                 case "$target" in
                     x86*)
                         case "$name" in
+                            obj_int_extract)
+                                tag Tool \
+                                    Name="VCLinkerTool" \
+                                    OutputFile="${name}.exe" \
+                                    GenerateDebugInformation="true" \
+                            ;;
                             *)
                                 tag Tool \
                                     Name="VCLinkerTool" \
@@ -406,10 +444,44 @@ generate_vcproj() {
         case "$target" in
             x86*)
                 case "$name" in
+                    obj_int_extract)
+                        tag Tool \
+                            Name="VCCLCompilerTool" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
+                            AdditionalIncludeDirectories="$incs" \
+                            PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE" \
+                            RuntimeLibrary="$release_runtime" \
+                            UsePrecompiledHeader="0" \
+                            WarningLevel="3" \
+                            Detect64BitPortabilityProblems="true" \
+                            DebugInformationFormat="0" \
+                    ;;
+                    vpx)
+                        tag Tool \
+                            Name="VCPreBuildEventTool" \
+                            CommandLine="call obj_int_extract.bat $src_path_bare" \
+
+                        tag Tool \
+                            Name="VCCLCompilerTool" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
+                            AdditionalIncludeDirectories="$incs" \
+                            PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
+                            RuntimeLibrary="$release_runtime" \
+                            UsePrecompiledHeader="0" \
+                            WarningLevel="3" \
+                            DebugInformationFormat="0" \
+                            Detect64BitPortabilityProblems="true" \
+
+                        $uses_asm && tag Tool Name="YASM"  IncludePaths="$incs"
+                    ;;
                     *)
                         tag Tool \
                             Name="VCCLCompilerTool" \
                             AdditionalIncludeDirectories="$incs" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
                             PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
                             RuntimeLibrary="$release_runtime" \
                             UsePrecompiledHeader="0" \
@@ -428,6 +500,12 @@ generate_vcproj() {
                 case "$target" in
                     x86*)
                         case "$name" in
+                            obj_int_extract)
+                                tag Tool \
+                                    Name="VCLinkerTool" \
+                                    OutputFile="${name}.exe" \
+                                    GenerateDebugInformation="true" \
+                            ;;
                             *)
                                 tag Tool \
                                     Name="VCLinkerTool" \

build/x86-msvs/obj_int_extract.bat

@@ -0,0 +1,15 @@
+REM   Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+REM
+REM   Use of this source code is governed by a BSD-style license
+REM   that can be found in the LICENSE file in the root of the source
+REM   tree. An additional intellectual property rights grant can be found
+REM   in the file PATENTS.  All contributing project authors may
+REM   be found in the AUTHORS file in the root of the source tree.
+echo on
+
+cl /I "./" /I "%1" /nologo /c "%1/vp8/common/asm_com_offsets.c"
+cl /I "./" /I "%1" /nologo /c "%1/vp8/decoder/asm_dec_offsets.c"
+cl /I "./" /I "%1" /nologo /c "%1/vp8/encoder/asm_enc_offsets.c"
+obj_int_extract.exe rvds "asm_com_offsets.obj" > "asm_com_offsets.asm"
+obj_int_extract.exe rvds "asm_dec_offsets.obj" > "asm_dec_offsets.asm"
+obj_int_extract.exe rvds "asm_enc_offsets.obj" > "asm_enc_offsets.asm"

configure

@@ -31,14 +31,16 @@ Advanced options:
   ${toggle_md5}                   support for output of checksum data
   ${toggle_static_msvcrt}         use static MSVCRT (VS builds only)
   ${toggle_vp8}                   VP8 codec support
-  ${toggle_psnr}                  output of PSNR data, if supported (encoders)
+  ${toggle_internal_stats}        output of encoder internal stats for debug, if supported (encoders)
   ${toggle_mem_tracker}           track memory usage
   ${toggle_postproc}              postprocessing
   ${toggle_multithread}           multithreaded encoding and decoding.
   ${toggle_spatial_resampling}    spatial sampling (scaling) support
   ${toggle_realtime_only}         enable this option while building for real-time encoding
+  ${toggle_error_concealment}     enable this option to get a decoder which is able to conceal losses
   ${toggle_runtime_cpu_detect}    runtime cpu detection
   ${toggle_shared}                shared library support
+  ${toggle_static}                static library support
   ${toggle_small}                 favor smaller size over speed
   ${toggle_postproc_visualizer}   macro block / block level visualizers
 
@@ -152,6 +154,7 @@ enabled doxygen && php -v >/dev/null 2>&1 && enable install_docs
 enable install_bins
 enable install_libs
 
+enable static
 enable optimizations
 enable fast_unaligned #allow unaligned accesses, if supported by hw
 enable md5
@@ -211,6 +214,7 @@ HAVE_LIST="
     alt_tree_layout
     pthread_h
     sys_mman_h
+    unistd_h
 "
 CONFIG_LIST="
     external_build
@@ -240,7 +244,7 @@ CONFIG_LIST="
     runtime_cpu_detect
     postproc
     multithread
-    psnr
+    internal_stats
     ${CODECS}
     ${CODEC_FAMILIES}
     encoders
@@ -248,7 +252,9 @@ CONFIG_LIST="
     static_msvcrt
     spatial_resampling
     realtime_only
+    error_concealment
     shared
+    static
     small
     postproc_visualizer
     os_support
@@ -281,14 +287,16 @@ CMDLINE_SELECT="
     dc_recon
     postproc
     multithread
-    psnr
+    internal_stats
     ${CODECS}
     ${CODEC_FAMILIES}
     static_msvcrt
     mem_tracker
     spatial_resampling
     realtime_only
+    error_concealment
     shared
+    static
     small
     postproc_visualizer
 "
@@ -377,6 +385,7 @@ process_targets() {
     if [ -f "${source_path}/build/make/version.sh" ]; then
         local ver=`"$source_path/build/make/version.sh" --bare $source_path`
         DIST_DIR="${DIST_DIR}-${ver}"
+        VERSION_STRING=${ver}
         ver=${ver%%-*}
         VERSION_PATCH=${ver##*.}
         ver=${ver%.*}
@@ -385,6 +394,8 @@ process_targets() {
         VERSION_MAJOR=${ver%.*}
     fi
     enabled child || cat <<EOF >> config.mk
+
+PREFIX=${prefix}
 ifeq (\$(MAKECMDGOALS),dist)
 DIST_DIR?=${DIST_DIR}
 else
@@ -392,6 +403,8 @@ DIST_DIR?=\$(DESTDIR)${prefix}
 endif
 LIBSUBDIR=${libdir##${prefix}/}
 
+VERSION_STRING=${VERSION_STRING}
+
 VERSION_MAJOR=${VERSION_MAJOR}
 VERSION_MINOR=${VERSION_MINOR}
 VERSION_PATCH=${VERSION_PATCH}
@@ -486,7 +499,7 @@ process_toolchain() {
         check_add_cflags -Wpointer-arith
         check_add_cflags -Wtype-limits
         check_add_cflags -Wcast-qual
-        enabled extra_warnings || check_add_cflags -Wno-unused
+        enabled extra_warnings || check_add_cflags -Wno-unused-function
     fi
 
     if enabled icc; then

examples.mk

@@ -77,6 +77,11 @@ GEN_EXAMPLES-$(CONFIG_ENCODERS) += decode_with_drops.c
 endif
 decode_with_drops.GUID           = CE5C53C4-8DDA-438A-86ED-0DDD3CDB8D26
 decode_with_drops.DESCRIPTION    = Drops frames while decoding
+ifeq ($(CONFIG_DECODERS),yes)
+GEN_EXAMPLES-$(CONFIG_ERROR_CONCEALMENT) += decode_with_partial_drops.c
+endif
+decode_with_partial_drops.GUID           = 61C2D026-5754-46AC-916F-1343ECC5537E
+decode_with_partial_drops.DESCRIPTION    = Drops parts of frames while decoding
 GEN_EXAMPLES-$(CONFIG_ENCODERS) += error_resilient.c
 error_resilient.GUID             = DF5837B9-4145-4F92-A031-44E4F832E00C
 error_resilient.DESCRIPTION      = Error Resiliency Feature
@@ -122,8 +127,8 @@ else
     LIB_PATH := $(call enabled,LIB_PATH)
     INC_PATH := $(call enabled,INC_PATH)
 endif
-CFLAGS += $(addprefix -I,$(INC_PATH))
-LDFLAGS += $(addprefix -L,$(LIB_PATH))
+INTERNAL_CFLAGS = $(addprefix -I,$(INC_PATH))
+INTERNAL_LDFLAGS += $(addprefix -L,$(LIB_PATH))
 
 
 # Expand list of selected examples to build (as specified above)
@@ -162,8 +167,10 @@ BINS-$(NOT_MSVS)           += $(addprefix $(BUILD_PFX),$(ALL_EXAMPLES:.c=))
 
 # Instantiate linker template for all examples.
 CODEC_LIB=$(if $(CONFIG_DEBUG_LIBS),vpx_g,vpx)
+CODEC_LIB_SUF=$(if $(CONFIG_SHARED),.so,.a)
 $(foreach bin,$(BINS-yes),\
-    $(if $(BUILD_OBJS),$(eval $(bin): $(LIB_PATH)/lib$(CODEC_LIB).a))\
+    $(if $(BUILD_OBJS),$(eval $(bin):\
+        $(LIB_PATH)/lib$(CODEC_LIB)$(CODEC_LIB_SUF)))\
     $(if $(BUILD_OBJS),$(eval $(call linker_template,$(bin),\
         $(call objs,$($(notdir $(bin)).SRCS)) \
         -l$(CODEC_LIB) $(addprefix -l,$(CODEC_EXTRA_LIBS))\
@@ -214,7 +221,8 @@ $(1): $($(1:.vcproj=).SRCS)
             --ver=$$(CONFIG_VS_VERSION)\
             --proj-guid=$$($$(@:.vcproj=).GUID)\
             $$(if $$(CONFIG_STATIC_MSVCRT),--static-crt) \
-            --out=$$@ $$(CFLAGS) $$(LDFLAGS) -l$$(CODEC_LIB) -lwinmm $$^
+            --out=$$@ $$(INTERNAL_CFLAGS) $$(CFLAGS) \
+            $$(INTERNAL_LDFLAGS) $$(LDFLAGS) -l$$(CODEC_LIB) -lwinmm $$^
 endef
 PROJECTS-$(CONFIG_MSVS) += $(ALL_EXAMPLES:.c=.vcproj)
 INSTALL-BINS-$(CONFIG_MSVS) += $(foreach p,$(VS_PLATFORMS),\

examples/decode_with_partial_drops.txt

@@ -0,0 +1,238 @@
+@TEMPLATE decoder_tmpl.c
+Decode With Partial Drops Example
+=========================
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ INTRODUCTION
+This is an example utility which drops a series of frames (or parts of frames),
+as specified on the command line. This is useful for observing the error
+recovery features of the codec.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ INTRODUCTION
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ EXTRA_INCLUDES
+#include <time.h>
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ EXTRA_INCLUDES
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ HELPERS
+struct parsed_header
+{
+    char key_frame;
+    int version;
+    char show_frame;
+    int first_part_size;
+};
+
+int next_packet(struct parsed_header* hdr, int pos, int length, int mtu)
+{
+    int size = 0;
+    int remaining = length - pos;
+    /* Uncompressed part is 3 bytes for P frames and 10 bytes for I frames */
+    int uncomp_part_size = (hdr->key_frame ? 10 : 3);
+    /* number of bytes yet to send from header and the first partition */
+    int remainFirst = uncomp_part_size + hdr->first_part_size - pos;
+    if (remainFirst > 0)
+    {
+        if (remainFirst <= mtu)
+        {
+            size = remainFirst;
+        }
+        else
+        {
+            size = mtu;
+        }
+
+        return size;
+    }
+
+    /* second partition; just slot it up according to MTU */
+    if (remaining <= mtu)
+    {
+        size = remaining;
+        return size;
+    }
+    return mtu;
+}
+
+void throw_packets(unsigned char* frame, int* size, int loss_rate,
+                   int* thrown, int* kept)
+{
+    unsigned char loss_frame[256*1024];
+    int pkg_size = 1;
+    int pos = 0;
+    int loss_pos = 0;
+    struct parsed_header hdr;
+    unsigned int tmp;
+    int mtu = 1500;
+
+    if (*size < 3)
+    {
+        return;
+    }
+    putc('|', stdout);
+    /* parse uncompressed 3 bytes */
+    tmp = (frame[2] << 16) | (frame[1] << 8) | frame[0];
+    hdr.key_frame = !(tmp & 0x1); /* inverse logic */
+    hdr.version = (tmp >> 1) & 0x7;
+    hdr.show_frame = (tmp >> 4) & 0x1;
+    hdr.first_part_size = (tmp >> 5) & 0x7FFFF;
+
+    /* don't drop key frames */
+    if (hdr.key_frame)
+    {
+        int i;
+        *kept = *size/mtu + ((*size % mtu > 0) ? 1 : 0); /* approximate */
+        for (i=0; i < *kept; i++)
+            putc('.', stdout);
+        return;
+    }
+
+    while ((pkg_size = next_packet(&hdr, pos, *size, mtu)) > 0)
+    {
+        int loss_event = ((rand() + 1.0)/(RAND_MAX + 1.0) < loss_rate/100.0);
+        if (*thrown == 0 && !loss_event)
+        {
+            memcpy(loss_frame + loss_pos, frame + pos, pkg_size);
+            loss_pos += pkg_size;
+            (*kept)++;
+            putc('.', stdout);
+        }
+        else
+        {
+            (*thrown)++;
+            putc('X', stdout);
+        }
+        pos += pkg_size;
+    }
+    memcpy(frame, loss_frame, loss_pos);
+    memset(frame + loss_pos, 0, *size - loss_pos);
+    *size = loss_pos;
+}
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ HELPERS
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DEC_INIT
+/* Initialize codec */
+flags = VPX_CODEC_USE_ERROR_CONCEALMENT;
+res = vpx_codec_dec_init(&codec, interface, &dec_cfg, flags);
+if(res)
+    die_codec(&codec, "Failed to initialize decoder");
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DEC_INIT
+
+Usage
+-----
+This example adds a single argument to the `simple_decoder` example,
+which specifies the range or pattern of frames to drop. The parameter is
+parsed as follows:
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ USAGE
+if(argc < 4 || argc > 6)
+    die("Usage: %s <infile> <outfile> [-t <num threads>] <N-M|N/M|L,S>\n",
+        argv[0]);
+{
+    char *nptr;
+    int arg_num = 3;
+    if (argc == 6 && strncmp(argv[arg_num++], "-t", 2) == 0)
+        dec_cfg.threads = strtol(argv[arg_num++], NULL, 0);
+    n = strtol(argv[arg_num], &nptr, 0);
+    mode = (*nptr == '\0' || *nptr == ',') ? 2 : (*nptr == '-') ? 1 : 0;
+
+    m = strtol(nptr+1, NULL, 0);
+    if((!n && !m) || (*nptr != '-' && *nptr != '/' &&
+        *nptr != '\0' && *nptr != ','))
+        die("Couldn't parse pattern %s\n", argv[3]);
+}
+seed = (m > 0) ? m : (unsigned int)time(NULL);
+srand(seed);thrown_frame = 0;
+printf("Seed: %u\n", seed);
+printf("Threads: %d\n", dec_cfg.threads);
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ USAGE
+
+
+Dropping A Range Of Frames
+--------------------------
+To drop a range of frames, specify the starting frame and the ending
+frame to drop, separated by a dash. The following command will drop
+frames 5 through 10 (base 1).
+
+  $ ./decode_with_partial_drops in.ivf out.i420 5-10
+
+
+Dropping A Pattern Of Frames
+----------------------------
+To drop a pattern of frames, specify the number of frames to drop and
+the number of frames after which to repeat the pattern, separated by
+a forward-slash. The following command will drop 3 of 7 frames.
+Specifically, it will decode 4 frames, then drop 3 frames, and then
+repeat.
+
+  $ ./decode_with_partial_drops in.ivf out.i420 3/7
+
+Dropping Random Parts Of Frames
+-------------------------------
+A third argument tuple is available to split the frame into 1500 bytes pieces
+and randomly drop pieces rather than frames. The frame will be split at
+partition boundaries where possible. The following example will seed the RNG
+with the seed 123 and drop approximately 5% of the pieces. Pieces which
+are depending on an already dropped piece will also be dropped.
+
+  $ ./decode_with_partial_drops in.ivf out.i420 5,123
+
+
+Extra Variables
+---------------
+This example maintains the pattern passed on the command line in the
+`n`, `m`, and `is_range` variables:
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ EXTRA_VARS
+int              n, m, mode;
+unsigned int     seed;
+int              thrown=0, kept=0;
+int              thrown_frame=0, kept_frame=0;
+vpx_codec_dec_cfg_t  dec_cfg = {0};
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ EXTRA_VARS
+
+
+Making The Drop Decision
+------------------------
+The example decides whether to drop the frame based on the current
+frame number, immediately before decoding the frame.
+
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ PRE_DECODE
+/* Decide whether to throw parts of the frame or the whole frame
+   depending on the drop mode */
+thrown_frame = 0;
+kept_frame = 0;
+switch (mode)
+{
+case 0:
+    if (m - (frame_cnt-1)%m <= n)
+    {
+        frame_sz = 0;
+    }
+    break;
+case 1:
+    if (frame_cnt >= n && frame_cnt <= m)
+    {
+        frame_sz = 0;
+    }
+    break;
+case 2:
+    throw_packets(frame, &frame_sz, n, &thrown_frame, &kept_frame);
+    break;
+default: break;
+}
+if (mode < 2)
+{
+    if (frame_sz == 0)
+    {
+        putc('X', stdout);
+        thrown_frame++;
+    }
+    else
+    {
+        putc('.', stdout);
+        kept_frame++;
+    }
+}
+thrown += thrown_frame;
+kept += kept_frame;
+fflush(stdout);
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ PRE_DECODE

examples/decoder_tmpl.c

@@ -42,6 +42,8 @@ static void die(const char *fmt, ...) {
 
 @DIE_CODEC
 
+@HELPERS
+
 int main(int argc, char **argv) {
     FILE            *infile, *outfile;
     vpx_codec_ctx_t  codec;

examples/encoder_tmpl.c

@@ -111,8 +111,6 @@ int main(int argc, char **argv) {
     vpx_codec_ctx_t      codec;
     vpx_codec_enc_cfg_t  cfg;
     int                  frame_cnt = 0;
-    unsigned char        file_hdr[IVF_FILE_HDR_SZ];
-    unsigned char        frame_hdr[IVF_FRAME_HDR_SZ];
     vpx_image_t          raw;
     vpx_codec_err_t      res;
     long                 width;

examples/postproc.txt

@@ -21,7 +21,7 @@ res = vpx_codec_dec_init(&codec, interface, NULL,
 if(res == VPX_CODEC_INCAPABLE) {
     printf("NOTICE: Postproc not supported by %s\n",
            vpx_codec_iface_name(interface));
-    res = vpx_codec_dec_init(&codec, interface, NULL, 0);
+    res = vpx_codec_dec_init(&codec, interface, NULL, flags);
 }
 if(res)
     die_codec(&codec, "Failed to initialize decoder");

libmkv/EbmlIDs.h

@@ -120,7 +120,7 @@ enum mkv
     //video
     Video = 0xE0,
     FlagInterlaced = 0x9A,
-//  StereoMode = 0x53B8,
+    StereoMode = 0x53B8,
     PixelWidth = 0xB0,
     PixelHeight = 0xBA,
     PixelCropBottom = 0x54AA,

libmkv/EbmlWriter.c

@@ -11,6 +11,7 @@
 #include <stdlib.h>
 #include <wchar.h>
 #include <string.h>
+#include <limits.h>
 #if defined(_MSC_VER)
 #define LITERALU64(n) n
 #else
@@ -33,7 +34,7 @@ void Ebml_WriteLen(EbmlGlobal *glob, long long val)
 
     val |= (LITERALU64(0x000000000000080) << ((size - 1) * 7));
 
-    Ebml_Serialize(glob, (void *) &val, size);
+    Ebml_Serialize(glob, (void *) &val, sizeof(val), size);
 }
 
 void Ebml_WriteString(EbmlGlobal *glob, const char *str)
@@ -60,21 +61,26 @@ void Ebml_WriteUTF8(EbmlGlobal *glob, const wchar_t *wstr)
 
 void Ebml_WriteID(EbmlGlobal *glob, unsigned long class_id)
 {
+    int len;
+
     if (class_id >= 0x01000000)
-        Ebml_Serialize(glob, (void *)&class_id, 4);
+        len = 4;
     else if (class_id >= 0x00010000)
-        Ebml_Serialize(glob, (void *)&class_id, 3);
+        len = 3;
     else if (class_id >= 0x00000100)
-        Ebml_Serialize(glob, (void *)&class_id, 2);
+        len = 2;
     else
-        Ebml_Serialize(glob, (void *)&class_id, 1);
+        len = 1;
+
+    Ebml_Serialize(glob, (void *)&class_id, sizeof(class_id), len);
 }
+
 void Ebml_SerializeUnsigned64(EbmlGlobal *glob, unsigned long class_id, uint64_t ui)
 {
     unsigned char sizeSerialized = 8 | 0x80;
     Ebml_WriteID(glob, class_id);
-    Ebml_Serialize(glob, &sizeSerialized, 1);
-    Ebml_Serialize(glob, &ui, 8);
+    Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
+    Ebml_Serialize(glob, &ui, sizeof(ui), 8);
 }
 
 void Ebml_SerializeUnsigned(EbmlGlobal *glob, unsigned long class_id, unsigned long ui)
@@ -97,8 +103,8 @@ void Ebml_SerializeUnsigned(EbmlGlobal *glob, unsigned long class_id, unsigned l
     }
 
     sizeSerialized = 0x80 | size;
-    Ebml_Serialize(glob, &sizeSerialized, 1);
-    Ebml_Serialize(glob, &ui, size);
+    Ebml_Serialize(glob, &sizeSerialized, sizeof(sizeSerialized), 1);
+    Ebml_Serialize(glob, &ui, sizeof(ui), size);
 }
 //TODO: perhaps this is a poor name for this id serializer helper function
 void Ebml_SerializeBinary(EbmlGlobal *glob, unsigned long class_id, unsigned long bin)
@@ -119,14 +125,14 @@ void Ebml_SerializeFloat(EbmlGlobal *glob, unsigned long class_id, double d)
     unsigned char len = 0x88;
 
     Ebml_WriteID(glob, class_id);
-    Ebml_Serialize(glob, &len, 1);
-    Ebml_Serialize(glob,  &d, 8);
+    Ebml_Serialize(glob, &len, sizeof(len), 1);
+    Ebml_Serialize(glob,  &d, sizeof(d), 8);
 }
 
 void Ebml_WriteSigned16(EbmlGlobal *glob, short val)
 {
     signed long out = ((val & 0x003FFFFF) | 0x00200000) << 8;
-    Ebml_Serialize(glob, &out, 3);
+    Ebml_Serialize(glob, &out, sizeof(out), 3);
 }
 
 void Ebml_SerializeString(EbmlGlobal *glob, unsigned long class_id, const char *s)
@@ -143,7 +149,6 @@ void Ebml_SerializeUTF8(EbmlGlobal *glob, unsigned long class_id, wchar_t *s)
 
 void Ebml_SerializeData(EbmlGlobal *glob, unsigned long class_id, unsigned char *data, unsigned long data_length)
 {
-    unsigned char size = 4;
     Ebml_WriteID(glob, class_id);
     Ebml_WriteLen(glob, data_length);
     Ebml_Write(glob,  data, data_length);

libmkv/EbmlWriter.h

@@ -15,7 +15,7 @@
 #include "vpx/vpx_integer.h"
 
 typedef struct EbmlGlobal EbmlGlobal;
-void  Ebml_Serialize(EbmlGlobal *glob, const void *, unsigned long);
+void  Ebml_Serialize(EbmlGlobal *glob, const void *, int, unsigned long);
 void  Ebml_Write(EbmlGlobal *glob, const void *, unsigned long);
 /////
 

libmkv/WebMElement.c

@@ -35,11 +35,11 @@ void writeSimpleBlock(EbmlGlobal *glob, unsigned char trackNumber, short timeCod
     Ebml_WriteID(glob, SimpleBlock);
     unsigned long blockLength = 4 + dataLength;
     blockLength |= 0x10000000; //TODO check length < 0x0FFFFFFFF
-    Ebml_Serialize(glob, &blockLength, 4);
+    Ebml_Serialize(glob, &blockLength, sizeof(blockLength), 4);
     trackNumber |= 0x80;  //TODO check track nubmer < 128
     Ebml_Write(glob, &trackNumber, 1);
     //Ebml_WriteSigned16(glob, timeCode,2); //this is 3 bytes
-    Ebml_Serialize(glob, &timeCode, 2);
+    Ebml_Serialize(glob, &timeCode, sizeof(timeCode), 2);
     unsigned char flags = 0x00 | (isKeyframe ? 0x80 : 0x00) | (lacingFlag << 1) | discardable;
     Ebml_Write(glob, &flags, 1);
     Ebml_Write(glob, data, dataLength);

libs.mk

@@ -9,7 +9,13 @@
 ##
 
 
-ASM:=$(if $(filter yes,$(CONFIG_GCC)),.asm.s,.asm)
+# ARM assembly files are written in RVCT-style. We use some make magic to
+# filter those files to allow GCC compilation
+ifeq ($(ARCH_ARM),yes)
+  ASM:=$(if $(filter yes,$(CONFIG_GCC)),.asm.s,.asm)
+else
+  ASM:=.asm
+endif
 
 CODEC_SRCS-yes += libs.mk
 
@@ -29,6 +35,7 @@ ifeq ($(CONFIG_VP8_ENCODER),yes)
   CODEC_SRCS-yes += $(addprefix $(VP8_PREFIX),$(call enabled,VP8_CX_SRCS))
   CODEC_EXPORTS-yes += $(addprefix $(VP8_PREFIX),$(VP8_CX_EXPORTS))
   CODEC_SRCS-yes += $(VP8_PREFIX)vp8cx.mk vpx/vp8.h vpx/vp8cx.h vpx/vp8e.h
+  CODEC_SRCS-$(ARCH_ARM) += $(VP8_PREFIX)vp8cx_arm.mk
   INSTALL-LIBS-yes += include/vpx/vp8.h include/vpx/vp8e.h include/vpx/vp8cx.h
   INSTALL_MAPS += include/vpx/% $(SRC_PATH_BARE)/$(VP8_PREFIX)/%
   CODEC_DOC_SRCS += vpx/vp8.h vpx/vp8cx.h
@@ -41,6 +48,7 @@ ifeq ($(CONFIG_VP8_DECODER),yes)
   CODEC_SRCS-yes += $(addprefix $(VP8_PREFIX),$(call enabled,VP8_DX_SRCS))
   CODEC_EXPORTS-yes += $(addprefix $(VP8_PREFIX),$(VP8_DX_EXPORTS))
   CODEC_SRCS-yes += $(VP8_PREFIX)vp8dx.mk vpx/vp8.h vpx/vp8dx.h
+  CODEC_SRCS-$(ARCH_ARM) += $(VP8_PREFIX)vp8dx_arm.mk
   INSTALL-LIBS-yes += include/vpx/vp8.h include/vpx/vp8dx.h
   INSTALL_MAPS += include/vpx/% $(SRC_PATH_BARE)/$(VP8_PREFIX)/%
   CODEC_DOC_SRCS += vpx/vp8.h vpx/vp8dx.h
@@ -83,6 +91,7 @@ $(eval $(if $(filter universal%,$(TOOLCHAIN)),LIPO_LIBVPX,BUILD_LIBVPX):=yes)
 
 CODEC_SRCS-$(BUILD_LIBVPX) += build/make/version.sh
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx/vpx_integer.h
+CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/asm_offsets.h
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/vpx_timer.h
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/mem.h
 CODEC_SRCS-$(BUILD_LIBVPX) += $(BUILD_PFX)vpx_config.c
@@ -94,7 +103,7 @@ CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86_abi_support.asm
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86_cpuid.c
 endif
 CODEC_SRCS-$(ARCH_ARM) += vpx_ports/arm_cpudetect.c
-CODEC_SRCS-$(ARCH_ARM) += $(BUILD_PFX)vpx_config.asm
+CODEC_SRCS-$(ARCH_ARM) += vpx_ports/arm.h
 CODEC_EXPORTS-$(BUILD_LIBVPX) += vpx/exports_com
 CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_enc
 CODEC_EXPORTS-$(CONFIG_DECODERS) += vpx/exports_dec
@@ -115,7 +124,7 @@ INSTALL-LIBS-$(CONFIG_SHARED) += $(foreach p,$(VS_PLATFORMS),$(LIBSUBDIR)/$(p)/v
 INSTALL-LIBS-$(CONFIG_SHARED) += $(foreach p,$(VS_PLATFORMS),$(LIBSUBDIR)/$(p)/vpx.exp)
 endif
 else
-INSTALL-LIBS-yes += $(LIBSUBDIR)/libvpx.a
+INSTALL-LIBS-$(CONFIG_STATIC) += $(LIBSUBDIR)/libvpx.a
 INSTALL-LIBS-$(CONFIG_DEBUG_LIBS) += $(LIBSUBDIR)/libvpx_g.a
 endif
 
@@ -123,9 +132,34 @@ CODEC_SRCS=$(call enabled,CODEC_SRCS)
 INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += $(CODEC_SRCS)
 INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += $(call enabled,CODEC_EXPORTS)
 
+
+# Generate a list of all enabled sources, in particular for exporting to gyp
+# based build systems.
+libvpx_srcs.txt:
+	@echo "    [CREATE] $@"
+	@echo $(CODEC_SRCS) | xargs -n1 echo | sort -u > $@
+
+
 ifeq ($(CONFIG_EXTERNAL_BUILD),yes)
 ifeq ($(CONFIG_MSVS),yes)
 
+obj_int_extract.vcproj: $(SRC_PATH_BARE)/build/make/obj_int_extract.c
+	@cp $(SRC_PATH_BARE)/build/x86-msvs/obj_int_extract.bat .
+	@echo "    [CREATE] $@"
+	$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
+    --exe \
+    --target=$(TOOLCHAIN) \
+    --name=obj_int_extract \
+    --ver=$(CONFIG_VS_VERSION) \
+    --proj-guid=E1360C65-D375-4335-8057-7ED99CC3F9B2 \
+    $(if $(CONFIG_STATIC_MSVCRT),--static-crt) \
+    --out=$@ $^ \
+    -I. \
+    -I"$(SRC_PATH_BARE)" \
+
+PROJECTS-$(BUILD_LIBVPX) += obj_int_extract.vcproj
+PROJECTS-$(BUILD_LIBVPX) += obj_int_extract.bat
+
 vpx.def: $(call enabled,CODEC_EXPORTS)
 	@echo "    [CREATE] $@"
 	$(SRC_PATH_BARE)/build/make/gen_msvs_def.sh\
@@ -135,15 +169,16 @@ CLEAN-OBJS += vpx.def
 
 vpx.vcproj: $(CODEC_SRCS) vpx.def
 	@echo "    [CREATE] $@"
-	$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh\
-			--lib\
-			--target=$(TOOLCHAIN)\
+	$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
+			--lib \
+			--target=$(TOOLCHAIN) \
             $(if $(CONFIG_STATIC_MSVCRT),--static-crt) \
-            --name=vpx\
-            --proj-guid=DCE19DAF-69AC-46DB-B14A-39F0FAA5DB74\
-            --module-def=vpx.def\
-            --ver=$(CONFIG_VS_VERSION)\
-            --out=$@ $(CFLAGS) $^\
+            --name=vpx \
+            --proj-guid=DCE19DAF-69AC-46DB-B14A-39F0FAA5DB74 \
+            --module-def=vpx.def \
+            --ver=$(CONFIG_VS_VERSION) \
+            --out=$@ $(CFLAGS) $^ \
+            --src-path-bare="$(SRC_PATH_BARE)" \
 
 PROJECTS-$(BUILD_LIBVPX) += vpx.vcproj
 
@@ -153,14 +188,15 @@ endif
 else
 LIBVPX_OBJS=$(call objs,$(CODEC_SRCS))
 OBJS-$(BUILD_LIBVPX) += $(LIBVPX_OBJS)
-LIBS-$(BUILD_LIBVPX) += $(BUILD_PFX)libvpx.a $(BUILD_PFX)libvpx_g.a
+LIBS-$(if $(BUILD_LIBVPX),$(CONFIG_STATIC)) += $(BUILD_PFX)libvpx.a $(BUILD_PFX)libvpx_g.a
 $(BUILD_PFX)libvpx_g.a: $(LIBVPX_OBJS)
 
 BUILD_LIBVPX_SO         := $(if $(BUILD_LIBVPX),$(CONFIG_SHARED))
 LIBVPX_SO               := libvpx.so.$(VERSION_MAJOR).$(VERSION_MINOR).$(VERSION_PATCH)
-LIBS-$(BUILD_LIBVPX_SO) += $(BUILD_PFX)$(LIBVPX_SO)
+LIBS-$(BUILD_LIBVPX_SO) += $(BUILD_PFX)$(LIBVPX_SO)\
+                           $(notdir $(LIBVPX_SO_SYMLINKS))
 $(BUILD_PFX)$(LIBVPX_SO): $(LIBVPX_OBJS) libvpx.ver
-$(BUILD_PFX)$(LIBVPX_SO): extralibs += -lm -pthread
+$(BUILD_PFX)$(LIBVPX_SO): extralibs += -lm
 $(BUILD_PFX)$(LIBVPX_SO): SONAME = libvpx.so.$(VERSION_MAJOR)
 $(BUILD_PFX)$(LIBVPX_SO): SO_VERSION_SCRIPT = libvpx.ver
 LIBVPX_SO_SYMLINKS      := $(addprefix $(LIBSUBDIR)/, \
@@ -174,12 +210,41 @@ libvpx.ver: $(call enabled,CODEC_EXPORTS)
 	$(qexec)echo "local: *; };" >> $@
 CLEAN-OBJS += libvpx.ver
 
-$(addprefix $(DIST_DIR)/,$(LIBVPX_SO_SYMLINKS)):
-	@echo "    [LN]      $@"
-	$(qexec)ln -sf $(LIBVPX_SO) $@
+define libvpx_symlink_template
+$(1): $(2)
+	@echo "    [LN]      $$@"
+	$(qexec)ln -sf $(LIBVPX_SO) $$@
+endef
+
+$(eval $(call libvpx_symlink_template,\
+    $(addprefix $(BUILD_PFX),$(notdir $(LIBVPX_SO_SYMLINKS))),\
+    $(BUILD_PFX)$(LIBVPX_SO)))
+$(eval $(call libvpx_symlink_template,\
+    $(addprefix $(DIST_DIR)/,$(LIBVPX_SO_SYMLINKS)),\
+    $(DIST_DIR)/$(LIBSUBDIR)/$(LIBVPX_SO)))
 
 INSTALL-LIBS-$(CONFIG_SHARED) += $(LIBVPX_SO_SYMLINKS)
 INSTALL-LIBS-$(CONFIG_SHARED) += $(LIBSUBDIR)/$(LIBVPX_SO)
+
+LIBS-$(BUILD_LIBVPX) += vpx.pc
+vpx.pc: config.mk libs.mk
+	@echo "    [CREATE] $@"
+	$(qexec)echo '# pkg-config file from libvpx $(VERSION_STRING)' > $@
+	$(qexec)echo 'prefix=$(PREFIX)' >> $@
+	$(qexec)echo 'exec_prefix=$${prefix}' >> $@
+	$(qexec)echo 'libdir=$${prefix}/lib' >> $@
+	$(qexec)echo 'includedir=$${prefix}/include' >> $@
+	$(qexec)echo '' >> $@
+	$(qexec)echo 'Name: vpx' >> $@
+	$(qexec)echo 'Description: WebM Project VPx codec implementation' >> $@
+	$(qexec)echo 'Version: $(VERSION_MAJOR).$(VERSION_MINOR).$(VERSION_PATCH)' >> $@
+	$(qexec)echo 'Requires:' >> $@
+	$(qexec)echo 'Conflicts:' >> $@
+	$(qexec)echo 'Libs: -L$${libdir} -lvpx' >> $@
+	$(qexec)echo 'Cflags: -I$${includedir}' >> $@
+INSTALL-LIBS-yes += $(LIBSUBDIR)/pkgconfig/vpx.pc
+INSTALL_MAPS += $(LIBSUBDIR)/pkgconfig/%.pc %.pc
+CLEAN-OBJS += vpx.pc
 endif
 
 LIBS-$(LIPO_LIBVPX) += libvpx.a
@@ -207,30 +272,44 @@ endif
 #
 # Add assembler dependencies for configuration and offsets
 #
-$(filter %.s.o,$(OBJS-yes)):   $(BUILD_PFX)vpx_config.asm
-$(filter %.asm.o,$(OBJS-yes)): $(BUILD_PFX)vpx_config.asm
+$(filter %.s.o,$(OBJS-yes)):     $(BUILD_PFX)vpx_config.asm
+$(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)vpx_config.asm
 
 #
 # Calculate platform- and compiler-specific offsets for hand coded assembly
 #
-ifeq ($(ARCH_ARM), yes)
-  asm_com_offsets.asm: obj_int_extract
-  asm_com_offsets.asm: $(VP8_PREFIX)common/asm_com_offsets.c.o
-	./obj_int_extract rvds $< $(ADS2GAS) > $@
-  OBJS-yes += $(VP8_PREFIX)common/asm_com_offsets.c.o
-  CLEAN-OBJS += asm_com_offsets.asm
-  $(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)asm_com_offsets.asm
 
-  ifeq ($(CONFIG_VP8_ENCODER), yes)
+ifeq ($(filter icc gcc,$(TGT_CC)), $(TGT_CC))
+    $(BUILD_PFX)asm_com_offsets.asm: $(BUILD_PFX)$(VP8_PREFIX)common/asm_com_offsets.c.S
+	grep EQU $< | tr -d '$$\#' $(ADS2GAS) > $@
+    $(BUILD_PFX)$(VP8_PREFIX)common/asm_com_offsets.c.S: $(VP8_PREFIX)common/asm_com_offsets.c
+    CLEAN-OBJS += $(BUILD_PFX)asm_com_offsets.asm $(BUILD_PFX)$(VP8_PREFIX)common/asm_com_offsets.c.S
+
+    $(BUILD_PFX)asm_enc_offsets.asm: $(BUILD_PFX)$(VP8_PREFIX)encoder/asm_enc_offsets.c.S
+	grep EQU $< | tr -d '$$\#' $(ADS2GAS) > $@
+    $(BUILD_PFX)$(VP8_PREFIX)encoder/asm_enc_offsets.c.S: $(VP8_PREFIX)encoder/asm_enc_offsets.c
+    CLEAN-OBJS += $(BUILD_PFX)asm_enc_offsets.asm $(BUILD_PFX)$(VP8_PREFIX)encoder/asm_enc_offsets.c.S
+
+    $(BUILD_PFX)asm_dec_offsets.asm: $(BUILD_PFX)$(VP8_PREFIX)decoder/asm_dec_offsets.c.S
+	grep EQU $< | tr -d '$$\#' $(ADS2GAS) > $@
+    $(BUILD_PFX)$(VP8_PREFIX)decoder/asm_dec_offsets.c.S: $(VP8_PREFIX)decoder/asm_dec_offsets.c
+    CLEAN-OBJS += $(BUILD_PFX)asm_dec_offsets.asm $(BUILD_PFX)$(VP8_PREFIX)decoder/asm_dec_offsets.c.S
+else
+  ifeq ($(filter rvct,$(TGT_CC)), $(TGT_CC))
+    asm_com_offsets.asm: obj_int_extract
+    asm_com_offsets.asm: $(VP8_PREFIX)common/asm_com_offsets.c.o
+	./obj_int_extract rvds $< $(ADS2GAS) > $@
+    OBJS-yes += $(VP8_PREFIX)common/asm_com_offsets.c.o
+    CLEAN-OBJS += asm_com_offsets.asm
+    $(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)asm_com_offsets.asm
+
     asm_enc_offsets.asm: obj_int_extract
     asm_enc_offsets.asm: $(VP8_PREFIX)encoder/asm_enc_offsets.c.o
 	./obj_int_extract rvds $< $(ADS2GAS) > $@
     OBJS-yes += $(VP8_PREFIX)encoder/asm_enc_offsets.c.o
     CLEAN-OBJS += asm_enc_offsets.asm
     $(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)asm_enc_offsets.asm
-  endif
 
-  ifeq ($(CONFIG_VP8_DECODER), yes)
     asm_dec_offsets.asm: obj_int_extract
     asm_dec_offsets.asm: $(VP8_PREFIX)decoder/asm_dec_offsets.c.o
 	./obj_int_extract rvds $< $(ADS2GAS) > $@

solution.mk

@@ -13,8 +13,9 @@ vpx.sln: $(wildcard *.vcproj)
 	@echo "    [CREATE] $@"
 	$(SRC_PATH_BARE)/build/make/gen_msvs_sln.sh \
             $(if $(filter %vpx.vcproj,$^),\
-                $(foreach vcp,$(filter-out %vpx.vcproj,$^),\
+                $(foreach vcp,$(filter-out %vpx.vcproj %obj_int_extract.vcproj,$^),\
                   --dep=$(vcp:.vcproj=):vpx)) \
+            --dep=vpx:obj_int_extract \
             --ver=$(CONFIG_VS_VERSION)\
             --out=$@ $^
 vpx.sln.mk: vpx.sln

vp8/common/alloccommon.c

@@ -20,13 +20,16 @@
 
 extern  void vp8_init_scan_order_mask();
 
-void vp8_update_mode_info_border(MODE_INFO *mi, int rows, int cols)
+static void update_mode_info_border(MODE_INFO *mi, int rows, int cols)
 {
     int i;
     vpx_memset(mi - cols - 2, 0, sizeof(MODE_INFO) * (cols + 1));
 
     for (i = 0; i < rows; i++)
     {
+        /* TODO(holmer): Bug? This updates the last element of each row
+         * rather than the border element!
+         */
         vpx_memset(&mi[i*cols-1], 0, sizeof(MODE_INFO));
     }
 }
@@ -43,9 +46,11 @@ void vp8_de_alloc_frame_buffers(VP8_COMMON *oci)
 
     vpx_free(oci->above_context);
     vpx_free(oci->mip);
+    vpx_free(oci->prev_mip);
 
     oci->above_context = 0;
     oci->mip = 0;
+    oci->prev_mip = 0;
 
 }
 
@@ -65,9 +70,9 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
 
     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)
+        oci->fb_idx_ref_cnt[i] = 0;
+        oci->yv12_fb[i].flags = 0;
+        if (vp8_yv12_alloc_frame_buffer(&oci->yv12_fb[i], width, height, VP8BORDERINPIXELS) < 0)
         {
             vp8_de_alloc_frame_buffers(oci);
             return 1;
@@ -110,6 +115,21 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
 
     oci->mi = oci->mip + oci->mode_info_stride + 1;
 
+    /* allocate memory for last frame MODE_INFO array */
+#if CONFIG_ERROR_CONCEALMENT
+    oci->prev_mip = vpx_calloc((oci->mb_cols + 1) * (oci->mb_rows + 1), sizeof(MODE_INFO));
+
+    if (!oci->prev_mip)
+    {
+        vp8_de_alloc_frame_buffers(oci);
+        return 1;
+    }
+
+    oci->prev_mi = oci->prev_mip + oci->mode_info_stride + 1;
+#else
+    oci->prev_mip = NULL;
+    oci->prev_mi = NULL;
+#endif
 
     oci->above_context = vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * oci->mb_cols, 1);
 
@@ -119,7 +139,10 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
         return 1;
     }
 
-    vp8_update_mode_info_border(oci->mi, oci->mb_rows, oci->mb_cols);
+    update_mode_info_border(oci->mi, oci->mb_rows, oci->mb_cols);
+#if CONFIG_ERROR_CONCEALMENT
+    update_mode_info_border(oci->prev_mi, oci->mb_rows, oci->mb_cols);
+#endif
 
     return 0;
 }
@@ -129,32 +152,32 @@ void vp8_setup_version(VP8_COMMON *cm)
     {
     case 0:
         cm->no_lpf = 0;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 0;
         cm->full_pixel = 0;
         break;
     case 1:
         cm->no_lpf = 0;
-        cm->simpler_lpf = 1;
+        cm->filter_type = SIMPLE_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 0;
         break;
     case 2:
         cm->no_lpf = 1;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 0;
         break;
     case 3:
         cm->no_lpf = 1;
-        cm->simpler_lpf = 1;
+        cm->filter_type = SIMPLE_LOOPFILTER;
         cm->use_bilinear_mc_filter = 1;
         cm->full_pixel = 1;
         break;
     default:
         /*4,5,6,7 are reserved for future use*/
         cm->no_lpf = 0;
-        cm->simpler_lpf = 0;
+        cm->filter_type = NORMAL_LOOPFILTER;
         cm->use_bilinear_mc_filter = 0;
         cm->full_pixel = 0;
         break;
@@ -169,7 +192,7 @@ void vp8_create_common(VP8_COMMON *oci)
 
     oci->mb_no_coeff_skip = 1;
     oci->no_lpf = 0;
-    oci->simpler_lpf = 0;
+    oci->filter_type = NORMAL_LOOPFILTER;
     oci->use_bilinear_mc_filter = 0;
     oci->full_pixel = 0;
     oci->multi_token_partition = ONE_PARTITION;

vp8/common/arm/arm_systemdependent.c

@@ -19,27 +19,22 @@
 #include "vp8/common/idct.h"
 #include "vp8/common/onyxc_int.h"
 
-extern void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_neon(MACROBLOCKD *x);
-
-extern void (*vp8_build_intra_predictors_mby_s_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x);
-
 void vp8_arch_arm_common_init(VP8_COMMON *ctx)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
     int flags = arm_cpu_caps();
-    int has_edsp = flags & HAS_EDSP;
-    int has_media = flags & HAS_MEDIA;
-    int has_neon = flags & HAS_NEON;
     rtcd->flags = flags;
 
     /* Override default functions with fastest ones for this CPU. */
+#if HAVE_ARMV5TE
+    if (flags & HAS_EDSP)
+    {
+    }
+#endif
+
 #if HAVE_ARMV6
-    if (has_media)
+    if (flags & HAS_MEDIA)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_armv6;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_armv6;
@@ -59,9 +54,11 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_armv6;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_armv6;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_armv6;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_armv6;
+        rtcd->loopfilter.simple_mb_v =
+                vp8_loop_filter_simple_vertical_edge_armv6;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_armv6;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_armv6;
+        rtcd->loopfilter.simple_mb_h =
+                vp8_loop_filter_simple_horizontal_edge_armv6;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_armv6;
 
         rtcd->recon.copy16x16   = vp8_copy_mem16x16_v6;
@@ -74,7 +71,7 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
 #endif
 
 #if HAVE_ARMV7
-    if (has_neon)
+    if (flags & HAS_NEON)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_neon;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_neon;
@@ -106,31 +103,12 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
         rtcd->recon.recon2      = vp8_recon2b_neon;
         rtcd->recon.recon4      = vp8_recon4b_neon;
         rtcd->recon.recon_mb    = vp8_recon_mb_neon;
-
+        rtcd->recon.build_intra_predictors_mby =
+            vp8_build_intra_predictors_mby_neon;
+        rtcd->recon.build_intra_predictors_mby_s =
+            vp8_build_intra_predictors_mby_s_neon;
     }
 #endif
 
 #endif
-
-#if HAVE_ARMV6
-#if CONFIG_RUNTIME_CPU_DETECT
-    if (has_media)
-#endif
-    {
-        vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby;
-        vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s;
-    }
-#endif
-
-#if HAVE_ARMV7
-#if CONFIG_RUNTIME_CPU_DETECT
-    if (has_neon)
-#endif
-    {
-        vp8_build_intra_predictors_mby_ptr =
-         vp8_build_intra_predictors_mby_neon;
-        vp8_build_intra_predictors_mby_s_ptr =
-         vp8_build_intra_predictors_mby_s_neon;
-    }
-#endif
 }

vp8/common/arm/armv6/bilinearfilter_v6.asm

@@ -30,11 +30,11 @@
     ldr     r4, [sp, #36]                   ; width
 
     mov     r12, r3                         ; outer-loop counter
-    sub     r2, r2, r4                      ; src increment for height loop
 
-    ;;IF ARCHITECTURE=6
-    pld     [r0]
-    ;;ENDIF
+    add     r7, r2, r4                      ; preload next row
+    pld     [r0, r7]
+
+    sub     r2, r2, r4                      ; src increment for height loop
 
     ldr     r5, [r11]                       ; load up filter coefficients
 
@@ -96,9 +96,8 @@
     add     r0, r0, r2                      ; move to next input row
     subs    r12, r12, #1
 
-    ;;IF ARCHITECTURE=6
-    pld     [r0]
-    ;;ENDIF
+    add     r9, r2, r4, lsl #1              ; adding back block width
+    pld     [r0, r9]                        ; preload next row
 
     add     r11, r11, #2                    ; move over to next column
     mov     r1, r11

vp8/common/arm/armv6/copymem16x16_v6.asm

@@ -22,9 +22,7 @@
     ;push   {r4-r7}
 
     ;preload
-    pld     [r0]
-    pld     [r0, r1]
-    pld     [r0, r1, lsl #1]
+    pld     [r0, #31]                ; preload for next 16x16 block
 
     ands    r4, r0, #15
     beq     copy_mem16x16_fast
@@ -90,6 +88,8 @@ copy_mem16x16_1_loop
     ldrneb  r6, [r0, #2]
     ldrneb  r7, [r0, #3]
 
+    pld     [r0, #31]               ; preload for next 16x16 block
+
     bne     copy_mem16x16_1_loop
 
     ldmia       sp!, {r4 - r7}
@@ -121,6 +121,8 @@ copy_mem16x16_4_loop
     ldrne   r6, [r0, #8]
     ldrne   r7, [r0, #12]
 
+    pld     [r0, #31]               ; preload for next 16x16 block
+
     bne     copy_mem16x16_4_loop
 
     ldmia       sp!, {r4 - r7}
@@ -148,6 +150,7 @@ copy_mem16x16_8_loop
 
     add     r2, r2, r3
 
+    pld     [r0, #31]               ; preload for next 16x16 block
     bne     copy_mem16x16_8_loop
 
     ldmia       sp!, {r4 - r7}
@@ -171,6 +174,7 @@ copy_mem16x16_fast_loop
     ;stm        r2, {r4-r7}
     add     r2, r2, r3
 
+    pld     [r0, #31]               ; preload for next 16x16 block
     bne     copy_mem16x16_fast_loop
 
     ldmia       sp!, {r4 - r7}

vp8/common/arm/armv6/filter_v6.asm

@@ -10,6 +10,8 @@
 
 
     EXPORT  |vp8_filter_block2d_first_pass_armv6|
+    EXPORT  |vp8_filter_block2d_first_pass_16x16_armv6|
+    EXPORT  |vp8_filter_block2d_first_pass_8x8_armv6|
     EXPORT  |vp8_filter_block2d_second_pass_armv6|
     EXPORT  |vp8_filter4_block2d_second_pass_armv6|
     EXPORT  |vp8_filter_block2d_first_pass_only_armv6|
@@ -40,11 +42,6 @@
     add     r12, r3, #16                    ; square off the output
     sub     sp, sp, #4
 
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, #-2]
-    ;;pld       [r0, #30]
-    ;;ENDIF
-
     ldr     r4, [r11]                       ; load up packed filter coefficients
     ldr     r5, [r11, #4]
     ldr     r6, [r11, #8]
@@ -101,15 +98,10 @@
 
     bne     width_loop_1st_6
 
-    ;;add       r9, r2, #30                 ; attempt to load 2 adjacent cache lines
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, r2]
-    ;;pld       [r0, r9]
-    ;;ENDIF
-
     ldr     r1, [sp]                        ; load and update dst address
     subs    r7, r7, #0x10000
     add     r0, r0, r2                      ; move to next input line
+
     add     r1, r1, #2                      ; move over to next column
     str     r1, [sp]
 
@@ -120,6 +112,192 @@
 
     ENDP
 
+; --------------------------
+; 16x16 version
+; -----------------------------
+|vp8_filter_block2d_first_pass_16x16_armv6| PROC
+    stmdb   sp!, {r4 - r11, lr}
+
+    ldr     r11, [sp, #40]                  ; vp8_filter address
+    ldr     r7, [sp, #36]                   ; output height
+
+    add     r4, r2, #18                     ; preload next low
+    pld     [r0, r4]
+
+    sub     r2, r2, r3                      ; inside loop increments input array,
+                                            ; so the height loop only needs to add
+                                            ; r2 - width to the input pointer
+
+    mov     r3, r3, lsl #1                  ; multiply width by 2 because using shorts
+    add     r12, r3, #16                    ; square off the output
+    sub     sp, sp, #4
+
+    ldr     r4, [r11]                       ; load up packed filter coefficients
+    ldr     r5, [r11, #4]
+    ldr     r6, [r11, #8]
+
+    str     r1, [sp]                        ; push destination to stack
+    mov     r7, r7, lsl #16                 ; height is top part of counter
+
+; six tap filter
+|height_loop_1st_16_6|
+    ldrb    r8, [r0, #-2]                   ; load source data
+    ldrb    r9, [r0, #-1]
+    ldrb    r10, [r0], #2
+    orr     r7, r7, r3, lsr #2              ; construct loop counter
+
+|width_loop_1st_16_6|
+    ldrb    r11, [r0, #-1]
+
+    pkhbt   lr, r8, r9, lsl #16             ; r9 | r8
+    pkhbt   r8, r9, r10, lsl #16            ; r10 | r9
+
+    ldrb    r9, [r0]
+
+    smuad   lr, lr, r4                      ; apply the filter
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+    smuad   r8, r8, r4
+    pkhbt   r11, r11, r9, lsl #16           ; r9 | r11
+
+    smlad   lr, r10, r5, lr
+    ldrb    r10, [r0, #1]
+    smlad   r8, r11, r5, r8
+    ldrb    r11, [r0, #2]
+
+    sub     r7, r7, #1
+
+    pkhbt   r9, r9, r10, lsl #16            ; r10 | r9
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+
+    smlad   lr, r9, r6, lr
+    smlad   r11, r10, r6, r8
+
+    ands    r10, r7, #0xff                  ; test loop counter
+
+    add     lr, lr, #0x40                   ; round_shift_and_clamp
+    ldrneb  r8, [r0, #-2]                   ; load data for next loop
+    usat    lr, #8, lr, asr #7
+    add     r11, r11, #0x40
+    ldrneb  r9, [r0, #-1]
+    usat    r11, #8, r11, asr #7
+
+    strh    lr, [r1], r12                   ; result is transposed and stored, which
+                                            ; will make second pass filtering easier.
+    ldrneb  r10, [r0], #2
+    strh    r11, [r1], r12
+
+    bne     width_loop_1st_16_6
+
+    ldr     r1, [sp]                        ; load and update dst address
+    subs    r7, r7, #0x10000
+    add     r0, r0, r2                      ; move to next input line
+
+    add     r11, r2, #34                    ; adding back block width(=16)
+    pld     [r0, r11]                       ; preload next low
+
+    add     r1, r1, #2                      ; move over to next column
+    str     r1, [sp]
+
+    bne     height_loop_1st_16_6
+
+    add     sp, sp, #4
+    ldmia   sp!, {r4 - r11, pc}
+
+    ENDP
+
+; --------------------------
+; 8x8 version
+; -----------------------------
+|vp8_filter_block2d_first_pass_8x8_armv6| PROC
+    stmdb   sp!, {r4 - r11, lr}
+
+    ldr     r11, [sp, #40]                  ; vp8_filter address
+    ldr     r7, [sp, #36]                   ; output height
+
+    add     r4, r2, #10                     ; preload next low
+    pld     [r0, r4]
+
+    sub     r2, r2, r3                      ; inside loop increments input array,
+                                            ; so the height loop only needs to add
+                                            ; r2 - width to the input pointer
+
+    mov     r3, r3, lsl #1                  ; multiply width by 2 because using shorts
+    add     r12, r3, #16                    ; square off the output
+    sub     sp, sp, #4
+
+    ldr     r4, [r11]                       ; load up packed filter coefficients
+    ldr     r5, [r11, #4]
+    ldr     r6, [r11, #8]
+
+    str     r1, [sp]                        ; push destination to stack
+    mov     r7, r7, lsl #16                 ; height is top part of counter
+
+; six tap filter
+|height_loop_1st_8_6|
+    ldrb    r8, [r0, #-2]                   ; load source data
+    ldrb    r9, [r0, #-1]
+    ldrb    r10, [r0], #2
+    orr     r7, r7, r3, lsr #2              ; construct loop counter
+
+|width_loop_1st_8_6|
+    ldrb    r11, [r0, #-1]
+
+    pkhbt   lr, r8, r9, lsl #16             ; r9 | r8
+    pkhbt   r8, r9, r10, lsl #16            ; r10 | r9
+
+    ldrb    r9, [r0]
+
+    smuad   lr, lr, r4                      ; apply the filter
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+    smuad   r8, r8, r4
+    pkhbt   r11, r11, r9, lsl #16           ; r9 | r11
+
+    smlad   lr, r10, r5, lr
+    ldrb    r10, [r0, #1]
+    smlad   r8, r11, r5, r8
+    ldrb    r11, [r0, #2]
+
+    sub     r7, r7, #1
+
+    pkhbt   r9, r9, r10, lsl #16            ; r10 | r9
+    pkhbt   r10, r10, r11, lsl #16          ; r11 | r10
+
+    smlad   lr, r9, r6, lr
+    smlad   r11, r10, r6, r8
+
+    ands    r10, r7, #0xff                  ; test loop counter
+
+    add     lr, lr, #0x40                   ; round_shift_and_clamp
+    ldrneb  r8, [r0, #-2]                   ; load data for next loop
+    usat    lr, #8, lr, asr #7
+    add     r11, r11, #0x40
+    ldrneb  r9, [r0, #-1]
+    usat    r11, #8, r11, asr #7
+
+    strh    lr, [r1], r12                   ; result is transposed and stored, which
+                                            ; will make second pass filtering easier.
+    ldrneb  r10, [r0], #2
+    strh    r11, [r1], r12
+
+    bne     width_loop_1st_8_6
+
+    ldr     r1, [sp]                        ; load and update dst address
+    subs    r7, r7, #0x10000
+    add     r0, r0, r2                      ; move to next input line
+
+    add     r11, r2, #18                    ; adding back block width(=8)
+    pld     [r0, r11]                       ; preload next low
+
+    add     r1, r1, #2                      ; move over to next column
+    str     r1, [sp]
+
+    bne     height_loop_1st_8_6
+
+    add     sp, sp, #4
+    ldmia   sp!, {r4 - r11, pc}
+
+    ENDP
+
 ;---------------------------------
 ; r0    short         *src_ptr,
 ; r1    unsigned char *output_ptr,
@@ -262,6 +440,10 @@
 |vp8_filter_block2d_first_pass_only_armv6| PROC
     stmdb   sp!, {r4 - r11, lr}
 
+    add     r7, r2, r3                      ; preload next low
+    add     r7, r7, #2
+    pld     [r0, r7]
+
     ldr     r4, [sp, #36]                   ; output pitch
     ldr     r11, [sp, #40]                  ; HFilter address
     sub     sp, sp, #8
@@ -330,16 +512,15 @@
 
     bne     width_loop_1st_only_6
 
-    ;;add       r9, r2, #30                 ; attempt to load 2 adjacent cache lines
-    ;;IF ARCHITECTURE=6
-    ;pld        [r0, r2]
-    ;;pld       [r0, r9]
-    ;;ENDIF
-
     ldr     lr, [sp]                        ; load back output pitch
     ldr     r12, [sp, #4]                   ; load back output pitch
     subs    r7, r7, #1
     add     r0, r0, r12                     ; updata src for next loop
+
+    add     r11, r12, r3                    ; preload next low
+    add     r11, r11, #2
+    pld     [r0, r11]
+
     add     r1, r1, lr                      ; update dst for next loop
 
     bne     height_loop_1st_only_6

vp8/common/arm/armv6/loopfilter_v6.asm

@@ -53,14 +53,11 @@ count       RN  r5
 
 ;r0     unsigned char *src_ptr,
 ;r1     int src_pixel_step,
-;r2     const char *flimit,
+;r2     const char *blimit,
 ;r3     const char *limit,
 ;stack  const char *thresh,
 ;stack  int  count
 
-;Note: 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.
-
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 |vp8_loop_filter_horizontal_edge_armv6| PROC
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
@@ -72,14 +69,18 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r9, [src], pstep            ; p3
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r10, [src], pstep           ; p2
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r11, [src], pstep           ; p1
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r6], #4                ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r6]                    ; thresh
+    orr         r2, r2, r2, lsl #8
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |Hnext8|
     ; vp8_filter_mask() function
@@ -253,12 +254,6 @@ count       RN  r5
 
     subs        count, count, #1
 
-    ;pld            [src]
-    ;pld            [src, pstep]
-    ;pld            [src, pstep, lsl #1]
-    ;pld            [src, pstep, lsl #2]
-    ;pld            [src, pstep, lsl #3]
-
     ldrne       r9, [src], pstep            ; p3
     ldrne       r10, [src], pstep           ; p2
     ldrne       r11, [src], pstep           ; p1
@@ -281,14 +276,18 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r9, [src], pstep            ; p3
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r10, [src], pstep           ; p2
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r11, [src], pstep           ; p1
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r6], #4                ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r6]                    ; thresh
+    orr         r2, r2, r2, lsl #8
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |MBHnext8|
 
@@ -590,15 +589,19 @@ count       RN  r5
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r6, [src], pstep            ; load source data
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
     ldr         r7, [src], pstep
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
     ldr         r8, [src], pstep
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r12], #4               ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r12]                   ; thresh
+    orr         r2, r2, r2, lsl #8
     ldr         lr, [src], pstep
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |Vnext8|
 
@@ -857,18 +860,26 @@ count       RN  r5
     sub         src, src, #4                ; move src pointer down by 4
     ldr         count, [sp, #40]            ; count for 8-in-parallel
     ldr         r12, [sp, #36]              ; load thresh address
+    pld         [src, #23]                  ; preload for next block
     sub         sp, sp, #16                 ; create temp buffer
 
     ldr         r6, [src], pstep            ; load source data
-    ldr         r4, [r2], #4                ; flimit
+    ldrb        r4, [r2]                    ; blimit
+    pld         [src, #23]
     ldr         r7, [src], pstep
-    ldr         r2, [r3], #4                ; limit
+    ldrb        r2, [r3]                    ; limit
+    pld         [src, #23]
     ldr         r8, [src], pstep
-    uadd8       r4, r4, r4                  ; flimit * 2
-    ldr         r3, [r12], #4               ; thresh
+    orr         r4, r4, r4, lsl #8
+    ldrb        r3, [r12]                   ; thresh
+    orr         r2, r2, r2, lsl #8
+    pld         [src, #23]
     ldr         lr, [src], pstep
     mov         count, count, lsl #1        ; 4-in-parallel
-    uadd8       r4, r4, r2                  ; flimit * 2 + limit
+    orr         r4, r4, r4, lsl #16
+    orr         r3, r3, r3, lsl #8
+    orr         r2, r2, r2, lsl #16
+    orr         r3, r3, r3, lsl #16
 
 |MBVnext8|
     ; vp8_filter_mask() function
@@ -908,6 +919,7 @@ count       RN  r5
     str         lr, [sp, #8]
     ldr         lr, [src], pstep
 
+
     TRANSPOSE_MATRIX r6, r7, r8, lr, r9, r10, r11, r12
 
     ldr         lr, [sp, #8]                ; load back (f)limit accumulator
@@ -956,6 +968,7 @@ count       RN  r5
     beq         mbvskip_filter               ; skip filtering
 
 
+
     ;vp8_hevmask() function
     ;calculate high edge variance
 
@@ -1123,6 +1136,7 @@ count       RN  r5
     smlabb      r8, r6, lr, r7
     smlatb      r6, r6, lr, r7
     smlabb      r9, r10, lr, r7
+
     smlatb      r10, r10, lr, r7
     ssat        r8, #8, r8, asr #7
     ssat        r6, #8, r6, asr #7
@@ -1242,9 +1256,13 @@ count       RN  r5
     sub         src, src, #4
     subs        count, count, #1
 
+    pld         [src, #23]                  ; preload for next block
     ldrne       r6, [src], pstep            ; load source data
+    pld         [src, #23]
     ldrne       r7, [src], pstep
+    pld         [src, #23]
     ldrne       r8, [src], pstep
+    pld         [src, #23]
     ldrne       lr, [src], pstep
 
     bne         MBVnext8

vp8/common/arm/armv6/simpleloopfilter_v6.asm

@@ -45,35 +45,28 @@
     MEND
 
 
+
 src         RN  r0
 pstep       RN  r1
 
 ;r0     unsigned char *src_ptr,
 ;r1     int src_pixel_step,
-;r2     const char *flimit,
-;r3     const char *limit,
-;stack  const char *thresh,
-;stack  int  count
-
-; All 16 elements in flimit are equal. So, in the code, only one load is needed
-; for flimit. Same applies to limit. thresh is not used in simple looopfilter
+;r2     const char *blimit
 
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 |vp8_loop_filter_simple_horizontal_edge_armv6| PROC
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    ldr         r12, [r3]                   ; limit
+    ldrb        r12, [r2]                   ; blimit
     ldr         r3, [src, -pstep, lsl #1]   ; p1
     ldr         r4, [src, -pstep]           ; p0
     ldr         r5, [src]                   ; q0
     ldr         r6, [src, pstep]            ; q1
-    ldr         r7, [r2]                    ; flimit
+    orr         r12, r12, r12, lsl #8       ; blimit
     ldr         r2, c0x80808080
-    ldr         r9, [sp, #40]               ; count for 8-in-parallel
-    uadd8       r7, r7, r7                  ; flimit * 2
-    mov         r9, r9, lsl #1              ; double the count. we're doing 4 at a time
-    uadd8       r12, r7, r12                ; flimit * 2 + limit
+    orr         r12, r12, r12, lsl #16      ; blimit
+    mov         r9, #4                      ; double the count. we're doing 4 at a time
     mov         lr, #0                      ; need 0 in a couple places
 
 |simple_hnext8|
@@ -148,30 +141,32 @@ pstep       RN  r1
 ;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
     stmdb       sp!, {r4 - r11, lr}
 
-    ldr         r12, [r2]                   ; r12: flimit
+    ldrb        r12, [r2]                   ; r12: blimit
     ldr         r2, c0x80808080
-    ldr         r7, [r3]                    ; limit
+    orr         r12, r12, r12, lsl #8
 
     ; load soure data to r7, r8, r9, r10
     ldrh        r3, [src, #-2]
+    pld         [src, #23]                  ; preload for next block
     ldrh        r4, [src], pstep
-    uadd8       r12, r12, r12               ; flimit * 2
+    orr         r12, r12, r12, lsl #16
 
     ldrh        r5, [src, #-2]
+    pld         [src, #23]
     ldrh        r6, [src], pstep
-    uadd8       r12, r12, r7                ; flimit * 2 + limit
 
     pkhbt       r7, r3, r4, lsl #16
 
     ldrh        r3, [src, #-2]
+    pld         [src, #23]
     ldrh        r4, [src], pstep
-    ldr         r11, [sp, #40]              ; count (r11) for 8-in-parallel
 
     pkhbt       r8, r5, r6, lsl #16
 
     ldrh        r5, [src, #-2]
+    pld         [src, #23]
     ldrh        r6, [src], pstep
-    mov         r11, r11, lsl #1            ; 4-in-parallel
+    mov         r11, #4                     ; double the count. we're doing 4 at a time
 
 |simple_vnext8|
     ; vp8_simple_filter_mask() function
@@ -259,19 +254,23 @@ pstep       RN  r1
 
     ; load soure data to r7, r8, r9, r10
     ldrneh      r3, [src, #-2]
+    pld         [src, #23]                  ; preload for next block
     ldrneh      r4, [src], pstep
 
     ldrneh      r5, [src, #-2]
+    pld         [src, #23]
     ldrneh      r6, [src], pstep
 
     pkhbt       r7, r3, r4, lsl #16
 
     ldrneh      r3, [src, #-2]
+    pld         [src, #23]
     ldrneh      r4, [src], pstep
 
     pkhbt       r8, r5, r6, lsl #16
 
     ldrneh      r5, [src, #-2]
+    pld         [src, #23]
     ldrneh      r6, [src], pstep
 
     bne         simple_vnext8

vp8/common/arm/armv6/sixtappredict8x4_v6.asm

@@ -32,9 +32,12 @@
     beq         skip_firstpass_filter
 
 ;first-pass filter
-    ldr         r12, _filter8_coeff_
+    adr         r12, filter8_coeff
     sub         r0, r0, r1, lsl #1
 
+    add         r3, r1, #10                 ; preload next low
+    pld         [r0, r3]
+
     add         r2, r12, r2, lsl #4         ;calculate filter location
     add         r0, r0, #3                  ;adjust src only for loading convinience
 
@@ -110,6 +113,9 @@
 
     add         r0, r0, r1                  ; move to next input line
 
+    add         r11, r1, #18                ; preload next low. adding back block width(=8), which is subtracted earlier
+    pld         [r0, r11]
+
     bne         first_pass_hloop_v6
 
 ;second pass filter
@@ -121,7 +127,7 @@ secondpass_filter
     cmp         r3, #0
     beq         skip_secondpass_filter
 
-    ldr         r12, _filter8_coeff_
+    adr         r12, filter8_coeff
     add         lr, r12, r3, lsl #4         ;calculate filter location
 
     mov         r2, #0x00080000
@@ -245,8 +251,6 @@ skip_secondpass_hloop
 ;-----------------
 ;One word each is reserved. Label filter_coeff can be used to access the data.
 ;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-_filter8_coeff_
-    DCD     filter8_coeff
 filter8_coeff
     DCD     0x00000000,     0x00000080,     0x00000000,     0x00000000
     DCD     0xfffa0000,     0x000c007b,     0x0000ffff,     0x00000000

vp8/common/arm/filter_arm.c

@@ -25,6 +25,28 @@ extern void vp8_filter_block2d_first_pass_armv6
     const short *vp8_filter
 );
 
+// 8x8
+extern void vp8_filter_block2d_first_pass_8x8_armv6
+(
+    unsigned char *src_ptr,
+    short         *output_ptr,
+    unsigned int src_pixels_per_line,
+    unsigned int output_width,
+    unsigned int output_height,
+    const short *vp8_filter
+);
+
+// 16x16
+extern void vp8_filter_block2d_first_pass_16x16_armv6
+(
+    unsigned char *src_ptr,
+    short         *output_ptr,
+    unsigned int src_pixels_per_line,
+    unsigned int output_width,
+    unsigned int output_height,
+    const short *vp8_filter
+);
+
 extern void vp8_filter_block2d_second_pass_armv6
 (
     short         *src_ptr,
@@ -143,12 +165,12 @@ void vp8_sixtap_predict8x8_armv6
     {
         if (yoffset & 0x1)
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 8, 11, HFilter);
+            vp8_filter_block2d_first_pass_8x8_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 8, 11, HFilter);
             vp8_filter4_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 8, VFilter);
         }
         else
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 8, 13, HFilter);
+            vp8_filter_block2d_first_pass_8x8_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 8, 13, HFilter);
             vp8_filter_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 8, VFilter);
         }
     }
@@ -185,12 +207,12 @@ void vp8_sixtap_predict16x16_armv6
     {
         if (yoffset & 0x1)
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 16, 19, HFilter);
+            vp8_filter_block2d_first_pass_16x16_armv6(src_ptr - src_pixels_per_line, FData + 1, src_pixels_per_line, 16, 19, HFilter);
             vp8_filter4_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 16, VFilter);
         }
         else
         {
-            vp8_filter_block2d_first_pass_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 16, 21, HFilter);
+            vp8_filter_block2d_first_pass_16x16_armv6(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 16, 21, HFilter);
             vp8_filter_block2d_second_pass_armv6(FData + 2, dst_ptr, dst_pitch, 16, VFilter);
         }
     }

vp8/common/arm/loopfilter_arm.c

@@ -9,135 +9,107 @@
  */
 
 
-#include "vpx_ports/config.h"
-#include <math.h>
+#include "vpx_config.h"
 #include "vp8/common/loopfilter.h"
 #include "vp8/common/onyxc_int.h"
 
+#if HAVE_ARMV6
 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);
 extern prototype_loopfilter(vp8_mbloop_filter_vertical_edge_armv6);
-extern prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_armv6);
-extern prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_armv6);
+#endif
 
-extern prototype_loopfilter(vp8_loop_filter_horizontal_edge_y_neon);
-extern prototype_loopfilter(vp8_loop_filter_vertical_edge_y_neon);
-extern prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_y_neon);
-extern prototype_loopfilter(vp8_mbloop_filter_vertical_edge_y_neon);
-extern prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_neon);
-extern prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_neon);
+#if HAVE_ARMV7
+typedef void loopfilter_y_neon(unsigned char *src, int pitch,
+        unsigned char blimit, unsigned char limit, unsigned char thresh);
+typedef void loopfilter_uv_neon(unsigned char *u, int pitch,
+        unsigned char blimit, unsigned char limit, unsigned char thresh,
+        unsigned char *v);
 
-extern loop_filter_uvfunction vp8_loop_filter_horizontal_edge_uv_neon;
-extern loop_filter_uvfunction vp8_loop_filter_vertical_edge_uv_neon;
-extern loop_filter_uvfunction vp8_mbloop_filter_horizontal_edge_uv_neon;
-extern loop_filter_uvfunction vp8_mbloop_filter_vertical_edge_uv_neon;
+extern loopfilter_y_neon vp8_loop_filter_horizontal_edge_y_neon;
+extern loopfilter_y_neon vp8_loop_filter_vertical_edge_y_neon;
+extern loopfilter_y_neon vp8_mbloop_filter_horizontal_edge_y_neon;
+extern loopfilter_y_neon vp8_mbloop_filter_vertical_edge_y_neon;
 
+extern loopfilter_uv_neon vp8_loop_filter_horizontal_edge_uv_neon;
+extern loopfilter_uv_neon vp8_loop_filter_vertical_edge_uv_neon;
+extern loopfilter_uv_neon vp8_mbloop_filter_horizontal_edge_uv_neon;
+extern loopfilter_uv_neon vp8_mbloop_filter_vertical_edge_uv_neon;
+#endif
 
 #if HAVE_ARMV6
 /*ARMV6 loopfilter functions*/
 /* Horizontal MB filtering */
 void vp8_loop_filter_mbh_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                               int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-void vp8_loop_filter_mbhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                                int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                               int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-void vp8_loop_filter_mbvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                                int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_armv6(unsigned char *y_ptr, int y_stride,
+                               const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, int y_stride,
+                               const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_armv6(y_ptr + 12, y_stride, blimit);
 }
 #endif
 
@@ -145,93 +117,60 @@ void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsig
 /* NEON loopfilter functions */
 /* Horizontal MB filtering */
 void vp8_loop_filter_mbh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    unsigned char mblim = *lfi->mblim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+    vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, mblim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
-}
-
-void vp8_loop_filter_mbhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, mblim, lim, hev_thr, v_ptr);
 }
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    unsigned char mblim = *lfi->mblim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, mblim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
-}
-
-void vp8_loop_filter_mbvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, mblim, lim, hev_thr, v_ptr);
 }
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    unsigned char blim = *lfi->blim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 4 * y_stride, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 8 * y_stride, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 12 * y_stride, y_stride, blim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4 * uv_stride);
-}
-
-void vp8_loop_filter_bhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_neon(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+        vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, blim, lim, hev_thr, v_ptr + 4 * uv_stride);
 }
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    unsigned char blim = *lfi->blim;
+    unsigned char lim = *lfi->lim;
+    unsigned char hev_thr = *lfi->hev_thr;
+
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 4, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 8, y_stride, blim, lim, hev_thr);
+    vp8_loop_filter_vertical_edge_y_neon(y_ptr + 12, y_stride, blim, lim, hev_thr);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4);
-}
-
-void vp8_loop_filter_bvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_neon(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+        vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, blim, lim, hev_thr, v_ptr + 4);
 }
 #endif

vp8/common/arm/loopfilter_arm.h

@@ -12,15 +12,17 @@
 #ifndef LOOPFILTER_ARM_H
 #define LOOPFILTER_ARM_H
 
+#include "vpx_config.h"
+
 #if HAVE_ARMV6
 extern prototype_loopfilter_block(vp8_loop_filter_mbv_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_armv6);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_armv6);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_armv6);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_armv6);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_lf_normal_mb_v
@@ -36,28 +38,29 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_armv6);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_armv6
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_armv6
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_armv6
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_armv6
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_armv6
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_armv6
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_armv6
-#endif
-#endif
+#endif /* !CONFIG_RUNTIME_CPU_DETECT */
+
+#endif /* HAVE_ARMV6 */
 
 #if HAVE_ARMV7
 extern prototype_loopfilter_block(vp8_loop_filter_mbv_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_neon);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_neon);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_mbvs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_mbhs_neon);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_neon);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_lf_normal_mb_v
@@ -83,7 +86,8 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_neon);
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_neon
-#endif
-#endif
+#endif /* !CONFIG_RUNTIME_CPU_DETECT */
 
-#endif
+#endif /* HAVE_ARMV7 */
+
+#endif /* LOOPFILTER_ARM_H */

vp8/common/arm/neon/bilinearpredict16x16_neon.asm

@@ -25,7 +25,7 @@
 |vp8_bilinear_predict16x16_neon| PROC
     push            {r4-r5, lr}
 
-    ldr             r12, _bifilter16_coeff_
+    adr             r12, bifilter16_coeff
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
 
@@ -351,8 +351,6 @@ filt_blk2d_spo16x16_loop_neon
 
 ;-----------------
 
-_bifilter16_coeff_
-    DCD     bifilter16_coeff
 bifilter16_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

vp8/common/arm/neon/bilinearpredict4x4_neon.asm

@@ -25,7 +25,7 @@
 |vp8_bilinear_predict4x4_neon| PROC
     push            {r4, lr}
 
-    ldr             r12, _bifilter4_coeff_
+    adr             r12, bifilter4_coeff
     ldr             r4, [sp, #8]            ;load parameters from stack
     ldr             lr, [sp, #12]           ;load parameters from stack
 
@@ -124,8 +124,6 @@ skip_secondpass_filter
 
 ;-----------------
 
-_bifilter4_coeff_
-    DCD     bifilter4_coeff
 bifilter4_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

vp8/common/arm/neon/bilinearpredict8x4_neon.asm

@@ -25,7 +25,7 @@
 |vp8_bilinear_predict8x4_neon| PROC
     push            {r4, lr}
 
-    ldr             r12, _bifilter8x4_coeff_
+    adr             r12, bifilter8x4_coeff
     ldr             r4, [sp, #8]            ;load parameters from stack
     ldr             lr, [sp, #12]           ;load parameters from stack
 
@@ -129,8 +129,6 @@ skip_secondpass_filter
 
 ;-----------------
 
-_bifilter8x4_coeff_
-    DCD     bifilter8x4_coeff
 bifilter8x4_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

vp8/common/arm/neon/bilinearpredict8x8_neon.asm

@@ -25,7 +25,7 @@
 |vp8_bilinear_predict8x8_neon| PROC
     push            {r4, lr}
 
-    ldr             r12, _bifilter8_coeff_
+    adr             r12, bifilter8_coeff
     ldr             r4, [sp, #8]            ;load parameters from stack
     ldr             lr, [sp, #12]           ;load parameters from stack
 
@@ -177,8 +177,6 @@ skip_secondpass_filter
 
 ;-----------------
 
-_bifilter8_coeff_
-    DCD     bifilter8_coeff
 bifilter8_coeff
     DCD     128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
 

vp8/common/arm/neon/iwalsh_neon.asm

@@ -20,19 +20,16 @@
 |vp8_short_inv_walsh4x4_neon| PROC
 
     ; read in all four lines of values: d0->d3
-    vldm.64 r0, {q0, q1}
+    vld1.i16 {q0-q1}, [r0@128]
 
     ; first for loop
-
     vadd.s16 d4, d0, d3 ;a = [0] + [12]
-    vadd.s16 d5, d1, d2 ;b = [4] + [8]
-    vsub.s16 d6, d1, d2 ;c = [4] - [8]
-    vsub.s16 d7, d0, d3 ;d = [0] - [12]
+    vadd.s16 d6, d1, d2 ;b = [4] + [8]
+    vsub.s16 d5, d0, d3 ;d = [0] - [12]
+    vsub.s16 d7, d1, d2 ;c = [4] - [8]
 
-    vadd.s16 d0, d4, d5 ;a + b
-    vadd.s16 d1, d6, d7 ;c + d
-    vsub.s16 d2, d4, d5 ;a - b
-    vsub.s16 d3, d7, d6 ;d - c
+    vadd.s16 q0, q2, q3 ; a+b d+c
+    vsub.s16 q1, q2, q3 ; a-b d-c
 
     vtrn.32 d0, d2 ;d0:  0  1  8  9
                    ;d2:  2  3 10 11
@@ -47,29 +44,22 @@
     ; second for loop
 
     vadd.s16 d4, d0, d3 ;a = [0] + [3]
-    vadd.s16 d5, d1, d2 ;b = [1] + [2]
-    vsub.s16 d6, d1, d2 ;c = [1] - [2]
-    vsub.s16 d7, d0, d3 ;d = [0] - [3]
+    vadd.s16 d6, d1, d2 ;b = [1] + [2]
+    vsub.s16 d5, d0, d3 ;d = [0] - [3]
+    vsub.s16 d7, d1, d2 ;c = [1] - [2]
 
-    vadd.s16 d0, d4, d5 ;e = a + b
-    vadd.s16 d1, d6, d7 ;f = c + d
-    vsub.s16 d2, d4, d5 ;g = a - b
-    vsub.s16 d3, d7, d6 ;h = d - c
+    vmov.i16 q8, #3
 
-    vmov.i16 q2, #3
-    vadd.i16 q0, q0, q2 ;e/f += 3
-    vadd.i16 q1, q1, q2 ;g/h += 3
+    vadd.s16 q0, q2, q3 ; a+b d+c
+    vsub.s16 q1, q2, q3 ; a-b d-c
+
+    vadd.i16 q0, q0, q8 ;e/f += 3
+    vadd.i16 q1, q1, q8 ;g/h += 3
 
     vshr.s16 q0, q0, #3 ;e/f >> 3
     vshr.s16 q1, q1, #3 ;g/h >> 3
 
-    vtrn.32 d0, d2
-    vtrn.32 d1, d3
-    vtrn.16 d0, d1
-    vtrn.16 d2, d3
-
-    vstmia.16 r1!, {q0}
-    vstmia.16 r1!, {q1}
+    vst4.i16 {d0,d1,d2,d3}, [r1@128]
 
     bx lr
     ENDP    ; |vp8_short_inv_walsh4x4_neon|
@@ -77,19 +67,13 @@
 
 ;short vp8_short_inv_walsh4x4_1_neon(short *input, short *output)
 |vp8_short_inv_walsh4x4_1_neon| PROC
-    ; load a full line into a neon register
-    vld1.16  {q0}, [r0]
-    ; extract first element and replicate
-    vdup.16 q1, d0[0]
-    ; add 3 to all values
-    vmov.i16 q2, #3
-    vadd.i16 q3, q1, q2
-    ; right shift
-    vshr.s16 q3, q3, #3
-    ; write it back
-    vstmia.16 r1!, {q3}
-    vstmia.16 r1!, {q3}
-
+    ldrsh r2, [r0]          ; load input[0]
+    add r3, r2, #3          ; add 3
+    add r2, r1, #16         ; base for last 8 output
+    asr r0, r3, #3          ; right shift 3
+    vdup.16 q0, r0          ; load and duplicate
+    vst1.16 {q0}, [r1@128]  ; write back 8
+    vst1.16 {q0}, [r2@128]  ; write back last 8
     bx lr
     ENDP    ; |vp8_short_inv_walsh4x4_1_neon|
 

vp8/common/arm/neon/loopfilter_neon.asm

@@ -14,109 +14,97 @@
     EXPORT  |vp8_loop_filter_vertical_edge_y_neon|
     EXPORT  |vp8_loop_filter_vertical_edge_uv_neon|
     ARM
-    REQUIRE8
-    PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
 
-; flimit, limit, and thresh should be positive numbers.
-; All 16 elements in these variables are equal.
-
-; void vp8_loop_filter_horizontal_edge_y_neon(unsigned char *src, int pitch,
-;                                             const signed char *flimit,
-;                                             const signed char *limit,
-;                                             const signed char *thresh,
-;                                             int count)
 ; r0    unsigned char *src
 ; r1    int pitch
-; r2    const signed char *flimit
-; r3    const signed char *limit
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_loop_filter_horizontal_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    sub         r2, r0, r1, lsl #2          ; move src pointer down by 4 lines
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    vdup.u8     q0, r2                     ; duplicate blimit
+    vdup.u8     q1, r3                     ; duplicate limit
+    sub         r2, r0, r1, lsl #2         ; move src pointer down by 4 lines
+    ldr         r3, [sp, #4]               ; load thresh
+    add         r12, r2, r1
+    add         r1, r1, r1
 
-    vld1.u8     {q3}, [r2], r1              ; p3
-    vld1.u8     {q4}, [r2], r1              ; p2
-    vld1.u8     {q5}, [r2], r1              ; p1
-    vld1.u8     {q6}, [r2], r1              ; p0
-    vld1.u8     {q7}, [r2], r1              ; q0
-    vld1.u8     {q8}, [r2], r1              ; q1
-    vld1.u8     {q9}, [r2], r1              ; q2
-    vld1.u8     {q10}, [r2]                 ; q3
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    sub         r0, r0, r1, lsl #1
+    vdup.u8     q2, r3                     ; duplicate thresh
+
+    vld1.u8     {q3}, [r2@128], r1              ; p3
+    vld1.u8     {q4}, [r12@128], r1             ; p2
+    vld1.u8     {q5}, [r2@128], r1              ; p1
+    vld1.u8     {q6}, [r12@128], r1             ; p0
+    vld1.u8     {q7}, [r2@128], r1              ; q0
+    vld1.u8     {q8}, [r12@128], r1             ; q1
+    vld1.u8     {q9}, [r2@128]                  ; q2
+    vld1.u8     {q10}, [r12@128]                ; q3
+
+    sub         r2, r2, r1, lsl #1
+    sub         r12, r12, r1, lsl #1
 
     bl          vp8_loop_filter_neon
 
-    vst1.u8     {q5}, [r0], r1              ; store op1
-    vst1.u8     {q6}, [r0], r1              ; store op0
-    vst1.u8     {q7}, [r0], r1              ; store oq0
-    vst1.u8     {q8}, [r0], r1              ; store oq1
+    vst1.u8     {q5}, [r2@128], r1              ; store op1
+    vst1.u8     {q6}, [r12@128], r1             ; store op0
+    vst1.u8     {q7}, [r2@128], r1              ; store oq0
+    vst1.u8     {q8}, [r12@128], r1             ; store oq1
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_horizontal_edge_y_neon|
 
-; void vp8_loop_filter_horizontal_edge_uv_neon(unsigned char *u, int pitch
-;                                              const signed char *flimit,
-;                                              const signed char *limit,
-;                                              const signed char *thresh,
-;                                              unsigned char *v)
+
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
 |vp8_loop_filter_horizontal_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
+    push        {lr}
+    vdup.u8     q0, r2                      ; duplicate blimit
+    vdup.u8     q1, r3                      ; duplicate limit
+    ldr         r12, [sp, #4]               ; load thresh
     ldr         r2, [sp, #8]                ; load v ptr
+    vdup.u8     q2, r12                     ; duplicate thresh
 
     sub         r3, r0, r1, lsl #2          ; move u pointer down by 4 lines
-    vld1.u8     {d6}, [r3], r1              ; p3
-    vld1.u8     {d8}, [r3], r1              ; p2
-    vld1.u8     {d10}, [r3], r1             ; p1
-    vld1.u8     {d12}, [r3], r1             ; p0
-    vld1.u8     {d14}, [r3], r1             ; q0
-    vld1.u8     {d16}, [r3], r1             ; q1
-    vld1.u8     {d18}, [r3], r1             ; q2
-    vld1.u8     {d20}, [r3]                 ; q3
-
-    ldr         r3, [sp, #4]                ; load thresh pointer
-
     sub         r12, r2, r1, lsl #2         ; move v pointer down by 4 lines
-    vld1.u8     {d7}, [r12], r1             ; p3
-    vld1.u8     {d9}, [r12], r1             ; p2
-    vld1.u8     {d11}, [r12], r1            ; p1
-    vld1.u8     {d13}, [r12], r1            ; p0
-    vld1.u8     {d15}, [r12], r1            ; q0
-    vld1.u8     {d17}, [r12], r1            ; q1
-    vld1.u8     {d19}, [r12], r1            ; q2
-    vld1.u8     {d21}, [r12]                ; q3
 
-    vld1.s8     {d4[], d5[]}, [r3]          ; thresh
+    vld1.u8     {d6}, [r3@64], r1              ; p3
+    vld1.u8     {d7}, [r12@64], r1             ; p3
+    vld1.u8     {d8}, [r3@64], r1              ; p2
+    vld1.u8     {d9}, [r12@64], r1             ; p2
+    vld1.u8     {d10}, [r3@64], r1             ; p1
+    vld1.u8     {d11}, [r12@64], r1            ; p1
+    vld1.u8     {d12}, [r3@64], r1             ; p0
+    vld1.u8     {d13}, [r12@64], r1            ; p0
+    vld1.u8     {d14}, [r3@64], r1             ; q0
+    vld1.u8     {d15}, [r12@64], r1            ; q0
+    vld1.u8     {d16}, [r3@64], r1             ; q1
+    vld1.u8     {d17}, [r12@64], r1            ; q1
+    vld1.u8     {d18}, [r3@64], r1             ; q2
+    vld1.u8     {d19}, [r12@64], r1            ; q2
+    vld1.u8     {d20}, [r3@64]                 ; q3
+    vld1.u8     {d21}, [r12@64]                ; q3
 
     bl          vp8_loop_filter_neon
 
     sub         r0, r0, r1, lsl #1
     sub         r2, r2, r1, lsl #1
 
-    vst1.u8     {d10}, [r0], r1             ; store u op1
-    vst1.u8     {d11}, [r2], r1             ; store v op1
-    vst1.u8     {d12}, [r0], r1             ; store u op0
-    vst1.u8     {d13}, [r2], r1             ; store v op0
-    vst1.u8     {d14}, [r0], r1             ; store u oq0
-    vst1.u8     {d15}, [r2], r1             ; store v oq0
-    vst1.u8     {d16}, [r0]                 ; store u oq1
-    vst1.u8     {d17}, [r2]                 ; store v oq1
+    vst1.u8     {d10}, [r0@64], r1             ; store u op1
+    vst1.u8     {d11}, [r2@64], r1             ; store v op1
+    vst1.u8     {d12}, [r0@64], r1             ; store u op0
+    vst1.u8     {d13}, [r2@64], r1             ; store v op0
+    vst1.u8     {d14}, [r0@64], r1             ; store u oq0
+    vst1.u8     {d15}, [r2@64], r1             ; store v oq0
+    vst1.u8     {d16}, [r0@64]                 ; store u oq1
+    vst1.u8     {d17}, [r2@64]                 ; store v oq1
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_horizontal_edge_uv_neon|
 
 ; void vp8_loop_filter_vertical_edge_y_neon(unsigned char *src, int pitch,
@@ -124,39 +112,38 @@
 ;                                           const signed char *limit,
 ;                                           const signed char *thresh,
 ;                                           int count)
-; r0    unsigned char *src,
-; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r0    unsigned char *src
+; r1    int pitch
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
+
 |vp8_loop_filter_vertical_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    sub         r2, r0, #4                  ; src ptr down by 4 columns
-    sub         r0, r0, #2                  ; dst ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    vdup.u8     q0, r2                     ; duplicate blimit
+    vdup.u8     q1, r3                     ; duplicate limit
+    sub         r2, r0, #4                 ; src ptr down by 4 columns
+    add         r1, r1, r1
+    ldr         r3, [sp, #4]               ; load thresh
+    add         r12, r2, r1, asr #1
 
-    vld1.u8     {d6}, [r2], r1              ; load first 8-line src data
-    vld1.u8     {d8}, [r2], r1
+    vld1.u8     {d6}, [r2], r1
+    vld1.u8     {d8}, [r12], r1
     vld1.u8     {d10}, [r2], r1
-    vld1.u8     {d12}, [r2], r1
+    vld1.u8     {d12}, [r12], r1
     vld1.u8     {d14}, [r2], r1
-    vld1.u8     {d16}, [r2], r1
+    vld1.u8     {d16}, [r12], r1
     vld1.u8     {d18}, [r2], r1
-    vld1.u8     {d20}, [r2], r1
-
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.u8     {d20}, [r12], r1
 
     vld1.u8     {d7}, [r2], r1              ; load second 8-line src data
-    vld1.u8     {d9}, [r2], r1
+    vld1.u8     {d9}, [r12], r1
     vld1.u8     {d11}, [r2], r1
-    vld1.u8     {d13}, [r2], r1
+    vld1.u8     {d13}, [r12], r1
     vld1.u8     {d15}, [r2], r1
-    vld1.u8     {d17}, [r2], r1
-    vld1.u8     {d19}, [r2], r1
-    vld1.u8     {d21}, [r2]
+    vld1.u8     {d17}, [r12], r1
+    vld1.u8     {d19}, [r2]
+    vld1.u8     {d21}, [r12]
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -164,6 +151,8 @@
     vtrn.32     q5, q9
     vtrn.32     q6, q10
 
+    vdup.u8     q2, r3                     ; duplicate thresh
+
     vtrn.16     q3, q5
     vtrn.16     q4, q6
     vtrn.16     q7, q9
@@ -178,28 +167,34 @@
 
     vswp        d12, d11
     vswp        d16, d13
+
+    sub         r0, r0, #2                 ; dst ptr
+
     vswp        d14, d12
     vswp        d16, d15
 
+    add         r12, r0, r1, asr #1
+
     ;store op1, op0, oq0, oq1
     vst4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
-    vst4.8      {d10[1], d11[1], d12[1], d13[1]}, [r0], r1
+    vst4.8      {d10[1], d11[1], d12[1], d13[1]}, [r12], r1
     vst4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r1
-    vst4.8      {d10[3], d11[3], d12[3], d13[3]}, [r0], r1
+    vst4.8      {d10[3], d11[3], d12[3], d13[3]}, [r12], r1
     vst4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r1
-    vst4.8      {d10[5], d11[5], d12[5], d13[5]}, [r0], r1
+    vst4.8      {d10[5], d11[5], d12[5], d13[5]}, [r12], r1
     vst4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r1
-    vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0], r1
-    vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r0], r1
-    vst4.8      {d14[1], d15[1], d16[1], d17[1]}, [r0], r1
-    vst4.8      {d14[2], d15[2], d16[2], d17[2]}, [r0], r1
-    vst4.8      {d14[3], d15[3], d16[3], d17[3]}, [r0], r1
-    vst4.8      {d14[4], d15[4], d16[4], d17[4]}, [r0], r1
-    vst4.8      {d14[5], d15[5], d16[5], d17[5]}, [r0], r1
-    vst4.8      {d14[6], d15[6], d16[6], d17[6]}, [r0], r1
-    vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r0]
+    vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r12], r1
 
-    ldmia       sp!, {pc}
+    vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r0], r1
+    vst4.8      {d14[1], d15[1], d16[1], d17[1]}, [r12], r1
+    vst4.8      {d14[2], d15[2], d16[2], d17[2]}, [r0], r1
+    vst4.8      {d14[3], d15[3], d16[3], d17[3]}, [r12], r1
+    vst4.8      {d14[4], d15[4], d16[4], d17[4]}, [r0], r1
+    vst4.8      {d14[5], d15[5], d16[5], d17[5]}, [r12], r1
+    vst4.8      {d14[6], d15[6], d16[6], d17[6]}, [r0]
+    vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r12]
+
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_vertical_edge_y_neon|
 
 ; void vp8_loop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch
@@ -209,38 +204,36 @@
 ;                                            unsigned char *v)
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
 |vp8_loop_filter_vertical_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r12, r0, #4                  ; move u pointer down by 4 columns
-    vld1.s8     {d0[], d1[]}, [r2]          ; flimit
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-
+    push        {lr}
+    vdup.u8     q0, r2                      ; duplicate blimit
+    sub         r12, r0, #4                 ; move u pointer down by 4 columns
     ldr         r2, [sp, #8]                ; load v ptr
-
-    vld1.u8     {d6}, [r12], r1              ;load u data
-    vld1.u8     {d8}, [r12], r1
-    vld1.u8     {d10}, [r12], r1
-    vld1.u8     {d12}, [r12], r1
-    vld1.u8     {d14}, [r12], r1
-    vld1.u8     {d16}, [r12], r1
-    vld1.u8     {d18}, [r12], r1
-    vld1.u8     {d20}, [r12]
-
+    vdup.u8     q1, r3                      ; duplicate limit
     sub         r3, r2, #4                  ; move v pointer down by 4 columns
+
+    vld1.u8     {d6}, [r12], r1             ;load u data
     vld1.u8     {d7}, [r3], r1              ;load v data
+    vld1.u8     {d8}, [r12], r1
     vld1.u8     {d9}, [r3], r1
+    vld1.u8     {d10}, [r12], r1
     vld1.u8     {d11}, [r3], r1
+    vld1.u8     {d12}, [r12], r1
     vld1.u8     {d13}, [r3], r1
+    vld1.u8     {d14}, [r12], r1
     vld1.u8     {d15}, [r3], r1
+    vld1.u8     {d16}, [r12], r1
     vld1.u8     {d17}, [r3], r1
+    vld1.u8     {d18}, [r12], r1
     vld1.u8     {d19}, [r3], r1
+    vld1.u8     {d20}, [r12]
     vld1.u8     {d21}, [r3]
 
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    ldr        r12, [sp, #4]               ; load thresh
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -248,6 +241,8 @@
     vtrn.32     q5, q9
     vtrn.32     q6, q10
 
+    vdup.u8     q2, r12                     ; duplicate thresh
+
     vtrn.16     q3, q5
     vtrn.16     q4, q6
     vtrn.16     q7, q9
@@ -258,18 +253,16 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-
     bl          vp8_loop_filter_neon
 
-    sub         r0, r0, #2
-    sub         r2, r2, #2
-
     vswp        d12, d11
     vswp        d16, d13
     vswp        d14, d12
     vswp        d16, d15
 
+    sub         r0, r0, #2
+    sub         r2, r2, #2
+
     ;store op1, op0, oq0, oq1
     vst4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
     vst4.8      {d14[0], d15[0], d16[0], d17[0]}, [r2], r1
@@ -288,7 +281,7 @@
     vst4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0]
     vst4.8      {d14[7], d15[7], d16[7], d17[7]}, [r2]
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_loop_filter_vertical_edge_uv_neon|
 
 ; void vp8_loop_filter_neon();
@@ -308,7 +301,6 @@
 ; q9    q2
 ; q10   q3
 |vp8_loop_filter_neon| PROC
-    ldr         r12, _lf_coeff_
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
@@ -317,42 +309,44 @@
     vabd.u8     q14, q8, q7                 ; abs(q1 - q0)
     vabd.u8     q3, q9, q8                  ; abs(q2 - q1)
     vabd.u8     q4, q10, q9                 ; abs(q3 - q2)
-    vabd.u8     q9, q6, q7                  ; abs(p0 - q0)
 
     vmax.u8     q11, q11, q12
     vmax.u8     q12, q13, q14
     vmax.u8     q3, q3, q4
     vmax.u8     q15, q11, q12
 
+    vabd.u8     q9, q6, q7                  ; abs(p0 - q0)
+
     ; vp8_hevmask
     vcgt.u8     q13, q13, q2                ; (abs(p1 - p0) > thresh)*-1
     vcgt.u8     q14, q14, q2                ; (abs(q1 - q0) > thresh)*-1
     vmax.u8     q15, q15, q3
 
-    vadd.u8     q0, q0, q0                  ; flimit * 2
-    vadd.u8     q0, q0, q1                  ; flimit * 2 + limit
-    vcge.u8     q15, q1, q15
+    vmov.u8     q10, #0x80                   ; 0x80
 
     vabd.u8     q2, q5, q8                  ; a = abs(p1 - q1)
     vqadd.u8    q9, q9, q9                  ; b = abs(p0 - q0) * 2
-    vshr.u8     q2, q2, #1                  ; a = a / 2
-    vqadd.u8    q9, q9, q2                  ; a = b + a
-    vcge.u8     q9, q0, q9                  ; (a > flimit * 2 + limit) * -1
 
-    vld1.u8     {q0}, [r12]!
+    vcge.u8     q15, q1, q15
 
     ; vp8_filter() function
     ; convert to signed
-    veor        q7, q7, q0                  ; qs0
-    veor        q6, q6, q0                  ; ps0
-    veor        q5, q5, q0                  ; ps1
-    veor        q8, q8, q0                  ; qs1
+    veor        q7, q7, q10                 ; qs0
+    vshr.u8     q2, q2, #1                  ; a = a / 2
+    veor        q6, q6, q10                 ; ps0
 
-    vld1.u8     {q10}, [r12]!
+    veor        q5, q5, q10                 ; ps1
+    vqadd.u8    q9, q9, q2                  ; a = b + a
+
+    veor        q8, q8, q10                 ; qs1
+
+    vmov.u8     q10, #3                     ; #3
 
     vsubl.s8    q2, d14, d12                ; ( qs0 - ps0)
     vsubl.s8    q11, d15, d13
 
+    vcge.u8     q9, q0, q9                  ; (a > flimit * 2 + limit) * -1
+
     vmovl.u8    q4, d20
 
     vqsub.s8    q1, q5, q8                  ; vp8_filter = clamp(ps1-qs1)
@@ -367,7 +361,7 @@
     vaddw.s8    q2, q2, d2
     vaddw.s8    q11, q11, d3
 
-    vld1.u8     {q9}, [r12]!
+    vmov.u8     q9, #4                      ; #4
 
     ; vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d2, q2
@@ -379,19 +373,20 @@
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
     vshr.s8     q1, q1, #3                  ; Filter1 >>= 3
 
+
     vqadd.s8    q11, q6, q2                 ; u = clamp(ps0 + Filter2)
     vqsub.s8    q10, q7, q1                 ; u = clamp(qs0 - Filter1)
 
     ; outer tap adjustments: ++vp8_filter >> 1
     vrshr.s8    q1, q1, #1
     vbic        q1, q1, q14                 ; vp8_filter &= ~hev
-
+    vmov.u8     q0, #0x80                   ; 0x80
     vqadd.s8    q13, q5, q1                 ; u = clamp(ps1 + vp8_filter)
     vqsub.s8    q12, q8, q1                 ; u = clamp(qs1 - vp8_filter)
 
-    veor        q5, q13, q0                 ; *op1 = u^0x80
     veor        q6, q11, q0                 ; *op0 = u^0x80
     veor        q7, q10, q0                 ; *oq0 = u^0x80
+    veor        q5, q13, q0                 ; *op1 = u^0x80
     veor        q8, q12, q0                 ; *oq1 = u^0x80
 
     bx          lr
@@ -399,12 +394,4 @@
 
 ;-----------------
 
-_lf_coeff_
-    DCD     lf_coeff
-lf_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
-    DCD     0x01010101, 0x01010101, 0x01010101, 0x01010101
-
     END

vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm

@@ -9,107 +9,109 @@
 ;
 
 
-    EXPORT  |vp8_loop_filter_simple_horizontal_edge_neon|
+    ;EXPORT  |vp8_loop_filter_simple_horizontal_edge_neon|
+    EXPORT  |vp8_loop_filter_bhs_neon|
+    EXPORT  |vp8_loop_filter_mbhs_neon|
     ARM
-    REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;Note: flimit, limit, and thresh shpuld be positive numbers. 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.
-; r0    unsigned char *s,
-; r1    int p, //pitch
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; stack(r4) const signed char *thresh,
-; //stack(r5)   int count --unused
+
+; r0    unsigned char *s, PRESERVE
+; r1    int p, PRESERVE
+; q1    limit, PRESERVE
 
 |vp8_loop_filter_simple_horizontal_edge_neon| PROC
-    sub         r0, r0, r1, lsl #1          ; move src pointer down by 2 lines
 
-    ldr         r12, _lfhy_coeff_
-    vld1.u8     {q5}, [r0], r1              ; p1
-    vld1.s8     {d2[], d3[]}, [r2]          ; flimit
-    vld1.s8     {d26[], d27[]}, [r3]        ; limit -> q13
-    vld1.u8     {q6}, [r0], r1              ; p0
-    vld1.u8     {q0}, [r12]!                ; 0x80
-    vld1.u8     {q7}, [r0], r1              ; q0
-    vld1.u8     {q10}, [r12]!               ; 0x03
-    vld1.u8     {q8}, [r0]                  ; q1
+    sub         r3, r0, r1, lsl #1          ; move src pointer down by 2 lines
+
+    vld1.u8     {q7}, [r0@128], r1          ; q0
+    vld1.u8     {q5}, [r3@128], r1          ; p0
+    vld1.u8     {q8}, [r0@128]              ; q1
+    vld1.u8     {q6}, [r3@128]              ; p1
 
-    ;vp8_filter_mask() function
     vabd.u8     q15, q6, q7                 ; abs(p0 - q0)
     vabd.u8     q14, q5, q8                 ; abs(p1 - q1)
+
     vqadd.u8    q15, q15, q15               ; abs(p0 - q0) * 2
     vshr.u8     q14, q14, #1                ; abs(p1 - q1) / 2
+    vmov.u8     q0, #0x80                   ; 0x80
+    vmov.s16    q13, #3
     vqadd.u8    q15, q15, q14               ; abs(p0 - q0) * 2 + abs(p1 - q1) / 2
 
-    ;vp8_filter() function
     veor        q7, q7, q0                  ; qs0: q0 offset to convert to a signed value
     veor        q6, q6, q0                  ; ps0: p0 offset to convert to a signed value
     veor        q5, q5, q0                  ; ps1: p1 offset to convert to a signed value
     veor        q8, q8, q0                  ; qs1: q1 offset to convert to a signed value
 
-    vadd.u8     q1, q1, q1                  ; flimit * 2
-    vadd.u8     q1, q1, q13                 ; flimit * 2 + limit
-    vcge.u8     q15, q1, q15                ; (abs(p0 - q0)*2 + abs(p1-q1)/2 > flimit*2 + limit)*-1
+    vcge.u8     q15, q1, q15                ; (abs(p0 - q0)*2 + abs(p1-q1)/2 > limit)*-1
 
-;;;;;;;;;;
-    ;vqsub.s8   q2, q7, q6                  ; ( qs0 - ps0)
     vsubl.s8    q2, d14, d12                ; ( qs0 - ps0)
     vsubl.s8    q3, d15, d13
 
     vqsub.s8    q4, q5, q8                  ; q4: vp8_filter = vp8_signed_char_clamp(ps1-qs1)
 
-    ;vmul.i8    q2, q2, q10                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q11, q2, q2                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q12, q3, q3
+    vmul.s16    q2, q2, q13                 ;  3 * ( qs0 - ps0)
+    vmul.s16    q3, q3, q13
 
-    vld1.u8     {q9}, [r12]!                ; 0x04
-
-    vadd.s16    q2, q2, q11
-    vadd.s16    q3, q3, q12
+    vmov.u8     q10, #0x03                  ; 0x03
+    vmov.u8     q9, #0x04                   ; 0x04
 
     vaddw.s8    q2, q2, d8                  ; vp8_filter + 3 * ( qs0 - ps0)
     vaddw.s8    q3, q3, d9
 
-    ;vqadd.s8   q4, q4, q2                  ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d8, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d9, q3
-;;;;;;;;;;;;;
 
-    vand        q4, q4, q15                 ; vp8_filter &= mask
+    vand        q14, q4, q15                ; vp8_filter &= mask
 
-    vqadd.s8    q2, q4, q10                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
-    vqadd.s8    q4, q4, q9                  ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
+    vqadd.s8    q2, q14, q10                ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
+    vqadd.s8    q3, q14, q9                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
-    vshr.s8     q4, q4, #3                  ; Filter1 >>= 3
+    vshr.s8     q4, q3, #3                  ; Filter1 >>= 3
 
-    sub         r0, r0, r1, lsl #1
+    sub         r0, r0, r1
 
     ;calculate output
     vqadd.s8    q11, q6, q2                 ; u = vp8_signed_char_clamp(ps0 + Filter2)
     vqsub.s8    q10, q7, q4                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
 
-    add         r3, r0, r1
-
     veor        q6, q11, q0                 ; *op0 = u^0x80
     veor        q7, q10, q0                 ; *oq0 = u^0x80
 
-    vst1.u8     {q6}, [r0]                  ; store op0
-    vst1.u8     {q7}, [r3]                  ; store oq0
+    vst1.u8     {q6}, [r3@128]              ; store op0
+    vst1.u8     {q7}, [r0@128]              ; store oq0
 
     bx          lr
     ENDP        ; |vp8_loop_filter_simple_horizontal_edge_neon|
 
-;-----------------
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
 
-_lfhy_coeff_
-    DCD     lfhy_coeff
-lfhy_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
+|vp8_loop_filter_bhs_neon| PROC
+    push        {r4, lr}
+    ldrb        r3, [r2]                    ; load blim from mem
+    vdup.s8     q1, r3                      ; duplicate blim
+
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 4 * y_stride
+    bl          vp8_loop_filter_simple_horizontal_edge_neon
+    ; vp8_loop_filter_simple_horizontal_edge_neon preserves r0, r1 and q1
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 8* y_stride
+    bl          vp8_loop_filter_simple_horizontal_edge_neon
+    add         r0, r0, r1, lsl #2          ; src = y_ptr + 12 * y_stride
+    pop         {r4, lr}
+    b           vp8_loop_filter_simple_horizontal_edge_neon
+    ENDP        ;|vp8_loop_filter_bhs_neon|
+
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
+
+|vp8_loop_filter_mbhs_neon| PROC
+    ldrb        r3, [r2]                   ; load blim from mem
+    vdup.s8     q1, r3                     ; duplicate mblim
+    b           vp8_loop_filter_simple_horizontal_edge_neon
+    ENDP        ;|vp8_loop_filter_bhs_neon|
 
     END

vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm

@@ -9,60 +9,54 @@
 ;
 
 
-    EXPORT  |vp8_loop_filter_simple_vertical_edge_neon|
+    ;EXPORT  |vp8_loop_filter_simple_vertical_edge_neon|
+    EXPORT |vp8_loop_filter_bvs_neon|
+    EXPORT |vp8_loop_filter_mbvs_neon|
     ARM
-    REQUIRE8
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
-;Note: flimit, limit, and thresh should be positive numbers. 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.
-; r0    unsigned char *s,
-; r1    int p, //pitch
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; stack(r4) const signed char *thresh,
-; //stack(r5)   int count --unused
+
+; r0    unsigned char *s, PRESERVE
+; r1    int p, PRESERVE
+; q1    limit, PRESERVE
 
 |vp8_loop_filter_simple_vertical_edge_neon| PROC
     sub         r0, r0, #2                  ; move src pointer down by 2 columns
+    add         r12, r1, r1
+    add         r3, r0, r1
 
-    vld4.8      {d6[0], d7[0], d8[0], d9[0]}, [r0], r1
-    vld1.s8     {d2[], d3[]}, [r2]          ; flimit
-    vld1.s8     {d26[], d27[]}, [r3]        ; limit -> q13
-    vld4.8      {d6[1], d7[1], d8[1], d9[1]}, [r0], r1
-    ldr         r12, _vlfy_coeff_
-    vld4.8      {d6[2], d7[2], d8[2], d9[2]}, [r0], r1
-    vld4.8      {d6[3], d7[3], d8[3], d9[3]}, [r0], r1
-    vld4.8      {d6[4], d7[4], d8[4], d9[4]}, [r0], r1
-    vld4.8      {d6[5], d7[5], d8[5], d9[5]}, [r0], r1
-    vld4.8      {d6[6], d7[6], d8[6], d9[6]}, [r0], r1
-    vld4.8      {d6[7], d7[7], d8[7], d9[7]}, [r0], r1
+    vld4.8      {d6[0], d7[0], d8[0], d9[0]}, [r0], r12
+    vld4.8      {d6[1], d7[1], d8[1], d9[1]}, [r3], r12
+    vld4.8      {d6[2], d7[2], d8[2], d9[2]}, [r0], r12
+    vld4.8      {d6[3], d7[3], d8[3], d9[3]}, [r3], r12
+    vld4.8      {d6[4], d7[4], d8[4], d9[4]}, [r0], r12
+    vld4.8      {d6[5], d7[5], d8[5], d9[5]}, [r3], r12
+    vld4.8      {d6[6], d7[6], d8[6], d9[6]}, [r0], r12
+    vld4.8      {d6[7], d7[7], d8[7], d9[7]}, [r3], r12
 
-    vld4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r1
-    vld1.u8     {q0}, [r12]!                ; 0x80
-    vld4.8      {d10[1], d11[1], d12[1], d13[1]}, [r0], r1
-    vld1.u8     {q11}, [r12]!               ; 0x03
-    vld4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r1
-    vld1.u8     {q12}, [r12]!               ; 0x04
-    vld4.8      {d10[3], d11[3], d12[3], d13[3]}, [r0], r1
-    vld4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r1
-    vld4.8      {d10[5], d11[5], d12[5], d13[5]}, [r0], r1
-    vld4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r1
-    vld4.8      {d10[7], d11[7], d12[7], d13[7]}, [r0], r1
+    vld4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r12
+    vld4.8      {d10[1], d11[1], d12[1], d13[1]}, [r3], r12
+    vld4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r12
+    vld4.8      {d10[3], d11[3], d12[3], d13[3]}, [r3], r12
+    vld4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r12
+    vld4.8      {d10[5], d11[5], d12[5], d13[5]}, [r3], r12
+    vld4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r12
+    vld4.8      {d10[7], d11[7], d12[7], d13[7]}, [r3]
 
     vswp        d7, d10
     vswp        d12, d9
-    ;vswp       q4, q5                      ; p1:q3, p0:q5, q0:q4, q1:q6
 
     ;vp8_filter_mask() function
     ;vp8_hevmask() function
     sub         r0, r0, r1, lsl #4
     vabd.u8     q15, q5, q4                 ; abs(p0 - q0)
     vabd.u8     q14, q3, q6                 ; abs(p1 - q1)
+
     vqadd.u8    q15, q15, q15               ; abs(p0 - q0) * 2
     vshr.u8     q14, q14, #1                ; abs(p1 - q1) / 2
+    vmov.u8     q0, #0x80                   ; 0x80
+    vmov.s16    q11, #3
     vqadd.u8    q15, q15, q14               ; abs(p0 - q0) * 2 + abs(p1 - q1) / 2
 
     veor        q4, q4, q0                  ; qs0: q0 offset to convert to a signed value
@@ -70,87 +64,91 @@
     veor        q3, q3, q0                  ; ps1: p1 offset to convert to a signed value
     veor        q6, q6, q0                  ; qs1: q1 offset to convert to a signed value
 
-    vadd.u8     q1, q1, q1                  ; flimit * 2
-    vadd.u8     q1, q1, q13                 ; flimit * 2 + limit
     vcge.u8     q15, q1, q15                ; abs(p0 - q0)*2 + abs(p1-q1)/2 > flimit*2 + limit)*-1
 
-    ;vp8_filter() function
-;;;;;;;;;;
-    ;vqsub.s8   q2, q5, q4                  ; ( qs0 - ps0)
     vsubl.s8    q2, d8, d10                 ; ( qs0 - ps0)
     vsubl.s8    q13, d9, d11
 
-    vqsub.s8    q1, q3, q6                  ; vp8_filter = vp8_signed_char_clamp(ps1-qs1)
+    vqsub.s8    q14, q3, q6                  ; vp8_filter = vp8_signed_char_clamp(ps1-qs1)
 
-    ;vmul.i8    q2, q2, q11                 ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
-    vadd.s16    q10, q2, q2                 ;  3 * ( qs0 - ps0)
-    vadd.s16    q14, q13, q13
-    vadd.s16    q2, q2, q10
-    vadd.s16    q13, q13, q14
+    vmul.s16    q2, q2, q11                 ;  3 * ( qs0 - ps0)
+    vmul.s16    q13, q13, q11
 
-    ;vqadd.s8   q1, q1, q2
-    vaddw.s8    q2, q2, d2                  ; vp8_filter + 3 * ( qs0 - ps0)
-    vaddw.s8    q13, q13, d3
+    vmov.u8     q11, #0x03                  ; 0x03
+    vmov.u8     q12, #0x04                  ; 0x04
 
-    vqmovn.s16  d2, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
-    vqmovn.s16  d3, q13
+    vaddw.s8    q2, q2, d28                  ; vp8_filter + 3 * ( qs0 - ps0)
+    vaddw.s8    q13, q13, d29
+
+    vqmovn.s16  d28, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
+    vqmovn.s16  d29, q13
 
     add         r0, r0, #1
-    add         r2, r0, r1
-;;;;;;;;;;;
+    add         r3, r0, r1
 
-    vand        q1, q1, q15                 ; vp8_filter &= mask
+    vand        q14, q14, q15                 ; vp8_filter &= mask
 
-    vqadd.s8    q2, q1, q11                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
-    vqadd.s8    q1, q1, q12                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
+    vqadd.s8    q2, q14, q11                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
+    vqadd.s8    q3, q14, q12                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
     vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
-    vshr.s8     q1, q1, #3                  ; Filter1 >>= 3
+    vshr.s8     q14, q3, #3                  ; Filter1 >>= 3
 
     ;calculate output
-    vqsub.s8    q10, q4, q1                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
     vqadd.s8    q11, q5, q2                 ; u = vp8_signed_char_clamp(ps0 + Filter2)
+    vqsub.s8    q10, q4, q14                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
 
-    veor        q7, q10, q0                 ; *oq0 = u^0x80
     veor        q6, q11, q0                 ; *op0 = u^0x80
-
-    add         r3, r2, r1
+    veor        q7, q10, q0                 ; *oq0 = u^0x80
+    add         r12, r1, r1
     vswp        d13, d14
-    add         r12, r3, r1
 
     ;store op1, op0, oq0, oq1
-    vst2.8      {d12[0], d13[0]}, [r0]
-    vst2.8      {d12[1], d13[1]}, [r2]
-    vst2.8      {d12[2], d13[2]}, [r3]
-    vst2.8      {d12[3], d13[3]}, [r12], r1
-    add         r0, r12, r1
-    vst2.8      {d12[4], d13[4]}, [r12]
-    vst2.8      {d12[5], d13[5]}, [r0], r1
-    add         r2, r0, r1
-    vst2.8      {d12[6], d13[6]}, [r0]
-    vst2.8      {d12[7], d13[7]}, [r2], r1
-    add         r3, r2, r1
-    vst2.8      {d14[0], d15[0]}, [r2]
-    vst2.8      {d14[1], d15[1]}, [r3], r1
-    add         r12, r3, r1
-    vst2.8      {d14[2], d15[2]}, [r3]
-    vst2.8      {d14[3], d15[3]}, [r12], r1
-    add         r0, r12, r1
-    vst2.8      {d14[4], d15[4]}, [r12]
-    vst2.8      {d14[5], d15[5]}, [r0], r1
-    add         r2, r0, r1
-    vst2.8      {d14[6], d15[6]}, [r0]
-    vst2.8      {d14[7], d15[7]}, [r2]
+    vst2.8      {d12[0], d13[0]}, [r0], r12
+    vst2.8      {d12[1], d13[1]}, [r3], r12
+    vst2.8      {d12[2], d13[2]}, [r0], r12
+    vst2.8      {d12[3], d13[3]}, [r3], r12
+    vst2.8      {d12[4], d13[4]}, [r0], r12
+    vst2.8      {d12[5], d13[5]}, [r3], r12
+    vst2.8      {d12[6], d13[6]}, [r0], r12
+    vst2.8      {d12[7], d13[7]}, [r3], r12
+    vst2.8      {d14[0], d15[0]}, [r0], r12
+    vst2.8      {d14[1], d15[1]}, [r3], r12
+    vst2.8      {d14[2], d15[2]}, [r0], r12
+    vst2.8      {d14[3], d15[3]}, [r3], r12
+    vst2.8      {d14[4], d15[4]}, [r0], r12
+    vst2.8      {d14[5], d15[5]}, [r3], r12
+    vst2.8      {d14[6], d15[6]}, [r0], r12
+    vst2.8      {d14[7], d15[7]}, [r3]
 
     bx          lr
     ENDP        ; |vp8_loop_filter_simple_vertical_edge_neon|
 
-;-----------------
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
 
-_vlfy_coeff_
-    DCD     vlfy_coeff
-vlfy_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
+|vp8_loop_filter_bvs_neon| PROC
+    push        {r4, lr}
+    ldrb        r3, [r2]                   ; load blim from mem
+    mov         r4, r0
+    add         r0, r0, #4
+    vdup.s8     q1, r3                     ; duplicate blim
+    bl          vp8_loop_filter_simple_vertical_edge_neon
+    ; vp8_loop_filter_simple_vertical_edge_neon preserves  r1 and q1
+    add         r0, r4, #8
+    bl          vp8_loop_filter_simple_vertical_edge_neon
+    add         r0, r4, #12
+    pop         {r4, lr}
+    b           vp8_loop_filter_simple_vertical_edge_neon
+    ENDP        ;|vp8_loop_filter_bvs_neon|
 
+; r0    unsigned char *y
+; r1    int ystride
+; r2    const unsigned char *blimit
+
+|vp8_loop_filter_mbvs_neon| PROC
+    ldrb        r3, [r2]                   ; load mblim from mem
+    vdup.s8     q1, r3                     ; duplicate mblim
+    b           vp8_loop_filter_simple_vertical_edge_neon
+    ENDP        ;|vp8_loop_filter_bvs_neon|
     END

vp8/common/arm/neon/mbloopfilter_neon.asm

@@ -14,155 +14,143 @@
     EXPORT  |vp8_mbloop_filter_vertical_edge_y_neon|
     EXPORT  |vp8_mbloop_filter_vertical_edge_uv_neon|
     ARM
-    REQUIRE8
-    PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
 
-; flimit, limit, and thresh should be positive numbers.
-; All 16 elements in these variables are equal.
-
 ; void vp8_mbloop_filter_horizontal_edge_y_neon(unsigned char *src, int pitch,
-;                                               const signed char *flimit,
-;                                               const signed char *limit,
-;                                               const signed char *thresh,
-;                                               int count)
+;                                               const unsigned char *blimit,
+;                                               const unsigned char *limit,
+;                                               const unsigned char *thresh)
 ; r0    unsigned char *src,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_mbloop_filter_horizontal_edge_y_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, r1, lsl #2          ; move src pointer down by 4 lines
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    push        {lr}
+    add         r1, r1, r1                  ; double stride
+    ldr         r12, [sp, #4]               ; load thresh
+    sub         r0, r0, r1, lsl #1          ; move src pointer down by 4 lines
+    vdup.u8     q2, r12                     ; thresh
+    add         r12, r0, r1,  lsr #1        ; move src pointer up by 1 line
 
-    vld1.u8     {q3}, [r0], r1              ; p3
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    vld1.u8     {q4}, [r0], r1              ; p2
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vld1.u8     {q5}, [r0], r1              ; p1
-    vld1.u8     {q6}, [r0], r1              ; p0
-    vld1.u8     {q7}, [r0], r1              ; q0
-    vld1.u8     {q8}, [r0], r1              ; q1
-    vld1.u8     {q9}, [r0], r1              ; q2
-    vld1.u8     {q10}, [r0], r1             ; q3
+    vld1.u8     {q3}, [r0@128], r1              ; p3
+    vld1.u8     {q4}, [r12@128], r1             ; p2
+    vld1.u8     {q5}, [r0@128], r1              ; p1
+    vld1.u8     {q6}, [r12@128], r1             ; p0
+    vld1.u8     {q7}, [r0@128], r1              ; q0
+    vld1.u8     {q8}, [r12@128], r1             ; q1
+    vld1.u8     {q9}, [r0@128], r1              ; q2
+    vld1.u8     {q10}, [r12@128], r1            ; q3
 
     bl          vp8_mbloop_filter_neon
 
-    sub         r0, r0, r1, lsl #3
-    add         r0, r0, r1
-    add         r2, r0, r1
-    add         r3, r2, r1
+    sub         r12, r12, r1, lsl #2
+    add         r0, r12, r1, lsr #1
 
-    vst1.u8     {q4}, [r0]                  ; store op2
-    vst1.u8     {q5}, [r2]                  ; store op1
-    vst1.u8     {q6}, [r3], r1              ; store op0
-    add         r12, r3, r1
-    vst1.u8     {q7}, [r3]                  ; store oq0
-    vst1.u8     {q8}, [r12], r1             ; store oq1
-    vst1.u8     {q9}, [r12]             ; store oq2
+    vst1.u8     {q4}, [r12@128],r1         ; store op2
+    vst1.u8     {q5}, [r0@128],r1          ; store op1
+    vst1.u8     {q6}, [r12@128], r1        ; store op0
+    vst1.u8     {q7}, [r0@128],r1          ; store oq0
+    vst1.u8     {q8}, [r12@128]            ; store oq1
+    vst1.u8     {q9}, [r0@128]             ; store oq2
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_horizontal_edge_y_neon|
 
 ; void vp8_mbloop_filter_horizontal_edge_uv_neon(unsigned char *u, int pitch,
-;                                                const signed char *flimit,
-;                                                const signed char *limit,
-;                                                const signed char *thresh,
+;                                                const unsigned char *blimit,
+;                                                const unsigned char *limit,
+;                                                const unsigned char *thresh,
 ;                                                unsigned char *v)
 ; r0    unsigned char *u,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 ; sp+4  unsigned char *v
+
 |vp8_mbloop_filter_horizontal_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, r1, lsl #2          ; move u pointer down by 4 lines
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    ldr         r3, [sp, #8]                ; load v ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
-    sub         r3, r3, r1, lsl #2          ; move v pointer down by 4 lines
+    push        {lr}
+    ldr         r12, [sp, #4]                 ; load thresh
+    sub         r0, r0, r1, lsl #2            ; move u pointer down by 4 lines
+    vdup.u8     q2, r12                       ; thresh
+    ldr         r12, [sp, #8]                 ; load v ptr
+    sub         r12, r12, r1, lsl #2          ; move v pointer down by 4 lines
 
-    vld1.u8     {d6}, [r0], r1              ; p3
-    vld1.u8     {d7}, [r3], r1              ; p3
-    vld1.u8     {d8}, [r0], r1              ; p2
-    vld1.u8     {d9}, [r3], r1              ; p2
-    vld1.u8     {d10}, [r0], r1             ; p1
-    vld1.u8     {d11}, [r3], r1             ; p1
-    vld1.u8     {d12}, [r0], r1             ; p0
-    vld1.u8     {d13}, [r3], r1             ; p0
-    vld1.u8     {d14}, [r0], r1             ; q0
-    vld1.u8     {d15}, [r3], r1             ; q0
-    vld1.u8     {d16}, [r0], r1             ; q1
-    vld1.u8     {d17}, [r3], r1             ; q1
-    vld1.u8     {d18}, [r0], r1             ; q2
-    vld1.u8     {d19}, [r3], r1             ; q2
-    vld1.u8     {d20}, [r0], r1             ; q3
-    vld1.u8     {d21}, [r3], r1             ; q3
-
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
+    vld1.u8     {d6}, [r0@64], r1              ; p3
+    vld1.u8     {d7}, [r12@64], r1              ; p3
+    vld1.u8     {d8}, [r0@64], r1              ; p2
+    vld1.u8     {d9}, [r12@64], r1              ; p2
+    vld1.u8     {d10}, [r0@64], r1             ; p1
+    vld1.u8     {d11}, [r12@64], r1             ; p1
+    vld1.u8     {d12}, [r0@64], r1             ; p0
+    vld1.u8     {d13}, [r12@64], r1             ; p0
+    vld1.u8     {d14}, [r0@64], r1             ; q0
+    vld1.u8     {d15}, [r12@64], r1             ; q0
+    vld1.u8     {d16}, [r0@64], r1             ; q1
+    vld1.u8     {d17}, [r12@64], r1             ; q1
+    vld1.u8     {d18}, [r0@64], r1             ; q2
+    vld1.u8     {d19}, [r12@64], r1             ; q2
+    vld1.u8     {d20}, [r0@64], r1             ; q3
+    vld1.u8     {d21}, [r12@64], r1             ; q3
 
     bl          vp8_mbloop_filter_neon
 
     sub         r0, r0, r1, lsl #3
-    sub         r3, r3, r1, lsl #3
+    sub         r12, r12, r1, lsl #3
 
     add         r0, r0, r1
-    add         r3, r3, r1
+    add         r12, r12, r1
 
-    vst1.u8     {d8}, [r0], r1              ; store u op2
-    vst1.u8     {d9}, [r3], r1              ; store v op2
-    vst1.u8     {d10}, [r0], r1             ; store u op1
-    vst1.u8     {d11}, [r3], r1             ; store v op1
-    vst1.u8     {d12}, [r0], r1             ; store u op0
-    vst1.u8     {d13}, [r3], r1             ; store v op0
-    vst1.u8     {d14}, [r0], r1             ; store u oq0
-    vst1.u8     {d15}, [r3], r1             ; store v oq0
-    vst1.u8     {d16}, [r0], r1             ; store u oq1
-    vst1.u8     {d17}, [r3], r1             ; store v oq1
-    vst1.u8     {d18}, [r0], r1             ; store u oq2
-    vst1.u8     {d19}, [r3], r1             ; store v oq2
+    vst1.u8     {d8}, [r0@64], r1              ; store u op2
+    vst1.u8     {d9}, [r12@64], r1              ; store v op2
+    vst1.u8     {d10}, [r0@64], r1             ; store u op1
+    vst1.u8     {d11}, [r12@64], r1             ; store v op1
+    vst1.u8     {d12}, [r0@64], r1             ; store u op0
+    vst1.u8     {d13}, [r12@64], r1             ; store v op0
+    vst1.u8     {d14}, [r0@64], r1             ; store u oq0
+    vst1.u8     {d15}, [r12@64], r1             ; store v oq0
+    vst1.u8     {d16}, [r0@64], r1             ; store u oq1
+    vst1.u8     {d17}, [r12@64], r1             ; store v oq1
+    vst1.u8     {d18}, [r0@64], r1             ; store u oq2
+    vst1.u8     {d19}, [r12@64], r1             ; store v oq2
 
-    ldmia       sp!, {pc}
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_horizontal_edge_uv_neon|
 
 ; void vp8_mbloop_filter_vertical_edge_y_neon(unsigned char *src, int pitch,
-;                                             const signed char *flimit,
-;                                             const signed char *limit,
-;                                             const signed char *thresh,
-;                                             int count)
+;                                             const unsigned char *blimit,
+;                                             const unsigned char *limit,
+;                                             const unsigned char *thresh)
 ; r0    unsigned char *src,
 ; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  int count (unused)
+; r2    unsigned char blimit
+; r3    unsigned char limit
+; sp    unsigned char thresh,
 |vp8_mbloop_filter_vertical_edge_y_neon| PROC
-    stmdb       sp!, {lr}
+    push        {lr}
+    ldr         r12, [sp, #4]               ; load thresh
     sub         r0, r0, #4                  ; move src pointer down by 4 columns
+    vdup.s8     q2, r12                     ; thresh
+    add         r12, r0, r1, lsl #3         ; move src pointer down by 8 lines
 
     vld1.u8     {d6}, [r0], r1              ; load first 8-line src data
-    ldr         r12, [sp, #4]               ; load thresh pointer
+    vld1.u8     {d7}, [r12], r1             ; load second 8-line src data
     vld1.u8     {d8}, [r0], r1
-    sub         sp, sp, #32
+    vld1.u8     {d9}, [r12], r1
     vld1.u8     {d10}, [r0], r1
+    vld1.u8     {d11}, [r12], r1
     vld1.u8     {d12}, [r0], r1
+    vld1.u8     {d13}, [r12], r1
     vld1.u8     {d14}, [r0], r1
+    vld1.u8     {d15}, [r12], r1
     vld1.u8     {d16}, [r0], r1
+    vld1.u8     {d17}, [r12], r1
     vld1.u8     {d18}, [r0], r1
+    vld1.u8     {d19}, [r12], r1
     vld1.u8     {d20}, [r0], r1
-
-    vld1.u8     {d7}, [r0], r1              ; load second 8-line src data
-    vld1.u8     {d9}, [r0], r1
-    vld1.u8     {d11}, [r0], r1
-    vld1.u8     {d13}, [r0], r1
-    vld1.u8     {d15}, [r0], r1
-    vld1.u8     {d17}, [r0], r1
-    vld1.u8     {d19}, [r0], r1
-    vld1.u8     {d21}, [r0], r1
+    vld1.u8     {d21}, [r12], r1
 
     ;transpose to 8x16 matrix
     vtrn.32     q3, q7
@@ -180,133 +168,11 @@
     vtrn.8      q7, q8
     vtrn.8      q9, q10
 
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    mov         r12, sp
-    vst1.u8     {q3}, [r12]!
-    vst1.u8     {q10}, [r12]!
-
-    bl          vp8_mbloop_filter_neon
-
-    sub         r0, r0, r1, lsl #4
-
-    add         r2, r0, r1
-
-    add         r3, r2, r1
-
-    vld1.u8     {q3}, [sp]!
-    vld1.u8     {q10}, [sp]!
-
-    ;transpose to 16x8 matrix
-    vtrn.32     q3, q7
-    vtrn.32     q4, q8
-    vtrn.32     q5, q9
-    vtrn.32     q6, q10
-    add         r12, r3, r1
-
-    vtrn.16     q3, q5
-    vtrn.16     q4, q6
-    vtrn.16     q7, q9
-    vtrn.16     q8, q10
-
-    vtrn.8      q3, q4
-    vtrn.8      q5, q6
-    vtrn.8      q7, q8
-    vtrn.8      q9, q10
-
-    ;store op2, op1, op0, oq0, oq1, oq2
-    vst1.8      {d6}, [r0]
-    vst1.8      {d8}, [r2]
-    vst1.8      {d10}, [r3]
-    vst1.8      {d12}, [r12], r1
-    add         r0, r12, r1
-    vst1.8      {d14}, [r12]
-    vst1.8      {d16}, [r0], r1
-    add         r2, r0, r1
-    vst1.8      {d18}, [r0]
-    vst1.8      {d20}, [r2], r1
-    add         r3, r2, r1
-    vst1.8      {d7}, [r2]
-    vst1.8      {d9}, [r3], r1
-    add         r12, r3, r1
-    vst1.8      {d11}, [r3]
-    vst1.8      {d13}, [r12], r1
-    add         r0, r12, r1
-    vst1.8      {d15}, [r12]
-    vst1.8      {d17}, [r0], r1
-    add         r2, r0, r1
-    vst1.8      {d19}, [r0]
-    vst1.8      {d21}, [r2]
-
-    ldmia       sp!, {pc}
-    ENDP        ; |vp8_mbloop_filter_vertical_edge_y_neon|
-
-; void vp8_mbloop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch,
-;                                              const signed char *flimit,
-;                                              const signed char *limit,
-;                                              const signed char *thresh,
-;                                              unsigned char *v)
-; r0    unsigned char *u,
-; r1    int pitch,
-; r2    const signed char *flimit,
-; r3    const signed char *limit,
-; sp    const signed char *thresh,
-; sp+4  unsigned char *v
-|vp8_mbloop_filter_vertical_edge_uv_neon| PROC
-    stmdb       sp!, {lr}
-    sub         r0, r0, #4                  ; move src pointer down by 4 columns
-    vld1.s8     {d2[], d3[]}, [r3]          ; limit
-    ldr         r3, [sp, #8]                ; load v ptr
-    ldr         r12, [sp, #4]               ; load thresh pointer
-
-    sub         r3, r3, #4                  ; move v pointer down by 4 columns
-
-    vld1.u8     {d6}, [r0], r1              ;load u data
-    vld1.u8     {d7}, [r3], r1              ;load v data
-    vld1.u8     {d8}, [r0], r1
-    vld1.u8     {d9}, [r3], r1
-    vld1.u8     {d10}, [r0], r1
-    vld1.u8     {d11}, [r3], r1
-    vld1.u8     {d12}, [r0], r1
-    vld1.u8     {d13}, [r3], r1
-    vld1.u8     {d14}, [r0], r1
-    vld1.u8     {d15}, [r3], r1
-    vld1.u8     {d16}, [r0], r1
-    vld1.u8     {d17}, [r3], r1
-    vld1.u8     {d18}, [r0], r1
-    vld1.u8     {d19}, [r3], r1
-    vld1.u8     {d20}, [r0], r1
-    vld1.u8     {d21}, [r3], r1
-
-    ;transpose to 8x16 matrix
-    vtrn.32     q3, q7
-    vtrn.32     q4, q8
-    vtrn.32     q5, q9
-    vtrn.32     q6, q10
-
-    vtrn.16     q3, q5
-    vtrn.16     q4, q6
-    vtrn.16     q7, q9
-    vtrn.16     q8, q10
-
-    vtrn.8      q3, q4
-    vtrn.8      q5, q6
-    vtrn.8      q7, q8
-    vtrn.8      q9, q10
-
-    sub         sp, sp, #32
-    vld1.s8     {d4[], d5[]}, [r12]         ; thresh
-    mov         r12, sp
-    vst1.u8     {q3}, [r12]!
-    vst1.u8     {q10}, [r12]!
-
-    bl          vp8_mbloop_filter_neon
-
     sub         r0, r0, r1, lsl #3
-    sub         r3, r3, r1, lsl #3
 
-    vld1.u8     {q3}, [sp]!
-    vld1.u8     {q10}, [sp]!
+    bl          vp8_mbloop_filter_neon
+
+    sub         r12, r12, r1, lsl #3
 
     ;transpose to 16x8 matrix
     vtrn.32     q3, q7
@@ -326,23 +192,118 @@
 
     ;store op2, op1, op0, oq0, oq1, oq2
     vst1.8      {d6}, [r0], r1
-    vst1.8      {d7}, [r3], r1
+    vst1.8      {d7}, [r12], r1
     vst1.8      {d8}, [r0], r1
-    vst1.8      {d9}, [r3], r1
+    vst1.8      {d9}, [r12], r1
     vst1.8      {d10}, [r0], r1
-    vst1.8      {d11}, [r3], r1
+    vst1.8      {d11}, [r12], r1
     vst1.8      {d12}, [r0], r1
-    vst1.8      {d13}, [r3], r1
+    vst1.8      {d13}, [r12], r1
     vst1.8      {d14}, [r0], r1
-    vst1.8      {d15}, [r3], r1
+    vst1.8      {d15}, [r12], r1
     vst1.8      {d16}, [r0], r1
-    vst1.8      {d17}, [r3], r1
+    vst1.8      {d17}, [r12], r1
     vst1.8      {d18}, [r0], r1
-    vst1.8      {d19}, [r3], r1
-    vst1.8      {d20}, [r0], r1
-    vst1.8      {d21}, [r3], r1
+    vst1.8      {d19}, [r12], r1
+    vst1.8      {d20}, [r0]
+    vst1.8      {d21}, [r12]
 
-    ldmia       sp!, {pc}
+    pop         {pc}
+    ENDP        ; |vp8_mbloop_filter_vertical_edge_y_neon|
+
+; void vp8_mbloop_filter_vertical_edge_uv_neon(unsigned char *u, int pitch,
+;                                              const unsigned char *blimit,
+;                                              const unsigned char *limit,
+;                                              const unsigned char *thresh,
+;                                              unsigned char *v)
+; r0    unsigned char *u,
+; r1    int pitch,
+; r2    const signed char *flimit,
+; r3    const signed char *limit,
+; sp    const signed char *thresh,
+; sp+4  unsigned char *v
+|vp8_mbloop_filter_vertical_edge_uv_neon| PROC
+    push        {lr}
+    ldr         r12, [sp, #4]               ; load thresh
+    sub         r0, r0, #4                  ; move u pointer down by 4 columns
+    vdup.u8     q2, r12                     ; thresh
+    ldr         r12, [sp, #8]               ; load v ptr
+    sub         r12, r12, #4                ; move v pointer down by 4 columns
+
+    vld1.u8     {d6}, [r0], r1              ;load u data
+    vld1.u8     {d7}, [r12], r1             ;load v data
+    vld1.u8     {d8}, [r0], r1
+    vld1.u8     {d9}, [r12], r1
+    vld1.u8     {d10}, [r0], r1
+    vld1.u8     {d11}, [r12], r1
+    vld1.u8     {d12}, [r0], r1
+    vld1.u8     {d13}, [r12], r1
+    vld1.u8     {d14}, [r0], r1
+    vld1.u8     {d15}, [r12], r1
+    vld1.u8     {d16}, [r0], r1
+    vld1.u8     {d17}, [r12], r1
+    vld1.u8     {d18}, [r0], r1
+    vld1.u8     {d19}, [r12], r1
+    vld1.u8     {d20}, [r0], r1
+    vld1.u8     {d21}, [r12], r1
+
+    ;transpose to 8x16 matrix
+    vtrn.32     q3, q7
+    vtrn.32     q4, q8
+    vtrn.32     q5, q9
+    vtrn.32     q6, q10
+
+    vtrn.16     q3, q5
+    vtrn.16     q4, q6
+    vtrn.16     q7, q9
+    vtrn.16     q8, q10
+
+    vtrn.8      q3, q4
+    vtrn.8      q5, q6
+    vtrn.8      q7, q8
+    vtrn.8      q9, q10
+
+    sub         r0, r0, r1, lsl #3
+
+    bl          vp8_mbloop_filter_neon
+
+    sub         r12, r12, r1, lsl #3
+
+    ;transpose to 16x8 matrix
+    vtrn.32     q3, q7
+    vtrn.32     q4, q8
+    vtrn.32     q5, q9
+    vtrn.32     q6, q10
+
+    vtrn.16     q3, q5
+    vtrn.16     q4, q6
+    vtrn.16     q7, q9
+    vtrn.16     q8, q10
+
+    vtrn.8      q3, q4
+    vtrn.8      q5, q6
+    vtrn.8      q7, q8
+    vtrn.8      q9, q10
+
+    ;store op2, op1, op0, oq0, oq1, oq2
+    vst1.8      {d6}, [r0], r1
+    vst1.8      {d7}, [r12], r1
+    vst1.8      {d8}, [r0], r1
+    vst1.8      {d9}, [r12], r1
+    vst1.8      {d10}, [r0], r1
+    vst1.8      {d11}, [r12], r1
+    vst1.8      {d12}, [r0], r1
+    vst1.8      {d13}, [r12], r1
+    vst1.8      {d14}, [r0], r1
+    vst1.8      {d15}, [r12], r1
+    vst1.8      {d16}, [r0], r1
+    vst1.8      {d17}, [r12], r1
+    vst1.8      {d18}, [r0], r1
+    vst1.8      {d19}, [r12], r1
+    vst1.8      {d20}, [r0]
+    vst1.8      {d21}, [r12]
+
+    pop         {pc}
     ENDP        ; |vp8_mbloop_filter_vertical_edge_uv_neon|
 
 ; void vp8_mbloop_filter_neon()
@@ -350,41 +311,33 @@
 ; functions do the necessary load, transpose (if necessary), preserve (if
 ; necessary) and store.
 
-; TODO:
-; The vertical filter writes p3/q3 back out because two 4 element writes are
-; much simpler than ordering and writing two 3 element sets (or three 2 elements
-; sets, or whichever other combinations are possible).
-; If we can preserve q3 and q10, the vertical filter will be able to avoid
-; storing those values on the stack and reading them back after the filter.
-
 ; r0,r1 PRESERVE
-; r2    flimit
-; r3    PRESERVE
-; q1    limit
+; r2    mblimit
+; r3    limit
+
 ; q2    thresh
-; q3    p3
+; q3    p3 PRESERVE
 ; q4    p2
 ; q5    p1
 ; q6    p0
 ; q7    q0
 ; q8    q1
 ; q9    q2
-; q10   q3
+; q10   q3 PRESERVE
 
 |vp8_mbloop_filter_neon| PROC
-    ldr         r12, _mblf_coeff_
 
     ; vp8_filter_mask
     vabd.u8     q11, q3, q4                 ; abs(p3 - p2)
     vabd.u8     q12, q4, q5                 ; abs(p2 - p1)
     vabd.u8     q13, q5, q6                 ; abs(p1 - p0)
     vabd.u8     q14, q8, q7                 ; abs(q1 - q0)
-    vabd.u8     q3, q9, q8                  ; abs(q2 - q1)
+    vabd.u8     q1, q9, q8                  ; abs(q2 - q1)
     vabd.u8     q0, q10, q9                 ; abs(q3 - q2)
 
     vmax.u8     q11, q11, q12
     vmax.u8     q12, q13, q14
-    vmax.u8     q3, q3, q0
+    vmax.u8     q1, q1, q0
     vmax.u8     q15, q11, q12
 
     vabd.u8     q12, q6, q7                 ; abs(p0 - q0)
@@ -392,51 +345,53 @@
     ; vp8_hevmask
     vcgt.u8     q13, q13, q2                ; (abs(p1 - p0) > thresh) * -1
     vcgt.u8     q14, q14, q2                ; (abs(q1 - q0) > thresh) * -1
-    vmax.u8     q15, q15, q3
+    vmax.u8     q15, q15, q1
 
-    vld1.s8     {d4[], d5[]}, [r2]          ; flimit
+    vdup.u8     q1, r3                      ; limit
+    vdup.u8     q2, r2                      ; mblimit
 
-    vld1.u8     {q0}, [r12]!
+    vmov.u8     q0, #0x80                   ; 0x80
 
-    vadd.u8     q2, q2, q2                  ; flimit * 2
-    vadd.u8     q2, q2, q1                  ; flimit * 2 +  limit
     vcge.u8     q15, q1, q15
 
     vabd.u8     q1, q5, q8                  ; a = abs(p1 - q1)
     vqadd.u8    q12, q12, q12               ; b = abs(p0 - q0) * 2
-    vshr.u8     q1, q1, #1                  ; a = a / 2
-    vqadd.u8    q12, q12, q1                ; a = b + a
-    vcge.u8     q12, q2, q12                ; (a > flimit * 2 + limit) * -1
+    vmov.u16    q11, #3                     ; #3
 
     ; vp8_filter
     ; convert to signed
     veor        q7, q7, q0                  ; qs0
+    vshr.u8     q1, q1, #1                  ; a = a / 2
     veor        q6, q6, q0                  ; ps0
     veor        q5, q5, q0                  ; ps1
+
+    vqadd.u8    q12, q12, q1                ; a = b + a
+
     veor        q8, q8, q0                  ; qs1
     veor        q4, q4, q0                  ; ps2
     veor        q9, q9, q0                  ; qs2
 
     vorr        q14, q13, q14               ; vp8_hevmask
 
+    vcge.u8     q12, q2, q12                ; (a > flimit * 2 + limit) * -1
+
     vsubl.s8    q2, d14, d12                ; qs0 - ps0
     vsubl.s8    q13, d15, d13
 
     vqsub.s8    q1, q5, q8                  ; vp8_filter = clamp(ps1-qs1)
 
-    vadd.s16    q10, q2, q2                 ; 3 * (qs0 - ps0)
-    vadd.s16    q11, q13, q13
+    vmul.i16    q2, q2, q11                 ; 3 * ( qs0 - ps0)
+
     vand        q15, q15, q12               ; vp8_filter_mask
 
-    vadd.s16    q2, q2, q10
-    vadd.s16    q13, q13, q11
+    vmul.i16    q13, q13, q11
 
-    vld1.u8     {q12}, [r12]!               ; #3
+    vmov.u8     q12, #3                     ; #3
 
     vaddw.s8    q2, q2, d2                  ; vp8_filter + 3 * ( qs0 - ps0)
     vaddw.s8    q13, q13, d3
 
-    vld1.u8     {q11}, [r12]!               ; #4
+    vmov.u8     q11, #4                     ; #4
 
     ; vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
     vqmovn.s16  d2, q2
@@ -444,27 +399,23 @@
 
     vand        q1, q1, q15                 ; vp8_filter &= mask
 
-    vld1.u8     {q15}, [r12]!               ; #63
-    ;
-    vand        q13, q1, q14                ; Filter2 &= hev
+    vmov.u16    q15, #63                    ; #63
 
-    vld1.u8     {d7}, [r12]!                ; #9
+    vand        q13, q1, q14                ; Filter2 &= hev
 
     vqadd.s8    q2, q13, q11                ; Filter1 = clamp(Filter2+4)
     vqadd.s8    q13, q13, q12               ; Filter2 = clamp(Filter2+3)
 
-    vld1.u8     {d6}, [r12]!                ; #18
+    vmov        q0, q15
 
     vshr.s8     q2, q2, #3                  ; Filter1 >>= 3
     vshr.s8     q13, q13, #3                ; Filter2 >>= 3
 
-    vmov        q10, q15
+    vmov        q11, q15
     vmov        q12, q15
 
     vqsub.s8    q7, q7, q2                  ; qs0 = clamp(qs0 - Filter1)
 
-    vld1.u8     {d5}, [r12]!                ; #27
-
     vqadd.s8    q6, q6, q13                 ; ps0 = clamp(ps0 + Filter2)
 
     vbic        q1, q1, q14                 ; vp8_filter &= ~hev
@@ -472,49 +423,47 @@
     ; roughly 1/7th difference across boundary
     ; roughly 2/7th difference across boundary
     ; roughly 3/7th difference across boundary
-    vmov        q11, q15
+
+    vmov.u8     d5, #9                      ; #9
+    vmov.u8     d4, #18                     ; #18
+
     vmov        q13, q15
     vmov        q14, q15
 
-    vmlal.s8    q10, d2, d7                 ; Filter2 * 9
-    vmlal.s8    q11, d3, d7
-    vmlal.s8    q12, d2, d6                 ; Filter2 * 18
-    vmlal.s8    q13, d3, d6
-    vmlal.s8    q14, d2, d5                 ; Filter2 * 27
+    vmlal.s8    q0, d2, d5                  ; 63 + Filter2 * 9
+    vmlal.s8    q11, d3, d5
+    vmov.u8     d5, #27                     ; #27
+    vmlal.s8    q12, d2, d4                 ; 63 + Filter2 * 18
+    vmlal.s8    q13, d3, d4
+    vmlal.s8    q14, d2, d5                 ; 63 + Filter2 * 27
     vmlal.s8    q15, d3, d5
-    vqshrn.s16  d20, q10, #7                ; u = clamp((63 + Filter2 * 9)>>7)
-    vqshrn.s16  d21, q11, #7
+
+    vqshrn.s16  d0, q0, #7                  ; u = clamp((63 + Filter2 * 9)>>7)
+    vqshrn.s16  d1, q11, #7
     vqshrn.s16  d24, q12, #7                ; u = clamp((63 + Filter2 * 18)>>7)
     vqshrn.s16  d25, q13, #7
     vqshrn.s16  d28, q14, #7                ; u = clamp((63 + Filter2 * 27)>>7)
     vqshrn.s16  d29, q15, #7
 
-    vqsub.s8    q11, q9, q10                ; s = clamp(qs2 - u)
-    vqadd.s8    q10, q4, q10                ; s = clamp(ps2 + u)
+    vmov.u8     q1, #0x80                   ; 0x80
+
+    vqsub.s8    q11, q9, q0                 ; s = clamp(qs2 - u)
+    vqadd.s8    q0, q4, q0                  ; s = clamp(ps2 + u)
     vqsub.s8    q13, q8, q12                ; s = clamp(qs1 - u)
     vqadd.s8    q12, q5, q12                ; s = clamp(ps1 + u)
     vqsub.s8    q15, q7, q14                ; s = clamp(qs0 - u)
     vqadd.s8    q14, q6, q14                ; s = clamp(ps0 + u)
-    veor        q9, q11, q0                 ; *oq2 = s^0x80
-    veor        q4, q10, q0                 ; *op2 = s^0x80
-    veor        q8, q13, q0                 ; *oq1 = s^0x80
-    veor        q5, q12, q0                 ; *op2 = s^0x80
-    veor        q7, q15, q0                 ; *oq0 = s^0x80
-    veor        q6, q14, q0                 ; *op0 = s^0x80
+
+    veor        q9, q11, q1                 ; *oq2 = s^0x80
+    veor        q4, q0, q1                  ; *op2 = s^0x80
+    veor        q8, q13, q1                 ; *oq1 = s^0x80
+    veor        q5, q12, q1                 ; *op2 = s^0x80
+    veor        q7, q15, q1                 ; *oq0 = s^0x80
+    veor        q6, q14, q1                 ; *op0 = s^0x80
 
     bx          lr
     ENDP        ; |vp8_mbloop_filter_neon|
 
 ;-----------------
 
-_mblf_coeff_
-    DCD     mblf_coeff
-mblf_coeff
-    DCD     0x80808080, 0x80808080, 0x80808080, 0x80808080
-    DCD     0x03030303, 0x03030303, 0x03030303, 0x03030303
-    DCD     0x04040404, 0x04040404, 0x04040404, 0x04040404
-    DCD     0x003f003f, 0x003f003f, 0x003f003f, 0x003f003f
-    DCD     0x09090909, 0x09090909, 0x12121212, 0x12121212
-    DCD     0x1b1b1b1b, 0x1b1b1b1b
-
     END

vp8/common/arm/neon/shortidct4x4llm_neon.asm

@@ -31,7 +31,7 @@
 ;result of the multiplication that is needed in IDCT.
 
 |vp8_short_idct4x4llm_neon| PROC
-    ldr             r12, _idct_coeff_
+    adr             r12, idct_coeff
     vld1.16         {q1, q2}, [r0]
     vld1.16         {d0}, [r12]
 
@@ -114,8 +114,6 @@
 
 ;-----------------
 
-_idct_coeff_
-    DCD     idct_coeff
 idct_coeff
     DCD     0x4e7b4e7b, 0x8a8c8a8c
 

vp8/common/arm/neon/sixtappredict16x16_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter16_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -33,7 +44,7 @@
 |vp8_sixtap_predict16x16_neon| PROC
     push            {r4-r5, lr}
 
-    ldr             r12, _filter16_coeff_
+    adr             r12, filter16_coeff
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
 
@@ -476,17 +487,4 @@ secondpass_only_inner_loop_neon
     ENDP
 
 ;-----------------
-
-_filter16_coeff_
-    DCD     filter16_coeff
-filter16_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
-
     END

vp8/common/arm/neon/sixtappredict4x4_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter4_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict_neon| PROC
     push            {r4, lr}
 
-    ldr             r12, _filter4_coeff_
+    adr             r12, filter4_coeff
     ldr             r4, [sp, #8]            ;load parameters from stack
     ldr             lr, [sp, #12]           ;load parameters from stack
 
@@ -408,16 +419,4 @@ secondpass_filter4x4_only
 
 ;-----------------
 
-_filter4_coeff_
-    DCD     filter4_coeff
-filter4_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
-
     END

vp8/common/arm/neon/sixtappredict8x4_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter8_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict8x4_neon| PROC
     push            {r4-r5, lr}
 
-    ldr             r12, _filter8_coeff_
+    adr             r12, filter8_coeff
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
 
@@ -459,16 +470,4 @@ secondpass_filter8x4_only
 
 ;-----------------
 
-_filter8_coeff_
-    DCD     filter8_coeff
-filter8_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
-
     END

vp8/common/arm/neon/sixtappredict8x8_neon.asm

@@ -15,6 +15,17 @@
     PRESERVE8
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
+
+filter8_coeff
+    DCD     0,  0,  128,    0,   0,  0,   0,  0
+    DCD     0, -6,  123,   12,  -1,  0,   0,  0
+    DCD     2, -11, 108,   36,  -8,  1,   0,  0
+    DCD     0, -9,   93,   50,  -6,  0,   0,  0
+    DCD     3, -16,  77,   77, -16,  3,   0,  0
+    DCD     0, -6,   50,   93,  -9,  0,   0,  0
+    DCD     1, -8,   36,  108, -11,  2,   0,  0
+    DCD     0, -1,   12,  123,  -6,   0,  0,  0
+
 ; r0    unsigned char  *src_ptr,
 ; r1    int  src_pixels_per_line,
 ; r2    int  xoffset,
@@ -25,7 +36,7 @@
 |vp8_sixtap_predict8x8_neon| PROC
     push            {r4-r5, lr}
 
-    ldr             r12, _filter8_coeff_
+    adr             r12, filter8_coeff
 
     ldr             r4, [sp, #12]           ;load parameters from stack
     ldr             r5, [sp, #16]           ;load parameters from stack
@@ -510,16 +521,4 @@ filt_blk2d_spo8x8_loop_neon
 
 ;-----------------
 
-_filter8_coeff_
-    DCD     filter8_coeff
-filter8_coeff
-    DCD     0,  0,  128,    0,   0,  0,   0,  0
-    DCD     0, -6,  123,   12,  -1,  0,   0,  0
-    DCD     2, -11, 108,   36,  -8,  1,   0,  0
-    DCD     0, -9,   93,   50,  -6,  0,   0,  0
-    DCD     3, -16,  77,   77, -16,  3,   0,  0
-    DCD     0, -6,   50,   93,  -9,  0,   0,  0
-    DCD     1, -8,   36,  108, -11,  2,   0,  0
-    DCD     0, -1,   12,  123,  -6,   0,  0,  0
-
     END

vp8/common/arm/recon_arm.h

@@ -53,6 +53,9 @@ extern prototype_copy_block(vp8_copy_mem16x16_neon);
 
 extern prototype_recon_macroblock(vp8_recon_mb_neon);
 
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mby_neon);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mby_s_neon);
+
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_recon_recon
 #define vp8_recon_recon vp8_recon_b_neon
@@ -74,6 +77,13 @@ extern prototype_recon_macroblock(vp8_recon_mb_neon);
 
 #undef  vp8_recon_recon_mb
 #define vp8_recon_recon_mb vp8_recon_mb_neon
+
+#undef  vp8_recon_build_intra_predictors_mby
+#define vp8_recon_build_intra_predictors_mby vp8_build_intra_predictors_mby_neon
+
+#undef  vp8_recon_build_intra_predictors_mby_s
+#define vp8_recon_build_intra_predictors_mby_s vp8_build_intra_predictors_mby_s_neon
+
 #endif
 #endif
 

vp8/common/asm_com_offsets.c

@@ -9,27 +9,14 @@
  */
 
 
-#include "vpx_ports/config.h"
-#include <stddef.h>
-
+#include "vpx_config.h"
+#include "vpx/vpx_codec.h"
+#include "vpx_ports/asm_offsets.h"
 #include "vpx_scale/yv12config.h"
 
-#define ct_assert(name,cond) \
-    static void assert_##name(void) UNUSED;\
-    static void assert_##name(void) {switch(0){case 0:case !!(cond):;}}
+BEGIN
 
-#define DEFINE(sym, val) int sym = val;
-
-/*
-#define BLANK() asm volatile("\n->" : : )
-*/
-
-/*
- * int main(void)
- * {
- */
-
-//vpx_scale
+/* vpx_scale */
 DEFINE(yv12_buffer_config_y_width,              offsetof(YV12_BUFFER_CONFIG, y_width));
 DEFINE(yv12_buffer_config_y_height,             offsetof(YV12_BUFFER_CONFIG, y_height));
 DEFINE(yv12_buffer_config_y_stride,             offsetof(YV12_BUFFER_CONFIG, y_stride));
@@ -40,10 +27,14 @@ DEFINE(yv12_buffer_config_y_buffer,             offsetof(YV12_BUFFER_CONFIG, y_b
 DEFINE(yv12_buffer_config_u_buffer,             offsetof(YV12_BUFFER_CONFIG, u_buffer));
 DEFINE(yv12_buffer_config_v_buffer,             offsetof(YV12_BUFFER_CONFIG, v_buffer));
 DEFINE(yv12_buffer_config_border,               offsetof(YV12_BUFFER_CONFIG, border));
+DEFINE(VP8BORDERINPIXELS_VAL,                   VP8BORDERINPIXELS);
 
-//add asserts for any offset that is not supported by assembly code
-//add asserts for any size that is not supported by assembly code
-/*
- * return 0;
- * }
- */
+END
+
+/* add asserts for any offset that is not supported by assembly code */
+/* add asserts for any size that is not supported by assembly code */
+
+#if HAVE_ARMV7
+/* vp8_yv12_extend_frame_borders_neon makes several assumptions based on this */
+ct_assert(VP8BORDERINPIXELS_VAL, VP8BORDERINPIXELS == 32)
+#endif

vp8/common/blockd.h

@@ -137,16 +137,11 @@ typedef enum
    modes for the Y blocks to the left and above us; for interframes, there
    is a single probability table. */
 
-typedef struct
+union b_mode_info
 {
-    B_PREDICTION_MODE mode;
-    union
-    {
-        int as_int;
-        MV  as_mv;
-    } mv;
-} B_MODE_INFO;
-
+    B_PREDICTION_MODE as_mode;
+    int_mv mv;
+};
 
 typedef enum
 {
@@ -161,38 +156,26 @@ typedef struct
 {
     MB_PREDICTION_MODE mode, uv_mode;
     MV_REFERENCE_FRAME ref_frame;
-    union
-    {
-        int as_int;
-        MV  as_mv;
-    } mv;
+    int_mv mv;
 
     unsigned char partitioning;
     unsigned char mb_skip_coeff;                                /* does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens */
-    unsigned char dc_diff;
     unsigned char need_to_clamp_mvs;
-
     unsigned char segment_id;                  /* Which set of segmentation parameters should be used for this MB */
-
-    unsigned char force_no_skip; /* encoder only */
 } MB_MODE_INFO;
 
-
 typedef struct
 {
     MB_MODE_INFO mbmi;
-    B_MODE_INFO bmi[16];
+    union b_mode_info bmi[16];
 } MODE_INFO;
 
-
 typedef struct
 {
     short *qcoeff;
     short *dqcoeff;
     unsigned char  *predictor;
     short *diff;
-    short *reference;
-
     short *dequant;
 
     /* 16 Y blocks, 4 U blocks, 4 V blocks each with 16 entries */
@@ -206,15 +189,13 @@ typedef struct
 
     int eob;
 
-    B_MODE_INFO bmi;
-
+    union b_mode_info bmi;
 } BLOCKD;
 
-typedef struct
+typedef struct MacroBlockD
 {
     DECLARE_ALIGNED(16, short, diff[400]);      /* from idct diff */
     DECLARE_ALIGNED(16, unsigned char,  predictor[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]);
@@ -271,6 +252,9 @@ typedef struct
     int mb_to_top_edge;
     int mb_to_bottom_edge;
 
+    int ref_frame_cost[MAX_REF_FRAMES];
+
+
     unsigned int frames_since_golden;
     unsigned int frames_till_alt_ref_frame;
     vp8_subpix_fn_t  subpixel_predict;
@@ -282,6 +266,14 @@ typedef struct
 
     int corrupted;
 
+#if ARCH_X86 || ARCH_X86_64
+    /* This is an intermediate buffer currently used in sub-pixel motion search
+     * to keep a copy of the reference area. This buffer can be used for other
+     * purpose.
+     */
+    DECLARE_ALIGNED(32, unsigned char, y_buf[22*32]);
+#endif
+
 #if CONFIG_RUNTIME_CPU_DETECT
     struct VP8_COMMON_RTCD  *rtcd;
 #endif
@@ -291,4 +283,20 @@ typedef struct
 extern void vp8_build_block_doffsets(MACROBLOCKD *x);
 extern void vp8_setup_block_dptrs(MACROBLOCKD *x);
 
+static void update_blockd_bmi(MACROBLOCKD *xd)
+{
+    int i;
+    int is_4x4;
+    is_4x4 = (xd->mode_info_context->mbmi.mode == SPLITMV) ||
+              (xd->mode_info_context->mbmi.mode == B_PRED);
+
+    if (is_4x4)
+    {
+        for (i = 0; i < 16; i++)
+        {
+            xd->block[i].bmi = xd->mode_info_context->bmi[i];
+        }
+    }
+}
+
 #endif  /* __INC_BLOCKD_H */

vp8/common/coefupdateprobs.h

@@ -12,7 +12,7 @@
 /* Update probabilities for the nodes in the token entropy tree.
    Generated file included by entropy.c */
 
-const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1] =
+const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES] =
 {
     {
         {

vp8/common/debugmodes.c

@@ -97,7 +97,7 @@ void vp8_print_modes_and_motion_vectors(MODE_INFO *mi, int rows, int cols, int f
                 bindex = (b_row & 3) * 4 + (b_col & 3);
 
                 if (mi[mb_index].mbmi.mode == B_PRED)
-                    fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].mode);
+                    fprintf(mvs, "%2d ", mi[mb_index].bmi[bindex].as_mode);
                 else
                     fprintf(mvs, "xx ");
 

vp8/common/defaultcoefcounts.c

@@ -0,0 +1,225 @@
+/*
+ *  Copyright (c) 2010 The WebM 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 "defaultcoefcounts.h"
+
+/* Generated file, included by entropy.c */
+
+const unsigned int vp8_default_coef_counts[BLOCK_TYPES]
+                                          [COEF_BANDS]
+                                          [PREV_COEF_CONTEXTS]
+                                          [MAX_ENTROPY_TOKENS] =
+{
+
+    {
+        /* Block Type ( 0 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {30190, 26544, 225,  24,   4,   0,   0,   0,   0,   0,   0, 4171593,},
+            {26846, 25157, 1241, 130,  26,   6,   1,   0,   0,   0,   0, 149987,},
+            {10484, 9538, 1006, 160,  36,  18,   0,   0,   0,   0,   0, 15104,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {25842, 40456, 1126,  83,  11,   2,   0,   0,   0,   0,   0,   0,},
+            {9338, 8010, 512,  73,   7,   3,   2,   0,   0,   0,   0, 43294,},
+            {1047, 751, 149,  31,  13,   6,   1,   0,   0,   0,   0, 879,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {26136, 9826, 252,  13,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {8134, 5574, 191,  14,   2,   0,   0,   0,   0,   0,   0, 35302,},
+            { 605, 677, 116,   9,   1,   0,   0,   0,   0,   0,   0, 611,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {10263, 15463, 283,  17,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {2773, 2191, 128,   9,   2,   2,   0,   0,   0,   0,   0, 10073,},
+            { 134, 125,  32,   4,   0,   2,   0,   0,   0,   0,   0,  50,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {10483, 2663,  23,   1,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {2137, 1251,  27,   1,   1,   0,   0,   0,   0,   0,   0, 14362,},
+            { 116, 156,  14,   2,   1,   0,   0,   0,   0,   0,   0, 190,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {40977, 27614, 412,  28,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {6113, 5213, 261,  22,   3,   0,   0,   0,   0,   0,   0, 26164,},
+            { 382, 312,  50,  14,   2,   0,   0,   0,   0,   0,   0, 345,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,  26,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,  13,   0,   0,   0,   0,   0,   0,   0,   0,   0, 319,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   8,},
+        },
+    },
+    {
+        /* Block Type ( 1 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {3268, 19382, 1043, 250,  93,  82,  49,  26,  17,   8,  25, 82289,},
+            {8758, 32110, 5436, 1832, 827, 668, 420, 153,  24,   0,   3, 52914,},
+            {9337, 23725, 8487, 3954, 2107, 1836, 1069, 399,  59,   0,   0, 18620,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {12419, 8420, 452,  62,   9,   1,   0,   0,   0,   0,   0,   0,},
+            {11715, 8705, 693,  92,  15,   7,   2,   0,   0,   0,   0, 53988,},
+            {7603, 8585, 2306, 778, 270, 145,  39,   5,   0,   0,   0, 9136,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {15938, 14335, 1207, 184,  55,  13,   4,   1,   0,   0,   0,   0,},
+            {7415, 6829, 1138, 244,  71,  26,   7,   0,   0,   0,   0, 9980,},
+            {1580, 1824, 655, 241,  89,  46,  10,   2,   0,   0,   0, 429,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {19453, 5260, 201,  19,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {9173, 3758, 213,  22,   1,   1,   0,   0,   0,   0,   0, 9820,},
+            {1689, 1277, 276,  51,  17,   4,   0,   0,   0,   0,   0, 679,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {12076, 10667, 620,  85,  19,   9,   5,   0,   0,   0,   0,   0,},
+            {4665, 3625, 423,  55,  19,   9,   0,   0,   0,   0,   0, 5127,},
+            { 415, 440, 143,  34,  20,   7,   2,   0,   0,   0,   0, 101,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {12183, 4846, 115,  11,   1,   0,   0,   0,   0,   0,   0,   0,},
+            {4226, 3149, 177,  21,   2,   0,   0,   0,   0,   0,   0, 7157,},
+            { 375, 621, 189,  51,  11,   4,   1,   0,   0,   0,   0, 198,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {61658, 37743, 1203,  94,  10,   3,   0,   0,   0,   0,   0,   0,},
+            {15514, 11563, 903, 111,  14,   5,   0,   0,   0,   0,   0, 25195,},
+            { 929, 1077, 291,  78,  14,   7,   1,   0,   0,   0,   0, 507,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0, 990,  15,   3,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0, 412,  13,   0,   0,   0,   0,   0,   0,   0,   0, 1641,},
+            {   0,  18,   7,   1,   0,   0,   0,   0,   0,   0,   0,  30,},
+        },
+    },
+    {
+        /* Block Type ( 2 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            { 953, 24519, 628, 120,  28,  12,   4,   0,   0,   0,   0, 2248798,},
+            {1525, 25654, 2647, 617, 239, 143,  42,   5,   0,   0,   0, 66837,},
+            {1180, 11011, 3001, 1237, 532, 448, 239,  54,   5,   0,   0, 7122,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {1356, 2220,  67,  10,   4,   1,   0,   0,   0,   0,   0,   0,},
+            {1450, 2544, 102,  18,   4,   3,   0,   0,   0,   0,   0, 57063,},
+            {1182, 2110, 470, 130,  41,  21,   0,   0,   0,   0,   0, 6047,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            { 370, 3378, 200,  30,   5,   4,   1,   0,   0,   0,   0,   0,},
+            { 293, 1006, 131,  29,  11,   0,   0,   0,   0,   0,   0, 5404,},
+            { 114, 387,  98,  23,   4,   8,   1,   0,   0,   0,   0, 236,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            { 579, 194,   4,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            { 395, 213,   5,   1,   0,   0,   0,   0,   0,   0,   0, 4157,},
+            { 119, 122,   4,   0,   0,   0,   0,   0,   0,   0,   0, 300,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {  38, 557,  19,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  21, 114,  12,   1,   0,   0,   0,   0,   0,   0,   0, 427,},
+            {   0,   5,   0,   0,   0,   0,   0,   0,   0,   0,   0,   7,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {  52,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  18,   6,   0,   0,   0,   0,   0,   0,   0,   0,   0, 652,},
+            {   1,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  30,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            { 640, 569,  10,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {  25,  77,   2,   0,   0,   0,   0,   0,   0,   0,   0, 517,},
+            {   4,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   3,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+        },
+    },
+    {
+        /* Block Type ( 3 ) */
+        {
+            /* Coeff Band ( 0 ) */
+            {2506, 20161, 2707, 767, 261, 178, 107,  30,  14,   3,   0, 100694,},
+            {8806, 36478, 8817, 3268, 1280, 850, 401, 114,  42,   0,   0, 58572,},
+            {11003, 27214, 11798, 5716, 2482, 2072, 1048, 175,  32,   0,   0, 19284,},
+        },
+        {
+            /* Coeff Band ( 1 ) */
+            {9738, 11313, 959, 205,  70,  18,  11,   1,   0,   0,   0,   0,},
+            {12628, 15085, 1507, 273,  52,  19,   9,   0,   0,   0,   0, 54280,},
+            {10701, 15846, 5561, 1926, 813, 570, 249,  36,   0,   0,   0, 6460,},
+        },
+        {
+            /* Coeff Band ( 2 ) */
+            {6781, 22539, 2784, 634, 182, 123,  20,   4,   0,   0,   0,   0,},
+            {6263, 11544, 2649, 790, 259, 168,  27,   5,   0,   0,   0, 20539,},
+            {3109, 4075, 2031, 896, 457, 386, 158,  29,   0,   0,   0, 1138,},
+        },
+        {
+            /* Coeff Band ( 3 ) */
+            {11515, 4079, 465,  73,   5,  14,   2,   0,   0,   0,   0,   0,},
+            {9361, 5834, 650,  96,  24,   8,   4,   0,   0,   0,   0, 22181,},
+            {4343, 3974, 1360, 415, 132,  96,  14,   1,   0,   0,   0, 1267,},
+        },
+        {
+            /* Coeff Band ( 4 ) */
+            {4787, 9297, 823, 168,  44,  12,   4,   0,   0,   0,   0,   0,},
+            {3619, 4472, 719, 198,  60,  31,   3,   0,   0,   0,   0, 8401,},
+            {1157, 1175, 483, 182,  88,  31,   8,   0,   0,   0,   0, 268,},
+        },
+        {
+            /* Coeff Band ( 5 ) */
+            {8299, 1226,  32,   5,   1,   0,   0,   0,   0,   0,   0,   0,},
+            {3502, 1568,  57,   4,   1,   1,   0,   0,   0,   0,   0, 9811,},
+            {1055, 1070, 166,  29,   6,   1,   0,   0,   0,   0,   0, 527,},
+        },
+        {
+            /* Coeff Band ( 6 ) */
+            {27414, 27927, 1989, 347,  69,  26,   0,   0,   0,   0,   0,   0,},
+            {5876, 10074, 1574, 341,  91,  24,   4,   0,   0,   0,   0, 21954,},
+            {1571, 2171, 778, 324, 124,  65,  16,   0,   0,   0,   0, 979,},
+        },
+        {
+            /* Coeff Band ( 7 ) */
+            {   0,  29,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
+            {   0,  23,   0,   0,   0,   0,   0,   0,   0,   0,   0, 459,},
+            {   0,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  13,},
+        },
+    },
+};

vp8/common/defaultcoefcounts.h

@@ -8,214 +8,14 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#ifndef __DEFAULTCOEFCOUNTS_H
+#define __DEFAULTCOEFCOUNTS_H
 
-/* Generated file, included by entropy.c */
+#include "entropy.h"
 
-static const unsigned int default_coef_counts [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens] =
-{
+extern const unsigned int vp8_default_coef_counts[BLOCK_TYPES]
+                                                 [COEF_BANDS]
+                                                 [PREV_COEF_CONTEXTS]
+                                                 [MAX_ENTROPY_TOKENS];
 
-    {
-        /* Block Type ( 0 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {30190, 26544, 225,  24,   4,   0,   0,   0,   0,   0,   0, 4171593,},
-            {26846, 25157, 1241, 130,  26,   6,   1,   0,   0,   0,   0, 149987,},
-            {10484, 9538, 1006, 160,  36,  18,   0,   0,   0,   0,   0, 15104,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {25842, 40456, 1126,  83,  11,   2,   0,   0,   0,   0,   0,   0,},
-            {9338, 8010, 512,  73,   7,   3,   2,   0,   0,   0,   0, 43294,},
-            {1047, 751, 149,  31,  13,   6,   1,   0,   0,   0,   0, 879,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {26136, 9826, 252,  13,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {8134, 5574, 191,  14,   2,   0,   0,   0,   0,   0,   0, 35302,},
-            { 605, 677, 116,   9,   1,   0,   0,   0,   0,   0,   0, 611,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {10263, 15463, 283,  17,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {2773, 2191, 128,   9,   2,   2,   0,   0,   0,   0,   0, 10073,},
-            { 134, 125,  32,   4,   0,   2,   0,   0,   0,   0,   0,  50,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {10483, 2663,  23,   1,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {2137, 1251,  27,   1,   1,   0,   0,   0,   0,   0,   0, 14362,},
-            { 116, 156,  14,   2,   1,   0,   0,   0,   0,   0,   0, 190,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {40977, 27614, 412,  28,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {6113, 5213, 261,  22,   3,   0,   0,   0,   0,   0,   0, 26164,},
-            { 382, 312,  50,  14,   2,   0,   0,   0,   0,   0,   0, 345,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,  26,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,  13,   0,   0,   0,   0,   0,   0,   0,   0,   0, 319,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   8,},
-        },
-    },
-    {
-        /* Block Type ( 1 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {3268, 19382, 1043, 250,  93,  82,  49,  26,  17,   8,  25, 82289,},
-            {8758, 32110, 5436, 1832, 827, 668, 420, 153,  24,   0,   3, 52914,},
-            {9337, 23725, 8487, 3954, 2107, 1836, 1069, 399,  59,   0,   0, 18620,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {12419, 8420, 452,  62,   9,   1,   0,   0,   0,   0,   0,   0,},
-            {11715, 8705, 693,  92,  15,   7,   2,   0,   0,   0,   0, 53988,},
-            {7603, 8585, 2306, 778, 270, 145,  39,   5,   0,   0,   0, 9136,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {15938, 14335, 1207, 184,  55,  13,   4,   1,   0,   0,   0,   0,},
-            {7415, 6829, 1138, 244,  71,  26,   7,   0,   0,   0,   0, 9980,},
-            {1580, 1824, 655, 241,  89,  46,  10,   2,   0,   0,   0, 429,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {19453, 5260, 201,  19,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {9173, 3758, 213,  22,   1,   1,   0,   0,   0,   0,   0, 9820,},
-            {1689, 1277, 276,  51,  17,   4,   0,   0,   0,   0,   0, 679,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {12076, 10667, 620,  85,  19,   9,   5,   0,   0,   0,   0,   0,},
-            {4665, 3625, 423,  55,  19,   9,   0,   0,   0,   0,   0, 5127,},
-            { 415, 440, 143,  34,  20,   7,   2,   0,   0,   0,   0, 101,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {12183, 4846, 115,  11,   1,   0,   0,   0,   0,   0,   0,   0,},
-            {4226, 3149, 177,  21,   2,   0,   0,   0,   0,   0,   0, 7157,},
-            { 375, 621, 189,  51,  11,   4,   1,   0,   0,   0,   0, 198,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {61658, 37743, 1203,  94,  10,   3,   0,   0,   0,   0,   0,   0,},
-            {15514, 11563, 903, 111,  14,   5,   0,   0,   0,   0,   0, 25195,},
-            { 929, 1077, 291,  78,  14,   7,   1,   0,   0,   0,   0, 507,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0, 990,  15,   3,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0, 412,  13,   0,   0,   0,   0,   0,   0,   0,   0, 1641,},
-            {   0,  18,   7,   1,   0,   0,   0,   0,   0,   0,   0,  30,},
-        },
-    },
-    {
-        /* Block Type ( 2 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            { 953, 24519, 628, 120,  28,  12,   4,   0,   0,   0,   0, 2248798,},
-            {1525, 25654, 2647, 617, 239, 143,  42,   5,   0,   0,   0, 66837,},
-            {1180, 11011, 3001, 1237, 532, 448, 239,  54,   5,   0,   0, 7122,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {1356, 2220,  67,  10,   4,   1,   0,   0,   0,   0,   0,   0,},
-            {1450, 2544, 102,  18,   4,   3,   0,   0,   0,   0,   0, 57063,},
-            {1182, 2110, 470, 130,  41,  21,   0,   0,   0,   0,   0, 6047,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            { 370, 3378, 200,  30,   5,   4,   1,   0,   0,   0,   0,   0,},
-            { 293, 1006, 131,  29,  11,   0,   0,   0,   0,   0,   0, 5404,},
-            { 114, 387,  98,  23,   4,   8,   1,   0,   0,   0,   0, 236,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            { 579, 194,   4,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            { 395, 213,   5,   1,   0,   0,   0,   0,   0,   0,   0, 4157,},
-            { 119, 122,   4,   0,   0,   0,   0,   0,   0,   0,   0, 300,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {  38, 557,  19,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  21, 114,  12,   1,   0,   0,   0,   0,   0,   0,   0, 427,},
-            {   0,   5,   0,   0,   0,   0,   0,   0,   0,   0,   0,   7,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {  52,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  18,   6,   0,   0,   0,   0,   0,   0,   0,   0,   0, 652,},
-            {   1,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  30,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            { 640, 569,  10,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {  25,  77,   2,   0,   0,   0,   0,   0,   0,   0,   0, 517,},
-            {   4,   7,   0,   0,   0,   0,   0,   0,   0,   0,   0,   3,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-        },
-    },
-    {
-        /* Block Type ( 3 ) */
-        {
-            /* Coeff Band ( 0 ) */
-            {2506, 20161, 2707, 767, 261, 178, 107,  30,  14,   3,   0, 100694,},
-            {8806, 36478, 8817, 3268, 1280, 850, 401, 114,  42,   0,   0, 58572,},
-            {11003, 27214, 11798, 5716, 2482, 2072, 1048, 175,  32,   0,   0, 19284,},
-        },
-        {
-            /* Coeff Band ( 1 ) */
-            {9738, 11313, 959, 205,  70,  18,  11,   1,   0,   0,   0,   0,},
-            {12628, 15085, 1507, 273,  52,  19,   9,   0,   0,   0,   0, 54280,},
-            {10701, 15846, 5561, 1926, 813, 570, 249,  36,   0,   0,   0, 6460,},
-        },
-        {
-            /* Coeff Band ( 2 ) */
-            {6781, 22539, 2784, 634, 182, 123,  20,   4,   0,   0,   0,   0,},
-            {6263, 11544, 2649, 790, 259, 168,  27,   5,   0,   0,   0, 20539,},
-            {3109, 4075, 2031, 896, 457, 386, 158,  29,   0,   0,   0, 1138,},
-        },
-        {
-            /* Coeff Band ( 3 ) */
-            {11515, 4079, 465,  73,   5,  14,   2,   0,   0,   0,   0,   0,},
-            {9361, 5834, 650,  96,  24,   8,   4,   0,   0,   0,   0, 22181,},
-            {4343, 3974, 1360, 415, 132,  96,  14,   1,   0,   0,   0, 1267,},
-        },
-        {
-            /* Coeff Band ( 4 ) */
-            {4787, 9297, 823, 168,  44,  12,   4,   0,   0,   0,   0,   0,},
-            {3619, 4472, 719, 198,  60,  31,   3,   0,   0,   0,   0, 8401,},
-            {1157, 1175, 483, 182,  88,  31,   8,   0,   0,   0,   0, 268,},
-        },
-        {
-            /* Coeff Band ( 5 ) */
-            {8299, 1226,  32,   5,   1,   0,   0,   0,   0,   0,   0,   0,},
-            {3502, 1568,  57,   4,   1,   1,   0,   0,   0,   0,   0, 9811,},
-            {1055, 1070, 166,  29,   6,   1,   0,   0,   0,   0,   0, 527,},
-        },
-        {
-            /* Coeff Band ( 6 ) */
-            {27414, 27927, 1989, 347,  69,  26,   0,   0,   0,   0,   0,   0,},
-            {5876, 10074, 1574, 341,  91,  24,   4,   0,   0,   0,   0, 21954,},
-            {1571, 2171, 778, 324, 124,  65,  16,   0,   0,   0,   0, 979,},
-        },
-        {
-            /* Coeff Band ( 7 ) */
-            {   0,  29,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,},
-            {   0,  23,   0,   0,   0,   0,   0,   0,   0,   0,   0, 459,},
-            {   0,   1,   0,   0,   0,   0,   0,   0,   0,   0,   0,  13,},
-        },
-    },
-};
+#endif //__DEFAULTCOEFCOUNTS_H

vp8/common/entropy.c

@@ -26,8 +26,32 @@ typedef vp8_prob Prob;
 
 #include "coefupdateprobs.h"
 
-DECLARE_ALIGNED(16, cuchar, vp8_coef_bands[16]) = { 0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7};
-DECLARE_ALIGNED(16, cuchar, vp8_prev_token_class[MAX_ENTROPY_TOKENS]) = { 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0};
+DECLARE_ALIGNED(16, const unsigned char, vp8_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,
+    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,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+};
+
+DECLARE_ALIGNED(16, cuchar, vp8_coef_bands[16]) =
+{ 0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7};
+
+DECLARE_ALIGNED(16, cuchar, vp8_prev_token_class[MAX_ENTROPY_TOKENS]) =
+{ 0, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 0};
+
 DECLARE_ALIGNED(16, const int, vp8_default_zig_zag1d[16]) =
 {
     0,  1,  4,  8,
@@ -65,7 +89,7 @@ const vp8_tree_index vp8_coef_tree[ 22] =     /* corresponding _CONTEXT_NODEs */
     -DCT_VAL_CATEGORY5, -DCT_VAL_CATEGORY6   /* 10 = CAT_FIVE */
 };
 
-struct vp8_token_struct vp8_coef_encodings[vp8_coef_tokens];
+struct vp8_token_struct vp8_coef_encodings[MAX_ENTROPY_TOKENS];
 
 /* Trees for extra bits.  Probabilities are constant and
    do not depend on previously encoded bits */
@@ -145,10 +169,12 @@ void vp8_default_coef_probs(VP8_COMMON *pc)
 
             do
             {
-                unsigned int branch_ct [vp8_coef_tokens-1] [2];
+                unsigned int branch_ct [ENTROPY_NODES] [2];
                 vp8_tree_probs_from_distribution(
-                    vp8_coef_tokens, vp8_coef_encodings, vp8_coef_tree,
-                    pc->fc.coef_probs [h][i][k], branch_ct, default_coef_counts [h][i][k],
+                    MAX_ENTROPY_TOKENS, vp8_coef_encodings, vp8_coef_tree,
+                    pc->fc.coef_probs[h][i][k],
+                    branch_ct,
+                    vp8_default_coef_counts[h][i][k],
                     256, 1);
 
             }

vp8/common/entropy.h

@@ -30,13 +30,12 @@
 #define DCT_VAL_CATEGORY6       10      /* 67+       Extra Bits 11+1 */
 #define DCT_EOB_TOKEN           11      /* EOB       Extra Bits 0+0 */
 
-#define vp8_coef_tokens 12
-#define MAX_ENTROPY_TOKENS vp8_coef_tokens
+#define MAX_ENTROPY_TOKENS 12
 #define ENTROPY_NODES 11
 
 extern const vp8_tree_index vp8_coef_tree[];
 
-extern struct vp8_token_struct vp8_coef_encodings[vp8_coef_tokens];
+extern struct vp8_token_struct vp8_coef_encodings[MAX_ENTROPY_TOKENS];
 
 typedef struct
 {
@@ -85,9 +84,9 @@ extern DECLARE_ALIGNED(16, const unsigned char, vp8_coef_bands[16]);
 /*# define DC_TOKEN_CONTEXTS        3*/ /* 00, 0!0, !0!0 */
 #   define PREV_COEF_CONTEXTS       3
 
-extern DECLARE_ALIGNED(16, const unsigned char, vp8_prev_token_class[vp8_coef_tokens]);
+extern DECLARE_ALIGNED(16, const unsigned char, vp8_prev_token_class[MAX_ENTROPY_TOKENS]);
 
-extern const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1];
+extern const vp8_prob vp8_coef_update_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
 
 
 struct VP8Common;

vp8/common/entropymode.c

@@ -33,11 +33,11 @@ typedef enum
     SUBMVREF_LEFT_ABOVE_ZED
 } sumvfref_t;
 
-int vp8_mv_cont(const MV *l, const MV *a)
+int vp8_mv_cont(const int_mv *l, const int_mv *a)
 {
-    int lez = (l->row == 0 && l->col == 0);
-    int aez = (a->row == 0 && a->col == 0);
-    int lea = (l->row == a->row && l->col == a->col);
+    int lez = (l->as_int == 0);
+    int aez = (a->as_int == 0);
+    int lea = (l->as_int == a->as_int);
 
     if (lea && lez)
         return SUBMVREF_LEFT_ABOVE_ZED;

vp8/common/entropymode.h

@@ -25,7 +25,7 @@ extern const int vp8_mbsplit_count [VP8_NUMMBSPLITS];    /* # of subsets */
 
 extern const vp8_prob vp8_mbsplit_probs [VP8_NUMMBSPLITS-1];
 
-extern int vp8_mv_cont(const MV *l, const MV *a);
+extern int vp8_mv_cont(const int_mv *l, const int_mv *a);
 #define SUBMVREF_COUNT 5
 extern const vp8_prob vp8_sub_mv_ref_prob2 [SUBMVREF_COUNT][VP8_SUBMVREFS-1];
 

vp8/common/entropymv.h

@@ -18,6 +18,8 @@ enum
 {
     mv_max  = 1023,              /* max absolute value of a MV component */
     MVvals = (2 * mv_max) + 1,   /* # possible values "" */
+    mvfp_max  = 255,              /* max absolute value of a full pixel MV component */
+    MVfpvals = (2 * mvfp_max) +1, /* # possible full pixel MV values */
 
     mvlong_width = 10,       /* Large MVs have 9 bit magnitudes */
     mvnum_short = 8,         /* magnitudes 0 through 7 */

vp8/common/extend.c

@@ -13,10 +13,12 @@
 #include "vpx_mem/vpx_mem.h"
 
 
-static void extend_plane_borders
+static void copy_and_extend_plane
 (
     unsigned char *s, /* source */
-    int sp,           /* pitch */
+    int sp,           /* source pitch */
+    unsigned char *d, /* destination */
+    int dp,           /* destination pitch */
     int h,            /* height */
     int w,            /* width */
     int et,           /* extend top border */
@@ -25,7 +27,6 @@ static void extend_plane_borders
     int er            /* extend right border */
 )
 {
-
     int i;
     unsigned char *src_ptr1, *src_ptr2;
     unsigned char *dest_ptr1, *dest_ptr2;
@@ -34,68 +35,73 @@ static void extend_plane_borders
     /* copy the left and right most columns out */
     src_ptr1 = s;
     src_ptr2 = s + w - 1;
-    dest_ptr1 = s - el;
-    dest_ptr2 = s + w;
+    dest_ptr1 = d - el;
+    dest_ptr2 = d + w;
 
-    for (i = 0; i < h - 0 + 1; i++)
+    for (i = 0; i < h; 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_ptr1, src_ptr1[0], el);
+        vpx_memcpy(dest_ptr1 + el, src_ptr1, w);
         vpx_memset(dest_ptr2, src_ptr2[0], er);
         src_ptr1  += sp;
         src_ptr2  += sp;
-        dest_ptr1 += sp;
-        dest_ptr2 += sp;
+        dest_ptr1 += dp;
+        dest_ptr2 += dp;
     }
 
-    /* Now copy the top and bottom source lines into each line of the respective borders */
-    src_ptr1 = s - el;
-    src_ptr2 = s + sp * (h - 1) - el;
-    dest_ptr1 = s + sp * (-et) - el;
-    dest_ptr2 = s + sp * (h) - el;
-    linesize = el + er + w + 1;
+    /* Now copy the top and bottom lines into each line of the respective
+     * borders
+     */
+    src_ptr1 = d - el;
+    src_ptr2 = d + dp * (h - 1) - el;
+    dest_ptr1 = d + dp * (-et) - el;
+    dest_ptr2 = d + dp * (h) - el;
+    linesize = el + er + w;
 
-    for (i = 0; i < (int)et; i++)
+    for (i = 0; i < et; i++)
     {
         vpx_memcpy(dest_ptr1, src_ptr1, linesize);
-        dest_ptr1 += sp;
+        dest_ptr1 += dp;
     }
 
-    for (i = 0; i < (int)eb; i++)
+    for (i = 0; i < eb; i++)
     {
         vpx_memcpy(dest_ptr2, src_ptr2, linesize);
-        dest_ptr2 += sp;
+        dest_ptr2 += dp;
     }
 }
 
 
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height)
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst)
 {
-    int er = 0xf & (16 - (width & 0xf));
-    int eb = 0xf & (16 - (height & 0xf));
+    int et = dst->border;
+    int el = dst->border;
+    int eb = dst->border + dst->y_height - src->y_height;
+    int er = dst->border + dst->y_width - src->y_width;
 
-    /* check for non multiples of 16 */
-    if (er != 0 || eb != 0)
-    {
-        extend_plane_borders(ybf->y_buffer, ybf->y_stride, height, width, 0, 0, eb, er);
+    copy_and_extend_plane(src->y_buffer, src->y_stride,
+                          dst->y_buffer, dst->y_stride,
+                          src->y_height, src->y_width,
+                          et, el, eb, er);
 
-        /* adjust for uv */
-        height = (height + 1) >> 1;
-        width  = (width  + 1) >> 1;
-        er = 0x7 & (8 - (width  & 0x7));
-        eb = 0x7 & (8 - (height & 0x7));
+    et = dst->border >> 1;
+    el = dst->border >> 1;
+    eb = (dst->border >> 1) + dst->uv_height - src->uv_height;
+    er = (dst->border >> 1) + dst->uv_width - src->uv_width;
 
-        if (er || eb)
-        {
-            extend_plane_borders(ybf->u_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-            extend_plane_borders(ybf->v_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-        }
-    }
+    copy_and_extend_plane(src->u_buffer, src->uv_stride,
+                          dst->u_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
+
+    copy_and_extend_plane(src->v_buffer, src->uv_stride,
+                          dst->v_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
 }
 
+
 /* note the extension is only for the last row, for intra prediction purpose */
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr)
 {

vp8/common/extend.h

@@ -14,8 +14,8 @@
 
 #include "vpx_scale/yv12config.h"
 
-void Extend(YV12_BUFFER_CONFIG *ybf);
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr);
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height);
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst);
 
 #endif

vp8/common/filter.c

@@ -38,7 +38,7 @@ DECLARE_ALIGNED(16, const short, vp8_sub_pel_filters[8][6]) =
     { 0, -1,   12,  123,  -6,  0 },
 };
 
-void vp8_filter_block2d_first_pass
+static void filter_block2d_first_pass
 (
     unsigned char *src_ptr,
     int *output_ptr,
@@ -82,7 +82,7 @@ void vp8_filter_block2d_first_pass
     }
 }
 
-void vp8_filter_block2d_second_pass
+static void filter_block2d_second_pass
 (
     int *src_ptr,
     unsigned char *output_ptr,
@@ -129,7 +129,7 @@ void vp8_filter_block2d_second_pass
 }
 
 
-void vp8_filter_block2d
+static void filter_block2d
 (
     unsigned char  *src_ptr,
     unsigned char  *output_ptr,
@@ -142,39 +142,13 @@ void vp8_filter_block2d
     int FData[9*4]; /* Temp data buffer used in filtering */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 4, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 4, HFilter);
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 8, output_ptr, output_pitch, 4, 4, 4, 4, VFilter);
+    filter_block2d_second_pass(FData + 8, output_ptr, output_pitch, 4, 4, 4, 4, VFilter);
 }
 
 
-void vp8_block_variation_c
-(
-    unsigned char  *src_ptr,
-    int   src_pixels_per_line,
-    int *HVar,
-    int *VVar
-)
-{
-    int i, j;
-    unsigned char *Ptr = src_ptr;
-
-    for (i = 0; i < 4; i++)
-    {
-        for (j = 0; j < 4; j++)
-        {
-            *HVar += abs((int)Ptr[j] - (int)Ptr[j+1]);
-            *VVar += abs((int)Ptr[j] - (int)Ptr[j+src_pixels_per_line]);
-        }
-
-        Ptr += src_pixels_per_line;
-    }
-}
-
-
-
-
 void vp8_sixtap_predict_c
 (
     unsigned char  *src_ptr,
@@ -191,7 +165,7 @@ void vp8_sixtap_predict_c
     HFilter = vp8_sub_pel_filters[xoffset];   /* 6 tap */
     VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
-    vp8_filter_block2d(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter);
+    filter_block2d(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter);
 }
 void vp8_sixtap_predict8x8_c
 (
@@ -211,11 +185,11 @@ void vp8_sixtap_predict8x8_c
     VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 13, 8, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 13, 8, HFilter);
 
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 8, 8, VFilter);
+    filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 8, 8, VFilter);
 
 }
 
@@ -237,11 +211,11 @@ void vp8_sixtap_predict8x4_c
     VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 8, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 9, 8, HFilter);
 
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 4, 8, VFilter);
+    filter_block2d_second_pass(FData + 16, dst_ptr, dst_pitch, 8, 8, 4, 8, VFilter);
 
 }
 
@@ -264,10 +238,10 @@ void vp8_sixtap_predict16x16_c
     VFilter = vp8_sub_pel_filters[yoffset];   /* 6 tap */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 21, 16, HFilter);
+    filter_block2d_first_pass(src_ptr - (2 * src_pixels_per_line), FData, src_pixels_per_line, 1, 21, 16, HFilter);
 
     /* then filter verticaly... */
-    vp8_filter_block2d_second_pass(FData + 32, dst_ptr, dst_pitch, 16, 16, 16, 16, VFilter);
+    filter_block2d_second_pass(FData + 32, dst_ptr, dst_pitch, 16, 16, 16, 16, VFilter);
 
 }
 
@@ -294,7 +268,7 @@ void vp8_sixtap_predict16x16_c
  *                  Two filter taps should sum to VP8_FILTER_WEIGHT.
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil_first_pass
+static void filter_block2d_bil_first_pass
 (
     unsigned char  *src_ptr,
     unsigned short *dst_ptr,
@@ -345,7 +319,7 @@ void vp8_filter_block2d_bil_first_pass
  *                  Two filter taps should sum to VP8_FILTER_WEIGHT.
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil_second_pass
+static void filter_block2d_bil_second_pass
 (
     unsigned short *src_ptr,
     unsigned char  *dst_ptr,
@@ -399,7 +373,7 @@ void vp8_filter_block2d_bil_second_pass
  *  SPECIAL NOTES : The largest block size can be handled here is 16x16
  *
  ****************************************************************************/
-void vp8_filter_block2d_bil
+static void filter_block2d_bil
 (
     unsigned char *src_ptr,
     unsigned char *dst_ptr,
@@ -415,10 +389,10 @@ void vp8_filter_block2d_bil
     unsigned short FData[17*16];    /* Temp data buffer used in filtering */
 
     /* First filter 1-D horizontally... */
-    vp8_filter_block2d_bil_first_pass(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
+    filter_block2d_bil_first_pass(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
 
     /* then 1-D vertically... */
-    vp8_filter_block2d_bil_second_pass(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
+    filter_block2d_bil_second_pass(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
 }
 
 
@@ -444,19 +418,19 @@ void vp8_bilinear_predict4x4_c
         unsigned char temp2[16];
 
         bilinear_predict4x4_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, temp1, 4);
-        vp8_filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
+        filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
 
         for (i = 0; i < 16; i++)
         {
             if (temp1[i] != temp2[i])
             {
                 bilinear_predict4x4_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, temp1, 4);
-                vp8_filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
+                filter_block2d_bil(src_ptr, temp2, src_pixels_per_line, 4, HFilter, VFilter, 4, 4);
             }
         }
     }
 #endif
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
 
 }
 
@@ -476,7 +450,7 @@ void vp8_bilinear_predict8x8_c
     HFilter = vp8_bilinear_filters[xoffset];
     VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
 
 }
 
@@ -496,7 +470,7 @@ void vp8_bilinear_predict8x4_c
     HFilter = vp8_bilinear_filters[xoffset];
     VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
 
 }
 
@@ -516,5 +490,5 @@ void vp8_bilinear_predict16x16_c
     HFilter = vp8_bilinear_filters[xoffset];
     VFilter = vp8_bilinear_filters[yoffset];
 
-    vp8_filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
+    filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
 }

vp8/common/findnearmv.c

@@ -11,6 +11,13 @@
 
 #include "findnearmv.h"
 
+const unsigned char vp8_mbsplit_offset[4][16] = {
+    { 0,  8,  0,  0,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  2,  0,  0,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  2,  8, 10,  0,  0,  0,  0,  0,  0,   0,  0,  0,  0,  0,  0},
+    { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9,  10, 11, 12, 13, 14, 15}
+};
+
 /* Predict motion vectors using those from already-decoded nearby blocks.
    Note that we only consider one 4x4 subblock from each candidate 16x16
    macroblock.   */
@@ -18,9 +25,9 @@ void vp8_find_near_mvs
 (
     MACROBLOCKD *xd,
     const MODE_INFO *here,
-    MV *nearest,
-    MV *nearby,
-    MV *best_mv,
+    int_mv *nearest,
+    int_mv *nearby,
+    int_mv *best_mv,
     int cnt[4],
     int refframe,
     int *ref_frame_sign_bias
@@ -124,13 +131,14 @@ void vp8_find_near_mvs
         near_mvs[CNT_INTRA] = near_mvs[CNT_NEAREST];
 
     /* Set up return values */
-    *best_mv = near_mvs[0].as_mv;
-    *nearest = near_mvs[CNT_NEAREST].as_mv;
-    *nearby = near_mvs[CNT_NEAR].as_mv;
+    best_mv->as_int = near_mvs[0].as_int;
+    nearest->as_int = near_mvs[CNT_NEAREST].as_int;
+    nearby->as_int = near_mvs[CNT_NEAR].as_int;
 
-    vp8_clamp_mv(nearest, xd);
-    vp8_clamp_mv(nearby, xd);
-    vp8_clamp_mv(best_mv, xd); /*TODO: move this up before the copy*/
+    //TODO: move clamp outside findnearmv
+    vp8_clamp_mv2(nearest, xd);
+    vp8_clamp_mv2(nearby, xd);
+    vp8_clamp_mv2(best_mv, xd);
 }
 
 vp8_prob *vp8_mv_ref_probs(
@@ -145,26 +153,3 @@ vp8_prob *vp8_mv_ref_probs(
     return p;
 }
 
-const B_MODE_INFO *vp8_left_bmi(const MODE_INFO *cur_mb, int b)
-{
-    if (!(b & 3))
-    {
-        /* On L edge, get from MB to left of us */
-        --cur_mb;
-        b += 4;
-    }
-
-    return cur_mb->bmi + b - 1;
-}
-
-const B_MODE_INFO *vp8_above_bmi(const MODE_INFO *cur_mb, int b, int mi_stride)
-{
-    if (!(b >> 2))
-    {
-        /* On top edge, get from MB above us */
-        cur_mb -= mi_stride;
-        b += 16;
-    }
-
-    return cur_mb->bmi + b - 4;
-}

vp8/common/findnearmv.h

@@ -17,11 +17,6 @@
 #include "modecont.h"
 #include "treecoder.h"
 
-typedef union
-{
-    unsigned int as_int;
-    MV           as_mv;
-} int_mv;        /* facilitates rapid equality tests */
 
 static void mv_bias(int refmb_ref_frame_sign_bias, int refframe, int_mv *mvp, const int *ref_frame_sign_bias)
 {
@@ -39,24 +34,48 @@ static void mv_bias(int refmb_ref_frame_sign_bias, int refframe, int_mv *mvp, co
 
 #define LEFT_TOP_MARGIN (16 << 3)
 #define RIGHT_BOTTOM_MARGIN (16 << 3)
-static void vp8_clamp_mv(MV *mv, const MACROBLOCKD *xd)
+static void vp8_clamp_mv2(int_mv *mv, const MACROBLOCKD *xd)
 {
-    if (mv->col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN))
-        mv->col = xd->mb_to_left_edge - LEFT_TOP_MARGIN;
-    else if (mv->col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN)
-        mv->col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN;
+    if (mv->as_mv.col < (xd->mb_to_left_edge - LEFT_TOP_MARGIN))
+        mv->as_mv.col = xd->mb_to_left_edge - LEFT_TOP_MARGIN;
+    else if (mv->as_mv.col > xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN)
+        mv->as_mv.col = xd->mb_to_right_edge + RIGHT_BOTTOM_MARGIN;
 
-    if (mv->row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN))
-        mv->row = xd->mb_to_top_edge - LEFT_TOP_MARGIN;
-    else if (mv->row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN)
-        mv->row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN;
+    if (mv->as_mv.row < (xd->mb_to_top_edge - LEFT_TOP_MARGIN))
+        mv->as_mv.row = xd->mb_to_top_edge - LEFT_TOP_MARGIN;
+    else if (mv->as_mv.row > xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN)
+        mv->as_mv.row = xd->mb_to_bottom_edge + RIGHT_BOTTOM_MARGIN;
+}
+
+static void vp8_clamp_mv(int_mv *mv, int mb_to_left_edge, int mb_to_right_edge,
+                         int mb_to_top_edge, int mb_to_bottom_edge)
+{
+    mv->as_mv.col = (mv->as_mv.col < mb_to_left_edge) ?
+        mb_to_left_edge : mv->as_mv.col;
+    mv->as_mv.col = (mv->as_mv.col > mb_to_right_edge) ?
+        mb_to_right_edge : mv->as_mv.col;
+    mv->as_mv.row = (mv->as_mv.row < mb_to_top_edge) ?
+        mb_to_top_edge : mv->as_mv.row;
+    mv->as_mv.row = (mv->as_mv.row > mb_to_bottom_edge) ?
+        mb_to_bottom_edge : mv->as_mv.row;
+}
+static unsigned int vp8_check_mv_bounds(int_mv *mv, int mb_to_left_edge,
+                                int mb_to_right_edge, int mb_to_top_edge,
+                                int mb_to_bottom_edge)
+{
+    unsigned int need_to_clamp;
+    need_to_clamp = (mv->as_mv.col < mb_to_left_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.col > mb_to_right_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.row < mb_to_top_edge) ? 1 : 0;
+    need_to_clamp |= (mv->as_mv.row > mb_to_bottom_edge) ? 1 : 0;
+    return need_to_clamp;
 }
 
 void vp8_find_near_mvs
 (
     MACROBLOCKD *xd,
     const MODE_INFO *here,
-    MV *nearest, MV *nearby, MV *best,
+    int_mv *nearest, int_mv *nearby, int_mv *best,
     int near_mv_ref_cts[4],
     int refframe,
     int *ref_frame_sign_bias
@@ -66,8 +85,89 @@ vp8_prob *vp8_mv_ref_probs(
     vp8_prob p[VP8_MVREFS-1], const int near_mv_ref_ct[4]
 );
 
-const B_MODE_INFO *vp8_left_bmi(const MODE_INFO *cur_mb, int b);
+extern const unsigned char vp8_mbsplit_offset[4][16];
 
-const B_MODE_INFO *vp8_above_bmi(const MODE_INFO *cur_mb, int b, int mi_stride);
+
+static int left_block_mv(const MODE_INFO *cur_mb, int b)
+{
+    if (!(b & 3))
+    {
+        /* On L edge, get from MB to left of us */
+        --cur_mb;
+
+        if(cur_mb->mbmi.mode != SPLITMV)
+            return cur_mb->mbmi.mv.as_int;
+        b += 4;
+    }
+
+    return (cur_mb->bmi + b - 1)->mv.as_int;
+}
+
+static int above_block_mv(const MODE_INFO *cur_mb, int b, int mi_stride)
+{
+    if (!(b >> 2))
+    {
+        /* On top edge, get from MB above us */
+        cur_mb -= mi_stride;
+
+        if(cur_mb->mbmi.mode != SPLITMV)
+            return cur_mb->mbmi.mv.as_int;
+        b += 16;
+    }
+
+    return (cur_mb->bmi + b - 4)->mv.as_int;
+}
+static B_PREDICTION_MODE left_block_mode(const MODE_INFO *cur_mb, int b)
+{
+    if (!(b & 3))
+    {
+        /* On L edge, get from MB to left of us */
+        --cur_mb;
+        switch (cur_mb->mbmi.mode)
+        {
+            case B_PRED:
+              return (cur_mb->bmi + b + 3)->as_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;
+            default:
+                return B_DC_PRED;
+        }
+    }
+
+    return (cur_mb->bmi + b - 1)->as_mode;
+}
+
+static B_PREDICTION_MODE above_block_mode(const MODE_INFO *cur_mb, int b, int mi_stride)
+{
+    if (!(b >> 2))
+    {
+        /* On top edge, get from MB above us */
+        cur_mb -= mi_stride;
+
+        switch (cur_mb->mbmi.mode)
+        {
+            case B_PRED:
+              return (cur_mb->bmi + b + 12)->as_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;
+            default:
+                return B_DC_PRED;
+        }
+    }
+
+    return (cur_mb->bmi + b - 4)->as_mode;
+}
 
 #endif

vp8/common/generic/systemdependent.c

@@ -17,14 +17,53 @@
 #include "vp8/common/idct.h"
 #include "vp8/common/onyxc_int.h"
 
+#if CONFIG_MULTITHREAD
+#if HAVE_UNISTD_H
+#include <unistd.h>
+#elif defined(_WIN32)
+#include <windows.h>
+typedef void (WINAPI *PGNSI)(LPSYSTEM_INFO);
+#endif
+#endif
+
 extern void vp8_arch_x86_common_init(VP8_COMMON *ctx);
 extern void vp8_arch_arm_common_init(VP8_COMMON *ctx);
 
-void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
+#if CONFIG_MULTITHREAD
+static int get_cpu_count()
+{
+    int core_count = 16;
 
-void (*vp8_build_intra_predictors_mby_s_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x);
+#if HAVE_UNISTD_H
+#if defined(_SC_NPROCESSORS_ONLN)
+    core_count = sysconf(_SC_NPROCESSORS_ONLN);
+#elif defined(_SC_NPROC_ONLN)
+    core_count = sysconf(_SC_NPROC_ONLN);
+#endif
+#elif defined(_WIN32)
+    {
+        PGNSI pGNSI;
+        SYSTEM_INFO sysinfo;
+
+        /* Call GetNativeSystemInfo if supported or
+         * GetSystemInfo otherwise. */
+
+        pGNSI = (PGNSI) GetProcAddress(
+                GetModuleHandle(TEXT("kernel32.dll")), "GetNativeSystemInfo");
+        if (pGNSI != NULL)
+            pGNSI(&sysinfo);
+        else
+            GetSystemInfo(&sysinfo);
+
+        core_count = sysinfo.dwNumberOfProcessors;
+    }
+#else
+    /* other platforms */
+#endif
+
+    return core_count > 0 ? core_count : 1;
+}
+#endif
 
 void vp8_machine_specific_config(VP8_COMMON *ctx)
 {
@@ -45,6 +84,16 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     rtcd->recon.recon4      = vp8_recon4b_c;
     rtcd->recon.recon_mb    = vp8_recon_mb_c;
     rtcd->recon.recon_mby   = vp8_recon_mby_c;
+    rtcd->recon.build_intra_predictors_mby =
+        vp8_build_intra_predictors_mby;
+    rtcd->recon.build_intra_predictors_mby_s =
+        vp8_build_intra_predictors_mby_s;
+    rtcd->recon.build_intra_predictors_mbuv =
+        vp8_build_intra_predictors_mbuv;
+    rtcd->recon.build_intra_predictors_mbuv_s =
+        vp8_build_intra_predictors_mbuv_s;
+    rtcd->recon.intra4x4_predict =
+        vp8_intra4x4_predict;
 
     rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_c;
     rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_c;
@@ -59,12 +108,12 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_c;
     rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_c;
     rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_c;
-    rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_c;
+    rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_c;
     rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_c;
-    rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_c;
+    rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_c;
     rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_c;
 
-#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_PSNR)
+#if CONFIG_POSTPROC || (CONFIG_VP8_ENCODER && CONFIG_INTERNAL_STATS)
     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;
@@ -75,9 +124,6 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
 #endif
 
 #endif
-    /* Pure C: */
-    vp8_build_intra_predictors_mby_ptr = vp8_build_intra_predictors_mby;
-    vp8_build_intra_predictors_mby_s_ptr = vp8_build_intra_predictors_mby_s;
 
 #if ARCH_X86 || ARCH_X86_64
     vp8_arch_x86_common_init(ctx);
@@ -87,4 +133,7 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
     vp8_arch_arm_common_init(ctx);
 #endif
 
+#if CONFIG_MULTITHREAD
+    ctx->processor_core_count = get_cpu_count();
+#endif /* CONFIG_MULTITHREAD */
 }

vp8/common/loopfilter.c

@@ -9,160 +9,149 @@
  */
 
 
-#include "vpx_ports/config.h"
+#include "vpx_config.h"
 #include "loopfilter.h"
 #include "onyxc_int.h"
+#include "vpx_mem/vpx_mem.h"
 
 typedef unsigned char uc;
 
-
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_c);
 prototype_loopfilter(vp8_loop_filter_vertical_edge_c);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_c);
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
+
+prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
+prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
 
 /* Horizontal MB filtering */
-void vp8_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_mbh_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                           unsigned char *v_ptr, int y_stride, int uv_stride,
+                           loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-void vp8_loop_filter_mbhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Vertical MB Filtering */
-void vp8_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_mbv_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                           unsigned char *v_ptr, int y_stride, int uv_stride,
+                           loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-void vp8_loop_filter_mbvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 /* Horizontal B Filtering */
-void vp8_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bh_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                          unsigned char *v_ptr, int y_stride, int uv_stride,
+                          loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_c(unsigned char *y_ptr, int y_stride,
+                           const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 /* Vertical B Filtering */
-void vp8_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bv_c(unsigned char *y_ptr, unsigned char *u_ptr,
+                          unsigned char *v_ptr, int y_stride, int uv_stride,
+                          loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
-void vp8_loop_filter_bvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                           int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_c(unsigned char *y_ptr, int y_stride,
+                           const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_c(y_ptr + 12, y_stride, blimit);
 }
 
-void vp8_init_loop_filter(VP8_COMMON *cm)
+static void lf_init_lut(loop_filter_info_n *lfi)
 {
-    loop_filter_info *lfi = cm->lf_info;
-    LOOPFILTERTYPE lft = cm->filter_type;
-    int sharpness_lvl = cm->sharpness_level;
-    int frame_type = cm->frame_type;
-    int i, j;
+    int filt_lvl;
 
-    int block_inside_limit = 0;
-    int HEVThresh;
-
-    /* For each possible value for the loop filter fill out a "loop_filter_info" entry. */
-    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    for (filt_lvl = 0; filt_lvl <= MAX_LOOP_FILTER; filt_lvl++)
     {
-        int filt_lvl = i;
-
-        if (frame_type == KEY_FRAME)
+        if (filt_lvl >= 40)
         {
-            if (filt_lvl >= 40)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 2;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 3;
+        }
+        else if (filt_lvl >= 20)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 2;
+        }
+        else if (filt_lvl >= 15)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 1;
         }
         else
         {
-            if (filt_lvl >= 40)
-                HEVThresh = 3;
-            else if (filt_lvl >= 20)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 0;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 0;
         }
+    }
+
+    lfi->mode_lf_lut[DC_PRED] = 1;
+    lfi->mode_lf_lut[V_PRED] = 1;
+    lfi->mode_lf_lut[H_PRED] = 1;
+    lfi->mode_lf_lut[TM_PRED] = 1;
+    lfi->mode_lf_lut[B_PRED]  = 0;
+
+    lfi->mode_lf_lut[ZEROMV]  = 1;
+    lfi->mode_lf_lut[NEARESTMV] = 2;
+    lfi->mode_lf_lut[NEARMV] = 2;
+    lfi->mode_lf_lut[NEWMV] = 2;
+    lfi->mode_lf_lut[SPLITMV] = 3;
+
+}
+
+void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
+                                      int sharpness_lvl)
+{
+    int i;
+
+    /* For each possible value for the loop filter fill out limits */
+    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    {
+        int filt_lvl = i;
+        int block_inside_limit = 0;
 
         /* Set loop filter paramaeters that control sharpness. */
         block_inside_limit = filt_lvl >> (sharpness_lvl > 0);
@@ -177,170 +166,143 @@ void vp8_init_loop_filter(VP8_COMMON *cm)
         if (block_inside_limit < 1)
             block_inside_limit = 1;
 
-        for (j = 0; j < 16; j++)
-        {
-            lfi[i].lim[j] = block_inside_limit;
-            lfi[i].mbflim[j] = filt_lvl + 2;
-            lfi[i].flim[j] = filt_lvl;
-            lfi[i].thr[j] = HEVThresh;
-        }
-
-    }
-
-    /* Set up the function pointers depending on the type of loop filtering selected */
-    if (lft == NORMAL_LOOPFILTER)
-    {
-        cm->lf_mbv = LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v);
-        cm->lf_bv  = LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v);
-        cm->lf_mbh = LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h);
-        cm->lf_bh  = LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h);
-    }
-    else
-    {
-        cm->lf_mbv = LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v);
-        cm->lf_bv  = LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v);
-        cm->lf_mbh = LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h);
-        cm->lf_bh  = LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h);
+        vpx_memset(lfi->lim[i], block_inside_limit, SIMD_WIDTH);
+        vpx_memset(lfi->blim[i], (2 * filt_lvl + block_inside_limit),
+                SIMD_WIDTH);
+        vpx_memset(lfi->mblim[i], (2 * (filt_lvl + 2) + block_inside_limit),
+                SIMD_WIDTH);
     }
 }
 
-/* Put vp8_init_loop_filter() in vp8dx_create_decompressor(). Only call vp8_frame_init_loop_filter() while decoding
- * each frame. Check last_frame_type to skip the function most of times.
- */
-void vp8_frame_init_loop_filter(loop_filter_info *lfi, int frame_type)
+void vp8_loop_filter_init(VP8_COMMON *cm)
 {
-    int HEVThresh;
-    int i, j;
+    loop_filter_info_n *lfi = &cm->lf_info;
+    int i;
 
-    /* For each possible value for the loop filter fill out a "loop_filter_info" entry. */
-    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    /* init limits for given sharpness*/
+    vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+    cm->last_sharpness_level = cm->sharpness_level;
+
+    /* init LUT for lvl  and hev thr picking */
+    lf_init_lut(lfi);
+
+    /* init hev threshold const vectors */
+    for(i = 0; i < 4 ; i++)
     {
-        int filt_lvl = i;
-
-        if (frame_type == KEY_FRAME)
-        {
-            if (filt_lvl >= 40)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
-        }
-        else
-        {
-            if (filt_lvl >= 40)
-                HEVThresh = 3;
-            else if (filt_lvl >= 20)
-                HEVThresh = 2;
-            else if (filt_lvl >= 15)
-                HEVThresh = 1;
-            else
-                HEVThresh = 0;
-        }
-
-        for (j = 0; j < 16; j++)
-        {
-            /*lfi[i].lim[j] = block_inside_limit;
-            lfi[i].mbflim[j] = filt_lvl+2;*/
-            /*lfi[i].flim[j] = filt_lvl;*/
-            lfi[i].thr[j] = HEVThresh;
-        }
+        vpx_memset(lfi->hev_thr[i], i, SIMD_WIDTH);
     }
 }
 
-
-int vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int filter_level)
+void vp8_loop_filter_frame_init(VP8_COMMON *cm,
+                                MACROBLOCKD *mbd,
+                                int default_filt_lvl)
 {
-    MB_MODE_INFO *mbmi = &mbd->mode_info_context->mbmi;
+    int seg,  /* segment number */
+        ref,  /* index in ref_lf_deltas */
+        mode; /* index in mode_lf_deltas */
 
-    if (mbd->mode_ref_lf_delta_enabled)
+    loop_filter_info_n *lfi = &cm->lf_info;
+
+    /* update limits if sharpness has changed */
+    if(cm->last_sharpness_level != cm->sharpness_level)
     {
+        vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+        cm->last_sharpness_level = cm->sharpness_level;
+    }
+
+    for(seg = 0; seg < MAX_MB_SEGMENTS; seg++)
+    {
+        int lvl_seg = default_filt_lvl;
+        int lvl_ref, lvl_mode;
+
+        /* Note the baseline filter values for each segment */
+        if (mbd->segmentation_enabled)
+        {
+            /* Abs value */
+            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
+            {
+                lvl_seg = mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+            }
+            else  /* Delta Value */
+            {
+                lvl_seg += mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+                lvl_seg = (lvl_seg > 0) ? ((lvl_seg > 63) ? 63: lvl_seg) : 0;
+            }
+        }
+
+        if (!mbd->mode_ref_lf_delta_enabled)
+        {
+            /* we could get rid of this if we assume that deltas are set to
+             * zero when not in use; encoder always uses deltas
+             */
+            vpx_memset(lfi->lvl[seg][0], lvl_seg, 4 * 4 );
+            continue;
+        }
+
+        lvl_ref = lvl_seg;
+
+        /* INTRA_FRAME */
+        ref = INTRA_FRAME;
+
         /* Apply delta for reference frame */
-        filter_level += mbd->ref_lf_deltas[mbmi->ref_frame];
+        lvl_ref += mbd->ref_lf_deltas[ref];
 
-        /* Apply delta for mode */
-        if (mbmi->ref_frame == INTRA_FRAME)
+        /* Apply delta for Intra modes */
+        mode = 0; /* B_PRED */
+        /* Only the split mode BPRED has a further special case */
+        lvl_mode = lvl_ref +  mbd->mode_lf_deltas[mode];
+        lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
+
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        mode = 1; /* all the rest of Intra modes */
+        lvl_mode = (lvl_ref > 0) ? (lvl_ref > 63 ? 63 : lvl_ref)  : 0; /* clamp */
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        /* LAST, GOLDEN, ALT */
+        for(ref = 1; ref < MAX_REF_FRAMES; ref++)
         {
-            /* Only the split mode BPRED has a further special case */
-            if (mbmi->mode == B_PRED)
-                filter_level +=  mbd->mode_lf_deltas[0];
+            int lvl_ref = lvl_seg;
+
+            /* Apply delta for reference frame */
+            lvl_ref += mbd->ref_lf_deltas[ref];
+
+            /* Apply delta for Inter modes */
+            for (mode = 1; mode < 4; mode++)
+            {
+                lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
+                lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0; /* clamp */
+
+                lfi->lvl[seg][ref][mode] = lvl_mode;
+            }
         }
-        else
-        {
-            /* Zero motion mode */
-            if (mbmi->mode == ZEROMV)
-                filter_level +=  mbd->mode_lf_deltas[1];
-
-            /* Split MB motion mode */
-            else if (mbmi->mode == SPLITMV)
-                filter_level +=  mbd->mode_lf_deltas[3];
-
-            /* All other inter motion modes (Nearest, Near, New) */
-            else
-                filter_level +=  mbd->mode_lf_deltas[2];
-        }
-
-        /* Range check */
-        if (filter_level > MAX_LOOP_FILTER)
-            filter_level = MAX_LOOP_FILTER;
-        else if (filter_level < 0)
-            filter_level = 0;
     }
-    return filter_level;
 }
 
-
 void vp8_loop_filter_frame
 (
     VP8_COMMON *cm,
-    MACROBLOCKD *mbd,
-    int default_filt_lvl
+    MACROBLOCKD *mbd
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
-    loop_filter_info *lfi = cm->lf_info;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
     FRAME_TYPE 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;
 
-    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)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-            /* Delta Value */
-            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;
-    }
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
 
     /* 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);
+    vp8_loop_filter_frame_init(cm, mbd, cm->filter_level);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer;
@@ -352,101 +314,108 @@ void vp8_loop_filter_frame
     {
         for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
         {
-            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                            mode_info_context->mbmi.mode != SPLITMV &&
+                            mode_info_context->mbmi.mb_skip_coeff);
 
-            filter_level = baseline_filter_level[Segment];
+            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
 
-            /* 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
-             */
-            filter_level = vp8_adjust_mb_lf_value(mbd, filter_level);
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
 
             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 (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                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);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
 
-                /* 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 (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
 
-                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);
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
             u_ptr += 8;
             v_ptr += 8;
 
-            mbd->mode_info_context++;     /* step to next MB */
+            mode_info_context++;     /* step to next MB */
         }
 
         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 */
+        mode_info_context++;         /* Skip border mb */
     }
 }
 
-
 void vp8_loop_filter_frame_yonly
 (
     VP8_COMMON *cm,
     MACROBLOCKD *mbd,
-    int default_filt_lvl,
-    int sharpness_lvl
+    int default_filt_lvl
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
 
-    int i;
     unsigned char *y_ptr;
     int mb_row;
     int mb_col;
 
-    loop_filter_info *lfi = cm->lf_info;
-    int baseline_filter_level[MAX_MB_SEGMENTS];
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
     int filter_level;
-    int alt_flt_enabled = mbd->segmentation_enabled;
     FRAME_TYPE frame_type = cm->frame_type;
 
-    (void) sharpness_lvl;
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
 
-    /*MODE_INFO * this_mb_mode_info = cm->mi;*/ /* Point at base of Mb MODE_INFO list */
-    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)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
-            /* Delta Value */
-            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;
-    }
+#if 0
+    if(default_filt_lvl == 0) /* no filter applied */
+        return;
+#endif
 
     /* 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);
+    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer;
@@ -456,72 +425,106 @@ void vp8_loop_filter_frame_yonly
     {
         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];
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                            mode_info_context->mbmi.mode != SPLITMV &&
+                            mode_info_context->mbmi.mb_skip_coeff);
 
-            /* Apply any context driven MB level adjustment */
-            filter_level = vp8_adjust_mb_lf_value(mbd, filter_level);
+            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
 
             if (filter_level)
             {
-                if (mb_col > 0)
-                    cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                /* don't apply across umv border */
-                if (mb_row > 0)
-                    cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
-            mbd->mode_info_context ++;        /* step to next MB */
+            mode_info_context ++;        /* step to next MB */
 
         }
 
         y_ptr += post->y_stride  * 16 - post->y_width;
-        mbd->mode_info_context ++;            /* Skip border mb */
+        mode_info_context ++;            /* Skip border mb */
     }
 
 }
 
-
 void vp8_loop_filter_partial_frame
 (
     VP8_COMMON *cm,
     MACROBLOCKD *mbd,
-    int default_filt_lvl,
-    int sharpness_lvl,
-    int Fraction
+    int default_filt_lvl
 )
 {
     YV12_BUFFER_CONFIG *post = cm->frame_to_show;
 
-    int i;
     unsigned char *y_ptr;
     int mb_row;
     int mb_col;
-    /*int mb_rows = post->y_height >> 4;*/
     int mb_cols = post->y_width  >> 4;
 
-    int linestocopy;
+    int linestocopy, i;
+
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
 
-    loop_filter_info *lfi = cm->lf_info;
-    int baseline_filter_level[MAX_MB_SEGMENTS];
     int filter_level;
     int alt_flt_enabled = mbd->segmentation_enabled;
     FRAME_TYPE frame_type = cm->frame_type;
 
-    (void) sharpness_lvl;
+    const MODE_INFO *mode_info_context;
 
-    /*MODE_INFO * this_mb_mode_info = cm->mi + (post->y_height>>5) * (mb_cols + 1);*/ /* Point at base of Mb MODE_INFO list */
-    mbd->mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);        /* Point at base of Mb MODE_INFO list */
+    int lvl_seg[MAX_MB_SEGMENTS];
 
-    linestocopy = (post->y_height >> (4 + Fraction));
+    mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);
+
+    /* 3 is a magic number. 4 is probably magic too */
+    linestocopy = (post->y_height >> (4 + 3));
 
     if (linestocopy < 1)
         linestocopy = 1;
@@ -529,32 +532,27 @@ void vp8_loop_filter_partial_frame
     linestocopy <<= 4;
 
     /* Note the baseline filter values for each segment */
+    /* See vp8_loop_filter_frame_init. Rather than call that for each change
+     * to default_filt_lvl, copy the relevant calculation here.
+     */
     if (alt_flt_enabled)
     {
         for (i = 0; i < MAX_MB_SEGMENTS; i++)
-        {
-            /* Abs value */
+        {    /* Abs value */
             if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-                baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+            {
+                lvl_seg[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+            }
             /* Delta Value */
             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 */
+                lvl_seg[i] = default_filt_lvl
+                        + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
+                lvl_seg[i] = (lvl_seg[i] > 0) ?
+                        ((lvl_seg[i] > 63) ? 63: lvl_seg[i]) : 0;
             }
         }
     }
-    else
-    {
-        for (i = 0; i < MAX_MB_SEGMENTS; i++)
-            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);
 
     /* Set up the buffer pointers */
     y_ptr = post->y_buffer + (post->y_height >> 5) * 16 * post->y_stride;
@@ -564,28 +562,64 @@ void vp8_loop_filter_partial_frame
     {
         for (mb_col = 0; mb_col < mb_cols; mb_col++)
         {
-            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
-            filter_level = baseline_filter_level[Segment];
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                           mode_info_context->mbmi.mode != SPLITMV &&
+                           mode_info_context->mbmi.mb_skip_coeff);
+
+            if (alt_flt_enabled)
+                filter_level = lvl_seg[mode_info_context->mbmi.segment_id];
+            else
+                filter_level = default_filt_lvl;
 
             if (filter_level)
             {
-                if (mb_col > 0)
-                    cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_v)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
-                    cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
+                    LF_INVOKE(&cm->rtcd.loopfilter, normal_mb_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, normal_b_h)
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_v)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_v)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    LF_INVOKE(&cm->rtcd.loopfilter, simple_mb_h)
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        LF_INVOKE(&cm->rtcd.loopfilter, simple_b_h)
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
             }
 
             y_ptr += 16;
-            mbd->mode_info_context += 1;      /* step to next MB */
+            mode_info_context += 1;      /* step to next MB */
         }
 
         y_ptr += post->y_stride  * 16 - post->y_width;
-        mbd->mode_info_context += 1;          /* Skip border mb */
+        mode_info_context += 1;          /* Skip border mb */
     }
 }

vp8/common/loopfilter.h

@@ -13,6 +13,7 @@
 #define loopfilter_h
 
 #include "vpx_ports/mem.h"
+#include "vpx_config.h"
 
 #define MAX_LOOP_FILTER 63
 
@@ -22,26 +23,45 @@ typedef enum
     SIMPLE_LOOPFILTER = 1
 } LOOPFILTERTYPE;
 
-/* FRK
- * Need to align this structure so when it is declared and
+#if ARCH_ARM
+#define SIMD_WIDTH 1
+#else
+#define SIMD_WIDTH 16
+#endif
+
+/* Need to align this structure so when it is declared and
  * passed it can be loaded into vector registers.
  */
 typedef struct
 {
-    DECLARE_ALIGNED(16, signed char, lim[16]);
-    DECLARE_ALIGNED(16, signed char, flim[16]);
-    DECLARE_ALIGNED(16, signed char, thr[16]);
-    DECLARE_ALIGNED(16, signed char, mbflim[16]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, mblim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, blim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, lim[MAX_LOOP_FILTER + 1][SIMD_WIDTH]);
+    DECLARE_ALIGNED(SIMD_WIDTH, unsigned char, hev_thr[4][SIMD_WIDTH]);
+    unsigned char lvl[4][4][4];
+    unsigned char hev_thr_lut[2][MAX_LOOP_FILTER + 1];
+    unsigned char mode_lf_lut[10];
+} loop_filter_info_n;
+
+typedef struct
+{
+    const unsigned char * mblim;
+    const unsigned char * blim;
+    const unsigned char * lim;
+    const unsigned char * hev_thr;
 } loop_filter_info;
 
 
 #define prototype_loopfilter(sym) \
-    void sym(unsigned char *src, int pitch, const signed char *flimit,\
-             const signed char *limit, const signed char *thresh, int count)
+    void sym(unsigned char *src, int pitch, const unsigned char *blimit,\
+             const unsigned char *limit, const unsigned char *thresh, int count)
 
 #define prototype_loopfilter_block(sym) \
-    void sym(unsigned char *y, unsigned char *u, unsigned char *v,\
-             int ystride, int uv_stride, loop_filter_info *lfi, int simpler)
+    void sym(unsigned char *y, unsigned char *u, unsigned char *v, \
+             int ystride, int uv_stride, loop_filter_info *lfi)
+
+#define prototype_simple_loopfilter(sym) \
+    void sym(unsigned char *y, int ystride, const unsigned char *blimit)
 
 #if ARCH_X86 || ARCH_X86_64
 #include "x86/loopfilter_x86.h"
@@ -71,38 +91,39 @@ extern prototype_loopfilter_block(vp8_lf_normal_mb_h);
 #endif
 extern prototype_loopfilter_block(vp8_lf_normal_b_h);
 
-
 #ifndef vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_c
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_mb_v);
+extern prototype_simple_loopfilter(vp8_lf_simple_mb_v);
 
 #ifndef vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_b_v);
+extern prototype_simple_loopfilter(vp8_lf_simple_b_v);
 
 #ifndef vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_c
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_mb_h);
+extern prototype_simple_loopfilter(vp8_lf_simple_mb_h);
 
 #ifndef vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_c
 #endif
-extern prototype_loopfilter_block(vp8_lf_simple_b_h);
+extern prototype_simple_loopfilter(vp8_lf_simple_b_h);
 
 typedef prototype_loopfilter_block((*vp8_lf_block_fn_t));
+typedef prototype_simple_loopfilter((*vp8_slf_block_fn_t));
+
 typedef struct
 {
     vp8_lf_block_fn_t  normal_mb_v;
     vp8_lf_block_fn_t  normal_b_v;
     vp8_lf_block_fn_t  normal_mb_h;
     vp8_lf_block_fn_t  normal_b_h;
-    vp8_lf_block_fn_t  simple_mb_v;
-    vp8_lf_block_fn_t  simple_b_v;
-    vp8_lf_block_fn_t  simple_mb_h;
-    vp8_lf_block_fn_t  simple_b_h;
+    vp8_slf_block_fn_t  simple_mb_v;
+    vp8_slf_block_fn_t  simple_b_v;
+    vp8_slf_block_fn_t  simple_mb_h;
+    vp8_slf_block_fn_t  simple_b_h;
 } vp8_loopfilter_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -115,10 +136,33 @@ 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,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     unsigned char *v
 );
 
+/* assorted loopfilter functions which get used elsewhere */
+struct VP8Common;
+struct MacroBlockD;
+
+void vp8_loop_filter_init(struct VP8Common *cm);
+
+void vp8_loop_filter_frame_init(struct VP8Common *cm,
+                                struct MacroBlockD *mbd,
+                                int default_filt_lvl);
+
+void vp8_loop_filter_frame(struct VP8Common *cm, struct MacroBlockD *mbd);
+
+void vp8_loop_filter_partial_frame(struct VP8Common *cm,
+                                   struct MacroBlockD *mbd,
+                                   int default_filt_lvl);
+
+void vp8_loop_filter_frame_yonly(struct VP8Common *cm,
+                                 struct MacroBlockD *mbd,
+                                 int default_filt_lvl);
+
+void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
+                                      int sharpness_lvl);
+
 #endif

vp8/common/loopfilter_filters.c

@@ -24,8 +24,9 @@ static __inline signed char vp8_signed_char_clamp(int t)
 
 
 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */
-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)
+static __inline signed char vp8_filter_mask(uc limit, uc blimit,
+                                     uc p3, uc p2, uc p1, uc p0,
+                                     uc q0, uc q1, uc q2, uc q3)
 {
     signed char mask = 0;
     mask |= (abs(p3 - p2) > limit) * -1;
@@ -34,13 +35,13 @@ static __inline signed char vp8_filter_mask(signed char limit, signed char flimi
     mask |= (abs(q1 - q0) > limit) * -1;
     mask |= (abs(q2 - q1) > limit) * -1;
     mask |= (abs(q3 - q2) > limit) * -1;
-    mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  > flimit * 2 + limit) * -1;
+    mask |= (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  > blimit) * -1;
     mask = ~mask;
     return mask;
 }
 
 /* is there high variance internal edge ( 11111111 yes, 00000000 no) */
-static __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0, uc q1)
+static __inline signed char vp8_hevmask(uc thresh, uc p1, uc p0, uc q0, uc q1)
 {
     signed char hev = 0;
     hev  |= (abs(p1 - p0) > thresh) * -1;
@@ -48,7 +49,8 @@ static __inline signed char vp8_hevmask(signed char thresh, uc p1, uc p0, uc q0,
     return hev;
 }
 
-static __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, uc hev, uc *op1,
+        uc *op0, uc *oq0, uc *oq1)
 
 {
     signed char ps0, qs0;
@@ -98,9 +100,9 @@ void vp8_loop_filter_horizontal_edge_c
 (
     unsigned char *s,
     int p, /* pitch */
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -113,11 +115,11 @@ void vp8_loop_filter_horizontal_edge_c
      */
     do
     {
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4*p], s[-3*p], s[-2*p], s[-1*p],
                                s[0*p], s[1*p], s[2*p], s[3*p]);
 
-        hev = vp8_hevmask(thresh[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        hev = vp8_hevmask(thresh[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
 
         vp8_filter(mask, hev, s - 2 * p, s - 1 * p, s, s + 1 * p);
 
@@ -130,9 +132,9 @@ void vp8_loop_filter_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -145,10 +147,10 @@ void vp8_loop_filter_vertical_edge_c
      */
     do
     {
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4], s[-3], s[-2], s[-1], s[0], s[1], s[2], s[3]);
 
-        hev = vp8_hevmask(thresh[i], s[-2], s[-1], s[0], s[1]);
+        hev = vp8_hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
 
         vp8_filter(mask, hev, s - 2, s - 1, s, s + 1);
 
@@ -157,7 +159,7 @@ void vp8_loop_filter_vertical_edge_c
     while (++i < count * 8);
 }
 
-static __inline void vp8_mbfilter(signed char mask, signed char hev,
+static __inline void vp8_mbfilter(signed char mask, uc hev,
                            uc *op2, uc *op1, uc *op0, uc *oq0, uc *oq1, uc *oq2)
 {
     signed char s, u;
@@ -216,9 +218,9 @@ void vp8_mbloop_filter_horizontal_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -232,11 +234,11 @@ void vp8_mbloop_filter_horizontal_edge_c
     do
     {
 
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4*p], s[-3*p], s[-2*p], s[-1*p],
                                s[0*p], s[1*p], s[2*p], s[3*p]);
 
-        hev = vp8_hevmask(thresh[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        hev = vp8_hevmask(thresh[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
 
         vp8_mbfilter(mask, hev, s - 3 * p, s - 2 * p, s - 1 * p, s, s + 1 * p, s + 2 * p);
 
@@ -251,9 +253,9 @@ void vp8_mbloop_filter_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
+    const unsigned char *blimit,
+    const unsigned char *limit,
+    const unsigned char *thresh,
     int count
 )
 {
@@ -264,10 +266,10 @@ void vp8_mbloop_filter_vertical_edge_c
     do
     {
 
-        mask = vp8_filter_mask(limit[i], flimit[i],
+        mask = vp8_filter_mask(limit[0], blimit[0],
                                s[-4], s[-3], s[-2], s[-1], s[0], s[1], s[2], s[3]);
 
-        hev = vp8_hevmask(thresh[i], s[-2], s[-1], s[0], s[1]);
+        hev = vp8_hevmask(thresh[0], s[-2], s[-1], s[0], s[1]);
 
         vp8_mbfilter(mask, hev, s - 3, s - 2, s - 1, s, s + 1, s + 2);
 
@@ -278,13 +280,13 @@ void vp8_mbloop_filter_vertical_edge_c
 }
 
 /* should we apply any filter at all ( 11111111 yes, 00000000 no) */
-static __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(uc blimit, uc p1, uc p0, uc q0, uc q1)
 {
 /* Why does this cause problems for win32?
  * error C2143: syntax error : missing ';' before 'type'
  *  (void) limit;
  */
-    signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  <= flimit * 2 + limit) * -1;
+    signed char mask = (abs(p0 - q0) * 2 + abs(p1 - q1) / 2  <= blimit) * -1;
     return mask;
 }
 
@@ -317,47 +319,37 @@ void vp8_loop_filter_simple_horizontal_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
-    int count
+    const unsigned char *blimit
 )
 {
     signed char mask = 0;
     int i = 0;
-    (void) thresh;
 
     do
     {
-        /*mask = vp8_simple_filter_mask( limit[i], flimit[i],s[-1*p],s[0*p]);*/
-        mask = vp8_simple_filter_mask(limit[i], flimit[i], s[-2*p], s[-1*p], s[0*p], s[1*p]);
+        mask = vp8_simple_filter_mask(blimit[0], s[-2*p], s[-1*p], s[0*p], s[1*p]);
         vp8_simple_filter(mask, s - 2 * p, s - 1 * p, s, s + 1 * p);
         ++s;
     }
-    while (++i < count * 8);
+    while (++i < 16);
 }
 
 void vp8_loop_filter_simple_vertical_edge_c
 (
     unsigned char *s,
     int p,
-    const signed char *flimit,
-    const signed char *limit,
-    const signed char *thresh,
-    int count
+    const unsigned char *blimit
 )
 {
     signed char mask = 0;
     int i = 0;
-    (void) thresh;
 
     do
     {
-        /*mask = vp8_simple_filter_mask( limit[i], flimit[i],s[-1],s[0]);*/
-        mask = vp8_simple_filter_mask(limit[i], flimit[i], s[-2], s[-1], s[0], s[1]);
+        mask = vp8_simple_filter_mask(blimit[0], s[-2], s[-1], s[0], s[1]);
         vp8_simple_filter(mask, s - 2, s - 1, s, s + 1);
         s += p;
     }
-    while (++i < count * 8);
+    while (++i < 16);
 
 }

vp8/common/mbpitch.c

@@ -17,7 +17,7 @@ typedef enum
     DEST = 1
 } BLOCKSET;
 
-void vp8_setup_block
+static void setup_block
 (
     BLOCKD *b,
     int mv_stride,
@@ -43,7 +43,8 @@ void vp8_setup_block
 
 }
 
-void vp8_setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
+
+static void setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
 {
     int block;
 
@@ -64,16 +65,16 @@ void vp8_setup_macroblock(MACROBLOCKD *x, BLOCKSET bs)
 
     for (block = 0; block < 16; block++) /* y blocks */
     {
-        vp8_setup_block(&x->block[block], x->dst.y_stride, y, x->dst.y_stride,
+        setup_block(&x->block[block], x->dst.y_stride, y, x->dst.y_stride,
                         (block >> 2) * 4 * x->dst.y_stride + (block & 3) * 4, bs);
     }
 
     for (block = 16; block < 20; block++) /* U and V blocks */
     {
-        vp8_setup_block(&x->block[block], x->dst.uv_stride, u, x->dst.uv_stride,
+        setup_block(&x->block[block], x->dst.uv_stride, u, x->dst.uv_stride,
                         ((block - 16) >> 1) * 4 * x->dst.uv_stride + (block & 1) * 4, bs);
 
-        vp8_setup_block(&x->block[block+4], x->dst.uv_stride, v, x->dst.uv_stride,
+        setup_block(&x->block[block+4], x->dst.uv_stride, v, x->dst.uv_stride,
                         ((block - 16) >> 1) * 4 * x->dst.uv_stride + (block & 1) * 4, bs);
     }
 }
@@ -124,6 +125,6 @@ void vp8_build_block_doffsets(MACROBLOCKD *x)
 {
 
     /* handle the destination pitch features */
-    vp8_setup_macroblock(x, DEST);
-    vp8_setup_macroblock(x, PRED);
+    setup_macroblock(x, DEST);
+    setup_macroblock(x, PRED);
 }

vp8/common/mv.h

@@ -11,6 +11,7 @@
 
 #ifndef __INC_MV_H
 #define __INC_MV_H
+#include "vpx/vpx_integer.h"
 
 typedef struct
 {
@@ -18,4 +19,10 @@ typedef struct
     short col;
 } MV;
 
+typedef union
+{
+    uint32_t  as_int;
+    MV        as_mv;
+} int_mv;        /* facilitates faster equality tests and copies */
+
 #endif

vp8/common/onyx.h

@@ -109,6 +109,7 @@ extern "C"
         int noise_sensitivity;   // parameter used for applying pre processing blur: recommendation 0
         int Sharpness;          // parameter used for sharpening output: recommendation 0:
         int cpu_used;
+        unsigned int rc_max_intra_bitrate_pct;
 
         // mode ->
         //(0)=Realtime/Live Encoding. This mode is optimized for realtim encoding (for example, capturing
@@ -139,8 +140,9 @@ extern "C"
 
         int end_usage; // vbr or cbr
 
-        // shoot to keep buffer full at all times by undershooting a bit 95 recommended
+        // buffer targeting aggressiveness
         int under_shoot_pct;
+        int over_shoot_pct;
 
         // buffering parameters
         int starting_buffer_level;  // in seconds
@@ -182,8 +184,11 @@ extern "C"
         int token_partitions; // how many token partitions to create for multi core decoding
         int encode_breakout;  // early breakout encode threshold : for video conf recommend 800
 
-        int error_resilient_mode;  // if running over udp networks provides decodable frames after a
-        // dropped packet
+        unsigned int error_resilient_mode; // Bitfield defining the error
+                                   // resiliency features to enable. Can provide
+                                   // decodable frames after losses in previous
+                                   // frames and decodable partitions after
+                                   // losses in the same frame.
 
         int arnr_max_frames;
         int arnr_strength ;
@@ -206,8 +211,8 @@ extern "C"
 
 // receive a frames worth of data caller can assume that a copy of this frame is made
 // and not just a copy of the pointer..
-    int vp8_receive_raw_frame(VP8_PTR comp, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, INT64 time_stamp, INT64 end_time_stamp);
-    int vp8_get_compressed_data(VP8_PTR comp, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, INT64 *time_stamp, INT64 *time_end, int flush);
+    int vp8_receive_raw_frame(VP8_PTR comp, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, int64_t time_stamp, int64_t end_time_stamp);
+    int vp8_get_compressed_data(VP8_PTR comp, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, int64_t *time_stamp, int64_t *time_end, int flush);
     int vp8_get_preview_raw_frame(VP8_PTR comp, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t *flags);
 
     int vp8_use_as_reference(VP8_PTR comp, int ref_frame_flags);

vp8/common/onyxc_int.h

@@ -19,7 +19,9 @@
 #include "entropy.h"
 #include "idct.h"
 #include "recon.h"
+#if CONFIG_POSTPROC
 #include "postproc.h"
+#endif
 
 /*#ifdef PACKET_TESTING*/
 #include "header.h"
@@ -35,13 +37,15 @@ void vp8_initialize_common(void);
 
 #define NUM_YV12_BUFFERS 4
 
+#define MAX_PARTITIONS 9
+
 typedef struct frame_contexts
 {
     vp8_prob bmode_prob [VP8_BINTRAMODES-1];
     vp8_prob ymode_prob [VP8_YMODES-1];   /* interframe intra mode probs */
     vp8_prob uv_mode_prob [VP8_UV_MODES-1];
     vp8_prob sub_mv_ref_prob [VP8_SUBMVREFS-1];
-    vp8_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [vp8_coef_tokens-1];
+    vp8_prob coef_probs [BLOCK_TYPES] [COEF_BANDS] [PREV_COEF_CONTEXTS] [ENTROPY_NODES];
     MV_CONTEXT mvc[2];
     MV_CONTEXT pre_mvc[2];  /* not to caculate the mvcost for the frame if mvc doesn't change. */
 } FRAME_CONTEXT;
@@ -73,7 +77,9 @@ typedef struct VP8_COMMON_RTCD
     vp8_recon_rtcd_vtable_t       recon;
     vp8_subpix_rtcd_vtable_t      subpix;
     vp8_loopfilter_rtcd_vtable_t  loopfilter;
+#if CONFIG_POSTPROC
     vp8_postproc_rtcd_vtable_t    postproc;
+#endif
     int                           flags;
 #else
     int unused;
@@ -81,6 +87,7 @@ typedef struct VP8_COMMON_RTCD
 } VP8_COMMON_RTCD;
 
 typedef struct VP8Common
+
 {
     struct vpx_internal_error_info  error;
 
@@ -105,7 +112,8 @@ typedef struct VP8Common
     YV12_BUFFER_CONFIG post_proc_buffer;
     YV12_BUFFER_CONFIG temp_scale_frame;
 
-    FRAME_TYPE last_frame_type;  /* Save last frame's frame type for loopfilter init checking and motion search. */
+
+    FRAME_TYPE last_frame_type;  /* Save last frame's frame type for motion search. */
     FRAME_TYPE frame_type;
 
     int show_frame;
@@ -119,7 +127,6 @@ typedef struct VP8Common
     /* profile settings */
     int mb_no_coeff_skip;
     int no_lpf;
-    int simpler_lpf;
     int use_bilinear_mc_filter;
     int full_pixel;
 
@@ -140,16 +147,15 @@ typedef struct VP8Common
 
     MODE_INFO *mip; /* Base of allocated array */
     MODE_INFO *mi;  /* Corresponds to upper left visible macroblock */
+    MODE_INFO *prev_mip; /* MODE_INFO array 'mip' from last decoded frame */
+    MODE_INFO *prev_mi;  /* 'mi' from last frame (points into prev_mip) */
 
 
     INTERPOLATIONFILTERTYPE mcomp_filter_type;
-    LOOPFILTERTYPE last_filter_type;
     LOOPFILTERTYPE filter_type;
-    loop_filter_info lf_info[MAX_LOOP_FILTER+1];
-    prototype_loopfilter_block((*lf_mbv));
-    prototype_loopfilter_block((*lf_mbh));
-    prototype_loopfilter_block((*lf_bv));
-    prototype_loopfilter_block((*lf_bh));
+
+    loop_filter_info_n lf_info;
+
     int filter_level;
     int last_sharpness_level;
     int sharpness_level;
@@ -196,13 +202,12 @@ typedef struct VP8Common
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD rtcd;
 #endif
+#if CONFIG_MULTITHREAD
+    int processor_core_count;
+#endif
+#if CONFIG_POSTPROC
     struct postproc_state  postproc_state;
+#endif
 } VP8_COMMON;
 
-
-int 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/onyxd.h

@@ -22,6 +22,7 @@ extern "C"
 #include "vpx_scale/yv12config.h"
 #include "ppflags.h"
 #include "vpx_ports/mem.h"
+#include "vpx/vpx_codec.h"
 
     typedef void   *VP8D_PTR;
     typedef struct
@@ -31,6 +32,8 @@ extern "C"
         int     Version;
         int     postprocess;
         int     max_threads;
+        int     error_concealment;
+        int     input_partition;
     } VP8D_CONFIG;
     typedef enum
     {
@@ -50,11 +53,11 @@ extern "C"
 
     int vp8dx_get_setting(VP8D_PTR comp, VP8D_SETTING oxst);
 
-    int vp8dx_receive_compressed_data(VP8D_PTR comp, unsigned long size, const unsigned char *dest, INT64 time_stamp);
-    int vp8dx_get_raw_frame(VP8D_PTR comp, YV12_BUFFER_CONFIG *sd, INT64 *time_stamp, INT64 *time_end_stamp, vp8_ppflags_t *flags);
+    int vp8dx_receive_compressed_data(VP8D_PTR comp, unsigned long size, const unsigned char *dest, int64_t time_stamp);
+    int vp8dx_get_raw_frame(VP8D_PTR comp, YV12_BUFFER_CONFIG *sd, int64_t *time_stamp, int64_t *time_end_stamp, vp8_ppflags_t *flags);
 
-    int vp8dx_get_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
-    int vp8dx_set_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_get_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_set_reference(VP8D_PTR comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
 
     VP8D_PTR vp8dx_create_decompressor(VP8D_CONFIG *oxcf);
 

vp8/common/postproc.c

@@ -211,7 +211,7 @@ void vp8_post_proc_down_and_across_c
     }
 }
 
-int vp8_q2mbl(int x)
+static int q2mbl(int x)
 {
     if (x < 20) x = 20;
 
@@ -314,8 +314,8 @@ static void vp8_deblock_and_de_macro_block(YV12_BUFFER_CONFIG         *source,
     (void) flag;
 
     POSTPROC_INVOKE(rtcd, downacross)(source->y_buffer, post->y_buffer, source->y_stride,  post->y_stride, source->y_height, source->y_width,  ppl);
-    POSTPROC_INVOKE(rtcd, across)(post->y_buffer, post->y_stride, post->y_height, post->y_width, vp8_q2mbl(q));
-    POSTPROC_INVOKE(rtcd, down)(post->y_buffer, post->y_stride, post->y_height, post->y_width, vp8_q2mbl(q));
+    POSTPROC_INVOKE(rtcd, across)(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
+    POSTPROC_INVOKE(rtcd, down)(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
 
     POSTPROC_INVOKE(rtcd, downacross)(source->u_buffer, post->u_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
     POSTPROC_INVOKE(rtcd, downacross)(source->v_buffer, post->v_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
@@ -804,11 +804,14 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
             for (j = 0; j < mb_cols; j++)
             {
                 char zz[4];
+                int dc_diff = !(mi[mb_index].mbmi.mode != B_PRED &&
+                              mi[mb_index].mbmi.mode != SPLITMV &&
+                              mi[mb_index].mbmi.mb_skip_coeff);
 
                 if (oci->frame_type == KEY_FRAME)
                     sprintf(zz, "a");
                 else
-                    sprintf(zz, "%c", mi[mb_index].mbmi.dc_diff + '0');
+                    sprintf(zz, "%c", dc_diff + '0');
 
                 vp8_blit_text(zz, y_ptr, post->y_stride);
                 mb_index ++;
@@ -834,7 +837,6 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
         YV12_BUFFER_CONFIG *post = &oci->post_proc_buffer;
         int width  = post->y_width;
         int height = post->y_height;
-        int mb_cols = width  >> 4;
         unsigned char *y_buffer = oci->post_proc_buffer.y_buffer;
         int y_stride = oci->post_proc_buffer.y_stride;
         MODE_INFO *mi = oci->mi;
@@ -858,7 +860,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                     {
                         case 0 :    /* mv_top_bottom */
                         {
-                            B_MODE_INFO *bmi = &mi->bmi[0];
+                            union b_mode_info *bmi = &mi->bmi[0];
                             MV *mv = &bmi->mv.as_mv;
 
                             x1 = x0 + 8 + (mv->col >> 3);
@@ -879,7 +881,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                         }
                         case 1 :    /* mv_left_right */
                         {
-                            B_MODE_INFO *bmi = &mi->bmi[0];
+                            union b_mode_info *bmi = &mi->bmi[0];
                             MV *mv = &bmi->mv.as_mv;
 
                             x1 = x0 + 4 + (mv->col >> 3);
@@ -900,7 +902,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                         }
                         case 2 :    /* mv_quarters   */
                         {
-                            B_MODE_INFO *bmi = &mi->bmi[0];
+                            union b_mode_info *bmi = &mi->bmi[0];
                             MV *mv = &bmi->mv.as_mv;
 
                             x1 = x0 + 4 + (mv->col >> 3);
@@ -936,7 +938,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                         }
                         default :
                         {
-                            B_MODE_INFO *bmi = mi->bmi;
+                            union b_mode_info *bmi = mi->bmi;
                             int bx0, by0;
 
                             for (by0 = y0; by0 < (y0+16); by0 += 4)
@@ -1009,7 +1011,7 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                 {
                     int by, bx;
                     unsigned char *yl, *ul, *vl;
-                    B_MODE_INFO *bmi = mi->bmi;
+                    union b_mode_info *bmi = mi->bmi;
 
                     yl = y_ptr + x;
                     ul = u_ptr + (x>>1);
@@ -1022,9 +1024,9 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
                             if ((ppflags->display_b_modes_flag & (1<<mi->mbmi.mode))
                                 || (ppflags->display_mb_modes_flag & B_PRED))
                             {
-                                Y = B_PREDICTION_MODE_colors[bmi->mode][0];
-                                U = B_PREDICTION_MODE_colors[bmi->mode][1];
-                                V = B_PREDICTION_MODE_colors[bmi->mode][2];
+                                Y = B_PREDICTION_MODE_colors[bmi->as_mode][0];
+                                U = B_PREDICTION_MODE_colors[bmi->as_mode][1];
+                                V = B_PREDICTION_MODE_colors[bmi->as_mode][2];
 
                                 POSTPROC_INVOKE(RTCD_VTABLE(oci), blend_b)
                                     (yl+bx, ul+(bx>>1), vl+(bx>>1), Y, U, V, 0xc000, y_stride);

vp8/common/ppc/loopfilter_altivec.c

@@ -53,9 +53,8 @@ loop_filter_function_s_ppc loop_filter_simple_vertical_edge_ppc;
 
 // Horizontal MB filtering
 void loop_filter_mbh_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                         int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                         int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     mbloop_filter_horizontal_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr);
 
     if (u_ptr)
@@ -63,9 +62,8 @@ void loop_filter_mbh_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned ch
 }
 
 void loop_filter_mbhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                          int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     (void)u_ptr;
     (void)v_ptr;
     (void)uv_stride;
@@ -74,9 +72,8 @@ void loop_filter_mbhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned c
 
 // Vertical MB Filtering
 void loop_filter_mbv_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                         int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                         int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     mbloop_filter_vertical_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr);
 
     if (u_ptr)
@@ -84,9 +81,8 @@ void loop_filter_mbv_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned ch
 }
 
 void loop_filter_mbvs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                          int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                          int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     (void)u_ptr;
     (void)v_ptr;
     (void)uv_stride;
@@ -95,9 +91,8 @@ void loop_filter_mbvs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned c
 
 // Horizontal B Filtering
 void loop_filter_bh_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                        int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                        int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     // These should all be done at once with one call, instead of 3
     loop_filter_horizontal_edge_y_ppc(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr);
     loop_filter_horizontal_edge_y_ppc(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr);
@@ -108,9 +103,8 @@ void loop_filter_bh_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned cha
 }
 
 void loop_filter_bhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                         int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                         int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     (void)u_ptr;
     (void)v_ptr;
     (void)uv_stride;
@@ -121,9 +115,8 @@ void loop_filter_bhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned ch
 
 // Vertical B Filtering
 void loop_filter_bv_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                        int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                        int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     loop_filter_vertical_edge_y_ppc(y_ptr, y_stride, lfi->flim, lfi->lim, lfi->thr);
 
     if (u_ptr)
@@ -131,9 +124,8 @@ void loop_filter_bv_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned cha
 }
 
 void loop_filter_bvs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                         int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                         int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void)simpler_lpf;
     (void)u_ptr;
     (void)v_ptr;
     (void)uv_stride;

vp8/common/recon.h

@@ -23,6 +23,12 @@
 #define prototype_recon_macroblock(sym) \
     void sym(const struct vp8_recon_rtcd_vtable *rtcd, MACROBLOCKD *x)
 
+#define prototype_build_intra_predictors(sym) \
+    void sym(MACROBLOCKD *x)
+
+#define prototype_intra4x4_predict(sym) \
+    void sym(BLOCKD *x, int b_mode, unsigned char *predictor)
+
 struct vp8_recon_rtcd_vtable;
 
 #if ARCH_X86 || ARCH_X86_64
@@ -73,9 +79,42 @@ extern prototype_recon_macroblock(vp8_recon_recon_mb);
 #endif
 extern prototype_recon_macroblock(vp8_recon_recon_mby);
 
+#ifndef vp8_recon_build_intra_predictors_mby
+#define vp8_recon_build_intra_predictors_mby vp8_build_intra_predictors_mby
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mby);
+
+#ifndef vp8_recon_build_intra_predictors_mby_s
+#define vp8_recon_build_intra_predictors_mby_s vp8_build_intra_predictors_mby_s
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mby_s);
+
+#ifndef vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mbuv);
+
+#ifndef vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s
+#endif
+extern prototype_build_intra_predictors\
+    (vp8_recon_build_intra_predictors_mbuv_s);
+
+#ifndef vp8_recon_intra4x4_predict
+#define vp8_recon_intra4x4_predict vp8_intra4x4_predict
+#endif
+extern prototype_intra4x4_predict\
+    (vp8_recon_intra4x4_predict);
+
+
 typedef prototype_copy_block((*vp8_copy_block_fn_t));
 typedef prototype_recon_block((*vp8_recon_fn_t));
 typedef prototype_recon_macroblock((*vp8_recon_mb_fn_t));
+typedef prototype_build_intra_predictors((*vp8_build_intra_pred_fn_t));
+typedef prototype_intra4x4_predict((*vp8_intra4x4_pred_fn_t));
 typedef struct vp8_recon_rtcd_vtable
 {
     vp8_copy_block_fn_t  copy16x16;
@@ -86,6 +125,11 @@ typedef struct vp8_recon_rtcd_vtable
     vp8_recon_fn_t       recon4;
     vp8_recon_mb_fn_t    recon_mb;
     vp8_recon_mb_fn_t    recon_mby;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mby_s;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mby;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mbuv_s;
+    vp8_build_intra_pred_fn_t  build_intra_predictors_mbuv;
+    vp8_intra4x4_pred_fn_t intra4x4_predict;
 } vp8_recon_rtcd_vtable_t;
 
 #if CONFIG_RUNTIME_CPU_DETECT

vp8/common/reconinter.c

@@ -10,6 +10,7 @@
 
 
 #include "vpx_ports/config.h"
+#include "vpx/vpx_integer.h"
 #include "recon.h"
 #include "subpixel.h"
 #include "blockd.h"
@@ -18,12 +19,6 @@
 #include "onyxc_int.h"
 #endif
 
-/* use this define on systems where unaligned int reads and writes are
- * not allowed, i.e. ARM architectures
- */
-/*#define MUST_BE_ALIGNED*/
-
-
 static const int bbb[4] = {0, 2, 8, 10};
 
 
@@ -39,7 +34,7 @@ void vp8_copy_mem16x16_c(
 
     for (r = 0; r < 16; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -58,10 +53,10 @@ void vp8_copy_mem16x16_c(
         dst[15] = src[15];
 
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
-        ((int *)dst)[2] = ((int *)src)[2] ;
-        ((int *)dst)[3] = ((int *)src)[3] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
+        ((uint32_t *)dst)[2] = ((uint32_t *)src)[2] ;
+        ((uint32_t *)dst)[3] = ((uint32_t *)src)[3] ;
 
 #endif
         src += src_stride;
@@ -81,7 +76,7 @@ void vp8_copy_mem8x8_c(
 
     for (r = 0; r < 8; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -91,8 +86,8 @@ void vp8_copy_mem8x8_c(
         dst[6] = src[6];
         dst[7] = src[7];
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
 #endif
         src += src_stride;
         dst += dst_stride;
@@ -111,7 +106,7 @@ void vp8_copy_mem8x4_c(
 
     for (r = 0; r < 4; r++)
     {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
         dst[0] = src[0];
         dst[1] = src[1];
         dst[2] = src[2];
@@ -121,8 +116,8 @@ void vp8_copy_mem8x4_c(
         dst[6] = src[6];
         dst[7] = src[7];
 #else
-        ((int *)dst)[0] = ((int *)src)[0] ;
-        ((int *)dst)[1] = ((int *)src)[1] ;
+        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
+        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
 #endif
         src += src_stride;
         dst += dst_stride;
@@ -154,13 +149,13 @@ void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
 
         for (r = 0; r < 4; r++)
         {
-#ifdef MUST_BE_ALIGNED
+#if !(CONFIG_FAST_UNALIGNED)
             pred_ptr[0]  = ptr[0];
             pred_ptr[1]  = ptr[1];
             pred_ptr[2]  = ptr[2];
             pred_ptr[3]  = ptr[3];
 #else
-            *(int *)pred_ptr = *(int *)ptr ;
+            *(uint32_t *)pred_ptr = *(uint32_t *)ptr ;
 #endif
             pred_ptr     += pitch;
             ptr         += d->pre_stride;
@@ -168,7 +163,7 @@ void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
     }
 }
 
-void vp8_build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
+static void build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 {
     unsigned char *ptr_base;
     unsigned char *ptr;
@@ -187,7 +182,7 @@ void vp8_build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
     }
 }
 
-void vp8_build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
+static void build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 {
     unsigned char *ptr_base;
     unsigned char *ptr;
@@ -207,12 +202,12 @@ void vp8_build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
 }
 
 
+/*encoder only*/
 void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 {
     int i;
 
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
         unsigned char *uptr, *vptr;
         unsigned char *upred_ptr = &x->predictor[256];
@@ -246,7 +241,7 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
             BLOCKD *d1 = &x->block[i+1];
 
             if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                vp8_build_inter_predictors2b(x, d0, 8);
+                build_inter_predictors2b(x, d0, 8);
             else
             {
                 vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
@@ -257,158 +252,132 @@ void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 }
 
 /*encoder only*/
-void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
+void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x)
 {
+    unsigned char *ptr_base;
+    unsigned char *ptr;
+    unsigned char *pred_ptr = x->predictor;
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+    int pre_stride = x->block[0].pre_stride;
 
-  if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-      x->mode_info_context->mbmi.mode != SPLITMV)
+    ptr_base = x->pre.y_buffer;
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
     {
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *pred_ptr = x->predictor;
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->block[0].pre_stride;
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
+    }
+}
 
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+void vp8_build_inter16x16_predictors_mb(MACROBLOCKD *x,
+                                        unsigned char *dst_y,
+                                        unsigned char *dst_u,
+                                        unsigned char *dst_v,
+                                        int dst_ystride,
+                                        int dst_uvstride)
+{
+    int offset;
+    unsigned char *ptr;
+    unsigned char *uptr, *vptr;
 
-        if ((mv_row | mv_col) & 7)
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+
+    unsigned char *ptr_base = x->pre.y_buffer;
+    int pre_stride = x->block[0].pre_stride;
+
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_y, dst_ystride);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_y, dst_ystride);
+    }
+
+    mv_row = x->block[16].bmi.mv.as_mv.row;
+    mv_col = x->block[16].bmi.mv.as_mv.col;
+    pre_stride >>= 1;
+    offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
+    uptr = x->pre.u_buffer + offset;
+    vptr = x->pre.v_buffer + offset;
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, dst_u, dst_uvstride);
+        x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, dst_v, dst_uvstride);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, dst_u, dst_uvstride);
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, dst_v, dst_uvstride);
+    }
+
+}
+
+void vp8_build_inter4x4_predictors_mb(MACROBLOCKD *x)
+{
+    int i;
+
+    if (x->mode_info_context->mbmi.partitioning < 3)
+    {
+        for (i = 0; i < 4; i++)
         {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
+            BLOCKD *d = &x->block[bbb[i]];
+            build_inter_predictors4b(x, d, 16);
         }
     }
     else
     {
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
+        for (i = 0; i < 16; i += 2)
         {
-            for (i = 0; i < 4; i++)
+            BLOCKD *d0 = &x->block[i];
+            BLOCKD *d1 = &x->block[i+1];
+
+            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
+                build_inter_predictors2b(x, d0, 16);
+            else
             {
-                BLOCKD *d = &x->block[bbb[i]];
-                vp8_build_inter_predictors4b(x, d, 16);
+                vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
+                vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
             }
 
         }
+
+    }
+
+    for (i = 16; i < 24; i += 2)
+    {
+        BLOCKD *d0 = &x->block[i];
+        BLOCKD *d1 = &x->block[i+1];
+
+        if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
+            build_inter_predictors2b(x, d0, 8);
         else
         {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                    vp8_build_inter_predictors2b(x, d0, 16);
-                else
-                {
-                    vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
-                    vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
-                }
-
-            }
+            vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
+            vp8_build_inter_predictors_b(d1, 8, x->subpixel_predict);
         }
     }
 }
 
 void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
 {
-
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
-        int offset;
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *uptr, *vptr;
-        unsigned char *pred_ptr = x->predictor;
-        unsigned char *upred_ptr = &x->predictor[256];
-        unsigned char *vpred_ptr = &x->predictor[320];
-
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        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;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
-        }
-
-        mv_row = x->block[16].bmi.mv.as_mv.row;
-        mv_col = x->block[16].bmi.mv.as_mv.col;
-        pre_stride >>= 1;
-        offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
-        uptr = x->pre.u_buffer + offset;
-        vptr = x->pre.v_buffer + offset;
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, upred_ptr, 8);
-            x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vpred_ptr, 8);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, upred_ptr, 8);
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vpred_ptr, 8);
-        }
+        vp8_build_inter16x16_predictors_mb(x, x->predictor, &x->predictor[256],
+                                           &x->predictor[320], 16, 8);
     }
     else
     {
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
-        {
-            for (i = 0; i < 4; i++)
-            {
-                BLOCKD *d = &x->block[bbb[i]];
-                vp8_build_inter_predictors4b(x, d, 16);
-            }
-        }
-        else
-        {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                    vp8_build_inter_predictors2b(x, d0, 16);
-                else
-                {
-                    vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
-                    vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
-                }
-
-            }
-
-        }
-
-        for (i = 16; i < 24; i += 2)
-        {
-            BLOCKD *d0 = &x->block[i];
-            BLOCKD *d1 = &x->block[i+1];
-
-            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                vp8_build_inter_predictors2b(x, d0, 8);
-            else
-            {
-                vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
-                vp8_build_inter_predictors_b(d1, 8, x->subpixel_predict);
-            }
-
-        }
-
+        vp8_build_inter4x4_predictors_mb(x);
     }
 }
 
@@ -492,202 +461,5 @@ void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel)
 }
 
 
-/* The following functions are wriiten for skip_recon_mb() to call. Since there is no recon in this
- * situation, we can write the result directly to dst buffer instead of writing it to predictor
- * buffer and then copying it to dst buffer.
- */
-static void vp8_build_inter_predictors_b_s(BLOCKD *d, unsigned char *dst_ptr, vp8_subpix_fn_t sppf)
-{
-    int r;
-    unsigned char *ptr_base;
-    unsigned char *ptr;
-    /*unsigned char *pred_ptr = d->predictor;*/
-    int dst_stride = d->dst_stride;
-    int pre_stride = d->pre_stride;
-
-    ptr_base = *(d->base_pre);
-
-    if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
-    {
-        ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-        sppf(ptr, pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, dst_stride);
-    }
-    else
-    {
-        ptr_base += d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-        ptr = ptr_base;
-
-        for (r = 0; r < 4; r++)
-        {
-#ifdef MUST_BE_ALIGNED
-            dst_ptr[0]   = ptr[0];
-            dst_ptr[1]   = ptr[1];
-            dst_ptr[2]   = ptr[2];
-            dst_ptr[3]   = ptr[3];
-#else
-            *(int *)dst_ptr = *(int *)ptr ;
-#endif
-            dst_ptr      += dst_stride;
-            ptr         += pre_stride;
-        }
-    }
-}
 
 
-
-void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
-{
-    /*unsigned char *pred_ptr = x->block[0].predictor;
-    unsigned char *dst_ptr = *(x->block[0].base_dst) + x->block[0].dst;*/
-    unsigned char *pred_ptr = x->predictor;
-    unsigned char *dst_ptr = x->dst.y_buffer;
-
-    if (x->mode_info_context->mbmi.mode != SPLITMV)
-    {
-        int offset;
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *uptr, *vptr;
-        /*unsigned char *pred_ptr = x->predictor;
-        unsigned char *upred_ptr = &x->predictor[256];
-        unsigned char *vpred_ptr = &x->predictor[320];*/
-        unsigned char *udst_ptr = x->dst.u_buffer;
-        unsigned char *vdst_ptr = x->dst.v_buffer;
-
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        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;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-
-        mv_row = x->block[16].bmi.mv.as_mv.row;
-        mv_col = x->block[16].bmi.mv.as_mv.col;
-        pre_stride >>= 1;
-        offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
-        uptr = x->pre.u_buffer + offset;
-        vptr = x->pre.v_buffer + offset;
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, udst_ptr, x->dst.uv_stride);
-            x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vdst_ptr, x->dst.uv_stride);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, udst_ptr, x->dst.uv_stride);
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vdst_ptr, x->dst.uv_stride);
-        }
-    }
-    else
-    {
-        /* note: this whole ELSE part is not executed at all. So, no way to test the correctness of my modification. Later,
-         * if sth is wrong, go back to what it is in build_inter_predictors_mb.
-         */
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
-        {
-            for (i = 0; i < 4; i++)
-            {
-                BLOCKD *d = &x->block[bbb[i]];
-                /*vp8_build_inter_predictors4b(x, d, 16);*/
-
-                {
-                    unsigned char *ptr_base;
-                    unsigned char *ptr;
-                    unsigned char *pred_ptr = d->predictor;
-
-                    ptr_base = *(d->base_pre);
-                    ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-
-                    if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
-                    {
-                        x->subpixel_predict8x8(ptr, d->pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
-                    else
-                    {
-                        RECON_INVOKE(&x->rtcd->recon, copy8x8)(ptr, d->pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
-                }
-            }
-        }
-        else
-        {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                {
-                    /*vp8_build_inter_predictors2b(x, d0, 16);*/
-                    unsigned char *ptr_base;
-                    unsigned char *ptr;
-                    unsigned char *pred_ptr = d0->predictor;
-
-                    ptr_base = *(d0->base_pre);
-                    ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
-
-                    if (d0->bmi.mv.as_mv.row & 7 || d0->bmi.mv.as_mv.col & 7)
-                    {
-                        x->subpixel_predict8x4(ptr, d0->pre_stride, d0->bmi.mv.as_mv.col & 7, d0->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride);
-                    }
-                    else
-                    {
-                        RECON_INVOKE(&x->rtcd->recon, copy8x4)(ptr, d0->pre_stride, dst_ptr, x->dst.y_stride);
-                    }
-                }
-                else
-                {
-                    vp8_build_inter_predictors_b_s(d0, dst_ptr, x->subpixel_predict);
-                    vp8_build_inter_predictors_b_s(d1, dst_ptr, x->subpixel_predict);
-                }
-            }
-        }
-
-        for (i = 16; i < 24; i += 2)
-        {
-            BLOCKD *d0 = &x->block[i];
-            BLOCKD *d1 = &x->block[i+1];
-
-            if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-            {
-                /*vp8_build_inter_predictors2b(x, d0, 8);*/
-                unsigned char *ptr_base;
-                unsigned char *ptr;
-                unsigned char *pred_ptr = d0->predictor;
-
-                ptr_base = *(d0->base_pre);
-                ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
-
-                if (d0->bmi.mv.as_mv.row & 7 || d0->bmi.mv.as_mv.col & 7)
-                {
-                    x->subpixel_predict8x4(ptr, d0->pre_stride,
-                        d0->bmi.mv.as_mv.col & 7,
-                        d0->bmi.mv.as_mv.row & 7,
-                        dst_ptr, x->dst.uv_stride);
-                }
-                else
-                {
-                    RECON_INVOKE(&x->rtcd->recon, copy8x4)(ptr,
-                        d0->pre_stride, dst_ptr, x->dst.uv_stride);
-                }
-            }
-            else
-            {
-                vp8_build_inter_predictors_b_s(d0, dst_ptr, x->subpixel_predict);
-                vp8_build_inter_predictors_b_s(d1, dst_ptr, x->subpixel_predict);
-            }
-        }
-    }
-}

vp8/common/reconinter.h

@@ -13,9 +13,15 @@
 #define __INC_RECONINTER_H
 
 extern void vp8_build_inter_predictors_mb(MACROBLOCKD *x);
-extern void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x);
+extern void vp8_build_inter16x16_predictors_mb(MACROBLOCKD *x,
+                                               unsigned char *dst_y,
+                                               unsigned char *dst_u,
+                                               unsigned char *dst_v,
+                                               int dst_ystride,
+                                               int dst_uvstride);
 
-extern void vp8_build_inter_predictors_mby(MACROBLOCKD *x);
+
+extern void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x);
 extern void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel);
 extern void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf);
 extern void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x);

vp8/common/reconintra.h

@@ -14,16 +14,4 @@
 
 extern void init_intra_left_above_pixels(MACROBLOCKD *x);
 
-extern void (*vp8_build_intra_predictors_mby_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_neon(MACROBLOCKD *x);
-extern void (*vp8_build_intra_predictors_mby_s_ptr)(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x);
-
-extern void vp8_build_intra_predictors_mbuv(MACROBLOCKD *x);
-extern void vp8_build_intra_predictors_mbuv_s(MACROBLOCKD *x);
-
-extern void vp8_predict_intra4x4(BLOCKD *x, int b_mode, unsigned char *Predictor);
-
 #endif

vp8/common/reconintra4x4.c

@@ -14,7 +14,7 @@
 #include "vpx_mem/vpx_mem.h"
 #include "reconintra.h"
 
-void vp8_predict_intra4x4(BLOCKD *x,
+void vp8_intra4x4_predict(BLOCKD *x,
                           int b_mode,
                           unsigned char *predictor)
 {

vp8/common/threading.h

@@ -12,8 +12,6 @@
 #ifndef _PTHREAD_EMULATION
 #define _PTHREAD_EMULATION
 
-#define VPXINFINITE 10000       /* 10second. */
-
 #if CONFIG_OS_SUPPORT && CONFIG_MULTITHREAD
 
 /* Thread management macros */
@@ -28,7 +26,7 @@
 #define pthread_t HANDLE
 #define pthread_attr_t DWORD
 #define pthread_create(thhandle,attr,thfunc,tharg) (int)((*thhandle=(HANDLE)_beginthreadex(NULL,0,(unsigned int (__stdcall *)(void *))thfunc,tharg,0,NULL))==NULL)
-#define pthread_join(thread, result) ((WaitForSingleObject((thread),VPXINFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
+#define pthread_join(thread, result) ((WaitForSingleObject((thread),INFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
 #define pthread_detach(thread) if(thread!=NULL)CloseHandle(thread)
 #define thread_sleep(nms) Sleep(nms)
 #define pthread_cancel(thread) terminate_thread(thread,0)
@@ -38,6 +36,7 @@
 #define pthread_self() GetCurrentThreadId()
 #else
 #ifdef __APPLE__
+#include <mach/mach_init.h>
 #include <mach/semaphore.h>
 #include <mach/task.h>
 #include <time.h>
@@ -60,9 +59,9 @@
 #ifdef _WIN32
 #define sem_t HANDLE
 #define pause(voidpara) __asm PAUSE
-#define sem_init(sem, sem_attr1, sem_init_value) (int)((*sem = CreateEvent(NULL,FALSE,FALSE,NULL))==NULL)
-#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,VPXINFINITE))
-#define sem_post(sem) SetEvent(*sem)
+#define sem_init(sem, sem_attr1, sem_init_value) (int)((*sem = CreateSemaphore(NULL,0,32768,NULL))==NULL)
+#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,INFINITE))
+#define sem_post(sem) ReleaseSemaphore(*sem,1,NULL)
 #define sem_destroy(sem) if(*sem)((int)(CloseHandle(*sem))==TRUE)
 #define thread_sleep(nms) Sleep(nms)
 

vp8/common/x86/boolcoder.cxx

@@ -1,494 +0,0 @@
-/*
- *  Copyright (c) 2010 The WebM 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.
- */
-
-
-
-/* Arithmetic bool coder with largish probability range.
-   Timothy S Murphy  6 August 2004 */
-
-#include <assert.h>
-#include <math.h>
-
-#include "bool_coder.h"
-
-#if tim_vp8
-    extern "C" {
-#       include "VP8cx/treewriter.h"
-    }
-#endif
-
-int_types::~int_types() {}
-
-void bool_coder_spec::check_prec() const {
-    assert( w  &&  (r==Up || w > 1)  &&  w < 24  &&  (ebias || w < 17));
-}
-
-bool bool_coder_spec::float_init( uint Ebits, uint Mbits) {
-    uint b = (ebits = Ebits) + (mbits = Mbits);
-    if( b) {
-        assert( ebits < 6  &&  w + mbits < 31);
-        assert( ebits + mbits  <  sizeof(Index) * 8);
-        ebias = (1 << ebits) + 1 + mbits;
-        mmask = (1 << mbits) - 1;
-        max_index = ( ( half_index = 1 << b ) << 1) - 1;
-    } else {
-        ebias = 0;
-        max_index = 255;
-        half_index = 128;
-    }
-    check_prec();
-    return b? 1:0;
-}
-
-void bool_coder_spec::cost_init()
-{
-    static cdouble c = -(1 << 20)/log( 2.);
-
-    FILE *f = fopen( "costs.txt", "w");
-    assert( f);
-
-    assert( sizeof(int) >= 4);  /* for C interface */
-    assert( max_index <= 255);   /* size of Ctbl */
-    uint i = 0;  do {
-        cdouble p = ( *this)( (Index) i);
-        Ctbl[i] = (uint32) ( log( p) * c);
-        fprintf(
-            f, "cost( %d -> %10.7f) = %10d = %12.5f bits\n",
-            i, p, Ctbl[i], (double) Ctbl[i] / (1<<20)
-        );
-    } while( ++i <= max_index);
-    fclose( f);
-}
-
-bool_coder_spec_explicit_table::bool_coder_spec_explicit_table(
-    cuint16 tbl[256], Rounding rr, uint prec
-)
-  : bool_coder_spec( prec, rr)
-{
-    check_prec();
-    uint i = 0;
-    if( tbl)
-        do { Ptbl[i] = tbl[i];}  while( ++i < 256);
-    else
-        do { Ptbl[i] = i << 8;}  while( ++i < 256);
-    cost_init();
-}
-
-
-bool_coder_spec_exponential_table::bool_coder_spec_exponential_table(
-    uint x, Rounding rr, uint prec
-)
-  : bool_coder_spec( prec, rr)
-{
-    assert( x > 1  &&  x <= 16);
-    check_prec();
-    Ptbl[128] = 32768u;
-    Ptbl[0] = (uint16) pow( 2., 16. - x);
-    --x;
-    int i=1;  do {
-        cdouble d = pow( .5, 1. + (1. - i/128.)*x) * 65536.;
-        uint16 v = (uint16) d;
-        if( v < i)
-            v = i;
-        Ptbl[256-i] = (uint16) ( 65536U - (Ptbl[i] = v));
-    } while( ++i < 128);
-    cost_init();
-}
-
-bool_coder_spec::bool_coder_spec( FILE *fp) {
-    fscanf( fp, "%d", &w);
-    int v;
-    fscanf( fp, "%d", &v);
-    assert( 0 <= v  &&  v <= 2);
-    r = (Rounding) v;
-    fscanf( fp, "%d", &ebits);
-    fscanf( fp, "%d", &mbits);
-    if( float_init( ebits, mbits))
-        return;
-    int i=0;  do {
-        uint v;
-        fscanf( fp, "%d", &v);
-        assert( 0 <=v  &&  v <= 65535U);
-        Ptbl[i] = v;
-    } while( ++i < 256);
-    cost_init();
-}
-
-void bool_coder_spec::dump( FILE *fp) const {
-    fprintf( fp, "%d %d %d %d\n", w, (int) r, ebits, mbits);
-    if( ebits  ||  mbits)
-        return;
-    int i=0;  do { fprintf( fp, "%d\n", Ptbl[i]);}  while( ++i < 256);
-}
-
-vp8bc_index_t bool_coder_spec::operator()( double p) const
-{
-    if( p <= 0.)
-        return 0;
-    if( p >= 1.)
-        return max_index;
-    if( ebias) {
-        if( p > .5)
-            return max_index - ( *this)( 1. - p);
-        int e;
-        uint m = (uint) ldexp( frexp( p, &e), mbits + 2);
-        uint x = 1 << (mbits + 1);
-        assert( x <= m  &&  m < x<<1);
-        if( (m = (m >> 1) + (m & 1)) >= x) {
-            m = x >> 1;
-            ++e;
-        }
-        int y = 1 << ebits;
-        if( (e += y) >= y)
-            return half_index - 1;
-        if( e < 0)
-            return 0;
-        return (Index) ( (e << mbits) + (m & mmask));
-    }
-
-    cuint16 v = (uint16) (p * 65536.);
-    int i = 128;
-    int j = 128;
-    uint16 w;
-    while( w = Ptbl[i], j >>= 1) {
-        if( w < v)
-            i += j;
-        else if( w == v)
-            return (uchar) i;
-        else
-            i -= j;
-    }
-    if( w > v) {
-        cuint16 x = Ptbl[i-1];
-        if( v <= x  ||  w - v > v - x)
-            --i;
-    } else if( w < v  &&  i < 255) {
-        cuint16 x = Ptbl[i+1];
-        if( x <= v  ||  x - v < v - w)
-            ++i;
-    }
-    return (Index) i;
-}
-
-double bool_coder_spec::operator()( Index i) const {
-    if( !ebias)
-        return Ptbl[i]/65536.;
-    if( i >= half_index)
-        return 1. - ( *this)( (Index) (max_index - i));
-    return ldexp( (double)mantissa( i), - (int) exponent( i));
-}
-
-
-
-void bool_writer::carry() {
-    uchar *p = B;
-    assert( p > Bstart);
-    while( *--p == 255) { assert( p > Bstart);  *p = 0;}
-    ++*p;
-}
-
-
-bool_writer::bool_writer( c_spec& s, uchar *Dest, size_t Len)
-  : bool_coder( s),
-    Bstart( Dest),
-    Bend( Len? Dest+Len : 0),
-    B( Dest)
-{
-    assert( Dest);
-    reset();
-}
-
-bool_writer::~bool_writer() { flush();}
-
-#if 1
-    extern "C" { int bc_v = 0;}
-#else
-#   define bc_v 0
-#endif
-
-
-void bool_writer::raw( bool value, uint32 s) {
-    uint32 L = Low;
-
-    assert( Range >= min_range  &&  Range <= spec.max_range());
-    assert( !is_toast  &&  s  &&  s < Range);
-
-    if( bc_v) printf(
-        "Writing a %d, B %x  Low %x  Range %x  s %x   blag %d ...\n",
-        value? 1:0, B-Bstart, Low, Range, s, bit_lag
-    );
-    if( value) {
-        L += s;
-        s = Range - s;
-    } else
-        s -= rinc;
-    if( s < min_range) {
-        int ct = bit_lag;  do {
-            if( !--ct) {
-                ct = 8;
-                if( L & (1 << 31))
-                    carry();
-                assert( !Bend  ||  B < Bend);
-                *B++ = (uchar) (L >> 23);
-                L &= (1<<23) - 1;
-            }
-        } while( L += L, (s += s + rinc) < min_range);
-        bit_lag = ct;
-    }
-    Low = L;
-    Range = s;
-    if( bc_v)
-        printf(
-            "...done, B %x  Low %x  Range %x  blag %d \n",
-                B-Bstart, Low, Range, bit_lag
-        );
-}
-
-bool_writer& bool_writer::flush() {
-    if( is_toast)
-        return *this;
-    int b = bit_lag;
-    uint32 L = Low;
-    assert( b);
-    if( L & (1 << (32 - b)))
-        carry();
-    L <<= b & 7;
-    b >>= 3;
-    while( --b >= 0)
-        L <<= 8;
-    b = 4;
-    assert( !Bend  ||  B + 4 <= Bend);
-    do {
-        *B++ = (uchar) (L >> 24);
-        L <<= 8;
-    } while( --b);
-    is_toast = 1;
-    return *this;
-}
-
-
-bool_reader::bool_reader( c_spec& s, cuchar *src, size_t Len)
-  : bool_coder( s),
-    Bstart( src),
-    B( src),
-    Bend( Len? src+Len : 0),
-    shf( 32 - s.w),
-    bct( 8)
-{
-    int i = 4;  do { Low <<= 8;  Low |= *B++;}  while( --i);
-}
-
-
-bool bool_reader::raw( uint32 s) {
-
-    bool val = 0;
-    uint32 L = Low;
-    cuint32 S = s << shf;
-
-    assert( Range >= min_range  &&  Range <= spec.max_range());
-    assert( s  &&  s < Range  &&  (L >> shf) < Range);
-
-    if( bc_v)
-        printf(
-            "Reading, B %x  Low %x  Range %x  s %x  bct %d ...\n",
-            B-Bstart, Low, Range, s, bct
-        );
-
-    if( L >= S) {
-        L -= S;
-        s = Range - s;
-        assert( L < (s << shf));
-        val = 1;
-    } else
-        s -= rinc;
-    if( s < min_range) {
-        int ct = bct;
-        do {
-            assert( ~L & (1 << 31));
-            L += L;
-            if( !--ct) {
-                ct = 8;
-                if( !Bend  ||  B < Bend)
-                    L |= *B++;
-            }
-        } while( (s += s + rinc) < min_range);
-        bct = ct;
-    }
-    Low = L;
-    Range = s;
-    if( bc_v)
-        printf(
-            "...done, val %d  B %x  Low %x  Range %x  bct %d\n",
-            val? 1:0, B-Bstart, Low, Range, bct
-        );
-    return val;
-}
-
-
-/* C interfaces */
-
-// spec interface
-
-struct NS : bool_coder_namespace {
-    static Rounding r( vp8bc_c_prec *p, Rounding rr =down_full) {
-        return p? (Rounding) p->r : rr;
-    }
-};
-
-bool_coder_spec *vp8bc_vp6spec() {
-    return new bool_coder_spec_explicit_table( 0, bool_coder_namespace::Down, 8);
-}
-bool_coder_spec *vp8bc_float_spec(
-    unsigned int Ebits, unsigned int Mbits, vp8bc_c_prec *p
-) {
-    return new bool_coder_spec_float( Ebits, Mbits, NS::r( p), p? p->prec : 12);
-}
-bool_coder_spec *vp8bc_literal_spec(
-    const unsigned short m[256], vp8bc_c_prec *p
-) {
-    return new bool_coder_spec_explicit_table( m, NS::r( p), p? p->prec : 16);
-}
-bool_coder_spec *vp8bc_exponential_spec( unsigned int x, vp8bc_c_prec *p)
-{
-    return new bool_coder_spec_exponential_table( x, NS::r( p), p? p->prec : 16);
-}
-bool_coder_spec *vp8bc_spec_from_file( FILE *fp) {
-    return new bool_coder_spec( fp);
-}
-void vp8bc_destroy_spec( c_bool_coder_spec *p) { delete p;}
-
-void vp8bc_spec_to_file( c_bool_coder_spec *p, FILE *fp) { p->dump( fp);}
-
-vp8bc_index_t vp8bc_index( c_bool_coder_spec *p, double x) {
-    return ( *p)( x);
-}
-
-vp8bc_index_t vp8bc_index_from_counts(
-    c_bool_coder_spec *p, unsigned int L, unsigned int R
-) {
-    return ( *p)( (R += L)? (double) L/R : .5);
-}
-
-double vp8bc_probability( c_bool_coder_spec *p, vp8bc_index_t i) {
-    return ( *p)( i);
-}
-
-vp8bc_index_t vp8bc_complement( c_bool_coder_spec *p, vp8bc_index_t i) {
-    return p->complement( i);
-}
-unsigned int vp8bc_cost_zero( c_bool_coder_spec *p, vp8bc_index_t i) {
-    return p->cost_zero( i);
-}
-unsigned int vp8bc_cost_one( c_bool_coder_spec *p, vp8bc_index_t i) {
-    return p->cost_one( i);
-}
-unsigned int vp8bc_cost_bit( c_bool_coder_spec *p, vp8bc_index_t i, int v) {
-    return p->cost_bit( i, v);
-}
-
-#if tim_vp8
-    extern "C" int tok_verbose;
-
-#   define dbg_l 1000000
-
-    static vp8bc_index_t dbg_i [dbg_l];
-    static char dbg_v [dbg_l];
-    static size_t dbg_w = 0, dbg_r = 0;
-#endif
-
-// writer interface
-
-bool_writer *vp8bc_create_writer(
-    c_bool_coder_spec *p, unsigned char *D, size_t L
-) {
-    return new bool_writer( *p, D, L);
-}
-
-size_t vp8bc_destroy_writer( bool_writer *p) {
-    const size_t s = p->flush().bytes_written();
-    delete p;
-    return s;
-}
-
-void vp8bc_write_bool( bool_writer *p, int v, vp8bc_index_t i)
-{
-#   if tim_vp8
-        // bc_v = dbg_w < 10;
-        if( bc_v = tok_verbose)
-            printf( " writing %d at prob %d\n", v? 1:0, i);
-        accum_entropy_bc( &p->Spec(), i, v);
-
-        ( *p)( i, (bool) v);
-
-        if( dbg_w < dbg_l) {
-            dbg_i [dbg_w] = i;
-            dbg_v [dbg_w++] = v? 1:0;
-        }
-#   else
-        ( *p)( i, (bool) v);
-#   endif
-}
-
-void vp8bc_write_bits( bool_writer *p, unsigned int v, int n)
-{
-#   if tim_vp8
-        {
-            c_bool_coder_spec * const s = & p->Spec();
-            const vp8bc_index_t i = s->half_index();
-            int m = n;
-            while( --m >= 0)
-                accum_entropy_bc( s, i, (v>>m) & 1);
-        }
-#   endif
-
-    p->write_bits( n, v);
-}
-
-c_bool_coder_spec *vp8bc_writer_spec( c_bool_writer *w) { return & w->Spec();}
-
-// reader interface
-
-bool_reader *vp8bc_create_reader(
-    c_bool_coder_spec *p, const unsigned char *S, size_t L
-) {
-    return new bool_reader( *p, S, L);
-}
-
-void vp8bc_destroy_reader( bool_reader * p) { delete p;}
-
-int vp8bc_read_bool( bool_reader *p, vp8bc_index_t i)
-{
-#   if tim_vp8
-        // bc_v = dbg_r < 10;
-        bc_v = tok_verbose;
-        const int v = ( *p)( i)? 1:0;
-        if( tok_verbose)
-            printf( " reading %d at prob %d\n", v, i);
-        if( dbg_r < dbg_l) {
-            assert( dbg_r <= dbg_w);
-            if( i != dbg_i[dbg_r]  ||  v != dbg_v[dbg_r]) {
-                printf(
-        "Position %d: INCORRECTLY READING %d  prob %d, wrote %d  prob %d\n",
-                    dbg_r, v, i, dbg_v[dbg_r], dbg_i[dbg_r]
-                );
-            }
-            ++dbg_r;
-        }
-        return v;
-#   else
-        return ( *p)( i)? 1:0;
-#   endif
-}
-
-unsigned int vp8bc_read_bits( bool_reader *p, int n) { return p->read_bits( n);}
-
-c_bool_coder_spec *vp8bc_reader_spec( c_bool_reader *r) { return & r->Spec();}
-
-#undef bc_v

vp8/common/x86/idctllm_mmx.asm

@@ -14,18 +14,18 @@
 ; /****************************************************************************
 ; * Notes:
 ; *
-; * This implementation makes use of 16 bit fixed point verio of two multiply
+; * This implementation makes use of 16 bit fixed point version of two multiply
 ; * constants:
 ; *        1.   sqrt(2) * cos (pi/8)
-; *         2.   sqrt(2) * sin (pi/8)
-; * Becuase the first constant is bigger than 1, to maintain the same 16 bit
-; * fixed point prrcision as the second one, we use a trick of
+; *        2.   sqrt(2) * sin (pi/8)
+; * Because the first constant is bigger than 1, to maintain the same 16 bit
+; * fixed point precision as the second one, we use a trick of
 ; *        x * a = x + x*(a-1)
 ; * so
 ; *        x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1).
 ; *
-; * For     the second constant, becuase of the 16bit version is 35468, which
-; * is bigger than 32768, in signed 16 bit multiply, it become a negative
+; * For the second constant, because of the 16bit version is 35468, which
+; * is bigger than 32768, in signed 16 bit multiply, it becomes a negative
 ; * number.
 ; *        (x * (unsigned)35468 >> 16) = x * (signed)35468 >> 16 + x
 ; *

vp8/common/x86/idctllm_sse2.asm

@@ -32,9 +32,6 @@ sym(idct_dequant_0_2x_sse2):
         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]
 
@@ -43,9 +40,12 @@ sym(idct_dequant_0_2x_sse2):
 
         pmullw      xmm4,           xmm5
 
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
+
     ; clear coeffs
-        movd        [rax],          xmm7
-        movd        [rax+32],       xmm7
+        movd        [rax],          xmm5
+        movd        [rax+32],       xmm5
 ;pshufb
         pshuflw     xmm4,           xmm4,       00000000b
         pshufhw     xmm4,           xmm4,       00000000b
@@ -62,10 +62,10 @@ sym(idct_dequant_0_2x_sse2):
         lea         rcx,            [3*rcx]
         movq        xmm3,           [rax+rcx]
 
-        punpcklbw   xmm0,           xmm7
-        punpcklbw   xmm1,           xmm7
-        punpcklbw   xmm2,           xmm7
-        punpcklbw   xmm3,           xmm7
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
         mov         rax,            arg(3) ; dst
         movsxd      rdx,            dword ptr arg(4) ; dst_stride
@@ -77,10 +77,10 @@ sym(idct_dequant_0_2x_sse2):
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; store blocks back out
         movq        [rax],          xmm0
@@ -102,6 +102,7 @@ sym(idct_dequant_full_2x_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -347,6 +348,7 @@ sym(idct_dequant_full_2x_sse2):
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -377,8 +379,8 @@ sym(idct_dequant_dc_0_2x_sse2):
         mov         rdi,            arg(3) ; dst
         mov         rdx,            arg(5) ; dc
 
-    ; Zero out xmm7, for use unpacking
-        pxor        xmm7,           xmm7
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
 
     ; load up 2 dc words here == 2*16 = doubleword
         movd        xmm4,           [rdx]
@@ -398,10 +400,10 @@ sym(idct_dequant_dc_0_2x_sse2):
         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
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
     ; Add to predict buffer
         paddw       xmm0,           xmm4
@@ -410,10 +412,10 @@ sym(idct_dequant_dc_0_2x_sse2):
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; Load destination stride before writing out,
     ;   doesn't need to persist
@@ -441,6 +443,7 @@ sym(idct_dequant_dc_full_2x_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -692,6 +695,7 @@ sym(idct_dequant_dc_full_2x_sse2):
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret

vp8/common/x86/iwalsh_sse2.asm

@@ -17,7 +17,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 2
-    SAVE_XMM
+    SAVE_XMM 6
     push        rsi
     push        rdi
     ; end prolog
@@ -41,7 +41,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     movdqa    xmm4, xmm0
     punpcklqdq  xmm0, xmm3          ;d1 a1
     punpckhqdq  xmm4, xmm3          ;c1 b1
-    movd    xmm7, eax
+    movd    xmm6, eax
 
     movdqa    xmm1, xmm4          ;c1 b1
     paddw   xmm4, xmm0          ;dl+cl a1+b1 aka op[4] op[0]
@@ -66,7 +66,7 @@ sym(vp8_short_inv_walsh4x4_sse2):
     pshufd    xmm2, xmm1, 4eh       ;ip[8] ip[12]
     movdqa    xmm3, xmm4          ;ip[4] ip[0]
 
-    pshufd    xmm7, xmm7, 0       ;03 03 03 03 03 03 03 03
+    pshufd    xmm6, xmm6, 0       ;03 03 03 03 03 03 03 03
 
     paddw   xmm4, xmm2          ;ip[4]+ip[8] ip[0]+ip[12] aka b1 a1
     psubw   xmm3, xmm2          ;ip[4]-ip[8] ip[0]-ip[12] aka c1 d1
@@ -90,8 +90,8 @@ sym(vp8_short_inv_walsh4x4_sse2):
     punpcklwd xmm5, xmm0          ; 31 21 11 01 30 20 10 00
     punpckhwd xmm1, xmm0          ; 33 23 13 03 32 22 12 02
 ;~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-    paddw   xmm5, xmm7
-    paddw   xmm1, xmm7
+    paddw   xmm5, xmm6
+    paddw   xmm1, xmm6
 
     psraw   xmm5, 3
     psraw   xmm1, 3

vp8/common/x86/loopfilter_mmx.asm

@@ -16,7 +16,7 @@
 ;(
 ;    unsigned char *src_ptr,
 ;    int src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int  count
@@ -122,12 +122,10 @@ next8_h:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm2              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx, arg(2) ;flimit           ; get flimit
-        movq        mm2, [rdx]            ; flimit mm2
-        paddb       mm2, mm2              ; flimit*2 (less than 255)
-        paddb       mm7, mm2              ; flimit * 2 + limit (less than 255)
+        mov         rdx, arg(2) ;blimit           ; get blimit
+        movq        mm7, [rdx]            ; blimit
 
-        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,    mm5
         pxor        mm5,    mm5
         pcmpeqb     mm1,    mm5           ; mask mm1
@@ -230,7 +228,7 @@ next8_h:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -406,9 +404,9 @@ next8_v:
         pand        mm5,        [GLOBAL(tfe)]               ; set lsb of each byte to zero
         psrlw       mm5,        1                           ; abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                      ;
+        mov         rdx,        arg(2) ;blimit                      ;
 
-        movq        mm2,        [rdx]                       ;flimit  mm2
+        movq        mm4,        [rdx]                       ;blimit
         movq        mm1,        mm3                         ; mm1=mm3=p0
 
         movq        mm7,        mm6                         ; mm7=mm6=q0
@@ -419,10 +417,7 @@ next8_v:
         paddusb     mm1,        mm1                         ; abs(q0-p0)*2
         paddusb     mm1,        mm5                         ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        paddb       mm2,        mm2                         ; flimit*2 (less than 255)
-        paddb       mm4,        mm2                         ; flimit * 2 + limit (less than 255)
-
-        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,        mm0;                        ; mask
 
         pxor        mm0,        mm0
@@ -603,7 +598,7 @@ next8_v:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -719,17 +714,15 @@ next8_mbh:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm2              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx, arg(2) ;flimit           ; get flimit
-        movq        mm2, [rdx]            ; flimit mm2
-        paddb       mm2, mm2              ; flimit*2 (less than 255)
-        paddb       mm7, mm2              ; flimit * 2 + limit (less than 255)
+        mov         rdx, arg(2) ;blimit           ; get blimit
+        movq        mm7, [rdx]            ; blimit
 
-        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,    mm7           ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,    mm5
         pxor        mm5,    mm5
         pcmpeqb     mm1,    mm5           ; mask mm1
 
-        ; mm1 = mask, mm0=q0,  mm7 = flimit, t0 = abs(q0-q1) t1 = abs(p1-p0)
+        ; mm1 = mask, mm0=q0,  mm7 = blimit, t0 = abs(q0-q1) t1 = abs(p1-p0)
         ; mm6 = p0,
 
         ; calculate high edge variance
@@ -922,7 +915,7 @@ next8_mbh:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
+;    const char *blimit,
 ;    const char *limit,
 ;    const char *thresh,
 ;    int count
@@ -1108,9 +1101,9 @@ next8_mbv:
         pand        mm5,        [GLOBAL(tfe)]               ; set lsb of each byte to zero
         psrlw       mm5,        1                           ; abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                      ;
+        mov         rdx,        arg(2) ;blimit                      ;
 
-        movq        mm2,        [rdx]                       ;flimit  mm2
+        movq        mm4,        [rdx]                       ;blimit
         movq        mm1,        mm3                         ; mm1=mm3=p0
 
         movq        mm7,        mm6                         ; mm7=mm6=q0
@@ -1121,10 +1114,7 @@ next8_mbv:
         paddusb     mm1,        mm1                         ; abs(q0-p0)*2
         paddusb     mm1,        mm5                         ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        paddb       mm2,        mm2                         ; flimit*2 (less than 255)
-        paddb       mm4,        mm2                         ; flimit * 2 + limit (less than 255)
-
-        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm1,        mm4                         ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         mm1,        mm0;                        ; mask
 
         pxor        mm0,        mm0
@@ -1392,16 +1382,13 @@ next8_mbv:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit
 ;)
 global sym(vp8_loop_filter_simple_horizontal_edge_mmx)
 sym(vp8_loop_filter_simple_horizontal_edge_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
+    SHADOW_ARGS_TO_STACK 3
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1410,14 +1397,10 @@ sym(vp8_loop_filter_simple_horizontal_edge_mmx):
         mov         rsi, arg(0) ;src_ptr
         movsxd      rax, dword ptr arg(1) ;src_pixel_step     ; destination pitch?
 
-        movsxd      rcx, dword ptr arg(5) ;count
+        mov         rcx, 2                ; count
 nexts8_h:
-        mov         rdx, arg(3) ;limit
-        movq        mm7, [rdx]
-        mov         rdx, arg(2) ;flimit           ; get flimit
+        mov         rdx, arg(2) ;blimit           ; get blimit
         movq        mm3, [rdx]            ;
-        paddb       mm3, mm3              ; flimit*2 (less than 255)
-        paddb       mm3, mm7              ; flimit * 2 + limit (less than 255)
 
         mov         rdi, rsi              ; rdi points to row +1 for indirect addressing
         add         rdi, rax
@@ -1445,7 +1428,7 @@ nexts8_h:
         paddusb     mm5, mm5              ; abs(p0-q0)*2
         paddusb     mm5, mm1              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        psubusb     mm5, mm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5, mm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        mm3, mm3
         pcmpeqb     mm5, mm3
 
@@ -1515,16 +1498,13 @@ nexts8_h:
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit
 ;)
 global sym(vp8_loop_filter_simple_vertical_edge_mmx)
 sym(vp8_loop_filter_simple_vertical_edge_mmx):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
+    SHADOW_ARGS_TO_STACK 3
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1539,7 +1519,7 @@ sym(vp8_loop_filter_simple_vertical_edge_mmx):
         movsxd      rax, dword ptr arg(1) ;src_pixel_step     ; destination pitch?
 
         lea         rsi, [rsi + rax*4- 2];  ;
-        movsxd      rcx, dword ptr arg(5) ;count
+        mov         rcx, 2                                      ; count
 nexts8_v:
 
         lea         rdi,        [rsi + rax];
@@ -1602,14 +1582,10 @@ nexts8_v:
         paddusb     mm5,        mm5                             ; abs(p0-q0)*2
         paddusb     mm5,        mm6                             ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx,        arg(2) ;flimit                          ; get flimit
+        mov         rdx,        arg(2) ;blimit                          ; get blimit
         movq        mm7,        [rdx]
-        mov         rdx,        arg(3)                          ; get limit
-        movq        mm6,        [rdx]
-        paddb       mm7,        mm7                             ; flimit*2 (less than 255)
-        paddb       mm7,        mm6                             ; flimit * 2 + limit (less than 255)
 
-        psubusb     mm5,        mm7                             ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     mm5,        mm7                             ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        mm7,        mm7
         pcmpeqb     mm5,        mm7                             ; mm5 = mask
 

vp8/common/x86/loopfilter_sse2.asm

@@ -110,7 +110,7 @@
         psubusb     xmm6,                   xmm5              ; p1-=p0
 
         por         xmm6,                   xmm4              ; abs(p1 - p0)
-        mov         rdx,                    arg(2)            ; get flimit
+        mov         rdx,                    arg(2)            ; get blimit
 
         movdqa        t1,                   xmm6              ; save to t1
 
@@ -123,7 +123,7 @@
         psubusb     xmm1,                   xmm7
         por         xmm2,                   xmm3              ; abs(p1-q1)
 
-        movdqa      xmm4,                   XMMWORD PTR [rdx] ; flimit
+        movdqa      xmm7,                   XMMWORD PTR [rdx] ; blimit
 
         movdqa      xmm3,                   xmm0              ; q0
         pand        xmm2,                   [GLOBAL(tfe)]     ; set lsb of each byte to zero
@@ -134,13 +134,11 @@
         psrlw       xmm2,                   1                 ; abs(p1-q1)/2
 
         psubusb     xmm5,                   xmm3              ; p0-=q0
-        paddb       xmm4,                   xmm4              ; flimit*2 (less than 255)
 
         psubusb     xmm3,                   xmm6              ; q0-=p0
         por         xmm5,                   xmm3              ; abs(p0 - q0)
 
         paddusb     xmm5,                   xmm5              ; abs(p0-q0)*2
-        paddb       xmm7,                   xmm4              ; flimit * 2 + limit (less than 255)
 
         movdqa      xmm4,                   t0                ; hev get abs (q1 - q0)
 
@@ -150,7 +148,7 @@
 
         movdqa      xmm2,                   XMMWORD PTR [rdx] ; hev
 
-        psubusb     xmm5,                   xmm7              ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5,                   xmm7              ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         psubusb     xmm4,                   xmm2              ; hev
 
         psubusb     xmm3,                   xmm2              ; hev
@@ -278,7 +276,7 @@
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -288,7 +286,7 @@ sym(vp8_loop_filter_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -328,7 +326,7 @@ sym(vp8_loop_filter_horizontal_edge_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -338,7 +336,7 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -574,7 +572,7 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -584,7 +582,7 @@ sym(vp8_mbloop_filter_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -624,7 +622,7 @@ sym(vp8_mbloop_filter_horizontal_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -634,7 +632,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -904,7 +902,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         movdqa      xmm4,               XMMWORD PTR [rdx]; limit
 
         pmaxub      xmm0,               xmm7
-        mov         rdx,                arg(2)          ; flimit
+        mov         rdx,                arg(2)          ; blimit
 
         psubusb     xmm0,               xmm4
         movdqa      xmm5,               xmm2            ; q1
@@ -921,12 +919,11 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         psrlw       xmm5,               1               ; abs(p1-q1)/2
         psubusb     xmm6,               xmm3            ; q0-p0
 
-        movdqa      xmm2,               XMMWORD PTR [rdx]; flimit
+        movdqa      xmm4,               XMMWORD PTR [rdx]; blimit
 
         mov         rdx,                arg(4)          ; get thresh
 
         por         xmm1,               xmm6            ; abs(q0-p0)
-        paddb       xmm2,               xmm2            ; flimit*2 (less than 255)
 
         movdqa      xmm6,               t0              ; get abs (q1 - q0)
 
@@ -939,10 +936,9 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
         paddusb     xmm1,               xmm5            ; abs (p0 - q0) *2 + abs(p1-q1)/2
         psubusb     xmm6,               xmm7            ; abs(q1 - q0) > thresh
 
-        paddb       xmm4,               xmm2            ; flimit * 2 + limit (less than 255)
         psubusb     xmm3,               xmm7            ; abs(p1 - p0)> thresh
 
-        psubusb     xmm1,               xmm4            ; abs (p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm1,               xmm4            ; abs (p0 - q0) *2 + abs(p1-q1)/2  > blimit
         por         xmm6,               xmm3            ; abs(q1 - q0) > thresh || abs(p1 - p0) > thresh
 
         por         xmm1,               xmm0            ; mask
@@ -1014,7 +1010,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -1024,7 +1020,7 @@ sym(vp8_loop_filter_vertical_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1081,7 +1077,7 @@ sym(vp8_loop_filter_vertical_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -1091,7 +1087,7 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1239,7 +1235,7 @@ sym(vp8_loop_filter_vertical_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    int            count
@@ -1249,7 +1245,7 @@ sym(vp8_mbloop_filter_vertical_edge_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1308,7 +1304,7 @@ sym(vp8_mbloop_filter_vertical_edge_sse2):
 ;(
 ;    unsigned char *u,
 ;    int            src_pixel_step,
-;    const char    *flimit,
+;    const char    *blimit,
 ;    const char    *limit,
 ;    const char    *thresh,
 ;    unsigned char *v
@@ -1318,7 +1314,7 @@ sym(vp8_mbloop_filter_vertical_edge_uv_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1376,17 +1372,14 @@ sym(vp8_mbloop_filter_vertical_edge_uv_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit,
 ;)
 global sym(vp8_loop_filter_simple_horizontal_edge_sse2)
 sym(vp8_loop_filter_simple_horizontal_edge_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1394,13 +1387,8 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
 
         mov         rsi, arg(0)             ;src_ptr
         movsxd      rax, dword ptr arg(1)   ;src_pixel_step     ; destination pitch?
-        mov         rdx, arg(2) ;flimit     ; get flimit
+        mov         rdx, arg(2)             ;blimit
         movdqa      xmm3, XMMWORD PTR [rdx]
-        mov         rdx, arg(3) ;limit
-        movdqa      xmm7, XMMWORD PTR [rdx]
-
-        paddb       xmm3, xmm3              ; flimit*2 (less than 255)
-        paddb       xmm3, xmm7              ; flimit * 2 + limit (less than 255)
 
         mov         rdi, rsi                ; rdi points to row +1 for indirect addressing
         add         rdi, rax
@@ -1428,7 +1416,7 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
         paddusb     xmm5, xmm5              ; abs(p0-q0)*2
         paddusb     xmm5, xmm1              ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        psubusb     xmm5, xmm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5, xmm3              ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        xmm3, xmm3
         pcmpeqb     xmm5, xmm3
 
@@ -1493,17 +1481,14 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
 ;(
 ;    unsigned char *src_ptr,
 ;    int  src_pixel_step,
-;    const char *flimit,
-;    const char *limit,
-;    const char *thresh,
-;    int count
+;    const char *blimit,
 ;)
 global sym(vp8_loop_filter_simple_vertical_edge_sse2)
 sym(vp8_loop_filter_simple_vertical_edge_sse2):
     push        rbp         ; save old base pointer value.
     mov         rbp, rsp    ; set new base pointer value.
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM 7
     GET_GOT     rbx         ; save callee-saved reg
     push        rsi
     push        rdi
@@ -1607,14 +1592,10 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
         paddusb     xmm5,       xmm5                            ; abs(p0-q0)*2
         paddusb     xmm5,       xmm6                            ; abs (p0 - q0) *2 + abs(p1-q1)/2
 
-        mov         rdx,        arg(2)                          ;flimit
+        mov         rdx,        arg(2)                          ;blimit
         movdqa      xmm7, XMMWORD PTR [rdx]
-        mov         rdx,        arg(3)                          ; get limit
-        movdqa      xmm6, XMMWORD PTR [rdx]
-        paddb       xmm7,        xmm7                           ; flimit*2 (less than 255)
-        paddb       xmm7,        xmm6                           ; flimit * 2 + limit (less than 255)
 
-        psubusb     xmm5,        xmm7                           ; abs(p0 - q0) *2 + abs(p1-q1)/2  > flimit * 2 + limit
+        psubusb     xmm5,        xmm7                           ; abs(p0 - q0) *2 + abs(p1-q1)/2  > blimit
         pxor        xmm7,        xmm7
         pcmpeqb     xmm5,        xmm7                           ; mm5 = mask
 

vp8/common/x86/loopfilter_x86.c

@@ -9,30 +9,18 @@
  */
 
 
-#include "vpx_ports/config.h"
+#include "vpx_config.h"
 #include "vp8/common/loopfilter.h"
 
-prototype_loopfilter(vp8_loop_filter_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_c);
-prototype_loopfilter(vp8_mbloop_filter_vertical_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_c);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_c);
-
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_mmx);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_mmx);
 prototype_loopfilter(vp8_loop_filter_vertical_edge_mmx);
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_mmx);
-prototype_loopfilter(vp8_loop_filter_simple_vertical_edge_mmx);
-prototype_loopfilter(vp8_loop_filter_simple_horizontal_edge_mmx);
 
 prototype_loopfilter(vp8_loop_filter_vertical_edge_sse2);
 prototype_loopfilter(vp8_loop_filter_horizontal_edge_sse2);
 prototype_loopfilter(vp8_mbloop_filter_vertical_edge_sse2);
 prototype_loopfilter(vp8_mbloop_filter_horizontal_edge_sse2);
-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;
@@ -42,113 +30,77 @@ 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,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-
-void vp8_loop_filter_mbhs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
+        vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
-}
-
-
-void vp8_loop_filter_mbvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                            int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
-void vp8_loop_filter_bhs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_mmx(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                            int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                            int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 
     if (v_ptr)
-        vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
+        vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, 1);
 }
 
 
-void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_mmx(y_ptr + 12, y_stride, blimit);
 }
 #endif
 
@@ -156,113 +108,65 @@ void vp8_loop_filter_bvs_mmx(unsigned char *y_ptr, unsigned char *u_ptr, unsigne
 /* Horizontal MB filtering */
 #if HAVE_SSE2
 void vp8_loop_filter_mbh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
-}
-
-
-void vp8_loop_filter_mbhs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, v_ptr);
 }
 
 
 /* Vertical MB Filtering */
 void vp8_loop_filter_mbv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                              int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+    vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mblim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
-}
-
-
-void vp8_loop_filter_mbvs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                               int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
-{
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
+        vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->mblim, lfi->lim, lfi->hev_thr, v_ptr);
 }
 
 
 /* Horizontal B Filtering */
 void vp8_loop_filter_bh_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4 * uv_stride);
+        vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, v_ptr + 4 * uv_stride);
 }
 
 
-void vp8_loop_filter_bhs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bhs_sse2(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 4 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 8 * y_stride, y_stride, blimit);
+    vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, blimit);
 }
 
 
 /* Vertical B Filtering */
 void vp8_loop_filter_bv_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                             int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+                             int y_stride, int uv_stride, loop_filter_info *lfi)
 {
-    (void) simpler_lpf;
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
+    vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->blim, lfi->lim, lfi->hev_thr, 2);
 
     if (u_ptr)
-        vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4);
+        vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->blim, lfi->lim, lfi->hev_thr, v_ptr + 4);
 }
 
 
-void vp8_loop_filter_bvs_sse2(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
-                              int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
+void vp8_loop_filter_bvs_sse2(unsigned char *y_ptr, int y_stride, const unsigned char *blimit)
 {
-    (void) u_ptr;
-    (void) v_ptr;
-    (void) uv_stride;
-    (void) simpler_lpf;
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, blimit);
+    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, blimit);
 }
 
 #endif
-
-#if 0
-void vp8_fast_loop_filter_vertical_edges_sse(unsigned char *y_ptr,
-        int y_stride,
-        loop_filter_info *lfi)
-{
-
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 4, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 8, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-    vp8_loop_filter_simple_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
-}
-#endif

vp8/common/x86/loopfilter_x86.h

@@ -24,10 +24,10 @@ extern prototype_loopfilter_block(vp8_loop_filter_mbv_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_mmx);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_mmx);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_mmx);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_mmx);
 
 
 #if !CONFIG_RUNTIME_CPU_DETECT
@@ -44,13 +44,13 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_mmx);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_mmx
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_mmx
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_mmx
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_mmx
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_mmx
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_mmx
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_mmx
@@ -63,10 +63,10 @@ extern prototype_loopfilter_block(vp8_loop_filter_mbv_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_bv_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_mbh_sse2);
 extern prototype_loopfilter_block(vp8_loop_filter_bh_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_mbvs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_bvs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_mbhs_sse2);
-extern prototype_loopfilter_block(vp8_loop_filter_bhs_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_vertical_edge_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_bvs_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_simple_horizontal_edge_sse2);
+extern prototype_simple_loopfilter(vp8_loop_filter_bhs_sse2);
 
 
 #if !CONFIG_RUNTIME_CPU_DETECT
@@ -83,13 +83,13 @@ extern prototype_loopfilter_block(vp8_loop_filter_bhs_sse2);
 #define vp8_lf_normal_b_h vp8_loop_filter_bh_sse2
 
 #undef  vp8_lf_simple_mb_v
-#define vp8_lf_simple_mb_v vp8_loop_filter_mbvs_sse2
+#define vp8_lf_simple_mb_v vp8_loop_filter_simple_vertical_edge_sse2
 
 #undef  vp8_lf_simple_b_v
 #define vp8_lf_simple_b_v vp8_loop_filter_bvs_sse2
 
 #undef  vp8_lf_simple_mb_h
-#define vp8_lf_simple_mb_h vp8_loop_filter_mbhs_sse2
+#define vp8_lf_simple_mb_h vp8_loop_filter_simple_horizontal_edge_sse2
 
 #undef  vp8_lf_simple_b_h
 #define vp8_lf_simple_b_h vp8_loop_filter_bhs_sse2

vp8/common/x86/postproc_sse2.asm

@@ -26,7 +26,7 @@ sym(vp8_post_proc_down_and_across_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -256,7 +256,7 @@ sym(vp8_mbpost_proc_down_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -456,7 +456,7 @@ sym(vp8_mbpost_proc_across_ip_xmm):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

vp8/common/x86/recon_sse2.asm

@@ -67,7 +67,7 @@ sym(vp8_recon4b_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
-    SAVE_XMM
+    SAVE_XMM 7
     push        rsi
     push        rdi
     ; end prolog
@@ -229,3 +229,460 @@ sym(vp8_copy_mem16x16_sse2):
     UNSHADOW_ARGS
     pop         rbp
     ret
+
+
+;void vp8_intra_pred_uv_dc_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dc_mmx2)
+sym(vp8_intra_pred_uv_dc_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from top
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        mm0,        mm0
+    movq        mm1,        [rsi]
+    psadbw      mm1,        mm0
+
+    ; from left
+    dec         rsi
+    lea         rdi,        [rax*3]
+    movzx       ecx,        byte [rsi+rax]
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    lea         rsi,        [rsi+rax*4]
+    movzx       edx,        byte [rsi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*4]
+    add         ecx,        edx
+
+    ; add up
+    pextrw      edx,        mm1, 0x0
+    lea         edx,        [edx+ecx+8]
+    sar         edx,        4
+    movd        mm1,        edx
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dctop_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dctop_mmx2)
+sym(vp8_intra_pred_uv_dctop_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from top
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        mm0,        mm0
+    movq        mm1,        [rsi]
+    psadbw      mm1,        mm0
+
+    ; add up
+    paddw       mm1,        [GLOBAL(dc_4)]
+    psraw       mm1,        3
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dcleft_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dcleft_mmx2)
+sym(vp8_intra_pred_uv_dcleft_mmx2):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; from left
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    dec         rsi
+    lea         rdi,        [rax*3]
+    movzx       ecx,        byte [rsi]
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    add         ecx,        edx
+    lea         rsi,        [rsi+rax*4]
+    movzx       edx,        byte [rsi]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rax*2]
+    add         ecx,        edx
+    movzx       edx,        byte [rsi+rdi]
+    lea         edx,        [ecx+edx+4]
+
+    ; add up
+    shr         edx,        3
+    movd        mm1,        edx
+    pshufw      mm1,        mm1, 0x0
+    packuswb    mm1,        mm1
+
+    ; write out
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+    lea         rax,        [rcx*3]
+
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+    lea         rdi,        [rdi+rcx*4]
+    movq [rdi      ],       mm1
+    movq [rdi+rcx  ],       mm1
+    movq [rdi+rcx*2],       mm1
+    movq [rdi+rax  ],       mm1
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_dc128_mmx(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_dc128_mmx)
+sym(vp8_intra_pred_uv_dc128_mmx):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    ; end prolog
+
+    ; write out
+    movq        mm1,        [GLOBAL(dc_128)]
+    mov         rax,        arg(0) ;dst;
+    movsxd      rdx,        dword ptr arg(1) ;dst_stride
+    lea         rcx,        [rdx*3]
+
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+    lea         rax,        [rax+rdx*4]
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+
+    ; begin epilog
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_tm_sse2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+%macro vp8_intra_pred_uv_tm 1
+global sym(vp8_intra_pred_uv_tm_%1)
+sym(vp8_intra_pred_uv_tm_%1):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    GET_GOT     rbx
+    push        rsi
+    push        rdi
+    ; end prolog
+
+    ; read top row
+    mov         edx,        4
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    sub         rsi,        rax
+    pxor        xmm0,       xmm0
+%ifidn %1, ssse3
+    movdqa      xmm2,       [GLOBAL(dc_1024)]
+%endif
+    movq        xmm1,       [rsi]
+    punpcklbw   xmm1,       xmm0
+
+    ; set up left ptrs ans subtract topleft
+    movd        xmm3,       [rsi-1]
+    lea         rsi,        [rsi+rax-1]
+%ifidn %1, sse2
+    punpcklbw   xmm3,       xmm0
+    pshuflw     xmm3,       xmm3, 0x0
+    punpcklqdq  xmm3,       xmm3
+%else
+    pshufb      xmm3,       xmm2
+%endif
+    psubw       xmm1,       xmm3
+
+    ; set up dest ptrs
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+
+vp8_intra_pred_uv_tm_%1_loop:
+    movd        xmm3,       [rsi]
+    movd        xmm5,       [rsi+rax]
+%ifidn %1, sse2
+    punpcklbw   xmm3,       xmm0
+    punpcklbw   xmm5,       xmm0
+    pshuflw     xmm3,       xmm3, 0x0
+    pshuflw     xmm5,       xmm5, 0x0
+    punpcklqdq  xmm3,       xmm3
+    punpcklqdq  xmm5,       xmm5
+%else
+    pshufb      xmm3,       xmm2
+    pshufb      xmm5,       xmm2
+%endif
+    paddw       xmm3,       xmm1
+    paddw       xmm5,       xmm1
+    packuswb    xmm3,       xmm5
+    movq  [rdi    ],        xmm3
+    movhps[rdi+rcx],        xmm3
+    lea         rsi,        [rsi+rax*2]
+    lea         rdi,        [rdi+rcx*2]
+    dec         edx
+    jnz vp8_intra_pred_uv_tm_%1_loop
+
+    ; begin epilog
+    pop         rdi
+    pop         rsi
+    RESTORE_GOT
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+%endmacro
+
+vp8_intra_pred_uv_tm sse2
+vp8_intra_pred_uv_tm ssse3
+
+;void vp8_intra_pred_uv_ve_mmx(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+global sym(vp8_intra_pred_uv_ve_mmx)
+sym(vp8_intra_pred_uv_ve_mmx):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    ; end prolog
+
+    ; read from top
+    mov         rax,        arg(2) ;src;
+    movsxd      rdx,        dword ptr arg(3) ;src_stride;
+    sub         rax,        rdx
+    movq        mm1,        [rax]
+
+    ; write out
+    mov         rax,        arg(0) ;dst;
+    movsxd      rdx,        dword ptr arg(1) ;dst_stride
+    lea         rcx,        [rdx*3]
+
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+    lea         rax,        [rax+rdx*4]
+    movq [rax      ],       mm1
+    movq [rax+rdx  ],       mm1
+    movq [rax+rdx*2],       mm1
+    movq [rax+rcx  ],       mm1
+
+    ; begin epilog
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+
+;void vp8_intra_pred_uv_ho_mmx2(
+;    unsigned char *dst,
+;    int dst_stride
+;    unsigned char *src,
+;    int src_stride,
+;    )
+%macro vp8_intra_pred_uv_ho 1
+global sym(vp8_intra_pred_uv_ho_%1)
+sym(vp8_intra_pred_uv_ho_%1):
+    push        rbp
+    mov         rbp, rsp
+    SHADOW_ARGS_TO_STACK 4
+    push        rsi
+    push        rdi
+%ifidn %1, ssse3
+%ifndef GET_GOT_SAVE_ARG
+    push        rbx
+%endif
+    GET_GOT     rbx
+%endif
+    ; end prolog
+
+    ; read from left and write out
+%ifidn %1, mmx2
+    mov         edx,        4
+%endif
+    mov         rsi,        arg(2) ;src;
+    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    mov         rdi,        arg(0) ;dst;
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
+%ifidn %1, ssse3
+    lea         rdx,        [rcx*3]
+    movdqa      xmm2,       [GLOBAL(dc_00001111)]
+    lea         rbx,        [rax*3]
+%endif
+    dec         rsi
+%ifidn %1, mmx2
+vp8_intra_pred_uv_ho_%1_loop:
+    movd        mm0,        [rsi]
+    movd        mm1,        [rsi+rax]
+    punpcklbw   mm0,        mm0
+    punpcklbw   mm1,        mm1
+    pshufw      mm0,        mm0, 0x0
+    pshufw      mm1,        mm1, 0x0
+    movq  [rdi    ],        mm0
+    movq  [rdi+rcx],        mm1
+    lea         rsi,        [rsi+rax*2]
+    lea         rdi,        [rdi+rcx*2]
+    dec         edx
+    jnz vp8_intra_pred_uv_ho_%1_loop
+%else
+    movd        xmm0,       [rsi]
+    movd        xmm3,       [rsi+rax]
+    movd        xmm1,       [rsi+rax*2]
+    movd        xmm4,       [rsi+rbx]
+    punpcklbw   xmm0,       xmm3
+    punpcklbw   xmm1,       xmm4
+    pshufb      xmm0,       xmm2
+    pshufb      xmm1,       xmm2
+    movq   [rdi    ],       xmm0
+    movhps [rdi+rcx],       xmm0
+    movq [rdi+rcx*2],       xmm1
+    movhps [rdi+rdx],       xmm1
+    lea         rsi,        [rsi+rax*4]
+    lea         rdi,        [rdi+rcx*4]
+    movd        xmm0,       [rsi]
+    movd        xmm3,       [rsi+rax]
+    movd        xmm1,       [rsi+rax*2]
+    movd        xmm4,       [rsi+rbx]
+    punpcklbw   xmm0,       xmm3
+    punpcklbw   xmm1,       xmm4
+    pshufb      xmm0,       xmm2
+    pshufb      xmm1,       xmm2
+    movq   [rdi    ],       xmm0
+    movhps [rdi+rcx],       xmm0
+    movq [rdi+rcx*2],       xmm1
+    movhps [rdi+rdx],       xmm1
+%endif
+
+    ; begin epilog
+%ifidn %1, ssse3
+    RESTORE_GOT
+%ifndef GET_GOT_SAVE_ARG
+    pop         rbx
+%endif
+%endif
+    pop         rdi
+    pop         rsi
+    UNSHADOW_ARGS
+    pop         rbp
+    ret
+%endmacro
+
+vp8_intra_pred_uv_ho mmx2
+vp8_intra_pred_uv_ho ssse3
+
+SECTION_RODATA
+dc_128:
+    times 8 db 128
+dc_4:
+    times 4 dw 4
+align 16
+dc_1024:
+    times 8 dw 0x400
+align 16
+dc_00001111:
+    times 8 db 0
+    times 8 db 1

vp8/common/x86/recon_wrapper_sse2.c

@@ -0,0 +1,96 @@
+/*
+ *  Copyright (c) 2010 The WebM 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 "vp8/common/recon.h"
+#include "recon_x86.h"
+#include "vpx_mem/vpx_mem.h"
+
+#define build_intra_predictors_mbuv_prototype(sym) \
+    void sym(unsigned char *dst, int dst_stride, \
+             const unsigned char *src, int src_stride)
+typedef build_intra_predictors_mbuv_prototype((*build_intra_predictors_mbuv_fn_t));
+
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dc_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dctop_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dcleft_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_dc128_mmx);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ho_mmx2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ho_ssse3);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_ve_mmx);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_tm_sse2);
+extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_uv_tm_ssse3);
+
+static void vp8_build_intra_predictors_mbuv_x86(MACROBLOCKD *x,
+                                                unsigned char *dst_u,
+                                                unsigned char *dst_v,
+                                                int dst_stride,
+                                                build_intra_predictors_mbuv_fn_t tm_func,
+                                                build_intra_predictors_mbuv_fn_t ho_func)
+{
+    int mode = x->mode_info_context->mbmi.uv_mode;
+    build_intra_predictors_mbuv_fn_t fn;
+    int src_stride = x->dst.uv_stride;
+
+    switch (mode) {
+        case  V_PRED: fn = vp8_intra_pred_uv_ve_mmx; break;
+        case  H_PRED: fn = ho_func; break;
+        case TM_PRED: fn = tm_func; break;
+        case DC_PRED:
+            if (x->up_available) {
+                if (x->left_available) {
+                    fn = vp8_intra_pred_uv_dc_mmx2; break;
+                } else {
+                    fn = vp8_intra_pred_uv_dctop_mmx2; break;
+                }
+            } else if (x->left_available) {
+                fn = vp8_intra_pred_uv_dcleft_mmx2; break;
+            } else {
+                fn = vp8_intra_pred_uv_dc128_mmx; break;
+            }
+            break;
+        default: return;
+    }
+
+    fn(dst_u, dst_stride, x->dst.u_buffer, src_stride);
+    fn(dst_v, dst_stride, x->dst.v_buffer, src_stride);
+}
+
+void vp8_build_intra_predictors_mbuv_sse2(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, &x->predictor[256],
+                                        &x->predictor[320], 8,
+                                        vp8_intra_pred_uv_tm_sse2,
+                                        vp8_intra_pred_uv_ho_mmx2);
+}
+
+void vp8_build_intra_predictors_mbuv_ssse3(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, &x->predictor[256],
+                                        &x->predictor[320], 8,
+                                        vp8_intra_pred_uv_tm_ssse3,
+                                        vp8_intra_pred_uv_ho_ssse3);
+}
+
+void vp8_build_intra_predictors_mbuv_s_sse2(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, x->dst.u_buffer,
+                                        x->dst.v_buffer, x->dst.uv_stride,
+                                        vp8_intra_pred_uv_tm_sse2,
+                                        vp8_intra_pred_uv_ho_mmx2);
+}
+
+void vp8_build_intra_predictors_mbuv_s_ssse3(MACROBLOCKD *x)
+{
+    vp8_build_intra_predictors_mbuv_x86(x, x->dst.u_buffer,
+                                        x->dst.v_buffer, x->dst.uv_stride,
+                                        vp8_intra_pred_uv_tm_ssse3,
+                                        vp8_intra_pred_uv_ho_ssse3);
+}

vp8/common/x86/recon_x86.h

@@ -46,6 +46,8 @@ extern prototype_copy_block(vp8_copy_mem16x16_mmx);
 extern prototype_recon_block(vp8_recon2b_sse2);
 extern prototype_recon_block(vp8_recon4b_sse2);
 extern prototype_copy_block(vp8_copy_mem16x16_sse2);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_sse2);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_s_sse2);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 #undef  vp8_recon_recon2
@@ -57,6 +59,26 @@ extern prototype_copy_block(vp8_copy_mem16x16_sse2);
 #undef  vp8_recon_copy16x16
 #define vp8_recon_copy16x16 vp8_copy_mem16x16_sse2
 
+#undef  vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv_sse2
+
+#undef  vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s_sse2
+
+#endif
+#endif
+
+#if HAVE_SSSE3
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_ssse3);
+extern prototype_build_intra_predictors(vp8_build_intra_predictors_mbuv_s_ssse3);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+#undef  vp8_recon_build_intra_predictors_mbuv
+#define vp8_recon_build_intra_predictors_mbuv vp8_build_intra_predictors_mbuv_ssse3
+
+#undef  vp8_recon_build_intra_predictors_mbuv_s
+#define vp8_recon_build_intra_predictors_mbuv_s vp8_build_intra_predictors_mbuv_s_ssse3
+
 #endif
 #endif
 #endif

vp8/common/x86/subpixel_mmx.asm

@@ -113,97 +113,6 @@ nextrow:
     ret
 
 
-;
-; THIS FUNCTION APPEARS TO BE UNUSED
-;
-;void vp8_filter_block1d_v6_mmx
-;(
-;   short *src_ptr,
-;   unsigned char *output_ptr,
-;   unsigned int pixels_per_line,
-;   unsigned int pixel_step,
-;   unsigned int output_height,
-;   unsigned int output_width,
-;   short * vp8_filter
-;)
-global sym(vp8_filter_block1d_v6_mmx)
-sym(vp8_filter_block1d_v6_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        movq      mm5, [GLOBAL(rd)]
-        push        rbx
-        mov         rbx, arg(6) ;vp8_filter
-        movq      mm1, [rbx + 16]             ; do both the negative taps first!!!
-        movq      mm2, [rbx + 32]         ;
-        movq      mm6, [rbx + 48]        ;
-        movq      mm7, [rbx + 64]        ;
-
-        movsxd      rdx, dword ptr arg(2) ;pixels_per_line
-        mov         rdi, arg(1) ;output_ptr
-        mov         rsi, arg(0) ;src_ptr
-        sub         rsi, rdx
-        sub         rsi, rdx
-        movsxd      rcx, DWORD PTR arg(4) ;output_height
-        movsxd      rax, DWORD PTR arg(5) ;output_width      ; destination pitch?
-        pxor        mm0, mm0              ; mm0 = 00000000
-
-
-nextrow_v:
-        movq        mm3, [rsi+rdx]        ; mm3 = p0..p8  = row -1
-        pmullw      mm3, mm1              ; mm3 *= kernel 1 modifiers.
-
-
-        movq        mm4, [rsi + 4*rdx]      ; mm4 = p0..p3  = row 2
-        pmullw      mm4, mm7              ; mm4 *= kernel 4 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 2*rdx]           ; mm4 = p0..p3  = row 0
-        pmullw      mm4, mm2              ; mm4 *= kernel 2 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi]            ; mm4 = p0..p3  = row -2
-        pmullw      mm4, [rbx]            ; mm4 *= kernel 0 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        add         rsi, rdx              ; move source forward 1 line to avoid 3 * pitch
-        movq        mm4, [rsi + 2*rdx]     ; mm4 = p0..p3  = row 1
-        pmullw      mm4, mm6              ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 4*rdx]    ; mm4 = p0..p3  = row 3
-        pmullw      mm4, [rbx +80]        ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        paddsw      mm3, mm5               ; mm3 += round value
-        psraw       mm3, VP8_FILTER_SHIFT     ; mm3 /= 128
-        packuswb    mm3, mm0              ; pack and saturate
-
-        movd        [rdi],mm3             ; store the results in the destination
-
-        add         rdi,rax;
-
-        dec         rcx                   ; decrement count
-        jnz         nextrow_v             ; next row
-
-        pop         rbx
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
 ;void vp8_filter_block1dc_v6_mmx
 ;(
 ;   short *src_ptr,

vp8/common/x86/subpixel_sse2.asm

@@ -37,7 +37,7 @@ sym(vp8_filter_block1d8_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -157,7 +157,7 @@ sym(vp8_filter_block1d16_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -333,7 +333,7 @@ sym(vp8_filter_block1d8_v6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -428,7 +428,7 @@ sym(vp8_filter_block1d16_v6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -538,7 +538,7 @@ sym(vp8_filter_block1d8_h6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -651,7 +651,7 @@ sym(vp8_filter_block1d16_h6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -816,7 +816,7 @@ sym(vp8_filter_block1d8_v6_only_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -908,7 +908,6 @@ sym(vp8_unpack_block1d16_h6_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    ;SAVE_XMM                          ;xmm6, xmm7 are not used here.
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -948,7 +947,6 @@ unpack_block1d16_h6_sse2_rowloop:
     pop rdi
     pop rsi
     RESTORE_GOT
-    ;RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -969,7 +967,7 @@ sym(vp8_bilinear_predict16x16_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1238,7 +1236,7 @@ sym(vp8_bilinear_predict8x8_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

vp8/common/x86/subpixel_ssse3.asm

@@ -39,6 +39,7 @@ sym(vp8_filter_block1d8_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -107,6 +108,7 @@ filter_block1d8_h6_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -162,6 +164,7 @@ filter_block1d8_h4_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -179,7 +182,7 @@ sym(vp8_filter_block1d16_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -194,10 +197,6 @@ sym(vp8_filter_block1d16_h6_ssse3):
 
     mov         rdi, arg(2)                     ;output_ptr
 
-;;
-;;    cmp         esi, DWORD PTR [rax]
-;;    je          vp8_filter_block1d16_h4_ssse3
-
     mov         rsi, arg(0)                     ;src_ptr
 
     movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
@@ -271,61 +270,7 @@ filter_block1d16_h6_rowloop_ssse3:
     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, [GLOBAL(shuf2b)]
-    pshufb      xmm2, [GLOBAL(shuf3b)]
-    pmaddubsw   xmm1, xmm5
-
-    movdqu      xmm3,   XMMWORD PTR [rsi + 6]
-
-    pmaddubsw   xmm2, xmm6
-    movdqa      xmm0, xmm3
-    pshufb      xmm3, [GLOBAL(shuf3b)]
-    pshufb      xmm0, [GLOBAL(shuf2b)]
-
-    paddsw      xmm1, [GLOBAL(rd)]
-    paddsw      xmm1, xmm2
-
-    pmaddubsw   xmm0, xmm5
-    pmaddubsw   xmm3, xmm6
-
-    psraw       xmm1, 7
-    packuswb    xmm1, xmm1
-    lea         rsi,    [rsi + rax]
-    paddsw      xmm3, xmm0
-    paddsw      xmm3, [GLOBAL(rd)]
-    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
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -344,6 +289,7 @@ sym(vp8_filter_block1d4_h6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -451,6 +397,7 @@ filter_block1d4_h4_rowloop_ssse3:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -471,6 +418,7 @@ sym(vp8_filter_block1d16_v6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -566,6 +514,7 @@ vp8_filter_block1d16_v6_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -638,6 +587,7 @@ vp8_filter_block1d16_v4_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -656,6 +606,7 @@ sym(vp8_filter_block1d8_v6_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -728,6 +679,7 @@ vp8_filter_block1d8_v6_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -776,6 +728,7 @@ vp8_filter_block1d8_v4_ssse3_loop:
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -932,7 +885,7 @@ sym(vp8_bilinear_predict16x16_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1195,7 +1148,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi

vp8/common/x86/x86_systemdependent.c

@@ -9,7 +9,7 @@
  */
 
 
-#include "vpx_ports/config.h"
+#include "vpx_config.h"
 #include "vpx_ports/x86.h"
 #include "vp8/common/g_common.h"
 #include "vp8/common/subpixel.h"
@@ -24,10 +24,6 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 #if CONFIG_RUNTIME_CPU_DETECT
     VP8_COMMON_RTCD *rtcd = &ctx->rtcd;
     int flags = x86_simd_caps();
-    int mmx_enabled = flags & HAS_MMX;
-    int xmm_enabled = flags & HAS_SSE;
-    int wmt_enabled = flags & HAS_SSE2;
-    int SSSE3Enabled = flags & HAS_SSSE3;
 
     /* Note:
      *
@@ -39,7 +35,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
     /* Override default functions with fastest ones for this CPU. */
 #if HAVE_MMX
 
-    if (mmx_enabled)
+    if (flags & HAS_MMX)
     {
         rtcd->idct.idct1        = vp8_short_idct4x4llm_1_mmx;
         rtcd->idct.idct16       = vp8_short_idct4x4llm_mmx;
@@ -67,9 +63,9 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_mmx;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_mmx;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_mmx;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_mmx;
+        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_mmx;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_mmx;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_mmx;
+        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_mmx;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_mmx;
 
 #if CONFIG_POSTPROC
@@ -83,11 +79,15 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 #endif
 #if HAVE_SSE2
 
-    if (wmt_enabled)
+    if (flags & HAS_SSE2)
     {
         rtcd->recon.recon2      = vp8_recon2b_sse2;
         rtcd->recon.recon4      = vp8_recon4b_sse2;
         rtcd->recon.copy16x16   = vp8_copy_mem16x16_sse2;
+        rtcd->recon.build_intra_predictors_mbuv =
+            vp8_build_intra_predictors_mbuv_sse2;
+        rtcd->recon.build_intra_predictors_mbuv_s =
+            vp8_build_intra_predictors_mbuv_s_sse2;
 
         rtcd->idct.iwalsh16     = vp8_short_inv_walsh4x4_sse2;
 
@@ -101,9 +101,9 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->loopfilter.normal_b_v  = vp8_loop_filter_bv_sse2;
         rtcd->loopfilter.normal_mb_h = vp8_loop_filter_mbh_sse2;
         rtcd->loopfilter.normal_b_h  = vp8_loop_filter_bh_sse2;
-        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_mbvs_sse2;
+        rtcd->loopfilter.simple_mb_v = vp8_loop_filter_simple_vertical_edge_sse2;
         rtcd->loopfilter.simple_b_v  = vp8_loop_filter_bvs_sse2;
-        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_mbhs_sse2;
+        rtcd->loopfilter.simple_mb_h = vp8_loop_filter_simple_horizontal_edge_sse2;
         rtcd->loopfilter.simple_b_h  = vp8_loop_filter_bhs_sse2;
 
 #if CONFIG_POSTPROC
@@ -118,7 +118,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
 
 #if HAVE_SSSE3
 
-    if (SSSE3Enabled)
+    if (flags & HAS_SSSE3)
     {
         rtcd->subpix.sixtap16x16   = vp8_sixtap_predict16x16_ssse3;
         rtcd->subpix.sixtap8x8     = vp8_sixtap_predict8x8_ssse3;
@@ -126,6 +126,11 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
         rtcd->subpix.sixtap4x4     = vp8_sixtap_predict4x4_ssse3;
         rtcd->subpix.bilinear16x16 = vp8_bilinear_predict16x16_ssse3;
         rtcd->subpix.bilinear8x8   = vp8_bilinear_predict8x8_ssse3;
+
+        rtcd->recon.build_intra_predictors_mbuv =
+            vp8_build_intra_predictors_mbuv_ssse3;
+        rtcd->recon.build_intra_predictors_mbuv_s =
+            vp8_build_intra_predictors_mbuv_s_ssse3;
     }
 #endif
 

vp8/decoder/arm/arm_dsystemdependent.c

@@ -13,7 +13,6 @@
 #include "vpx_ports/arm.h"
 #include "vp8/common/blockd.h"
 #include "vp8/common/pragmas.h"
-#include "vp8/common/postproc.h"
 #include "vp8/decoder/dequantize.h"
 #include "vp8/decoder/onyxd_int.h"
 
@@ -21,12 +20,15 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
 {
 #if CONFIG_RUNTIME_CPU_DETECT
     int flags = pbi->common.rtcd.flags;
-    int has_edsp = flags & HAS_EDSP;
-    int has_media = flags & HAS_MEDIA;
-    int has_neon = flags & HAS_NEON;
+
+#if HAVE_ARMV5TE
+    if (flags & HAS_EDSP)
+    {
+    }
+#endif
 
 #if HAVE_ARMV6
-    if (has_media)
+    if (flags & HAS_MEDIA)
     {
         pbi->dequant.block               = vp8_dequantize_b_v6;
         pbi->dequant.idct_add            = vp8_dequant_idct_add_v6;
@@ -38,7 +40,7 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
 #endif
 
 #if HAVE_ARMV7
-    if (has_neon)
+    if (flags & HAS_NEON)
     {
         pbi->dequant.block               = vp8_dequantize_b_neon;
         pbi->dequant.idct_add            = vp8_dequant_idct_add_neon;

vp8/decoder/arm/armv5/dequantize_v5.asm

@@ -1,52 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM 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_dequantize_b_armv5|
-
-    AREA    |.text|, CODE, READONLY  ; name this block of code
-
-q       RN  r0
-dqc     RN  r1
-cnt     RN  r2
-
-;void dequantize_b_armv5(short *Q, short *DQC)
-|vp8_dequantize_b_armv5| PROC
-    stmdb   sp!, {r4, lr}
-    ldr     r3, [q]
-    ldr     r4, [dqc], #8
-
-    mov     cnt, #4
-dequant_loop
-    smulbb  lr, r3, r4
-    smultt  r12, r3, r4
-
-    ldr     r3, [q, #4]
-    ldr     r4, [dqc, #-4]
-
-    strh    lr, [q], #2
-    strh    r12, [q], #2
-
-    smulbb  lr, r3, r4
-    smultt  r12, r3, r4
-
-    subs    cnt, cnt, #1
-    ldrne   r3, [q, #4]
-    ldrne   r4, [dqc], #8
-
-    strh    lr, [q], #2
-    strh    r12, [q], #2
-
-    bne     dequant_loop
-
-    ldmia   sp!, {r4, pc}
-    ENDP    ;|vp8_dequantize_b_arm|
-
-    END

vp8/decoder/arm/dequantize_arm.c

@@ -26,7 +26,6 @@ extern void vp8_dequantize_b_loop_v6(short *Q, short *DQC, short *DQ);
 
 void vp8_dequantize_b_neon(BLOCKD *d)
 {
-    int i;
     short *DQ  = d->dqcoeff;
     short *Q   = d->qcoeff;
     short *DQC = d->dequant;
@@ -38,7 +37,6 @@ void vp8_dequantize_b_neon(BLOCKD *d)
 #if HAVE_ARMV6
 void vp8_dequantize_b_v6(BLOCKD *d)
 {
-    int i;
     short *DQ  = d->dqcoeff;
     short *Q   = d->qcoeff;
     short *DQC = d->dequant;