Update CHANGELOG for v1.2.0 release

Change-Id: Ie0525d300c94bfde2947b1cc2572bb7d37e17998

.gitignore

@@ -32,6 +32,8 @@
 /ivfdec.dox
 /ivfenc
 /ivfenc.dox
+/libvpx.so*
+/libvpx.ver
 /obj_int_extract
 /postproc
 /postproc.c
@@ -43,6 +45,7 @@
 /simple_encoder
 /simple_encoder.c
 /simple_encoder.dox
+/test_libvpx
 /twopass_encoder
 /twopass_encoder.c
 /twopass_encoder.dox
@@ -55,7 +58,14 @@
 /vp8cx_set_ref
 /vp8cx_set_ref.c
 /vp8cx_set_ref.dox
+/vpx.pc
 /vpx_config.c
 /vpx_config.h
+/vpx_rtcd.h
 /vpx_version.h
+/vpxdec
+/vpxenc
 TAGS
+.cproject
+.project
+.settings

CHANGELOG

@@ -1,3 +1,32 @@
+2012-12-21 v1.2.0
+  This release acts as a checkpoint for a large amount of internal refactoring
+  and testing. It also contains a number of small bugfixes, so all users are
+  encouraged to upgrade.
+
+  - Upgrading:
+    This release is ABI and API compatible with Duclair (v1.0.0). Users
+    of older releases should refer to the Upgrading notes in this
+    document for that release.
+
+  - Enhancements:
+      VP8 optimizations for MIPS dspr2
+      vpxenc: add -quiet option
+
+  - Speed:
+      Encoder and decoder speed is consistent with the Eider release.
+
+  - Quality:
+      In general, quality is consistent with the Eider release.
+
+      Minor tweaks to ARNR filtering
+      Minor improvements to real time encoding with multiple temporal layers
+
+  - Bug Fixes:
+      Fixes multithreaded encoder race condition in loopfilter
+      Fixes multi-resolution threaded encoding
+      Fix potential encoder dead-lock after picture resize
+
+
 2012-05-09 v1.1.0 "Eider"
   This introduces a number of enhancements, mostly focused on real-time
   encoding. In addition, it fixes a decoder bug (first introduced in

README

@@ -1,5 +1,5 @@
 vpx Multi-Format Codec SDK
-README - 19 May 2010
+README - 21 June 2012
 
 Welcome to the WebM VP8 Codec SDK!
 
@@ -15,11 +15,19 @@ COMPILING THE APPLICATIONS/LIBRARIES:
     * Building the documentation requires PHP[3] and Doxygen[4]. If you do not
       have these packages, you must pass --disable-install-docs to the
       configure script.
+    * Downloading the data for the unit tests requires curl[5] and sha1sum.
+      sha1sum is provided via the GNU coreutils, installed by default on
+      many *nix platforms, as well as MinGW and Cygwin. If coreutils is not
+      available, a compatible version of sha1sum can be built from
+      source[6]. These requirements are optional if not running the unit
+      tests.
 
     [1]: http://www.tortall.net/projects/yasm
     [2]: http://www.cygwin.com
     [3]: http://php.net
     [4]: http://www.doxygen.org
+    [5]: http://curl.haxx.se
+    [6]: http://www.microbrew.org/tools/md5sha1sum/
 
   2. Out-of-tree builds
   Out of tree builds are a supported method of building the application. For
@@ -94,5 +102,5 @@ COMPILING THE APPLICATIONS/LIBRARIES:
 
 SUPPORT
   This library is an open source project supported by its community. Please
-  please email webm-users@webmproject.org for help.
+  please email webm-discuss@webmproject.org for help.
 

build/make/Android.mk

@@ -27,15 +27,22 @@
 # Android.mk file in the libvpx directory:
 # LOCAL_PATH := $(call my-dir)
 # include $(CLEAR_VARS)
-# include libvpx/build/make/Android.mk
+# include jni/libvpx/build/make/Android.mk
 #
 # There are currently two TARGET_ARCH_ABI targets for ARM.
 # armeabi and armeabi-v7a.  armeabi-v7a is selected by creating an
 # Application.mk in the jni directory that contains:
 # APP_ABI := armeabi-v7a
 #
+# By default libvpx will detect at runtime the existance of NEON extension.
+# For this we import the 'cpufeatures' module from the NDK sources.
+# libvpx can also be configured without this runtime detection method.
+# Configuring with --disable-runtime-cpu-detect will assume presence of NEON.
+# Configuring with --disable-runtime-cpu-detect --disable-neon will remove any
+# NEON dependency.
+
 # To change to building armeabi, run ./libvpx/configure again, but with
-# --target=arm5te-android-gcc and and modify the Application.mk file to
+# --target=arm5te-android-gcc and modify the Application.mk file to
 # set APP_ABI := armeabi
 #
 # Running ndk-build will build libvpx and include it in your project.
@@ -166,7 +173,9 @@ LOCAL_MODULE := libvpx
 
 LOCAL_LDLIBS := -llog
 
-LOCAL_STATIC_LIBRARIES := cpufeatures
+ifeq ($(CONFIG_RUNTIME_CPU_DETECT),yes)
+  LOCAL_STATIC_LIBRARIES := cpufeatures
+endif
 
 $(foreach file, $(LOCAL_SRC_FILES), $(LOCAL_PATH)/$(file)): vpx_rtcd.h
 
@@ -196,4 +205,7 @@ ifeq ($(CONFIG_VP8_ENCODER), yes)
     $(LIBVPX_PATH)/vp8/encoder/asm_enc_offsets.c))
 endif
 
+ifeq ($(CONFIG_RUNTIME_CPU_DETECT),yes)
 $(call import-module,cpufeatures)
+endif
+

build/make/Makefile

@@ -21,6 +21,7 @@ all: .DEFAULT
 clean:: .DEFAULT
 install:: .DEFAULT
 test:: .DEFAULT
+testdata:: .DEFAULT
 
 
 # Note: md5sum is not installed on OS X, but openssl is. Openssl may not be
@@ -66,6 +67,7 @@ endif
 BUILD_ROOT?=.
 VPATH=$(SRC_PATH_BARE)
 CFLAGS+=-I$(BUILD_PFX)$(BUILD_ROOT) -I$(SRC_PATH)
+CXXFLAGS+=-I$(BUILD_PFX)$(BUILD_ROOT) -I$(SRC_PATH)
 ASFLAGS+=-I$(BUILD_PFX)$(BUILD_ROOT)/ -I$(SRC_PATH)/
 DIST_DIR?=dist
 HOSTCC?=gcc
@@ -98,6 +100,8 @@ dist:
 install::
 .PHONY: test
 test::
+.PHONY: testdata
+testdata::
 
 $(BUILD_PFX)%.c.d: %.c
 	$(if $(quiet),@echo "    [DEP] $@")
@@ -111,11 +115,11 @@ $(BUILD_PFX)%.c.o: %.c
 $(BUILD_PFX)%.cc.d: %.cc
 	$(if $(quiet),@echo "    [DEP] $@")
 	$(qexec)mkdir -p $(dir $@)
-	$(qexec)g++ $(INTERNAL_CFLAGS) $(CFLAGS) -M $< | $(fmt_deps) > $@
+	$(qexec)$(CXX) $(INTERNAL_CFLAGS) $(CXXFLAGS) -M $< | $(fmt_deps) > $@
 
 $(BUILD_PFX)%.cc.o: %.cc
 	$(if $(quiet),@echo "    [CXX] $@")
-	$(qexec)g++ $(INTERNAL_CFLAGS) $(CFLAGS) -c -o $@ $<
+	$(qexec)$(CXX) $(INTERNAL_CFLAGS) $(CXXFLAGS) -c -o $@ $<
 
 $(BUILD_PFX)%.asm.d: %.asm
 	$(if $(quiet),@echo "    [DEP] $@")
@@ -213,7 +217,7 @@ define linkerxx_template
 $(1): $(filter-out -%,$(2))
 $(1):
 	$(if $(quiet),@echo    "    [LD] $$@")
-	$(qexec)g++ $$(strip $$(INTERNAL_LDFLAGS) $$(LDFLAGS) -o $$@ $(2) $(3) $$(extralibs))
+	$(qexec)$$(CXX) $$(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

build/make/ads2gas.pl

@@ -26,12 +26,22 @@ print "\t.equ DO1STROUNDING, 0\n";
 
 while (<STDIN>)
 {
+    undef $comment;
+    undef $line;
+    $comment_char = ";";
+    $comment_sub = "@";
+
+    # Handle comments.
+    if (/$comment_char/)
+    {
+      $comment = "";
+      ($line, $comment) = /(.*?)$comment_char(.*)/;
+      $_ = $line;
+    }
+
     # Load and store alignment
     s/@/,:/g;
 
-    # Comment character
-    s/;/@/g;
-
     # Hexadecimal constants prefaced by 0x
     s/#&/#0x/g;
 
@@ -51,16 +61,27 @@ while (<STDIN>)
     s/:SHR:/ >> /g;
 
     # Convert ELSE to .else
-    s/ELSE/.else/g;
+    s/\bELSE\b/.else/g;
 
     # Convert ENDIF to .endif
-    s/ENDIF/.endif/g;
+    s/\bENDIF\b/.endif/g;
 
     # Convert ELSEIF to .elseif
-    s/ELSEIF/.elseif/g;
+    s/\bELSEIF\b/.elseif/g;
 
     # Convert LTORG to .ltorg
-    s/LTORG/.ltorg/g;
+    s/\bLTORG\b/.ltorg/g;
+
+    # Convert endfunc to nothing.
+    s/\bendfunc\b//ig;
+
+    # Convert FUNCTION to nothing.
+    s/\bFUNCTION\b//g;
+    s/\bfunction\b//g;
+
+    s/\bENTRY\b//g;
+    s/\bMSARMASM\b/0/g;
+    s/^\s+end\s+$//g;
 
     # Convert IF :DEF:to .if
     # gcc doesn't have the ability to do a conditional
@@ -106,6 +127,7 @@ while (<STDIN>)
     if (s/RN\s+([Rr]\d+|lr)/.req $1/)
     {
         print;
+        print "$comment_sub$comment\n" if defined $comment;
         next;
     }
 
@@ -114,6 +136,9 @@ while (<STDIN>)
     s/EXPORT\s+\|([\$\w]*)\|/.global $1 \n\t.type $1, function/;
     s/IMPORT\s+\|([\$\w]*)\|/.global $1/;
 
+    s/EXPORT\s+([\$\w]*)/.global $1/;
+    s/export\s+([\$\w]*)/.global $1/;
+
     # No vertical bars required; make additional symbol with prepended
     # underscore
     s/^\|(\$?\w+)\|/_$1\n\t$1:/g;
@@ -124,11 +149,19 @@ while (<STDIN>)
     s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
 
     # ALIGN directive
-    s/ALIGN/.balign/g;
+    s/\bALIGN\b/.balign/g;
 
     # ARM code
     s/\sARM/.arm/g;
 
+    # push/pop
+    s/(push\s+)(r\d+)/stmdb sp\!, \{$2\}/g;
+    s/(pop\s+)(r\d+)/ldmia sp\!, \{$2\}/g;
+
+    # NEON code
+    s/(vld1.\d+\s+)(q\d+)/$1\{$2\}/g;
+    s/(vtbl.\d+\s+[^,]+),([^,]+)/$1,\{$2\}/g;
+
     # eabi_attributes numerical equivalents can be found in the
     # "ARM IHI 0045C" document.
 
@@ -157,10 +190,10 @@ while (<STDIN>)
     }
 
     # EQU directive
-    s/(.*)EQU(.*)/.equ $1, $2/;
+    s/(\S+\s+)EQU(\s+\S+)/.equ $1, $2/;
 
     # Begin macro definition
-    if (/MACRO/) {
+    if (/\bMACRO\b/) {
         $_ = <STDIN>;
         s/^/.macro/;
         s/\$//g;                # remove formal param reference
@@ -169,9 +202,10 @@ while (<STDIN>)
 
     # For macros, use \ to reference formal params
     s/\$/\\/g;                  # End macro definition
-    s/MEND/.endm/;              # No need to tell it where to stop assembling
+    s/\bMEND\b/.endm/;              # No need to tell it where to stop assembling
     next if /^\s*END\s*$/;
     print;
+    print "$comment_sub$comment\n" if defined $comment;
 }
 
 # Mark that this object doesn't need an executable stack.

build/make/configure.sh

@@ -166,6 +166,17 @@ is_in(){
 
 add_cflags() {
     CFLAGS="${CFLAGS} $@"
+    CXXFLAGS="${CXXFLAGS} $@"
+}
+
+
+add_cflags_only() {
+    CFLAGS="${CFLAGS} $@"
+}
+
+
+add_cxxflags_only() {
+    CXXFLAGS="${CXXFLAGS} $@"
 }
 
 
@@ -277,6 +288,13 @@ check_cc() {
     check_cmd ${CC} ${CFLAGS} "$@" -c -o ${TMP_O} ${TMP_C}
 }
 
+check_cxx() {
+    log check_cxx "$@"
+    cat >${TMP_C}
+    log_file ${TMP_C}
+    check_cmd ${CXX} ${CXXFLAGS} "$@" -c -o ${TMP_O} ${TMP_C}
+}
+
 check_cpp() {
     log check_cpp "$@"
     cat > ${TMP_C}
@@ -310,8 +328,25 @@ int x;
 EOF
 }
 
+check_cxxflags() {
+    log check_cxxflags "$@"
+
+    # Catch CFLAGS that trigger CXX warnings
+    case "$CXX" in
+      *g++*) check_cxx -Werror "$@" <<EOF
+int x;
+EOF
+      ;;
+      *) check_cxx "$@" <<EOF
+int x;
+EOF
+      ;;
+    esac
+}
+
 check_add_cflags() {
-    check_cflags "$@" && add_cflags "$@"
+    check_cxxflags "$@" && add_cxxflags_only "$@"
+    check_cflags "$@" && add_cflags_only "$@"
 }
 
 check_add_asflags() {
@@ -367,7 +402,9 @@ true
 
 write_common_target_config_mk() {
     local CC=${CC}
+    local CXX=${CXX}
     enabled ccache && CC="ccache ${CC}"
+    enabled ccache && CXX="ccache ${CXX}"
     print_webm_license $1 "##" ""
 
     cat >> $1 << EOF
@@ -379,6 +416,7 @@ TOOLCHAIN=${toolchain}
 ASM_CONVERSION=${asm_conversion_cmd:-${source_path}/build/make/ads2gas.pl}
 
 CC=${CC}
+CXX=${CXX}
 AR=${AR}
 LD=${LD}
 AS=${AS}
@@ -386,6 +424,7 @@ STRIP=${STRIP}
 NM=${NM}
 
 CFLAGS  = ${CFLAGS}
+CXXFLAGS  = ${CXXFLAGS}
 ARFLAGS = -rus\$(if \$(quiet),c,v)
 LDFLAGS = ${LDFLAGS}
 ASFLAGS = ${ASFLAGS}
@@ -538,6 +577,7 @@ post_process_cmdline() {
 
 setup_gnu_toolchain() {
         CC=${CC:-${CROSS}gcc}
+        CXX=${CXX:-${CROSS}g++}
         AR=${AR:-${CROSS}ar}
         LD=${LD:-${CROSS}${link_with_cc:-ld}}
         AS=${AS:-${CROSS}as}
@@ -549,10 +589,19 @@ setup_gnu_toolchain() {
 
 process_common_toolchain() {
     if [ -z "$toolchain" ]; then
-        gcctarget="$(gcc -dumpmachine 2> /dev/null)"
+        gcctarget="${CHOST:-$(gcc -dumpmachine 2> /dev/null)}"
 
         # detect tgt_isa
         case "$gcctarget" in
+            armv6*)
+                tgt_isa=armv6
+                ;;
+            armv7*)
+                tgt_isa=armv7
+                ;;
+            armv5te*)
+                tgt_isa=armv5te
+                ;;
             *x86_64*|*amd64*)
                 tgt_isa=x86_64
                 ;;
@@ -718,6 +767,7 @@ process_common_toolchain() {
             ;;
         armv5te)
             soft_enable edsp
+            disable fast_unaligned
             ;;
         esac
 
@@ -733,17 +783,23 @@ process_common_toolchain() {
             check_add_asflags --defsym ARCHITECTURE=${arch_int}
             tune_cflags="-mtune="
             if [ ${tgt_isa} == "armv7" ]; then
+                check_add_cflags  -march=armv7-a -mfloat-abi=softfp
+                check_add_asflags -march=armv7-a -mfloat-abi=softfp
+
                 if enabled neon
                 then
                     check_add_cflags -mfpu=neon #-ftree-vectorize
                     check_add_asflags -mfpu=neon
                 fi
-                check_add_cflags -march=armv7-a -mcpu=cortex-a8 -mfloat-abi=softfp
-                check_add_asflags -mcpu=cortex-a8 -mfloat-abi=softfp  #-march=armv7-a
+
+                if [ -z "${tune_cpu}" ]; then
+                    tune_cpu=cortex-a8
+                fi
             else
                 check_add_cflags -march=${tgt_isa}
                 check_add_asflags -march=${tgt_isa}
             fi
+
             enabled debug && add_asflags -g
             asm_conversion_cmd="${source_path}/build/make/ads2gas.pl"
             ;;
@@ -792,6 +848,7 @@ process_common_toolchain() {
                                -name "arm-linux-androideabi-gcc*" -print -quit`
             TOOLCHAIN_PATH=${COMPILER_LOCATION%/*}/arm-linux-androideabi-
             CC=${TOOLCHAIN_PATH}gcc
+            CXX=${TOOLCHAIN_PATH}g++
             AR=${TOOLCHAIN_PATH}ar
             LD=${TOOLCHAIN_PATH}gcc
             AS=${TOOLCHAIN_PATH}as
@@ -810,12 +867,17 @@ process_common_toolchain() {
             add_cflags "--sysroot=${alt_libc}"
             add_ldflags "--sysroot=${alt_libc}"
 
-            add_cflags "-I${SDK_PATH}/sources/android/cpufeatures/"
+            # linker flag that routes around a CPU bug in some
+            # Cortex-A8 implementations (NDK Dev Guide)
+            add_ldflags "-Wl,--fix-cortex-a8"
 
             enable pic
             soft_enable realtime_only
             if [ ${tgt_isa} == "armv7" ]; then
-                enable runtime_cpu_detect
+                soft_enable runtime_cpu_detect
+            fi
+            if enabled runtime_cpu_detect; then
+                add_cflags "-I${SDK_PATH}/sources/android/cpufeatures"
             fi
           ;;
 
@@ -827,6 +889,7 @@ process_common_toolchain() {
                 SDK_PATH=${sdk_path}
             fi
             TOOLCHAIN_PATH=${SDK_PATH}/usr/bin
+            CXX=${TOOLCHAIN_PATH}/g++
             CC=${TOOLCHAIN_PATH}/gcc
             AR=${TOOLCHAIN_PATH}/ar
             LD=${TOOLCHAIN_PATH}/arm-apple-darwin10-llvm-gcc-4.2
@@ -890,13 +953,16 @@ process_common_toolchain() {
         esac
     ;;
     mips*)
-        CROSS=${CROSS:-mipsel-linux-uclibc-}
         link_with_cc=gcc
         setup_gnu_toolchain
         tune_cflags="-mtune="
+        if enabled dspr2; then
+            check_add_cflags -mips32r2 -mdspr2
+            disable fast_unaligned
+        fi
         check_add_cflags -march=${tgt_isa}
-    check_add_asflags -march=${tgt_isa}
-    check_add_asflags -KPIC
+        check_add_asflags -march=${tgt_isa}
+        check_add_asflags -KPIC
     ;;
     ppc*)
         enable ppc
@@ -924,6 +990,11 @@ process_common_toolchain() {
     x86*)
         bits=32
         enabled x86_64 && bits=64
+        check_cpp <<EOF && bits=x32
+#ifndef __ILP32__
+#error "not x32"
+#endif
+EOF
         soft_enable runtime_cpu_detect
         soft_enable mmx
         soft_enable sse
@@ -938,6 +1009,7 @@ process_common_toolchain() {
                 ;;
             solaris*)
                 CC=${CC:-${CROSS}gcc}
+                CXX=${CXX:-${CROSS}g++}
                 LD=${LD:-${CROSS}gcc}
                 CROSS=${CROSS:-g}
                 ;;
@@ -965,7 +1037,7 @@ process_common_toolchain() {
                         tune_cflags="-march="
                     ;;
                 esac
-                ;;
+            ;;
             gcc*)
                 add_cflags -m${bits}
                 add_ldflags -m${bits}
@@ -974,7 +1046,13 @@ process_common_toolchain() {
             setup_gnu_toolchain
                 #for 32 bit x86 builds, -O3 did not turn on this flag
                 enabled optimizations && check_add_cflags -fomit-frame-pointer
-                ;;
+            ;;
+            vs*)
+                # When building with Microsoft Visual Studio the assembler is
+                # invoked directly. Checking at configure time is unnecessary.
+                # Skip the check by setting AS arbitrarily
+                AS=msvs
+            ;;
         esac
 
         case "${AS}" in
@@ -983,7 +1061,7 @@ process_common_toolchain() {
                 which yasm >/dev/null 2>&1 && AS=yasm
                 [ "${AS}" = auto -o -z "${AS}" ] \
                     && die "Neither yasm nor nasm have been found"
-                ;;
+            ;;
         esac
         log_echo "  using $AS"
         [ "${AS##*/}" = nasm ] && add_asflags -Ox
@@ -1065,7 +1143,7 @@ process_common_toolchain() {
 
     # 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
+    enabled linux && check_add_cflags -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0
 
     # Check for strip utility variant
     ${STRIP} -V 2>/dev/null | grep GNU >/dev/null && enable gnu_strip
@@ -1080,12 +1158,24 @@ EOF
     # Almost every platform uses pthreads.
     if enabled multithread; then
         case ${toolchain} in
-            *-win*);;
+            *-win*-vs*);;
             *-android-gcc);;
             *) check_header pthread.h && add_extralibs -lpthread
         esac
     fi
 
+    # only for MIPS platforms
+    case ${toolchain} in
+        mips*)
+            if enabled dspr2; then
+                if enabled big_endian; then
+                    echo "dspr2 optimizations are available only for little endian platforms"
+                    disable dspr2
+                fi
+            fi
+        ;;
+    esac
+
     # for sysconf(3) and friends.
     check_header unistd.h
 

build/make/gen_asm_deps.sh

@@ -42,7 +42,7 @@ done
 
 [ -n "$srcfile" ] || show_help
 sfx=${sfx:-asm}
-includes=$(LC_ALL=C egrep -i "include +\"?+[a-z0-9_/]+\.${sfx}" $srcfile |
+includes=$(LC_ALL=C egrep -i "include +\"?[a-z0-9_/]+\.${sfx}" $srcfile |
            perl -p -e "s;.*?([a-z0-9_/]+.${sfx}).*;\1;")
 #" restore editor state
 for inc in ${includes}; do

build/make/rtcd.sh

@@ -278,6 +278,26 @@ EOF
 }
 
 
+mips() {
+  determine_indirection c $ALL_ARCHS
+  cat <<EOF
+$(common_top)
+#include "vpx_config.h"
+
+#ifdef RTCD_C
+static void setup_rtcd_internal(void)
+{
+$(set_function_pointers c $ALL_ARCHS)
+#if HAVE_DSPR2
+void dsputil_static_init();
+dsputil_static_init();
+#endif
+}
+#endif
+$(common_bottom)
+EOF
+}
+
 unoptimized() {
   determine_indirection c
   cat <<EOF
@@ -309,6 +329,15 @@ case $arch in
     require $(filter $REQUIRES)
     x86
     ;;
+  mips32)
+    ALL_ARCHS=$(filter mips32)
+    dspr2=$([ -f "$config_file" ] && eval echo $(grep HAVE_DSPR2 "$config_file"))
+    HAVE_DSPR2="${dspr2#*=}"
+    if [ "$HAVE_DSPR2" = "yes" ]; then
+        ALL_ARCHS=$(filter mips32 dspr2)
+    fi
+    mips
+    ;;
   armv5te)
     ALL_ARCHS=$(filter edsp)
     arm

configure

@@ -20,9 +20,10 @@ show_help(){
     show_help_pre
     cat << EOF
 Advanced options:
-  ${toggle_libs}                  don't build libraries
-  ${toggle_examples}              don't build examples
-  ${toggle_unit_tests}            build unit tests
+  ${toggle_libs}                  libraries
+  ${toggle_examples}              examples
+  ${toggle_docs}                  documentation
+  ${toggle_unit_tests}            unit tests
   --libc=PATH                     path to alternate libc
   --as={yasm|nasm|auto}           use specified assembler [auto, yasm preferred]
   --sdk-path=PATH                 path to root of sdk (iOS, android builds only)
@@ -209,6 +210,7 @@ ARCH_EXT_LIST="
     neon
 
     mips32
+    dspr2
 
     mmx
     sse
@@ -292,6 +294,7 @@ CMDLINE_SELECT="
 
     libs
     examples
+    docs
     libc
     as
     fast_unaligned
@@ -453,7 +456,13 @@ process_detect() {
         # Can only build shared libs on a subset of platforms. Doing this check
         # here rather than at option parse time because the target auto-detect
         # magic happens after the command line has been parsed.
-        enabled linux || die "--enable-shared only supported on ELF for now"
+        if ! enabled linux; then
+            if enabled gnu; then
+                echo "--enable-shared is only supported on ELF; assuming this is OK"
+            else
+                die "--enable-shared only supported on ELF for now"
+            fi
+        fi
     fi
     if [ -z "$CC" ]; then
         echo "Bypassing toolchain for environment detection."
@@ -586,6 +595,25 @@ process_toolchain() {
     if enabled postproc_visualizer; then
         enabled postproc || die "postproc_visualizer requires postproc to be enabled"
     fi
+
+    # Enable unit tests if we have a working C++ compiler
+    case "$toolchain" in
+        *-vs*)
+            soft_enable unit_tests
+        ;;
+        *-android-*)
+            # GTestLog must be modified to use Android logging utilities.
+        ;;
+        *-darwin-*)
+            # iOS/ARM builds do not work with gtest. This does not match
+            # x86 targets.
+        ;;
+        *)
+            check_cxx "$@" <<EOF && soft_enable unit_tests
+int z;
+EOF
+        ;;
+    esac
 }
 
 

examples.mk

@@ -98,7 +98,7 @@ vp8cx_set_ref.GUID                  = C5E31F7F-96F6-48BD-BD3E-10EBF6E8057A
 vp8cx_set_ref.DESCRIPTION           = VP8 set encoder reference frame
 
 # C file is provided, not generated automatically.
-GEN_EXAMPLES-$(CONFIG_MULTI_RES_ENCODING) += vp8_multi_resolution_encoder.c
+UTILS-$(CONFIG_MULTI_RES_ENCODING) += vp8_multi_resolution_encoder.c
 vp8_multi_resolution_encoder.SRCS  \
                          += third_party/libyuv/include/libyuv/basic_types.h  \
                             third_party/libyuv/include/libyuv/cpu_id.h  \

examples/decoder_tmpl.txt

@@ -48,8 +48,8 @@ for(plane=0; plane < 3; plane++) {
     unsigned char *buf =img->planes[plane];
 
     for(y=0; y < (plane ? (img->d_h + 1) >> 1 : img->d_h); y++) {
-        if(fwrite(buf, 1, (plane ? (img->d_w + 1) >> 1 : img->d_w),
-           outfile));
+        (void) fwrite(buf, 1, (plane ? (img->d_w + 1) >> 1 : img->d_w),
+                      outfile);
         buf += img->stride[plane];
     }
 }

examples/encoder_tmpl.c

@@ -85,7 +85,7 @@ static void write_ivf_file_header(FILE *outfile,
     mem_put_le32(header+24, frame_cnt);           /* length */
     mem_put_le32(header+28, 0);                   /* unused */
 
-    if(fwrite(header, 1, 32, outfile));
+    (void) fwrite(header, 1, 32, outfile);
 }
 
 
@@ -103,7 +103,7 @@ static void write_ivf_frame_header(FILE *outfile,
     mem_put_le32(header+4, pts&0xFFFFFFFF);
     mem_put_le32(header+8, pts >> 32);
 
-    if(fwrite(header, 1, 12, outfile));
+    (void) fwrite(header, 1, 12, outfile);
 }
 
 int main(int argc, char **argv) {

examples/encoder_tmpl.txt

@@ -61,13 +61,14 @@ if(vpx_codec_encode(&codec, frame_avail? &raw : NULL, frame_cnt,
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ PROCESS_FRAME
 case VPX_CODEC_CX_FRAME_PKT:
     write_ivf_frame_header(outfile, pkt);
-    if(fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
-              outfile));
+    (void) fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz,
+                  outfile);
     break;
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ PROCESS_FRAME
 
 
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DESTROY
+vpx_img_free(&raw);
 if(vpx_codec_destroy(&codec))
     die_codec(&codec, "Failed to destroy codec");
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DESTROY

examples/twopass_encoder.txt

@@ -71,5 +71,17 @@ Pass Progress Reporting
 It's sometimes helpful to see when each pass completes.
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ TWOPASS_LOOP_END
     printf("Pass %d complete.\n", pass+1);
+    if(vpx_codec_destroy(&codec))
+        die_codec(&codec, "Failed to destroy codec");
 }
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ TWOPASS_LOOP_END
+
+
+Clean-up
+-----------------------------
+Destruction of the encoder instance must be done on each pass. The
+raw image should be destroyed at the end as usual.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DESTROY
+vpx_img_free(&raw);
+free(stats.buf);
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ DESTROY

libmkv/EbmlIDs.h

@@ -1,16 +1,16 @@
-// 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.
-
-
+/*
+ *  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.
+ */
 #ifndef MKV_DEFS_HPP
 #define MKV_DEFS_HPP 1
 
-//Commenting out values not available in webm, but available in matroska
+/* Commenting out values not available in webm, but available in matroska */
 
 enum mkv
 {
@@ -22,7 +22,7 @@ enum mkv
     DocType = 0x4282,
     DocTypeVersion = 0x4287,
     DocTypeReadVersion = 0x4285,
-//  CRC_32 = 0xBF,
+/* CRC_32 = 0xBF, */
     Void = 0xEC,
     SignatureSlot = 0x1B538667,
     SignatureAlgo = 0x7E8A,
@@ -32,61 +32,61 @@ enum mkv
     SignatureElements = 0x7E5B,
     SignatureElementList = 0x7E7B,
     SignedElement = 0x6532,
-    //segment
+    /* segment */
     Segment = 0x18538067,
-    //Meta Seek Information
+    /* Meta Seek Information */
     SeekHead = 0x114D9B74,
     Seek = 0x4DBB,
     SeekID = 0x53AB,
     SeekPosition = 0x53AC,
-    //Segment Information
+    /* Segment Information */
     Info = 0x1549A966,
-//  SegmentUID = 0x73A4,
-//  SegmentFilename = 0x7384,
-//  PrevUID = 0x3CB923,
-//  PrevFilename = 0x3C83AB,
-//  NextUID = 0x3EB923,
-//  NextFilename = 0x3E83BB,
-//  SegmentFamily = 0x4444,
-//  ChapterTranslate = 0x6924,
-//  ChapterTranslateEditionUID = 0x69FC,
-//  ChapterTranslateCodec = 0x69BF,
-//  ChapterTranslateID = 0x69A5,
+/* SegmentUID = 0x73A4, */
+/* SegmentFilename = 0x7384, */
+/* PrevUID = 0x3CB923, */
+/* PrevFilename = 0x3C83AB, */
+/* NextUID = 0x3EB923, */
+/* NextFilename = 0x3E83BB, */
+/* SegmentFamily = 0x4444, */
+/* ChapterTranslate = 0x6924, */
+/* ChapterTranslateEditionUID = 0x69FC, */
+/* ChapterTranslateCodec = 0x69BF, */
+/* ChapterTranslateID = 0x69A5, */
     TimecodeScale = 0x2AD7B1,
     Segment_Duration = 0x4489,
     DateUTC = 0x4461,
-//  Title = 0x7BA9,
+/* Title = 0x7BA9, */
     MuxingApp = 0x4D80,
     WritingApp = 0x5741,
-    //Cluster
+    /* Cluster */
     Cluster = 0x1F43B675,
     Timecode = 0xE7,
-//  SilentTracks = 0x5854,
-//  SilentTrackNumber = 0x58D7,
-//  Position = 0xA7,
+/* SilentTracks = 0x5854, */
+/* SilentTrackNumber = 0x58D7, */
+/* Position = 0xA7, */
     PrevSize = 0xAB,
     BlockGroup = 0xA0,
     Block = 0xA1,
-//  BlockVirtual = 0xA2,
-//  BlockAdditions = 0x75A1,
-//  BlockMore = 0xA6,
-//  BlockAddID = 0xEE,
-//  BlockAdditional = 0xA5,
+/* BlockVirtual = 0xA2, */
+/* BlockAdditions = 0x75A1, */
+/* BlockMore = 0xA6, */
+/* BlockAddID = 0xEE, */
+/* BlockAdditional = 0xA5, */
     BlockDuration = 0x9B,
-//  ReferencePriority = 0xFA,
+/* ReferencePriority = 0xFA, */
     ReferenceBlock = 0xFB,
-//  ReferenceVirtual = 0xFD,
-//  CodecState = 0xA4,
-//  Slices = 0x8E,
-//  TimeSlice = 0xE8,
+/* ReferenceVirtual = 0xFD, */
+/* CodecState = 0xA4, */
+/* Slices = 0x8E, */
+/* TimeSlice = 0xE8, */
     LaceNumber = 0xCC,
-//  FrameNumber = 0xCD,
-//  BlockAdditionID = 0xCB,
-//  MkvDelay = 0xCE,
-//  Cluster_Duration = 0xCF,
+/* FrameNumber = 0xCD, */
+/* BlockAdditionID = 0xCB, */
+/* MkvDelay = 0xCE, */
+/* Cluster_Duration = 0xCF, */
     SimpleBlock = 0xA3,
-//  EncryptedBlock = 0xAF,
-    //Track
+/* EncryptedBlock = 0xAF, */
+    /* Track */
     Tracks = 0x1654AE6B,
     TrackEntry = 0xAE,
     TrackNumber = 0xD7,
@@ -96,28 +96,28 @@ enum mkv
     FlagDefault = 0x88,
     FlagForced = 0x55AA,
     FlagLacing = 0x9C,
-//  MinCache = 0x6DE7,
-//  MaxCache = 0x6DF8,
+/* MinCache = 0x6DE7, */
+/* MaxCache = 0x6DF8, */
     DefaultDuration = 0x23E383,
-//  TrackTimecodeScale = 0x23314F,
-//  TrackOffset = 0x537F,
-//  MaxBlockAdditionID = 0x55EE,
+/* TrackTimecodeScale = 0x23314F, */
+/* TrackOffset = 0x537F, */
+/* MaxBlockAdditionID = 0x55EE, */
     Name = 0x536E,
     Language = 0x22B59C,
     CodecID = 0x86,
     CodecPrivate = 0x63A2,
     CodecName = 0x258688,
-//  AttachmentLink = 0x7446,
-//  CodecSettings = 0x3A9697,
-//  CodecInfoURL = 0x3B4040,
-//  CodecDownloadURL = 0x26B240,
-//  CodecDecodeAll = 0xAA,
-//  TrackOverlay = 0x6FAB,
-//  TrackTranslate = 0x6624,
-//  TrackTranslateEditionUID = 0x66FC,
-//  TrackTranslateCodec = 0x66BF,
-//  TrackTranslateTrackID = 0x66A5,
-    //video
+/* AttachmentLink = 0x7446, */
+/* CodecSettings = 0x3A9697, */
+/* CodecInfoURL = 0x3B4040, */
+/* CodecDownloadURL = 0x26B240, */
+/* CodecDecodeAll = 0xAA, */
+/* TrackOverlay = 0x6FAB, */
+/* TrackTranslate = 0x6624, */
+/* TrackTranslateEditionUID = 0x66FC, */
+/* TrackTranslateCodec = 0x66BF, */
+/* TrackTranslateTrackID = 0x66A5, */
+    /* video */
     Video = 0xE0,
     FlagInterlaced = 0x9A,
     StereoMode = 0x53B8,
@@ -131,101 +131,101 @@ enum mkv
     DisplayHeight = 0x54BA,
     DisplayUnit = 0x54B2,
     AspectRatioType = 0x54B3,
-//  ColourSpace = 0x2EB524,
-//  GammaValue = 0x2FB523,
+/* ColourSpace = 0x2EB524, */
+/* GammaValue = 0x2FB523, */
     FrameRate = 0x2383E3,
-    //end video
-    //audio
+    /* end video */
+    /* audio */
     Audio = 0xE1,
     SamplingFrequency = 0xB5,
     OutputSamplingFrequency = 0x78B5,
     Channels = 0x9F,
-//  ChannelPositions = 0x7D7B,
+/* ChannelPositions = 0x7D7B, */
     BitDepth = 0x6264,
-    //end audio
-    //content encoding
-//  ContentEncodings = 0x6d80,
-//  ContentEncoding = 0x6240,
-//  ContentEncodingOrder = 0x5031,
-//  ContentEncodingScope = 0x5032,
-//  ContentEncodingType = 0x5033,
-//  ContentCompression = 0x5034,
-//  ContentCompAlgo = 0x4254,
-//  ContentCompSettings = 0x4255,
-//  ContentEncryption = 0x5035,
-//  ContentEncAlgo = 0x47e1,
-//  ContentEncKeyID = 0x47e2,
-//  ContentSignature = 0x47e3,
-//  ContentSigKeyID = 0x47e4,
-//  ContentSigAlgo = 0x47e5,
-//  ContentSigHashAlgo = 0x47e6,
-    //end content encoding
-    //Cueing Data
+    /* end audio */
+    /* content encoding */
+/* ContentEncodings = 0x6d80, */
+/* ContentEncoding = 0x6240, */
+/* ContentEncodingOrder = 0x5031, */
+/* ContentEncodingScope = 0x5032, */
+/* ContentEncodingType = 0x5033, */
+/* ContentCompression = 0x5034, */
+/* ContentCompAlgo = 0x4254, */
+/* ContentCompSettings = 0x4255, */
+/* ContentEncryption = 0x5035, */
+/* ContentEncAlgo = 0x47e1, */
+/* ContentEncKeyID = 0x47e2, */
+/* ContentSignature = 0x47e3, */
+/* ContentSigKeyID = 0x47e4, */
+/* ContentSigAlgo = 0x47e5, */
+/* ContentSigHashAlgo = 0x47e6, */
+    /* end content encoding */
+    /* Cueing Data */
     Cues = 0x1C53BB6B,
     CuePoint = 0xBB,
     CueTime = 0xB3,
     CueTrackPositions = 0xB7,
     CueTrack = 0xF7,
     CueClusterPosition = 0xF1,
-    CueBlockNumber = 0x5378,
-//  CueCodecState = 0xEA,
-//  CueReference = 0xDB,
-//  CueRefTime = 0x96,
-//  CueRefCluster = 0x97,
-//  CueRefNumber = 0x535F,
-//  CueRefCodecState = 0xEB,
-    //Attachment
-//  Attachments = 0x1941A469,
-//  AttachedFile = 0x61A7,
-//  FileDescription = 0x467E,
-//  FileName = 0x466E,
-//  FileMimeType = 0x4660,
-//  FileData = 0x465C,
-//  FileUID = 0x46AE,
-//  FileReferral = 0x4675,
-    //Chapters
-//  Chapters = 0x1043A770,
-//  EditionEntry = 0x45B9,
-//  EditionUID = 0x45BC,
-//  EditionFlagHidden = 0x45BD,
-//  EditionFlagDefault = 0x45DB,
-//  EditionFlagOrdered = 0x45DD,
-//  ChapterAtom = 0xB6,
-//  ChapterUID = 0x73C4,
-//  ChapterTimeStart = 0x91,
-//  ChapterTimeEnd = 0x92,
-//  ChapterFlagHidden = 0x98,
-//  ChapterFlagEnabled = 0x4598,
-//  ChapterSegmentUID = 0x6E67,
-//  ChapterSegmentEditionUID = 0x6EBC,
-//  ChapterPhysicalEquiv = 0x63C3,
-//  ChapterTrack = 0x8F,
-//  ChapterTrackNumber = 0x89,
-//  ChapterDisplay = 0x80,
-//  ChapString = 0x85,
-//  ChapLanguage = 0x437C,
-//  ChapCountry = 0x437E,
-//  ChapProcess = 0x6944,
-//  ChapProcessCodecID = 0x6955,
-//  ChapProcessPrivate = 0x450D,
-//  ChapProcessCommand = 0x6911,
-//  ChapProcessTime = 0x6922,
-//  ChapProcessData = 0x6933,
-    //Tagging
-//  Tags = 0x1254C367,
-//  Tag = 0x7373,
-//  Targets = 0x63C0,
-//  TargetTypeValue = 0x68CA,
-//  TargetType = 0x63CA,
-//  Tagging_TrackUID = 0x63C5,
-//  Tagging_EditionUID = 0x63C9,
-//  Tagging_ChapterUID = 0x63C4,
-//  AttachmentUID = 0x63C6,
-//  SimpleTag = 0x67C8,
-//  TagName = 0x45A3,
-//  TagLanguage = 0x447A,
-//  TagDefault = 0x4484,
-//  TagString = 0x4487,
-//  TagBinary = 0x4485,
+    CueBlockNumber = 0x5378
+/* CueCodecState = 0xEA, */
+/* CueReference = 0xDB, */
+/* CueRefTime = 0x96, */
+/* CueRefCluster = 0x97, */
+/* CueRefNumber = 0x535F, */
+/* CueRefCodecState = 0xEB, */
+    /* Attachment */
+/* Attachments = 0x1941A469, */
+/* AttachedFile = 0x61A7, */
+/* FileDescription = 0x467E, */
+/* FileName = 0x466E, */
+/* FileMimeType = 0x4660, */
+/* FileData = 0x465C, */
+/* FileUID = 0x46AE, */
+/* FileReferral = 0x4675, */
+    /* Chapters */
+/* Chapters = 0x1043A770, */
+/* EditionEntry = 0x45B9, */
+/* EditionUID = 0x45BC, */
+/* EditionFlagHidden = 0x45BD, */
+/* EditionFlagDefault = 0x45DB, */
+/* EditionFlagOrdered = 0x45DD, */
+/* ChapterAtom = 0xB6, */
+/* ChapterUID = 0x73C4, */
+/* ChapterTimeStart = 0x91, */
+/* ChapterTimeEnd = 0x92, */
+/* ChapterFlagHidden = 0x98, */
+/* ChapterFlagEnabled = 0x4598, */
+/* ChapterSegmentUID = 0x6E67, */
+/* ChapterSegmentEditionUID = 0x6EBC, */
+/* ChapterPhysicalEquiv = 0x63C3, */
+/* ChapterTrack = 0x8F, */
+/* ChapterTrackNumber = 0x89, */
+/* ChapterDisplay = 0x80, */
+/* ChapString = 0x85, */
+/* ChapLanguage = 0x437C, */
+/* ChapCountry = 0x437E, */
+/* ChapProcess = 0x6944, */
+/* ChapProcessCodecID = 0x6955, */
+/* ChapProcessPrivate = 0x450D, */
+/* ChapProcessCommand = 0x6911, */
+/* ChapProcessTime = 0x6922, */
+/* ChapProcessData = 0x6933, */
+    /* Tagging */
+/* Tags = 0x1254C367, */
+/* Tag = 0x7373, */
+/* Targets = 0x63C0, */
+/* TargetTypeValue = 0x68CA, */
+/* TargetType = 0x63CA, */
+/* Tagging_TrackUID = 0x63C5, */
+/* Tagging_EditionUID = 0x63C9, */
+/* Tagging_ChapterUID = 0x63C4, */
+/* AttachmentUID = 0x63C6, */
+/* SimpleTag = 0x67C8, */
+/* TagName = 0x45A3, */
+/* TagLanguage = 0x447A, */
+/* TagDefault = 0x4484, */
+/* TagString = 0x4487, */
+/* TagBinary = 0x4485, */
 };
 #endif

libmkv/EbmlWriter.c

@@ -1,12 +1,12 @@
-// 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.
-
-
+/*
+ *  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 "EbmlWriter.h"
 #include <stdlib.h>
 #include <wchar.h>
@@ -18,11 +18,13 @@
 #define LITERALU64(n) n##LLU
 #endif
 
-void Ebml_WriteLen(EbmlGlobal *glob, long long val)
+void Ebml_WriteLen(EbmlGlobal *glob, int64_t val)
 {
-    //TODO check and make sure we are not > than 0x0100000000000000LLU
-    unsigned char size = 8; //size in bytes to output
-    unsigned long long minVal = LITERALU64(0x00000000000000ff); //mask to compare for byte size
+    /* TODO check and make sure we are not > than 0x0100000000000000LLU */
+    unsigned char size = 8; /* size in bytes to output */
+
+    /* mask to compare for byte size */
+    int64_t minVal = 0xff;
 
     for (size = 1; size < 8; size ++)
     {
@@ -32,7 +34,7 @@ void Ebml_WriteLen(EbmlGlobal *glob, long long val)
         minVal = (minVal << 7);
     }
 
-    val |= (LITERALU64(0x000000000000080) << ((size - 1) * 7));
+    val |= (((uint64_t)0x80) << ((size - 1) * 7));
 
     Ebml_Serialize(glob, (void *) &val, sizeof(val), size);
 }
@@ -40,23 +42,25 @@ void Ebml_WriteLen(EbmlGlobal *glob, long long val)
 void Ebml_WriteString(EbmlGlobal *glob, const char *str)
 {
     const size_t size_ = strlen(str);
-    const unsigned long long  size = size_;
+    const uint64_t  size = size_;
     Ebml_WriteLen(glob, size);
-    //TODO: it's not clear from the spec whether the nul terminator
-    //should be serialized too.  For now we omit the null terminator.
-    Ebml_Write(glob, str, size);
+    /* TODO: it's not clear from the spec whether the nul terminator
+     * should be serialized too.  For now we omit the null terminator.
+     */
+    Ebml_Write(glob, str, (unsigned long)size);
 }
 
 void Ebml_WriteUTF8(EbmlGlobal *glob, const wchar_t *wstr)
 {
     const size_t strlen = wcslen(wstr);
 
-    //TODO: it's not clear from the spec whether the nul terminator
-    //should be serialized too.  For now we include it.
-    const unsigned long long  size = strlen;
+    /* TODO: it's not clear from the spec whether the nul terminator
+     * should be serialized too.  For now we include it.
+     */
+    const uint64_t  size = strlen;
 
     Ebml_WriteLen(glob, size);
-    Ebml_Write(glob, wstr, size);
+    Ebml_Write(glob, wstr, (unsigned long)size);
 }
 
 void Ebml_WriteID(EbmlGlobal *glob, unsigned long class_id)
@@ -85,12 +89,12 @@ void Ebml_SerializeUnsigned64(EbmlGlobal *glob, unsigned long class_id, uint64_t
 
 void Ebml_SerializeUnsigned(EbmlGlobal *glob, unsigned long class_id, unsigned long ui)
 {
-    unsigned char size = 8; //size in bytes to output
+    unsigned char size = 8; /* size in bytes to output */
     unsigned char sizeSerialized = 0;
     unsigned long minVal;
 
     Ebml_WriteID(glob, class_id);
-    minVal = 0x7fLU; //mask to compare for byte size
+    minVal = 0x7fLU; /* mask to compare for byte size */
 
     for (size = 1; size < 4; size ++)
     {
@@ -106,7 +110,7 @@ void Ebml_SerializeUnsigned(EbmlGlobal *glob, unsigned long class_id, unsigned l
     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
+/* 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)
 {
     int size;
@@ -168,4 +172,4 @@ void Ebml_WriteVoid(EbmlGlobal *glob, unsigned long vSize)
     }
 }
 
-//TODO Serialize Date
+/* TODO Serialize Date */

libmkv/EbmlWriter.h

@@ -1,26 +1,30 @@
+/*
+ *  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.
+ */
 #ifndef EBMLWRITER_HPP
 #define EBMLWRITER_HPP
-
-// 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.
-
-//note: you must define write and serialize functions as well as your own EBML_GLOBAL
-//These functions MUST be implemented
 #include <stddef.h>
 #include "vpx/vpx_integer.h"
 
+/* note: you must define write and serialize functions as well as your own
+ * EBML_GLOBAL
+ *
+ * These functions MUST be implemented
+ */
+
 typedef struct EbmlGlobal EbmlGlobal;
 void  Ebml_Serialize(EbmlGlobal *glob, const void *, int, unsigned long);
 void  Ebml_Write(EbmlGlobal *glob, const void *, unsigned long);
-/////
 
+/*****/
 
-void Ebml_WriteLen(EbmlGlobal *glob, long long val);
+void Ebml_WriteLen(EbmlGlobal *glob, int64_t val);
 void Ebml_WriteString(EbmlGlobal *glob, const char *str);
 void Ebml_WriteUTF8(EbmlGlobal *glob, const wchar_t *wstr);
 void Ebml_WriteID(EbmlGlobal *glob, unsigned long class_id);
@@ -28,11 +32,11 @@ void Ebml_SerializeUnsigned64(EbmlGlobal *glob, unsigned long class_id, uint64_t
 void Ebml_SerializeUnsigned(EbmlGlobal *glob, unsigned long class_id, unsigned long ui);
 void Ebml_SerializeBinary(EbmlGlobal *glob, unsigned long class_id, unsigned long ui);
 void Ebml_SerializeFloat(EbmlGlobal *glob, unsigned long class_id, double d);
-//TODO make this more generic to signed
+/* TODO make this more generic to signed */
 void Ebml_WriteSigned16(EbmlGlobal *glob, short val);
 void Ebml_SerializeString(EbmlGlobal *glob, unsigned long class_id, const char *s);
 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);
 void Ebml_WriteVoid(EbmlGlobal *glob, unsigned long vSize);
-//TODO need date function
+/* TODO need date function */
 #endif

libs.mk

@@ -20,8 +20,16 @@ endif
 CODEC_SRCS-yes += CHANGELOG
 CODEC_SRCS-yes += libs.mk
 
+# If this is a universal (fat) binary, then all the subarchitectures have
+# already been built and our job is to stitch them together. The
+# BUILD_LIBVPX variable indicates whether we should be building
+# (compiling, linking) the library. The LIPO_LIBVPX variable indicates
+# that we're stitching.
+$(eval $(if $(filter universal%,$(TOOLCHAIN)),LIPO_LIBVPX,BUILD_LIBVPX):=yes)
+
 include $(SRC_PATH_BARE)/vpx/vpx_codec.mk
 CODEC_SRCS-yes += $(addprefix vpx/,$(call enabled,API_SRCS))
+CODEC_DOC_SRCS += $(addprefix vpx/,$(call enabled,API_DOC_SRCS))
 
 include $(SRC_PATH_BARE)/vpx_mem/vpx_mem.mk
 CODEC_SRCS-yes += $(addprefix vpx_mem/,$(call enabled,MEM_SRCS))
@@ -29,17 +37,17 @@ CODEC_SRCS-yes += $(addprefix vpx_mem/,$(call enabled,MEM_SRCS))
 include $(SRC_PATH_BARE)/vpx_scale/vpx_scale.mk
 CODEC_SRCS-yes += $(addprefix vpx_scale/,$(call enabled,SCALE_SRCS))
 
+include $(SRC_PATH_BARE)/vpx_ports/vpx_ports.mk
+CODEC_SRCS-yes += $(addprefix vpx_ports/,$(call enabled,PORTS_SRCS))
+
 
 ifeq ($(CONFIG_VP8_ENCODER),yes)
   VP8_PREFIX=vp8/
   include $(SRC_PATH_BARE)/$(VP8_PREFIX)vp8cx.mk
   CODEC_SRCS-yes += $(addprefix $(VP8_PREFIX),$(call enabled,VP8_CX_SRCS))
   CODEC_EXPORTS-yes += $(addprefix $(VP8_PREFIX),$(VP8_CX_EXPORTS))
-  CODEC_SRCS-yes += $(VP8_PREFIX)vp8cx.mk vpx/vp8.h vpx/vp8cx.h
-  CODEC_SRCS-$(ARCH_ARM) += $(VP8_PREFIX)vp8cx_arm.mk
   INSTALL-LIBS-yes += include/vpx/vp8.h include/vpx/vp8cx.h
   INSTALL_MAPS += include/vpx/% $(SRC_PATH_BARE)/$(VP8_PREFIX)/%
-  CODEC_DOC_SRCS += vpx/vp8.h vpx/vp8cx.h
   CODEC_DOC_SECTIONS += vp8 vp8_encoder
 endif
 
@@ -48,10 +56,8 @@ ifeq ($(CONFIG_VP8_DECODER),yes)
   include $(SRC_PATH_BARE)/$(VP8_PREFIX)vp8dx.mk
   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
   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
   CODEC_DOC_SECTIONS += vp8 vp8_decoder
 endif
 
@@ -66,6 +72,7 @@ endif
 
 ifeq ($(CONFIG_MSVS),yes)
 CODEC_LIB=$(if $(CONFIG_STATIC_MSVCRT),vpxmt,vpxmd)
+GTEST_LIB=$(if $(CONFIG_STATIC_MSVCRT),gtestmt,gtestmd)
 # This variable uses deferred expansion intentionally, since the results of
 # $(wildcard) may change during the course of the Make.
 VS_PLATFORMS = $(foreach d,$(wildcard */Release/$(CODEC_LIB).lib),$(word 1,$(subst /, ,$(d))))
@@ -82,29 +89,10 @@ INSTALL_MAPS += $(foreach p,$(VS_PLATFORMS),$(LIBSUBDIR)/$(p)/%  $(p)/Release/%)
 INSTALL_MAPS += $(foreach p,$(VS_PLATFORMS),$(LIBSUBDIR)/$(p)/%  $(p)/Debug/%)
 endif
 
-# If this is a universal (fat) binary, then all the subarchitectures have
-# already been built and our job is to stitch them together. The
-# BUILD_LIBVPX variable indicates whether we should be building
-# (compiling, linking) the library. The LIPO_LIBVPX variable indicates
-# that we're stitching.
-$(eval $(if $(filter universal%,$(TOOLCHAIN)),LIPO_LIBVPX,BUILD_LIBVPX):=yes)
-
 CODEC_SRCS-$(BUILD_LIBVPX) += build/make/version.sh
 CODEC_SRCS-$(BUILD_LIBVPX) += build/make/rtcd.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
 INSTALL-SRCS-no += $(BUILD_PFX)vpx_config.c
-ifeq ($(ARCH_X86)$(ARCH_X86_64),yes)
-CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/emms.asm
-CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/x86.h
-CODEC_SRCS-$(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-$(ARCH_ARM) += vpx_ports/arm.h
 CODEC_EXPORTS-$(BUILD_LIBVPX) += vpx/exports_com
 CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_enc
 CODEC_EXPORTS-$(CONFIG_DECODERS) += vpx/exports_dec
@@ -146,7 +134,7 @@ 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 \
+	$(qexec)$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
     --exe \
     --target=$(TOOLCHAIN) \
     --name=obj_int_extract \
@@ -162,14 +150,14 @@ PROJECTS-$(BUILD_LIBVPX) += obj_int_extract.bat
 
 vpx.def: $(call enabled,CODEC_EXPORTS)
 	@echo "    [CREATE] $@"
-	$(SRC_PATH_BARE)/build/make/gen_msvs_def.sh\
+	$(qexec)$(SRC_PATH_BARE)/build/make/gen_msvs_def.sh\
             --name=vpx\
             --out=$@ $^
 CLEAN-OBJS += vpx.def
 
 vpx.vcproj: $(CODEC_SRCS) vpx.def
 	@echo "    [CREATE] $@"
-	$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
+	$(qexec)$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
 			--lib \
 			--target=$(TOOLCHAIN) \
             $(if $(CONFIG_STATIC_MSVCRT),--static-crt) \
@@ -242,6 +230,7 @@ vpx.pc: config.mk libs.mk
 	$(qexec)echo 'Requires:' >> $@
 	$(qexec)echo 'Conflicts:' >> $@
 	$(qexec)echo 'Libs: -L$${libdir} -lvpx' >> $@
+	$(qexec)echo 'Libs.private: -lm -lpthread' >> $@
 	$(qexec)echo 'Cflags: -I$${includedir}' >> $@
 INSTALL-LIBS-yes += $(LIBSUBDIR)/pkgconfig/vpx.pc
 INSTALL_MAPS += $(LIBSUBDIR)/pkgconfig/%.pc %.pc
@@ -284,38 +273,44 @@ OFFSET_PATTERN:='^[a-zA-Z0-9_]* EQU'
 
 ifeq ($(filter icc gcc,$(TGT_CC)), $(TGT_CC))
     $(BUILD_PFX)asm_com_offsets.asm: $(BUILD_PFX)$(VP8_PREFIX)common/asm_com_offsets.c.S
-	LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(ADS2GAS) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(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
-	LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(ADS2GAS) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(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
-	LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(ADS2GAS) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)LC_ALL=C grep $(OFFSET_PATTERN) $< | tr -d '$$\#' $(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) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)./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) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)./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
 
     asm_dec_offsets.asm: obj_int_extract
     asm_dec_offsets.asm: $(VP8_PREFIX)decoder/asm_dec_offsets.c.o
-	./obj_int_extract rvds $< $(ADS2GAS) > $@
+	@echo "    [CREATE] $@"
+	$(qexec)./obj_int_extract rvds $< $(ADS2GAS) > $@
     OBJS-yes += $(VP8_PREFIX)decoder/asm_dec_offsets.c.o
     CLEAN-OBJS += asm_dec_offsets.asm
     $(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)asm_dec_offsets.asm
@@ -328,7 +323,6 @@ CLEAN-OBJS += $(BUILD_PFX)vpx_version.h
 #
 # Rule to generate runtime cpu detection files
 #
-$(OBJS-yes:.o=.d): $(BUILD_PFX)vpx_rtcd.h
 $(BUILD_PFX)vpx_rtcd.h: $(SRC_PATH_BARE)/$(sort $(filter %rtcd_defs.sh,$(CODEC_SRCS)))
 	@echo "    [CREATE] $@"
 	$(qexec)$(SRC_PATH_BARE)/build/make/rtcd.sh --arch=$(TGT_ISA) \
@@ -337,25 +331,43 @@ $(BUILD_PFX)vpx_rtcd.h: $(SRC_PATH_BARE)/$(sort $(filter %rtcd_defs.sh,$(CODEC_S
           $(RTCD_OPTIONS) $^ > $@
 CLEAN-OBJS += $(BUILD_PFX)vpx_rtcd.h
 
-CODEC_DOC_SRCS += vpx/vpx_codec.h \
-                  vpx/vpx_decoder.h \
-                  vpx/vpx_encoder.h \
-                  vpx/vpx_image.h
-
 ##
 ## libvpx test directives
 ##
-
 ifeq ($(CONFIG_UNIT_TESTS),yes)
+LIBVPX_TEST_DATA_PATH ?= .
+
+include $(SRC_PATH_BARE)/test/test.mk
+LIBVPX_TEST_SRCS=$(addprefix test/,$(call enabled,LIBVPX_TEST_SRCS))
+LIBVPX_TEST_BINS=./test_libvpx
+LIBVPX_TEST_DATA=$(addprefix $(LIBVPX_TEST_DATA_PATH)/,\
+                     $(call enabled,LIBVPX_TEST_DATA))
+libvpx_test_data_url=http://downloads.webmproject.org/test_data/libvpx/$(1)
+
+$(LIBVPX_TEST_DATA):
+	@echo "    [DOWNLOAD] $@"
+	$(qexec)trap 'rm -f $@' INT TERM &&\
+            curl -L -o $@ $(call libvpx_test_data_url,$(@F))
+
+testdata:: $(LIBVPX_TEST_DATA)
+	$(qexec)if [ -x "$$(which sha1sum)" ]; then\
+            echo "Checking test data:";\
+            if [ -n "$(LIBVPX_TEST_DATA)" ]; then\
+                for f in $(call enabled,LIBVPX_TEST_DATA); do\
+                    grep $$f $(SRC_PATH_BARE)/test/test-data.sha1 |\
+                        (cd $(LIBVPX_TEST_DATA_PATH); sha1sum -c);\
+                done; \
+            fi; \
+        else\
+            echo "Skipping test data integrity check, sha1sum not found.";\
+        fi
+
 ifeq ($(CONFIG_EXTERNAL_BUILD),yes)
 ifeq ($(CONFIG_MSVS),yes)
 
-LIBVPX_TEST_SRCS=$(filter %_test.cc,$(call enabled,CODEC_SRCS))
-LIBVPX_TEST_BINS=$(sort $(LIBVPX_TEST_SRCS:.cc.o=))
-
 gtest.vcproj: $(SRC_PATH_BARE)/third_party/googletest/src/src/gtest-all.cc
 	@echo "    [CREATE] $@"
-	$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
+	$(qexec)$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
             --lib \
             --target=$(TOOLCHAIN) \
             $(if $(CONFIG_STATIC_MSVCRT),--static-crt) \
@@ -368,27 +380,22 @@ gtest.vcproj: $(SRC_PATH_BARE)/third_party/googletest/src/src/gtest-all.cc
 
 PROJECTS-$(CONFIG_MSVS) += gtest.vcproj
 
-define unit_test_vcproj_template
-$(notdir $(1:.cc=.vcproj)): $(SRC_PATH_BARE)/$(1)
-	@echo "    [vcproj] $$@"
-	$$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh\
-            --exe\
-            --target=$$(TOOLCHAIN)\
-            --name=$(notdir $(1:.cc=))\
-            --ver=$$(CONFIG_VS_VERSION)\
-            $$(if $$(CONFIG_STATIC_MSVCRT),--static-crt) \
-            --out=$$@ $$(INTERNAL_CFLAGS) $$(CFLAGS) \
+test_libvpx.vcproj: $(LIBVPX_TEST_SRCS)
+	@echo "    [CREATE] $@"
+	$(qexec)$(SRC_PATH_BARE)/build/make/gen_msvs_proj.sh \
+            --exe \
+            --target=$(TOOLCHAIN) \
+            --name=test_libvpx \
+            --proj-guid=CD837F5F-52D8-4314-A370-895D614166A7 \
+            --ver=$(CONFIG_VS_VERSION) \
+            $(if $(CONFIG_STATIC_MSVCRT),--static-crt) \
+            --out=$@ $(INTERNAL_CFLAGS) $(CFLAGS) \
             -I. -I"$(SRC_PATH_BARE)/third_party/googletest/src/include" \
-            -L. -lvpxmt -lwinmm -lgtestmt $$^
-endef
+            -L. -l$(CODEC_LIB) -lwinmm -l$(GTEST_LIB) $^
 
-$(foreach proj,$(LIBVPX_TEST_BINS),\
-    $(eval $(call unit_test_vcproj_template,$(proj))))
+PROJECTS-$(CONFIG_MSVS) += test_libvpx.vcproj
 
-PROJECTS-$(CONFIG_MSVS) += $(foreach proj,$(LIBVPX_TEST_BINS),\
-     $(notdir $(proj:.cc=.vcproj)))
-
-test::
+test:: testdata
 	@set -e; for t in $(addprefix Win32/Release/,$(notdir $(LIBVPX_TEST_BINS:.cc=.exe))); do $$t; done
 endif
 else
@@ -396,28 +403,35 @@ else
 include $(SRC_PATH_BARE)/third_party/googletest/gtest.mk
 GTEST_SRCS := $(addprefix third_party/googletest/src/,$(call enabled,GTEST_SRCS))
 GTEST_OBJS=$(call objs,$(GTEST_SRCS))
-$(GTEST_OBJS) $(GTEST_OBJS:.o=.d): CFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src
-$(GTEST_OBJS) $(GTEST_OBJS:.o=.d): CFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src/include
+$(GTEST_OBJS) $(GTEST_OBJS:.o=.d): CXXFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src
+$(GTEST_OBJS) $(GTEST_OBJS:.o=.d): CXXFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src/include
 OBJS-$(BUILD_LIBVPX) += $(GTEST_OBJS)
 LIBS-$(BUILD_LIBVPX) += $(BUILD_PFX)libgtest.a $(BUILD_PFX)libgtest_g.a
 $(BUILD_PFX)libgtest_g.a: $(GTEST_OBJS)
 
-LIBVPX_TEST_SRCS=$(filter %_test.cc,$(call enabled,CODEC_SRCS))
-LIBVPX_TEST_OBJS=$(call objs,$(LIBVPX_TEST_SRCS))
-$(LIBVPX_TEST_OBJS) $(LIBVPX_TEST_OBJS:.o=.d): CFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src
-$(LIBVPX_TEST_OBJS) $(LIBVPX_TEST_OBJS:.o=.d): CFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src/include
-LIBVPX_TEST_BINS=$(sort $(LIBVPX_TEST_OBJS:.cc.o=))
+LIBVPX_TEST_OBJS=$(sort $(call objs,$(LIBVPX_TEST_SRCS)))
+$(LIBVPX_TEST_OBJS) $(LIBVPX_TEST_OBJS:.o=.d): CXXFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src
+$(LIBVPX_TEST_OBJS) $(LIBVPX_TEST_OBJS:.o=.d): CXXFLAGS += -I$(SRC_PATH_BARE)/third_party/googletest/src/include
 OBJS-$(BUILD_LIBVPX) += $(LIBVPX_TEST_OBJS)
+BINS-$(BUILD_LIBVPX) += $(LIBVPX_TEST_BINS)
 
+# Install test sources only if codec source is included
+INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += $(patsubst $(SRC_PATH_BARE)/%,%,\
+    $(shell find $(SRC_PATH_BARE)/third_party/googletest -type f))
+INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += $(LIBVPX_TEST_SRCS)
+
+CODEC_LIB=$(if $(CONFIG_DEBUG_LIBS),vpx_g,vpx)
+CODEC_LIB_SUF=$(if $(CONFIG_SHARED),.so,.a)
 $(foreach bin,$(LIBVPX_TEST_BINS),\
-    $(if $(BUILD_LIBVPX),$(eval $(bin): libvpx.a libgtest.a ))\
+    $(if $(BUILD_LIBVPX),$(eval $(bin): \
+        lib$(CODEC_LIB)$(CODEC_LIB_SUF) libgtest.a ))\
     $(if $(BUILD_LIBVPX),$(eval $(call linkerxx_template,$(bin),\
-        $(bin).cc.o \
+        $(LIBVPX_TEST_OBJS) \
         -L. -lvpx -lgtest -lpthread -lm)\
         )))\
     $(if $(LIPO_LIBS),$(eval $(call lipo_bin_template,$(bin))))\
 
-test:: $(LIBVPX_TEST_BINS)
+test:: $(LIBVPX_TEST_BINS) testdata
 	@set -e; for t in $(LIBVPX_TEST_BINS); do $$t; done
 
 endif
@@ -435,3 +449,6 @@ libs.doxy: $(CODEC_DOC_SRCS)
 	@echo "PREDEFINED = VPX_CODEC_DISABLE_COMPAT" >> $@
 	@echo "INCLUDE_PATH += ." >> $@;
 	@echo "ENABLED_SECTIONS += $(sort $(CODEC_DOC_SECTIONS))" >> $@
+
+## Generate vpx_rtcd.h for all objects
+$(OBJS-yes:.o=.d): $(BUILD_PFX)vpx_rtcd.h

nestegg/src/nestegg.c

@@ -1272,7 +1272,7 @@ ne_read_block(nestegg * ctx, uint64_t block_id, uint64_t block_size, nestegg_pac
   if (total > block_size)
     return -1;
 
-  entry = ne_find_track_entry(ctx, track - 1);
+  entry = ne_find_track_entry(ctx, (unsigned int)(track - 1));
   if (!entry)
     return -1;
 
@@ -1291,7 +1291,7 @@ ne_read_block(nestegg * ctx, uint64_t block_id, uint64_t block_size, nestegg_pac
 
   pkt = ne_alloc(sizeof(*pkt));
   pkt->track = track - 1;
-  pkt->timecode = abs_timecode * tc_scale * track_scale;
+  pkt->timecode = (uint64_t)(abs_timecode * tc_scale * track_scale);
 
   ctx->log(ctx, NESTEGG_LOG_DEBUG, "%sblock t %lld pts %f f %llx frames: %llu",
            block_id == ID_BLOCK ? "" : "simple", pkt->track, pkt->timecode / 1e9, flags, frames);
@@ -1774,35 +1774,35 @@ nestegg_track_video_params(nestegg * ctx, unsigned int track,
 
   if (ne_get_uint(entry->video.pixel_width, &value) != 0)
     return -1;
-  params->width = value;
+  params->width = (unsigned int)value;
 
   if (ne_get_uint(entry->video.pixel_height, &value) != 0)
     return -1;
-  params->height = value;
+  params->height = (unsigned int)value;
 
   value = 0;
   ne_get_uint(entry->video.pixel_crop_bottom, &value);
-  params->crop_bottom = value;
+  params->crop_bottom = (unsigned int)value;
 
   value = 0;
   ne_get_uint(entry->video.pixel_crop_top, &value);
-  params->crop_top = value;
+  params->crop_top = (unsigned int)value;
 
   value = 0;
   ne_get_uint(entry->video.pixel_crop_left, &value);
-  params->crop_left = value;
+  params->crop_left = (unsigned int)value;
 
   value = 0;
   ne_get_uint(entry->video.pixel_crop_right, &value);
-  params->crop_right = value;
+  params->crop_right = (unsigned int)value;
 
   value = params->width;
   ne_get_uint(entry->video.display_width, &value);
-  params->display_width = value;
+  params->display_width = (unsigned int)value;
 
   value = params->height;
   ne_get_uint(entry->video.display_height, &value);
-  params->display_height = value;
+  params->display_height = (unsigned int)value;
 
   return 0;
 }
@@ -1828,11 +1828,11 @@ nestegg_track_audio_params(nestegg * ctx, unsigned int track,
 
   value = 1;
   ne_get_uint(entry->audio.channels, &value);
-  params->channels = value;
+  params->channels = (unsigned int)value;
 
   value = 16;
   ne_get_uint(entry->audio.bit_depth, &value);
-  params->depth = value;
+  params->depth = (unsigned int)value;
 
   return 0;
 }
@@ -1888,7 +1888,7 @@ nestegg_free_packet(nestegg_packet * pkt)
 int
 nestegg_packet_track(nestegg_packet * pkt, unsigned int * track)
 {
-  *track = pkt->track;
+  *track = (unsigned int)pkt->track;
   return 0;
 }
 

solution.mk

@@ -8,18 +8,19 @@
 ##  be found in the AUTHORS file in the root of the source tree.
 ##
 
+# libvpx reverse dependencies (targets that depend on libvpx)
+VPX_NONDEPS=$(addsuffix .vcproj,vpx gtest obj_int_extract)
+VPX_RDEPS=$(foreach vcp,\
+              $(filter-out $(VPX_NONDEPS),$^), --dep=$(vcp:.vcproj=):vpx)
 
 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 %gtest.vcproj %obj_int_extract.vcproj,$^),\
-                  --dep=$(vcp:.vcproj=):vpx) \
-                $(foreach vcp,$(filter %_test.vcproj,$^),\
-                  --dep=$(vcp:.vcproj=):gtest)) \
-                  --dep=vpx:obj_int_extract \
-                  --ver=$(CONFIG_VS_VERSION)\
-                  --out=$@ $^
+            $(if $(filter vpx.vcproj,$^),$(VPX_RDEPS)) \
+            --dep=vpx:obj_int_extract \
+            --dep=test_libvpx:gtest \
+            --ver=$(CONFIG_VS_VERSION)\
+            --out=$@ $^
 vpx.sln.mk: vpx.sln
 	@true
 

test/acm_random.h

@@ -0,0 +1,53 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+#ifndef LIBVPX_TEST_ACM_RANDOM_H_
+#define LIBVPX_TEST_ACM_RANDOM_H_
+
+#include <stdlib.h>
+
+#include "vpx/vpx_integer.h"
+
+namespace libvpx_test {
+
+class ACMRandom {
+ public:
+  ACMRandom() {
+    Reset(DeterministicSeed());
+  }
+
+  explicit ACMRandom(int seed) {
+    Reset(seed);
+  }
+
+  void Reset(int seed) {
+    srand(seed);
+  }
+
+  uint8_t Rand8(void) {
+    return (rand() >> 8) & 0xff;
+  }
+
+  int PseudoUniform(int range) {
+    return (rand() >> 8) % range;
+  }
+
+  int operator()(int n) {
+    return PseudoUniform(n);
+  }
+
+  static int DeterministicSeed(void) {
+    return 0xbaba;
+  }
+};
+
+}  // namespace libvpx_test
+
+#endif  // LIBVPX_TEST_ACM_RANDOM_H_

test/altref_test.cc

@@ -0,0 +1,71 @@
+/*
+ *  Copyright (c) 2012 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 "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+
+namespace {
+
+// lookahead range: [kLookAheadMin, kLookAheadMax).
+const int kLookAheadMin = 5;
+const int kLookAheadMax = 26;
+
+class AltRefTest : public libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<int> {
+ protected:
+  AltRefTest() : altref_count_(0) {}
+  virtual ~AltRefTest() {}
+
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(libvpx_test::kTwoPassGood);
+  }
+
+  virtual void BeginPassHook(unsigned int pass) {
+    altref_count_ = 0;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
+                                  libvpx_test::Encoder *encoder) {
+    if (video->frame() == 1) {
+      encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
+      encoder->Control(VP8E_SET_CPUUSED, 3);
+    }
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    if (pkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE) ++altref_count_;
+  }
+
+  int altref_count() const { return altref_count_; }
+
+ private:
+  int altref_count_;
+};
+
+TEST_P(AltRefTest, MonotonicTimestamps) {
+  const vpx_rational timebase = { 33333333, 1000000000 };
+  cfg_.g_timebase = timebase;
+  cfg_.rc_target_bitrate = 1000;
+  cfg_.g_lag_in_frames = GetParam();
+
+  libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                     timebase.den, timebase.num, 0, 30);
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  EXPECT_GE(altref_count(), 1);
+}
+
+INSTANTIATE_TEST_CASE_P(NonZeroLag, AltRefTest,
+                        ::testing::Range(kLookAheadMin, kLookAheadMax));
+}  // namespace

test/boolcoder_test.cc

@@ -0,0 +1,90 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+extern "C" {
+#include "vp8/encoder/boolhuff.h"
+#include "vp8/decoder/dboolhuff.h"
+}
+
+#include <math.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+
+#include "test/acm_random.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "vpx/vpx_integer.h"
+
+namespace {
+const int num_tests = 10;
+}  // namespace
+
+using libvpx_test::ACMRandom;
+
+TEST(VP8, TestBitIO) {
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  for (int n = 0; n < num_tests; ++n) {
+    for (int method = 0; method <= 7; ++method) {   // we generate various proba
+      const int bits_to_test = 1000;
+      uint8_t probas[bits_to_test];
+
+      for (int i = 0; i < bits_to_test; ++i) {
+        const int parity = i & 1;
+        probas[i] =
+            (method == 0) ? 0 : (method == 1) ? 255 :
+            (method == 2) ? 128 :
+            (method == 3) ? rnd.Rand8() :
+            (method == 4) ? (parity ? 0 : 255) :
+            // alternate between low and high proba:
+            (method == 5) ? (parity ? rnd(128) : 255 - rnd(128)) :
+            (method == 6) ?
+                (parity ? rnd(64) : 255 - rnd(64)) :
+                (parity ? rnd(32) : 255 - rnd(32));
+      }
+      for (int bit_method = 0; bit_method <= 3; ++bit_method) {
+        const int random_seed = 6432;
+        const int buffer_size = 10000;
+        ACMRandom bit_rnd(random_seed);
+        BOOL_CODER bw;
+        uint8_t bw_buffer[buffer_size];
+        vp8_start_encode(&bw, bw_buffer, bw_buffer + buffer_size);
+
+        int bit = (bit_method == 0) ? 0 : (bit_method == 1) ? 1 : 0;
+        for (int i = 0; i < bits_to_test; ++i) {
+          if (bit_method == 2) {
+            bit = (i & 1);
+          } else if (bit_method == 3) {
+            bit = bit_rnd(2);
+          }
+          vp8_encode_bool(&bw, bit, static_cast<int>(probas[i]));
+        }
+
+        vp8_stop_encode(&bw);
+
+        BOOL_DECODER br;
+        vp8dx_start_decode(&br, bw_buffer, buffer_size);
+        bit_rnd.Reset(random_seed);
+        for (int i = 0; i < bits_to_test; ++i) {
+          if (bit_method == 2) {
+            bit = (i & 1);
+          } else if (bit_method == 3) {
+            bit = bit_rnd(2);
+          }
+          GTEST_ASSERT_EQ(vp8dx_decode_bool(&br, probas[i]), bit)
+              << "pos: "<< i << " / " << bits_to_test
+              << " bit_method: " << bit_method
+              << " method: " << method;
+        }
+      }
+    }
+  }
+}

test/config_test.cc

@@ -0,0 +1,61 @@
+/*
+ *  Copyright (c) 2012 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 "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/encode_test_driver.h"
+#include "test/video_source.h"
+
+namespace {
+
+class ConfigTest : public ::libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<enum libvpx_test::TestMode> {
+ public:
+  ConfigTest() : frame_count_in_(0), frame_count_out_(0), frame_count_max_(0) {}
+
+ protected:
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(GetParam());
+  }
+
+  virtual void BeginPassHook(unsigned int /*pass*/) {
+    frame_count_in_ = 0;
+    frame_count_out_ = 0;
+  }
+
+  virtual void PreEncodeFrameHook(libvpx_test::VideoSource* /*video*/) {
+    ++frame_count_in_;
+    abort_ |= (frame_count_in_ >= frame_count_max_);
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t* /*pkt*/) {
+    ++frame_count_out_;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  unsigned int frame_count_in_;
+  unsigned int frame_count_out_;
+  unsigned int frame_count_max_;
+};
+
+TEST_P(ConfigTest, LagIsDisabled) {
+  frame_count_max_ = 2;
+  cfg_.g_lag_in_frames = 15;
+
+  libvpx_test::DummyVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  EXPECT_EQ(frame_count_in_, frame_count_out_);
+}
+
+INSTANTIATE_TEST_CASE_P(OnePassModes, ConfigTest, ONE_PASS_TEST_MODES);
+}  // namespace

test/cq_test.cc

@@ -0,0 +1,106 @@
+/*
+ *  Copyright (c) 2012 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 <cmath>
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+
+// CQ level range: [kCQLevelMin, kCQLevelMax).
+const int kCQLevelMin = 4;
+const int kCQLevelMax = 63;
+const int kCQLevelStep = 8;
+const int kCQTargetBitrate = 2000;
+
+namespace {
+
+class CQTest : public libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<int> {
+ protected:
+  CQTest() : cq_level_(GetParam()) { init_flags_ = VPX_CODEC_USE_PSNR; }
+  virtual ~CQTest() {}
+
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(libvpx_test::kTwoPassGood);
+  }
+
+  virtual void BeginPassHook(unsigned int /*pass*/) {
+    file_size_ = 0;
+    psnr_ = 0.0;
+    n_frames_ = 0;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
+                                  libvpx_test::Encoder *encoder) {
+    if (video->frame() == 1) {
+      if (cfg_.rc_end_usage == VPX_CQ) {
+        encoder->Control(VP8E_SET_CQ_LEVEL, cq_level_);
+      }
+      encoder->Control(VP8E_SET_CPUUSED, 3);
+    }
+  }
+
+  virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
+    psnr_ += pow(10.0, pkt->data.psnr.psnr[0] / 10.0);
+    n_frames_++;
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    file_size_ += pkt->data.frame.sz;
+  }
+
+  double GetLinearPSNROverBitrate() const {
+    double avg_psnr = log10(psnr_ / n_frames_) * 10.0;
+    return pow(10.0, avg_psnr / 10.0) / file_size_;
+  }
+
+  int file_size() const { return file_size_; }
+  int n_frames() const { return n_frames_; }
+
+ private:
+  int cq_level_;
+  int file_size_;
+  double psnr_;
+  int n_frames_;
+};
+
+int prev_actual_bitrate = kCQTargetBitrate;
+TEST_P(CQTest, LinearPSNRIsHigherForCQLevel) {
+  const vpx_rational timebase = { 33333333, 1000000000 };
+  cfg_.g_timebase = timebase;
+  cfg_.rc_target_bitrate = kCQTargetBitrate;
+  cfg_.g_lag_in_frames = 25;
+
+  cfg_.rc_end_usage = VPX_CQ;
+  libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                     timebase.den, timebase.num, 0, 30);
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  const double cq_psnr_lin = GetLinearPSNROverBitrate();
+  const int cq_actual_bitrate = file_size() * 8 * 30 / (n_frames() * 1000);
+  EXPECT_LE(cq_actual_bitrate, kCQTargetBitrate);
+  EXPECT_LE(cq_actual_bitrate, prev_actual_bitrate);
+  prev_actual_bitrate = cq_actual_bitrate;
+
+  // try targeting the approximate same bitrate with VBR mode
+  cfg_.rc_end_usage = VPX_VBR;
+  cfg_.rc_target_bitrate = cq_actual_bitrate;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  const double vbr_psnr_lin = GetLinearPSNROverBitrate();
+  EXPECT_GE(cq_psnr_lin, vbr_psnr_lin);
+}
+
+INSTANTIATE_TEST_CASE_P(CQLevelRange, CQTest,
+                        ::testing::Range(kCQLevelMin, kCQLevelMax,
+                                         kCQLevelStep));
+}  // namespace

test/datarate_test.cc

@@ -0,0 +1,178 @@
+/*
+ *  Copyright (c) 2012 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 "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+namespace {
+
+class DatarateTest : public ::libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<enum libvpx_test::TestMode> {
+ protected:
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(GetParam());
+    ResetModel();
+  }
+
+  virtual void ResetModel() {
+    last_pts_ = 0;
+    bits_in_buffer_model_ = cfg_.rc_target_bitrate * cfg_.rc_buf_initial_sz;
+    frame_number_ = 0;
+    first_drop_ = 0;
+    bits_total_ = 0;
+    duration_ = 0.0;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
+                                  ::libvpx_test::Encoder *encoder) {
+    const vpx_rational_t tb = video->timebase();
+    timebase_ = static_cast<double>(tb.num) / tb.den;
+    duration_ = 0;
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    // Time since last timestamp = duration.
+    vpx_codec_pts_t duration = pkt->data.frame.pts - last_pts_;
+
+    // TODO(jimbankoski): Remove these lines when the issue:
+    // http://code.google.com/p/webm/issues/detail?id=496 is fixed.
+    // For now the codec assumes buffer starts at starting buffer rate
+    // plus one frame's time.
+    if (last_pts_ == 0)
+      duration = 1;
+
+    // Add to the buffer the bits we'd expect from a constant bitrate server.
+    bits_in_buffer_model_ += duration * timebase_ * cfg_.rc_target_bitrate
+        * 1000;
+
+    /* Test the buffer model here before subtracting the frame. Do so because
+     * the way the leaky bucket model works in libvpx is to allow the buffer to
+     * empty - and then stop showing frames until we've got enough bits to
+     * show one. As noted in comment below (issue 495), this does not currently
+     * apply to key frames. For now exclude key frames in condition below. */
+    bool key_frame = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) ? true: false;
+    if (!key_frame) {
+      ASSERT_GE(bits_in_buffer_model_, 0) << "Buffer Underrun at frame "
+          << pkt->data.frame.pts;
+    }
+
+    const int frame_size_in_bits = pkt->data.frame.sz * 8;
+
+    // Subtract from the buffer the bits associated with a played back frame.
+    bits_in_buffer_model_ -= frame_size_in_bits;
+
+    // Update the running total of bits for end of test datarate checks.
+    bits_total_ += frame_size_in_bits ;
+
+    // If first drop not set and we have a drop set it to this time.
+    if (!first_drop_ && duration > 1)
+      first_drop_ = last_pts_ + 1;
+
+    // Update the most recent pts.
+    last_pts_ = pkt->data.frame.pts;
+
+    // We update this so that we can calculate the datarate minus the last
+    // frame encoded in the file.
+    bits_in_last_frame_ = frame_size_in_bits;
+
+    ++frame_number_;
+  }
+
+  virtual void EndPassHook(void) {
+    if (bits_total_) {
+      const double file_size_in_kb = bits_total_ / 1000;  /* bits per kilobit */
+
+      duration_ = (last_pts_ + 1) * timebase_;
+
+      // Effective file datarate includes the time spent prebuffering.
+      effective_datarate_ = (bits_total_ - bits_in_last_frame_) / 1000.0
+          / (cfg_.rc_buf_initial_sz / 1000.0 + duration_);
+
+      file_datarate_ = file_size_in_kb / duration_;
+    }
+  }
+
+  vpx_codec_pts_t last_pts_;
+  int bits_in_buffer_model_;
+  double timebase_;
+  int frame_number_;
+  vpx_codec_pts_t first_drop_;
+  int64_t bits_total_;
+  double duration_;
+  double file_datarate_;
+  double effective_datarate_;
+  int bits_in_last_frame_;
+};
+
+TEST_P(DatarateTest, BasicBufferModel) {
+  cfg_.rc_buf_initial_sz = 500;
+  cfg_.rc_dropframe_thresh = 1;
+  cfg_.rc_max_quantizer = 56;
+  cfg_.rc_end_usage = VPX_CBR;
+  // 2 pass cbr datarate control has a bug hidden by the small # of
+  // frames selected in this encode. The problem is that even if the buffer is
+  // negative we produce a keyframe on a cutscene. Ignoring datarate
+  // constraints
+  // TODO(jimbankoski): ( Fix when issue
+  // http://code.google.com/p/webm/issues/detail?id=495 is addressed. )
+  ::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                       30, 1, 0, 140);
+
+  // There is an issue for low bitrates in real-time mode, where the
+  // effective_datarate slightly overshoots the target bitrate.
+  // This is same the issue as noted about (#495).
+  // TODO(jimbankoski/marpan): Update test to run for lower bitrates (< 100),
+  // when the issue is resolved.
+  for (int i = 100; i < 800; i += 200) {
+    cfg_.rc_target_bitrate = i;
+    ResetModel();
+    ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+    ASSERT_GE(cfg_.rc_target_bitrate, effective_datarate_)
+        << " The datarate for the file exceeds the target!";
+
+    ASSERT_LE(cfg_.rc_target_bitrate, file_datarate_ * 1.3)
+        << " The datarate for the file missed the target!";
+  }
+}
+
+TEST_P(DatarateTest, ChangingDropFrameThresh) {
+  cfg_.rc_buf_initial_sz = 500;
+  cfg_.rc_max_quantizer = 36;
+  cfg_.rc_end_usage = VPX_CBR;
+  cfg_.rc_target_bitrate = 200;
+  cfg_.kf_mode = VPX_KF_DISABLED;
+
+  const int frame_count = 40;
+  ::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                       30, 1, 0, frame_count);
+
+  // Here we check that the first dropped frame gets earlier and earlier
+  // as the drop frame threshold is increased.
+
+  const int kDropFrameThreshTestStep = 30;
+  vpx_codec_pts_t last_drop = frame_count;
+  for (int i = 1; i < 91; i += kDropFrameThreshTestStep) {
+    cfg_.rc_dropframe_thresh = i;
+    ResetModel();
+    ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+    ASSERT_LE(first_drop_, last_drop)
+        << " The first dropped frame for drop_thresh " << i
+        << " > first dropped frame for drop_thresh "
+        << i - kDropFrameThreshTestStep;
+    last_drop = first_drop_;
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(AllModes, DatarateTest, ALL_TEST_MODES);
+}  // namespace

test/decode_test_driver.cc

@@ -0,0 +1,48 @@
+/*
+ *  Copyright (c) 2012 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 "test/decode_test_driver.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/register_state_check.h"
+#include "test/video_source.h"
+
+namespace libvpx_test {
+#if CONFIG_VP8_DECODER
+void Decoder::DecodeFrame(const uint8_t *cxdata, int size) {
+  if (!decoder_.priv) {
+    const vpx_codec_err_t res_init = vpx_codec_dec_init(&decoder_,
+                                                        &vpx_codec_vp8_dx_algo,
+                                                        &cfg_, 0);
+    ASSERT_EQ(VPX_CODEC_OK, res_init) << DecodeError();
+  }
+
+  vpx_codec_err_t res_dec;
+  REGISTER_STATE_CHECK(res_dec = vpx_codec_decode(&decoder_,
+                                                  cxdata, size, NULL, 0));
+  ASSERT_EQ(VPX_CODEC_OK, res_dec) << DecodeError();
+}
+
+void DecoderTest::RunLoop(CompressedVideoSource *video) {
+  vpx_codec_dec_cfg_t dec_cfg = {0};
+  Decoder decoder(dec_cfg, 0);
+
+  // Decode frames.
+  for (video->Begin(); video->cxdata(); video->Next()) {
+    decoder.DecodeFrame(video->cxdata(), video->frame_size());
+
+    DxDataIterator dec_iter = decoder.GetDxData();
+    const vpx_image_t *img = NULL;
+
+    // Get decompressed data
+    while ((img = dec_iter.Next()))
+      DecompressedFrameHook(*img, video->frame_number());
+  }
+}
+#endif
+}  // namespace libvpx_test

test/decode_test_driver.h

@@ -0,0 +1,97 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+#ifndef TEST_DECODE_TEST_DRIVER_H_
+#define TEST_DECODE_TEST_DRIVER_H_
+#include <cstring>
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "vpx_config.h"
+#include "vpx/vpx_decoder.h"
+#include "vpx/vp8dx.h"
+
+namespace libvpx_test {
+
+class CompressedVideoSource;
+
+// Provides an object to handle decoding output
+class DxDataIterator {
+ public:
+  explicit DxDataIterator(vpx_codec_ctx_t *decoder)
+      : decoder_(decoder), iter_(NULL) {}
+
+  const vpx_image_t *Next() {
+    return vpx_codec_get_frame(decoder_, &iter_);
+  }
+
+ private:
+  vpx_codec_ctx_t  *decoder_;
+  vpx_codec_iter_t  iter_;
+};
+
+// Provides a simplified interface to manage one video decoding.
+//
+// TODO: similar to Encoder class, the exact services should be
+// added as more tests are added.
+class Decoder {
+ public:
+  Decoder(vpx_codec_dec_cfg_t cfg, unsigned long deadline)
+      : cfg_(cfg), deadline_(deadline) {
+    memset(&decoder_, 0, sizeof(decoder_));
+  }
+
+  ~Decoder() {
+    vpx_codec_destroy(&decoder_);
+  }
+
+  void DecodeFrame(const uint8_t *cxdata, int size);
+
+  DxDataIterator GetDxData() {
+    return DxDataIterator(&decoder_);
+  }
+
+  void set_deadline(unsigned long deadline) {
+    deadline_ = deadline;
+  }
+
+  void Control(int ctrl_id, int arg) {
+    const vpx_codec_err_t res = vpx_codec_control_(&decoder_, ctrl_id, arg);
+    ASSERT_EQ(VPX_CODEC_OK, res) << DecodeError();
+  }
+
+ protected:
+  const char *DecodeError() {
+    const char *detail = vpx_codec_error_detail(&decoder_);
+    return detail ? detail : vpx_codec_error(&decoder_);
+  }
+
+  vpx_codec_ctx_t     decoder_;
+  vpx_codec_dec_cfg_t cfg_;
+  unsigned int        deadline_;
+};
+
+// Common test functionality for all Decoder tests.
+class DecoderTest {
+ public:
+  // Main loop.
+  virtual void RunLoop(CompressedVideoSource *video);
+
+  // Hook to be called on every decompressed frame.
+  virtual void DecompressedFrameHook(const vpx_image_t& img,
+                                     const unsigned int frame_number) {}
+
+ protected:
+  DecoderTest() {}
+
+  virtual ~DecoderTest() {}
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_DECODE_TEST_DRIVER_H_

test/encode_test_driver.cc

@@ -0,0 +1,206 @@
+/*
+ *  Copyright (c) 2012 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_config.h"
+#include "test/encode_test_driver.h"
+#if CONFIG_VP8_DECODER
+#include "test/decode_test_driver.h"
+#endif
+#include "test/register_state_check.h"
+#include "test/video_source.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace libvpx_test {
+void Encoder::EncodeFrame(VideoSource *video, const unsigned long frame_flags) {
+  if (video->img())
+    EncodeFrameInternal(*video, frame_flags);
+  else
+    Flush();
+
+  // Handle twopass stats
+  CxDataIterator iter = GetCxData();
+
+  while (const vpx_codec_cx_pkt_t *pkt = iter.Next()) {
+    if (pkt->kind != VPX_CODEC_STATS_PKT)
+      continue;
+
+    stats_->Append(*pkt);
+  }
+}
+
+void Encoder::EncodeFrameInternal(const VideoSource &video,
+                                  const unsigned long frame_flags) {
+  vpx_codec_err_t res;
+  const vpx_image_t *img = video.img();
+
+  // Handle first frame initialization
+  if (!encoder_.priv) {
+    cfg_.g_w = img->d_w;
+    cfg_.g_h = img->d_h;
+    cfg_.g_timebase = video.timebase();
+    cfg_.rc_twopass_stats_in = stats_->buf();
+    res = vpx_codec_enc_init(&encoder_, &vpx_codec_vp8_cx_algo, &cfg_,
+                             init_flags_);
+    ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+  }
+
+  // Handle frame resizing
+  if (cfg_.g_w != img->d_w || cfg_.g_h != img->d_h) {
+    cfg_.g_w = img->d_w;
+    cfg_.g_h = img->d_h;
+    res = vpx_codec_enc_config_set(&encoder_, &cfg_);
+    ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+  }
+
+  // Encode the frame
+  REGISTER_STATE_CHECK(
+      res = vpx_codec_encode(&encoder_,
+                             video.img(), video.pts(), video.duration(),
+                             frame_flags, deadline_));
+  ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+}
+
+void Encoder::Flush() {
+  const vpx_codec_err_t res = vpx_codec_encode(&encoder_, NULL, 0, 0, 0,
+                                               deadline_);
+  ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+}
+
+void EncoderTest::SetMode(TestMode mode) {
+  switch (mode) {
+    case kRealTime:
+      deadline_ = VPX_DL_REALTIME;
+      break;
+
+    case kOnePassGood:
+    case kTwoPassGood:
+      deadline_ = VPX_DL_GOOD_QUALITY;
+      break;
+
+    case kOnePassBest:
+    case kTwoPassBest:
+      deadline_ = VPX_DL_BEST_QUALITY;
+      break;
+
+    default:
+      ASSERT_TRUE(false) << "Unexpected mode " << mode;
+  }
+
+  if (mode == kTwoPassGood || mode == kTwoPassBest)
+    passes_ = 2;
+  else
+    passes_ = 1;
+}
+// The function should return "true" most of the time, therefore no early
+// break-out is implemented within the match checking process.
+static bool compare_img(const vpx_image_t *img1,
+                        const vpx_image_t *img2) {
+  bool match = (img1->fmt == img2->fmt) &&
+               (img1->d_w == img2->d_w) &&
+               (img1->d_h == img2->d_h);
+
+  const unsigned int width_y  = img1->d_w;
+  const unsigned int height_y = img1->d_h;
+  unsigned int i;
+  for (i = 0; i < height_y; ++i)
+    match = ( memcmp(img1->planes[VPX_PLANE_Y] + i * img1->stride[VPX_PLANE_Y],
+                     img2->planes[VPX_PLANE_Y] + i * img2->stride[VPX_PLANE_Y],
+                     width_y) == 0) && match;
+  const unsigned int width_uv  = (img1->d_w + 1) >> 1;
+  const unsigned int height_uv = (img1->d_h + 1) >> 1;
+  for (i = 0; i <  height_uv; ++i)
+    match = ( memcmp(img1->planes[VPX_PLANE_U] + i * img1->stride[VPX_PLANE_U],
+                     img2->planes[VPX_PLANE_U] + i * img2->stride[VPX_PLANE_U],
+                     width_uv) == 0) && match;
+  for (i = 0; i < height_uv; ++i)
+    match = ( memcmp(img1->planes[VPX_PLANE_V] + i * img1->stride[VPX_PLANE_V],
+                     img2->planes[VPX_PLANE_V] + i * img2->stride[VPX_PLANE_V],
+                     width_uv) == 0) && match;
+  return match;
+}
+
+void EncoderTest::RunLoop(VideoSource *video) {
+#if CONFIG_VP8_DECODER
+  vpx_codec_dec_cfg_t dec_cfg = {0};
+#endif
+
+  stats_.Reset();
+
+  for (unsigned int pass = 0; pass < passes_; pass++) {
+    last_pts_ = 0;
+
+    if (passes_ == 1)
+      cfg_.g_pass = VPX_RC_ONE_PASS;
+    else if (pass == 0)
+      cfg_.g_pass = VPX_RC_FIRST_PASS;
+    else
+      cfg_.g_pass = VPX_RC_LAST_PASS;
+
+    BeginPassHook(pass);
+    Encoder encoder(cfg_, deadline_, init_flags_, &stats_);
+#if CONFIG_VP8_DECODER
+    Decoder decoder(dec_cfg, 0);
+    bool has_cxdata = false;
+#endif
+    bool again;
+    for (again = true, video->Begin(); again; video->Next()) {
+      again = video->img() != NULL;
+
+      PreEncodeFrameHook(video);
+      PreEncodeFrameHook(video, &encoder);
+      encoder.EncodeFrame(video, frame_flags_);
+
+      CxDataIterator iter = encoder.GetCxData();
+
+      while (const vpx_codec_cx_pkt_t *pkt = iter.Next()) {
+        again = true;
+
+        switch (pkt->kind) {
+          case VPX_CODEC_CX_FRAME_PKT:
+#if CONFIG_VP8_DECODER
+            has_cxdata = true;
+            decoder.DecodeFrame((const uint8_t*)pkt->data.frame.buf,
+                                pkt->data.frame.sz);
+#endif
+            ASSERT_GE(pkt->data.frame.pts, last_pts_);
+            last_pts_ = pkt->data.frame.pts;
+            FramePktHook(pkt);
+            break;
+
+          case VPX_CODEC_PSNR_PKT:
+            PSNRPktHook(pkt);
+            break;
+
+          default:
+            break;
+        }
+      }
+
+#if CONFIG_VP8_DECODER
+      if (has_cxdata) {
+        const vpx_image_t *img_enc = encoder.GetPreviewFrame();
+        DxDataIterator dec_iter = decoder.GetDxData();
+        const vpx_image_t *img_dec = dec_iter.Next();
+        if(img_enc && img_dec) {
+          const bool res = compare_img(img_enc, img_dec);
+          ASSERT_TRUE(res)<< "Encoder/Decoder mismatch found.";
+        }
+      }
+#endif
+      if (!Continue())
+        break;
+    }
+
+    EndPassHook();
+
+    if (!Continue())
+      break;
+  }
+}
+}  // namespace libvpx_test

test/encode_test_driver.h

@@ -0,0 +1,197 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+#ifndef TEST_ENCODE_TEST_DRIVER_H_
+#define TEST_ENCODE_TEST_DRIVER_H_
+#include <string>
+#include <vector>
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "vpx/vpx_encoder.h"
+#include "vpx/vp8cx.h"
+
+namespace libvpx_test {
+
+class VideoSource;
+
+enum TestMode {
+  kRealTime,
+  kOnePassGood,
+  kOnePassBest,
+  kTwoPassGood,
+  kTwoPassBest
+};
+#define ALL_TEST_MODES ::testing::Values(::libvpx_test::kRealTime, \
+                                         ::libvpx_test::kOnePassGood, \
+                                         ::libvpx_test::kOnePassBest, \
+                                         ::libvpx_test::kTwoPassGood, \
+                                         ::libvpx_test::kTwoPassBest)
+
+#define ONE_PASS_TEST_MODES ::testing::Values(::libvpx_test::kRealTime, \
+                                              ::libvpx_test::kOnePassGood, \
+                                              ::libvpx_test::kOnePassBest)
+
+
+// Provides an object to handle the libvpx get_cx_data() iteration pattern
+class CxDataIterator {
+ public:
+  explicit CxDataIterator(vpx_codec_ctx_t *encoder)
+    : encoder_(encoder), iter_(NULL) {}
+
+  const vpx_codec_cx_pkt_t *Next() {
+    return vpx_codec_get_cx_data(encoder_, &iter_);
+  }
+
+ private:
+  vpx_codec_ctx_t  *encoder_;
+  vpx_codec_iter_t  iter_;
+};
+
+// Implements an in-memory store for libvpx twopass statistics
+class TwopassStatsStore {
+ public:
+  void Append(const vpx_codec_cx_pkt_t &pkt) {
+    buffer_.append(reinterpret_cast<char *>(pkt.data.twopass_stats.buf),
+                   pkt.data.twopass_stats.sz);
+  }
+
+  vpx_fixed_buf_t buf() {
+    const vpx_fixed_buf_t buf = { &buffer_[0], buffer_.size() };
+    return buf;
+  }
+
+  void Reset() {
+    buffer_.clear();
+  }
+
+ protected:
+  std::string  buffer_;
+};
+
+
+// Provides a simplified interface to manage one video encoding pass, given
+// a configuration and video source.
+//
+// TODO(jkoleszar): The exact services it provides and the appropriate
+// level of abstraction will be fleshed out as more tests are written.
+class Encoder {
+ public:
+  Encoder(vpx_codec_enc_cfg_t cfg, unsigned long deadline,
+          const unsigned long init_flags, TwopassStatsStore *stats)
+    : cfg_(cfg), deadline_(deadline), init_flags_(init_flags), stats_(stats) {
+    memset(&encoder_, 0, sizeof(encoder_));
+  }
+
+  ~Encoder() {
+    vpx_codec_destroy(&encoder_);
+  }
+
+  CxDataIterator GetCxData() {
+    return CxDataIterator(&encoder_);
+  }
+
+  const vpx_image_t *GetPreviewFrame() {
+    return vpx_codec_get_preview_frame(&encoder_);
+  }
+  // This is a thin wrapper around vpx_codec_encode(), so refer to
+  // vpx_encoder.h for its semantics.
+  void EncodeFrame(VideoSource *video, const unsigned long frame_flags);
+
+  // Convenience wrapper for EncodeFrame()
+  void EncodeFrame(VideoSource *video) {
+    EncodeFrame(video, 0);
+  }
+
+  void Control(int ctrl_id, int arg) {
+    const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
+    ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
+  }
+
+  void set_deadline(unsigned long deadline) {
+    deadline_ = deadline;
+  }
+
+ protected:
+  const char *EncoderError() {
+    const char *detail = vpx_codec_error_detail(&encoder_);
+    return detail ? detail : vpx_codec_error(&encoder_);
+  }
+
+  // Encode an image
+  void EncodeFrameInternal(const VideoSource &video,
+                           const unsigned long frame_flags);
+
+  // Flush the encoder on EOS
+  void Flush();
+
+  vpx_codec_ctx_t      encoder_;
+  vpx_codec_enc_cfg_t  cfg_;
+  unsigned long        deadline_;
+  unsigned long        init_flags_;
+  TwopassStatsStore   *stats_;
+};
+
+// Common test functionality for all Encoder tests.
+//
+// This class is a mixin which provides the main loop common to all
+// encoder tests. It provides hooks which can be overridden by subclasses
+// to implement each test's specific behavior, while centralizing the bulk
+// of the boilerplate. Note that it doesn't inherit the gtest testing
+// classes directly, so that tests can be parameterized differently.
+class EncoderTest {
+ protected:
+  EncoderTest() : abort_(false), init_flags_(0), frame_flags_(0),
+                  last_pts_(0) {}
+
+  virtual ~EncoderTest() {}
+
+  // Initialize the cfg_ member with the default configuration.
+  void InitializeConfig() {
+    const vpx_codec_err_t res = vpx_codec_enc_config_default(
+                                    &vpx_codec_vp8_cx_algo, &cfg_, 0);
+    ASSERT_EQ(VPX_CODEC_OK, res);
+  }
+
+  // Map the TestMode enum to the deadline_ and passes_ variables.
+  void SetMode(TestMode mode);
+
+  // Main loop.
+  virtual void RunLoop(VideoSource *video);
+
+  // Hook to be called at the beginning of a pass.
+  virtual void BeginPassHook(unsigned int pass) {}
+
+  // Hook to be called at the end of a pass.
+  virtual void EndPassHook() {}
+
+  // Hook to be called before encoding a frame.
+  virtual void PreEncodeFrameHook(VideoSource *video) {}
+  virtual void PreEncodeFrameHook(VideoSource *video, Encoder *encoder) {}
+
+  // Hook to be called on every compressed data packet.
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {}
+
+  // Hook to be called on every PSNR packet.
+  virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {}
+
+  // Hook to determine whether the encode loop should continue.
+  virtual bool Continue() const { return !abort_; }
+
+  bool                 abort_;
+  vpx_codec_enc_cfg_t  cfg_;
+  unsigned int         passes_;
+  unsigned long        deadline_;
+  TwopassStatsStore    stats_;
+  unsigned long        init_flags_;
+  unsigned long        frame_flags_;
+  vpx_codec_pts_t      last_pts_;
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_ENCODE_TEST_DRIVER_H_

test/error_resilience_test.cc

@@ -0,0 +1,90 @@
+/*
+  Copyright (c) 2012 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 "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+
+namespace {
+
+class ErrorResilienceTest : public libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<int> {
+ protected:
+  ErrorResilienceTest() {
+    psnr_ = 0.0;
+    nframes_ = 0;
+    encoding_mode_ = static_cast<libvpx_test::TestMode>(GetParam());
+  }
+  virtual ~ErrorResilienceTest() {}
+
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(encoding_mode_);
+  }
+
+  virtual void BeginPassHook(unsigned int /*pass*/) {
+    psnr_ = 0.0;
+    nframes_ = 0;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
+    psnr_ += pkt->data.psnr.psnr[0];
+    nframes_++;
+  }
+
+  double GetAveragePsnr() const {
+    if (nframes_)
+      return psnr_ / nframes_;
+    return 0.0;
+  }
+
+ private:
+  double psnr_;
+  unsigned int nframes_;
+  libvpx_test::TestMode encoding_mode_;
+};
+
+TEST_P(ErrorResilienceTest, OnVersusOff) {
+  const vpx_rational timebase = { 33333333, 1000000000 };
+  cfg_.g_timebase = timebase;
+  cfg_.rc_target_bitrate = 2000;
+  cfg_.g_lag_in_frames = 25;
+
+  init_flags_ = VPX_CODEC_USE_PSNR;
+
+  libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                     timebase.den, timebase.num, 0, 30);
+
+  // Error resilient mode OFF.
+  cfg_.g_error_resilient = 0;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  const double psnr_resilience_off = GetAveragePsnr();
+  EXPECT_GT(psnr_resilience_off, 25.0);
+
+  // Error resilient mode ON.
+  cfg_.g_error_resilient = 1;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+  const double psnr_resilience_on = GetAveragePsnr();
+  EXPECT_GT(psnr_resilience_on, 25.0);
+
+  // Test that turning on error resilient mode hurts by 10% at most.
+  if (psnr_resilience_off > 0.0) {
+    const double psnr_ratio = psnr_resilience_on / psnr_resilience_off;
+    EXPECT_GE(psnr_ratio, 0.9);
+    EXPECT_LE(psnr_ratio, 1.1);
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(OnOffTest, ErrorResilienceTest,
+                        ONE_PASS_TEST_MODES);
+}  // namespace

test/fdct4x4_test.cc

@@ -0,0 +1,169 @@
+/*
+*  Copyright (c) 2012 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 <math.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+
+
+extern "C" {
+#include "vpx_rtcd.h"
+}
+
+#include "test/acm_random.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "vpx/vpx_integer.h"
+
+
+namespace {
+
+const int cospi8sqrt2minus1 = 20091;
+const int sinpi8sqrt2 = 35468;
+
+void reference_idct4x4(const int16_t *input, int16_t *output) {
+  const int16_t *ip = input;
+  int16_t *op = output;
+
+  for (int i = 0; i < 4; ++i) {
+    const int a1 = ip[0] + ip[8];
+    const int b1 = ip[0] - ip[8];
+    const int temp1 = (ip[4] * sinpi8sqrt2) >> 16;
+    const int temp2 = ip[12] + ((ip[12] * cospi8sqrt2minus1) >> 16);
+    const int c1 = temp1 - temp2;
+    const int temp3 = ip[4] + ((ip[4] * cospi8sqrt2minus1) >> 16);
+    const int temp4 = (ip[12] * sinpi8sqrt2) >> 16;
+    const int d1 = temp3 + temp4;
+    op[0] = a1 + d1;
+    op[12] = a1 - d1;
+    op[4] = b1 + c1;
+    op[8] = b1 - c1;
+    ++ip;
+    ++op;
+  }
+  ip = output;
+  op = output;
+  for (int i = 0; i < 4; ++i) {
+    const int a1 = ip[0] + ip[2];
+    const int b1 = ip[0] - ip[2];
+    const int temp1 = (ip[1] * sinpi8sqrt2) >> 16;
+    const int temp2 = ip[3] + ((ip[3] * cospi8sqrt2minus1) >> 16);
+    const int c1 = temp1 - temp2;
+    const int temp3 = ip[1] + ((ip[1] * cospi8sqrt2minus1) >> 16);
+    const int temp4 = (ip[3] * sinpi8sqrt2) >> 16;
+    const int d1 = temp3 + temp4;
+    op[0] = (a1 + d1 + 4) >> 3;
+    op[3] = (a1 - d1 + 4) >> 3;
+    op[1] = (b1 + c1 + 4) >> 3;
+    op[2] = (b1 - c1 + 4) >> 3;
+    ip += 4;
+    op += 4;
+  }
+}
+
+using libvpx_test::ACMRandom;
+
+TEST(Vp8FdctTest, SignBiasCheck) {
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  int16_t test_input_block[16];
+  int16_t test_output_block[16];
+  const int pitch = 8;
+  int count_sign_block[16][2];
+  const int count_test_block = 1000000;
+
+  memset(count_sign_block, 0, sizeof(count_sign_block));
+
+  for (int i = 0; i < count_test_block; ++i) {
+    // Initialize a test block with input range [-255, 255].
+    for (int j = 0; j < 16; ++j)
+      test_input_block[j] = rnd.Rand8() - rnd.Rand8();
+
+    vp8_short_fdct4x4_c(test_input_block, test_output_block, pitch);
+
+    for (int j = 0; j < 16; ++j) {
+      if (test_output_block[j] < 0)
+        ++count_sign_block[j][0];
+      else if (test_output_block[j] > 0)
+        ++count_sign_block[j][1];
+    }
+  }
+
+  bool bias_acceptable = true;
+  for (int j = 0; j < 16; ++j)
+    bias_acceptable = bias_acceptable &&
+    (abs(count_sign_block[j][0] - count_sign_block[j][1]) < 10000);
+
+  EXPECT_EQ(true, bias_acceptable)
+    << "Error: 4x4 FDCT has a sign bias > 1% for input range [-255, 255]";
+
+  memset(count_sign_block, 0, sizeof(count_sign_block));
+
+  for (int i = 0; i < count_test_block; ++i) {
+    // Initialize a test block with input range [-15, 15].
+    for (int j = 0; j < 16; ++j)
+      test_input_block[j] = (rnd.Rand8() >> 4) - (rnd.Rand8() >> 4);
+
+    vp8_short_fdct4x4_c(test_input_block, test_output_block, pitch);
+
+    for (int j = 0; j < 16; ++j) {
+      if (test_output_block[j] < 0)
+        ++count_sign_block[j][0];
+      else if (test_output_block[j] > 0)
+        ++count_sign_block[j][1];
+    }
+  }
+
+  bias_acceptable = true;
+  for (int j = 0; j < 16; ++j)
+    bias_acceptable = bias_acceptable &&
+    (abs(count_sign_block[j][0] - count_sign_block[j][1]) < 100000);
+
+  EXPECT_EQ(true, bias_acceptable)
+    << "Error: 4x4 FDCT has a sign bias > 10% for input range [-15, 15]";
+};
+
+TEST(Vp8FdctTest, RoundTripErrorCheck) {
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  int max_error = 0;
+  double total_error = 0;
+  const int count_test_block = 1000000;
+  for (int i = 0; i < count_test_block; ++i) {
+    int16_t test_input_block[16];
+    int16_t test_temp_block[16];
+    int16_t test_output_block[16];
+
+    // Initialize a test block with input range [-255, 255].
+    for (int j = 0; j < 16; ++j)
+      test_input_block[j] = rnd.Rand8() - rnd.Rand8();
+
+    const int pitch = 8;
+    vp8_short_fdct4x4_c(test_input_block, test_temp_block, pitch);
+    reference_idct4x4(test_temp_block, test_output_block);
+
+    for (int j = 0; j < 16; ++j) {
+      const int diff = test_input_block[j] - test_output_block[j];
+      const int error = diff * diff;
+      if (max_error < error)
+        max_error = error;
+      total_error += error;
+    }
+  }
+
+  EXPECT_GE(1, max_error )
+    << "Error: FDCT/IDCT has an individual roundtrip error > 1";
+
+  EXPECT_GE(count_test_block, total_error)
+    << "Error: FDCT/IDCT has average roundtrip error > 1 per block";
+};
+
+}  // namespace

test/i420_video_source.h

@@ -0,0 +1,117 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+#ifndef TEST_I420_VIDEO_SOURCE_H_
+#define TEST_I420_VIDEO_SOURCE_H_
+#include <cstdio>
+#include <cstdlib>
+
+#include "test/video_source.h"
+
+namespace libvpx_test {
+
+// This class extends VideoSource to allow parsing of raw yv12
+// so that we can do actual file encodes.
+class I420VideoSource : public VideoSource {
+ public:
+  I420VideoSource(const std::string &file_name,
+                  unsigned int width, unsigned int height,
+                  int rate_numerator, int rate_denominator,
+                  unsigned int start, int limit)
+      : file_name_(file_name),
+        input_file_(NULL),
+        img_(NULL),
+        start_(start),
+        limit_(limit),
+        frame_(0),
+        width_(0),
+        height_(0),
+        framerate_numerator_(rate_numerator),
+        framerate_denominator_(rate_denominator) {
+
+    // This initializes raw_sz_, width_, height_ and allocates an img.
+    SetSize(width, height);
+  }
+
+  virtual ~I420VideoSource() {
+    vpx_img_free(img_);
+    if (input_file_)
+      fclose(input_file_);
+  }
+
+  virtual void Begin() {
+    if (input_file_)
+      fclose(input_file_);
+    input_file_ = OpenTestDataFile(file_name_);
+    ASSERT_TRUE(input_file_) << "Input file open failed. Filename: "
+        << file_name_;
+    if (start_) {
+      fseek(input_file_, raw_sz_ * start_, SEEK_SET);
+    }
+
+    frame_ = start_;
+    FillFrame();
+  }
+
+  virtual void Next() {
+    ++frame_;
+    FillFrame();
+  }
+
+  virtual vpx_image_t *img() const { return (frame_ < limit_) ? img_ : NULL;  }
+
+  // Models a stream where Timebase = 1/FPS, so pts == frame.
+  virtual vpx_codec_pts_t pts() const { return frame_; }
+
+  virtual unsigned long duration() const { return 1; }
+
+  virtual vpx_rational_t timebase() const {
+    const vpx_rational_t t = { framerate_denominator_, framerate_numerator_ };
+    return t;
+  }
+
+  virtual unsigned int frame() const { return frame_; }
+
+  virtual unsigned int limit() const { return limit_; }
+
+  void SetSize(unsigned int width, unsigned int height) {
+    if (width != width_ || height != height_) {
+      vpx_img_free(img_);
+      img_ = vpx_img_alloc(NULL, VPX_IMG_FMT_VPXI420, width, height, 1);
+      ASSERT_TRUE(img_ != NULL);
+      width_ = width;
+      height_ = height;
+      raw_sz_ = width * height * 3 / 2;
+    }
+  }
+
+  virtual void FillFrame() {
+    // Read a frame from input_file.
+    if (fread(img_->img_data, raw_sz_, 1, input_file_) == 0) {
+      limit_ = frame_;
+    }
+  }
+
+ protected:
+  std::string file_name_;
+  FILE *input_file_;
+  vpx_image_t *img_;
+  size_t raw_sz_;
+  unsigned int start_;
+  unsigned int limit_;
+  unsigned int frame_;
+  unsigned int width_;
+  unsigned int height_;
+  unsigned int framerate_numerator_;
+  unsigned int framerate_denominator_;
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_I420_VIDEO_SOURCE_H_

test/idctllm_test.cc

@@ -9,10 +9,16 @@
  */
 
 
- #include "third_party/googletest/src/include/gtest/gtest.h"
+extern "C" {
+#include "vpx_config.h"
+#include "vpx_rtcd.h"
+}
+#include "test/register_state_check.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
 typedef void (*idct_fn_t)(short *input, unsigned char *pred_ptr,
-                            int pred_stride, unsigned char *dst_ptr,
-                            int dst_stride);
+                          int pred_stride, unsigned char *dst_ptr,
+                          int dst_stride);
 namespace {
 class IDCTTest : public ::testing::TestWithParam<idct_fn_t>
 {
@@ -49,7 +55,7 @@ TEST_P(IDCTTest, TestAllZeros)
 {
     int i;
 
-    UUT(input, output, 16, output, 16);
+    REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
 
     for(i=0; i<256; i++)
         if((i&0xF) < 4 && i<64)
@@ -63,7 +69,7 @@ TEST_P(IDCTTest, TestAllOnes)
     int i;
 
     input[0] = 4;
-    UUT(input, output, 16, output, 16);
+    REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
 
     for(i=0; i<256; i++)
         if((i&0xF) < 4 && i<64)
@@ -80,7 +86,7 @@ TEST_P(IDCTTest, TestAddOne)
         predict[i] = i;
 
     input[0] = 4;
-    UUT(input, predict, 16, output, 16);
+    REGISTER_STATE_CHECK(UUT(input, predict, 16, output, 16));
 
     for(i=0; i<256; i++)
         if((i&0xF) < 4 && i<64)
@@ -96,7 +102,7 @@ TEST_P(IDCTTest, TestWithData)
     for(i=0; i<16; i++)
         input[i] = i;
 
-    UUT(input, output, 16, output, 16);
+    REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
 
     for(i=0; i<256; i++)
         if((i&0xF) > 3 || i>63)
@@ -110,4 +116,11 @@ TEST_P(IDCTTest, TestWithData)
         else
             EXPECT_EQ(0, output[i]) << "i==" << i;
 }
+
+INSTANTIATE_TEST_CASE_P(C, IDCTTest,
+                        ::testing::Values(vp8_short_idct4x4llm_c));
+#if HAVE_MMX
+INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
+                        ::testing::Values(vp8_short_idct4x4llm_mmx));
+#endif
 }

test/intrapred_test.cc

@@ -0,0 +1,357 @@
+/*
+ *  Copyright (c) 2012 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 <string.h>
+#include "test/acm_random.h"
+#include "test/register_state_check.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+extern "C" {
+#include "vpx_config.h"
+#include "vpx_rtcd.h"
+#include "vp8/common/blockd.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+namespace {
+
+using libvpx_test::ACMRandom;
+
+class IntraPredBase {
+ protected:
+  void SetupMacroblock(uint8_t *data, int block_size, int stride,
+                       int num_planes) {
+    memset(&mb_, 0, sizeof(mb_));
+    memset(&mi_, 0, sizeof(mi_));
+    mb_.up_available = 1;
+    mb_.left_available = 1;
+    mb_.mode_info_context = &mi_;
+    stride_ = stride;
+    block_size_ = block_size;
+    num_planes_ = num_planes;
+    for (int p = 0; p < num_planes; p++)
+      data_ptr_[p] = data + stride * (block_size + 1) * p +
+                     stride + block_size;
+  }
+
+  void FillRandom() {
+    // Fill edges with random data
+    ACMRandom rnd(ACMRandom::DeterministicSeed());
+    for (int p = 0; p < num_planes_; p++) {
+      for (int x = -1 ; x <= block_size_; x++)
+        data_ptr_[p][x - stride_] = rnd.Rand8();
+      for (int y = 0; y < block_size_; y++)
+        data_ptr_[p][y * stride_ - 1] = rnd.Rand8();
+    }
+  }
+
+  virtual void Predict(MB_PREDICTION_MODE mode) = 0;
+
+  void SetLeftUnavailable() {
+    mb_.left_available = 0;
+    for (int p = 0; p < num_planes_; p++)
+      for (int i = -1; i < block_size_; ++i)
+        data_ptr_[p][stride_ * i - 1] = 129;
+  }
+
+  void SetTopUnavailable() {
+    mb_.up_available = 0;
+    for (int p = 0; p < num_planes_; p++)
+      memset(&data_ptr_[p][-1 - stride_], 127, block_size_ + 2);
+  }
+
+  void SetTopLeftUnavailable() {
+    SetLeftUnavailable();
+    SetTopUnavailable();
+  }
+
+  int BlockSizeLog2Min1() const {
+    switch (block_size_) {
+      case 16:
+        return 3;
+      case 8:
+        return 2;
+      default:
+        return 0;
+    }
+  }
+
+  // check DC prediction output against a reference
+  void CheckDCPrediction() const {
+    for (int p = 0; p < num_planes_; p++) {
+      // calculate expected DC
+      int expected;
+      if (mb_.up_available || mb_.left_available) {
+        int sum = 0, shift = BlockSizeLog2Min1() + mb_.up_available +
+                             mb_.left_available;
+        if (mb_.up_available)
+          for (int x = 0; x < block_size_; x++)
+            sum += data_ptr_[p][x - stride_];
+        if (mb_.left_available)
+          for (int y = 0; y < block_size_; y++)
+            sum += data_ptr_[p][y * stride_ - 1];
+        expected = (sum + (1 << (shift - 1))) >> shift;
+      } else
+        expected = 0x80;
+
+      // check that all subsequent lines are equal to the first
+      for (int y = 1; y < block_size_; ++y)
+        ASSERT_EQ(0, memcmp(data_ptr_[p], &data_ptr_[p][y * stride_],
+                            block_size_));
+      // within the first line, ensure that each pixel has the same value
+      for (int x = 1; x < block_size_; ++x)
+        ASSERT_EQ(data_ptr_[p][0], data_ptr_[p][x]);
+      // now ensure that that pixel has the expected (DC) value
+      ASSERT_EQ(expected, data_ptr_[p][0]);
+    }
+  }
+
+  // check V prediction output against a reference
+  void CheckVPrediction() const {
+    // check that all lines equal the top border
+    for (int p = 0; p < num_planes_; p++)
+      for (int y = 0; y < block_size_; y++)
+        ASSERT_EQ(0, memcmp(&data_ptr_[p][-stride_],
+                            &data_ptr_[p][y * stride_], block_size_));
+  }
+
+  // check H prediction output against a reference
+  void CheckHPrediction() const {
+    // for each line, ensure that each pixel is equal to the left border
+    for (int p = 0; p < num_planes_; p++)
+      for (int y = 0; y < block_size_; y++)
+        for (int x = 0; x < block_size_; x++)
+          ASSERT_EQ(data_ptr_[p][-1 + y * stride_],
+                    data_ptr_[p][x + y * stride_]);
+  }
+
+  static int ClipByte(int value) {
+    if (value > 255)
+      return 255;
+    else if (value < 0)
+      return 0;
+    return value;
+  }
+
+  // check TM prediction output against a reference
+  void CheckTMPrediction() const {
+    for (int p = 0; p < num_planes_; p++)
+      for (int y = 0; y < block_size_; y++)
+        for (int x = 0; x < block_size_; x++) {
+          const int expected = ClipByte(data_ptr_[p][x - stride_]
+                                      + data_ptr_[p][stride_ * y - 1]
+                                      - data_ptr_[p][-1 - stride_]);
+          ASSERT_EQ(expected, data_ptr_[p][y * stride_ + x]);
+       }
+  }
+
+  // Actual test
+  void RunTest() {
+    {
+      SCOPED_TRACE("DC_PRED");
+      FillRandom();
+      Predict(DC_PRED);
+      CheckDCPrediction();
+    }
+    {
+      SCOPED_TRACE("DC_PRED LEFT");
+      FillRandom();
+      SetLeftUnavailable();
+      Predict(DC_PRED);
+      CheckDCPrediction();
+    }
+    {
+      SCOPED_TRACE("DC_PRED TOP");
+      FillRandom();
+      SetTopUnavailable();
+      Predict(DC_PRED);
+      CheckDCPrediction();
+    }
+    {
+      SCOPED_TRACE("DC_PRED TOP_LEFT");
+      FillRandom();
+      SetTopLeftUnavailable();
+      Predict(DC_PRED);
+      CheckDCPrediction();
+    }
+    {
+      SCOPED_TRACE("H_PRED");
+      FillRandom();
+      Predict(H_PRED);
+      CheckHPrediction();
+    }
+    {
+      SCOPED_TRACE("V_PRED");
+      FillRandom();
+      Predict(V_PRED);
+      CheckVPrediction();
+    }
+    {
+      SCOPED_TRACE("TM_PRED");
+      FillRandom();
+      Predict(TM_PRED);
+      CheckTMPrediction();
+    }
+  }
+
+  MACROBLOCKD mb_;
+  MODE_INFO mi_;
+  uint8_t *data_ptr_[2];  // in the case of Y, only [0] is used
+  int stride_;
+  int block_size_;
+  int num_planes_;
+};
+
+typedef void (*intra_pred_y_fn_t)(MACROBLOCKD *x,
+                                  uint8_t *yabove_row,
+                                  uint8_t *yleft,
+                                  int left_stride,
+                                  uint8_t *ypred_ptr,
+                                  int y_stride);
+
+class IntraPredYTest : public ::testing::TestWithParam<intra_pred_y_fn_t>,
+    protected IntraPredBase {
+ public:
+  static void SetUpTestCase() {
+    data_array_ = reinterpret_cast<uint8_t*>(
+        vpx_memalign(kDataAlignment, kDataBufferSize));
+  }
+
+  static void TearDownTestCase() {
+    vpx_free(data_array_);
+    data_array_ = NULL;
+  }
+
+ protected:
+  static const int kBlockSize = 16;
+  static const int kDataAlignment = 16;
+  static const int kStride = kBlockSize * 3;
+  // We use 48 so that the data pointer of the first pixel in each row of
+  // each macroblock is 16-byte aligned, and this gives us access to the
+  // top-left and top-right corner pixels belonging to the top-left/right
+  // macroblocks.
+  // We use 17 lines so we have one line above us for top-prediction.
+  static const int kDataBufferSize = kStride * (kBlockSize + 1);
+
+  virtual void SetUp() {
+    pred_fn_ = GetParam();
+    SetupMacroblock(data_array_, kBlockSize, kStride, 1);
+  }
+
+  virtual void Predict(MB_PREDICTION_MODE mode) {
+    mb_.mode_info_context->mbmi.mode = mode;
+    REGISTER_STATE_CHECK(pred_fn_(&mb_,
+                                  data_ptr_[0] - kStride,
+                                  data_ptr_[0] - 1, kStride,
+                                  data_ptr_[0], kStride));
+  }
+
+  intra_pred_y_fn_t pred_fn_;
+  static uint8_t* data_array_;
+};
+
+uint8_t* IntraPredYTest::data_array_ = NULL;
+
+TEST_P(IntraPredYTest, IntraPredTests) {
+  RunTest();
+}
+
+INSTANTIATE_TEST_CASE_P(C, IntraPredYTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mby_s_c));
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, IntraPredYTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mby_s_sse2));
+#endif
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(SSSE3, IntraPredYTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mby_s_ssse3));
+#endif
+
+typedef void (*intra_pred_uv_fn_t)(MACROBLOCKD *x,
+                                   uint8_t *uabove_row,
+                                   uint8_t *vabove_row,
+                                   uint8_t *uleft,
+                                   uint8_t *vleft,
+                                   int left_stride,
+                                   uint8_t *upred_ptr,
+                                   uint8_t *vpred_ptr,
+                                   int pred_stride);
+
+class IntraPredUVTest : public ::testing::TestWithParam<intra_pred_uv_fn_t>,
+    protected IntraPredBase {
+ public:
+  static void SetUpTestCase() {
+    data_array_ = reinterpret_cast<uint8_t*>(
+        vpx_memalign(kDataAlignment, kDataBufferSize));
+  }
+
+  static void TearDownTestCase() {
+    vpx_free(data_array_);
+    data_array_ = NULL;
+  }
+
+ protected:
+  static const int kBlockSize = 8;
+  static const int kDataAlignment = 8;
+  static const int kStride = kBlockSize * 3;
+  // We use 24 so that the data pointer of the first pixel in each row of
+  // each macroblock is 8-byte aligned, and this gives us access to the
+  // top-left and top-right corner pixels belonging to the top-left/right
+  // macroblocks.
+  // We use 9 lines so we have one line above us for top-prediction.
+  // [0] = U, [1] = V
+  static const int kDataBufferSize = 2 * kStride * (kBlockSize + 1);
+
+  virtual void SetUp() {
+    pred_fn_ = GetParam();
+    SetupMacroblock(data_array_, kBlockSize, kStride, 2);
+  }
+
+  virtual void Predict(MB_PREDICTION_MODE mode) {
+    mb_.mode_info_context->mbmi.uv_mode = mode;
+    pred_fn_(&mb_, data_ptr_[0] - kStride, data_ptr_[1] - kStride,
+             data_ptr_[0] - 1, data_ptr_[1] - 1, kStride,
+             data_ptr_[0], data_ptr_[1], kStride);
+  }
+
+  intra_pred_uv_fn_t pred_fn_;
+  // We use 24 so that the data pointer of the first pixel in each row of
+  // each macroblock is 8-byte aligned, and this gives us access to the
+  // top-left and top-right corner pixels belonging to the top-left/right
+  // macroblocks.
+  // We use 9 lines so we have one line above us for top-prediction.
+  // [0] = U, [1] = V
+  static uint8_t* data_array_;
+};
+
+uint8_t* IntraPredUVTest::data_array_ = NULL;
+
+TEST_P(IntraPredUVTest, IntraPredTests) {
+  RunTest();
+}
+
+INSTANTIATE_TEST_CASE_P(C, IntraPredUVTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mbuv_s_c));
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, IntraPredUVTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mbuv_s_sse2));
+#endif
+#if HAVE_SSSE3
+INSTANTIATE_TEST_CASE_P(SSSE3, IntraPredUVTest,
+                        ::testing::Values(
+                            vp8_build_intra_predictors_mbuv_s_ssse3));
+#endif
+
+}  // namespace

test/ivf_video_source.h

@@ -0,0 +1,109 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+#ifndef TEST_IVF_VIDEO_SOURCE_H_
+#define TEST_IVF_VIDEO_SOURCE_H_
+#include <cstdio>
+#include <cstdlib>
+#include <new>
+#include <string>
+#include "test/video_source.h"
+
+namespace libvpx_test {
+const unsigned int kCodeBufferSize = 256 * 1024;
+const unsigned int kIvfFileHdrSize = 32;
+const unsigned int kIvfFrameHdrSize = 12;
+
+static unsigned int MemGetLe32(const uint8_t *mem) {
+  return (mem[3] << 24) | (mem[2] << 16) | (mem[1] << 8) | (mem[0]);
+}
+
+// This class extends VideoSource to allow parsing of ivf files,
+// so that we can do actual file decodes.
+class IVFVideoSource : public CompressedVideoSource {
+ public:
+  IVFVideoSource(const std::string &file_name)
+      : file_name_(file_name),
+        input_file_(NULL),
+        compressed_frame_buf_(NULL),
+        frame_sz_(0),
+        frame_(0),
+        end_of_file_(false) {
+  }
+
+  virtual ~IVFVideoSource() {
+    delete[] compressed_frame_buf_;
+
+    if (input_file_)
+      fclose(input_file_);
+  }
+
+  virtual void Init() {
+    // Allocate a buffer for read in the compressed video frame.
+    compressed_frame_buf_ = new uint8_t[libvpx_test::kCodeBufferSize];
+    ASSERT_TRUE(compressed_frame_buf_) << "Allocate frame buffer failed";
+  }
+
+  virtual void Begin() {
+    input_file_ = OpenTestDataFile(file_name_);
+    ASSERT_TRUE(input_file_) << "Input file open failed. Filename: "
+        << file_name_;
+
+    // Read file header
+    uint8_t file_hdr[kIvfFileHdrSize];
+    ASSERT_EQ(kIvfFileHdrSize, fread(file_hdr, 1, kIvfFileHdrSize, input_file_))
+        << "File header read failed.";
+    // Check file header
+    ASSERT_TRUE(file_hdr[0] == 'D' && file_hdr[1] == 'K' && file_hdr[2] == 'I'
+                && file_hdr[3] == 'F') << "Input is not an IVF file.";
+
+    FillFrame();
+  }
+
+  virtual void Next() {
+    ++frame_;
+    FillFrame();
+  }
+
+  void FillFrame() {
+    uint8_t frame_hdr[kIvfFrameHdrSize];
+    // Check frame header and read a frame from input_file.
+    if (fread(frame_hdr, 1, kIvfFrameHdrSize, input_file_)
+        != kIvfFrameHdrSize) {
+      end_of_file_ = true;
+    } else {
+      end_of_file_ = false;
+
+      frame_sz_ = MemGetLe32(frame_hdr);
+      ASSERT_LE(frame_sz_, kCodeBufferSize)
+          << "Frame is too big for allocated code buffer";
+      ASSERT_EQ(frame_sz_,
+                fread(compressed_frame_buf_, 1, frame_sz_, input_file_))
+          << "Failed to read complete frame";
+    }
+  }
+
+  virtual const uint8_t *cxdata() const {
+    return end_of_file_ ? NULL : compressed_frame_buf_;
+  }
+  virtual const unsigned int frame_size() const { return frame_sz_; }
+  virtual const unsigned int frame_number() const { return frame_; }
+
+ protected:
+  std::string file_name_;
+  FILE *input_file_;
+  uint8_t *compressed_frame_buf_;
+  unsigned int frame_sz_;
+  unsigned int frame_;
+  bool end_of_file_;
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_IVF_VIDEO_SOURCE_H_

test/keyframe_test.cc

@@ -0,0 +1,145 @@
+/*
+ *  Copyright (c) 2012 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 <climits>
+#include <vector>
+#include "test/encode_test_driver.h"
+#include "test/i420_video_source.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace {
+
+class KeyframeTest : public ::libvpx_test::EncoderTest,
+    public ::testing::TestWithParam<enum libvpx_test::TestMode> {
+ protected:
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(GetParam());
+    kf_count_ = 0;
+    kf_count_max_ = INT_MAX;
+    kf_do_force_kf_ = false;
+    set_cpu_used_ = 0;
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
+                                  ::libvpx_test::Encoder *encoder) {
+    if (kf_do_force_kf_)
+      frame_flags_ = (video->frame() % 3) ? 0 : VPX_EFLAG_FORCE_KF;
+    if (set_cpu_used_ && video->frame() == 1)
+      encoder->Control(VP8E_SET_CPUUSED, set_cpu_used_);
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    if (pkt->data.frame.flags & VPX_FRAME_IS_KEY) {
+      kf_pts_list_.push_back(pkt->data.frame.pts);
+      kf_count_++;
+      abort_ |= kf_count_ > kf_count_max_;
+    }
+  }
+
+  bool kf_do_force_kf_;
+  int kf_count_;
+  int kf_count_max_;
+  std::vector<vpx_codec_pts_t> kf_pts_list_;
+  int set_cpu_used_;
+};
+
+TEST_P(KeyframeTest, TestRandomVideoSource) {
+  // Validate that encoding the RandomVideoSource produces multiple keyframes.
+  // This validates the results of the TestDisableKeyframes test.
+  kf_count_max_ = 2;  // early exit successful tests.
+
+  ::libvpx_test::RandomVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  // In realtime mode - auto placed keyframes are exceedingly rare,  don't
+  // bother with this check   if(GetParam() > 0)
+  if(GetParam() > 0)
+    EXPECT_GT(kf_count_, 1);
+}
+
+TEST_P(KeyframeTest, TestDisableKeyframes) {
+  cfg_.kf_mode = VPX_KF_DISABLED;
+  kf_count_max_ = 1;  // early exit failed tests.
+
+  ::libvpx_test::RandomVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  EXPECT_EQ(1, kf_count_);
+}
+
+TEST_P(KeyframeTest, TestForceKeyframe) {
+  cfg_.kf_mode = VPX_KF_DISABLED;
+  kf_do_force_kf_ = true;
+
+  ::libvpx_test::DummyVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  // verify that every third frame is a keyframe.
+  for (std::vector<vpx_codec_pts_t>::const_iterator iter = kf_pts_list_.begin();
+       iter != kf_pts_list_.end(); ++iter) {
+    ASSERT_EQ(0, *iter % 3) << "Unexpected keyframe at frame " << *iter;
+  }
+}
+
+TEST_P(KeyframeTest, TestKeyframeMaxDistance) {
+  cfg_.kf_max_dist = 25;
+
+  ::libvpx_test::DummyVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  // verify that keyframe interval matches kf_max_dist
+  for (std::vector<vpx_codec_pts_t>::const_iterator iter = kf_pts_list_.begin();
+       iter != kf_pts_list_.end(); ++iter) {
+    ASSERT_EQ(0, *iter % 25) << "Unexpected keyframe at frame " << *iter;
+  }
+}
+
+TEST_P(KeyframeTest, TestAutoKeyframe) {
+  cfg_.kf_mode = VPX_KF_AUTO;
+  kf_do_force_kf_ = false;
+
+  // Force a deterministic speed step in Real Time mode, as the faster modes
+  // may not produce a keyframe like we expect. This is necessary when running
+  // on very slow environments (like Valgrind). The step -11 was determined
+  // experimentally as the fastest mode that still throws the keyframe.
+  if (deadline_ == VPX_DL_REALTIME)
+    set_cpu_used_ = -11;
+
+  // This clip has a cut scene every 30 frames -> Frame 0, 30, 60, 90, 120.
+  // I check only the first 40 frames to make sure there's a keyframe at frame
+  // 0 and 30.
+  ::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
+                                       30, 1, 0, 40);
+
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  // In realtime mode - auto placed keyframes are exceedingly rare,  don't
+  // bother with this check
+  if(GetParam() > 0)
+    EXPECT_EQ(2u, kf_pts_list_.size()) << " Not the right number of keyframes ";
+
+  // Verify that keyframes match the file keyframes in the file.
+  for (std::vector<vpx_codec_pts_t>::const_iterator iter = kf_pts_list_.begin();
+       iter != kf_pts_list_.end(); ++iter) {
+
+    if (deadline_ == VPX_DL_REALTIME && *iter > 0)
+      EXPECT_EQ(0, (*iter - 1) % 30) << "Unexpected keyframe at frame "
+        << *iter;
+    else
+      EXPECT_EQ(0, *iter % 30) << "Unexpected keyframe at frame " << *iter;
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(AllModes, KeyframeTest, ALL_TEST_MODES);
+}  // namespace

test/pp_filter_test.cc

@@ -0,0 +1,107 @@
+/*
+ *  Copyright (c) 2012 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 "test/register_state_check.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+extern "C" {
+#include "vpx_config.h"
+#include "vpx_rtcd.h"
+#include "vpx/vpx_integer.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+typedef void (*post_proc_func_t)(unsigned char *src_ptr,
+                                 unsigned char *dst_ptr,
+                                 int src_pixels_per_line,
+                                 int dst_pixels_per_line,
+                                 int cols,
+                                 unsigned char *flimit,
+                                 int size);
+
+namespace {
+
+class Vp8PostProcessingFilterTest
+    : public ::testing::TestWithParam<post_proc_func_t> {};
+
+// Test routine for the VP8 post-processing function
+// vp8_post_proc_down_and_across_mb_row_c.
+
+TEST_P(Vp8PostProcessingFilterTest, FilterOutputCheck) {
+  // Size of the underlying data block that will be filtered.
+  const int block_width  = 16;
+  const int block_height = 16;
+
+  // 5-tap filter needs 2 padding rows above and below the block in the input.
+  const int input_width = block_width;
+  const int input_height = block_height + 4;
+  const int input_stride = input_width;
+  const int input_size = input_width * input_height;
+
+  // Filter extends output block by 8 samples at left and right edges.
+  const int output_width = block_width + 16;
+  const int output_height = block_height;
+  const int output_stride = output_width;
+  const int output_size = output_width * output_height;
+
+  uint8_t *const src_image =
+      reinterpret_cast<uint8_t*>(vpx_calloc(input_size, 1));
+  uint8_t *const dst_image =
+      reinterpret_cast<uint8_t*>(vpx_calloc(output_size, 1));
+
+  // Pointers to top-left pixel of block in the input and output images.
+  uint8_t *const src_image_ptr = src_image + (input_stride << 1);
+  uint8_t *const dst_image_ptr = dst_image + 8;
+  uint8_t *const flimits = reinterpret_cast<uint8_t *>(vpx_memalign(16, block_width));
+  (void)vpx_memset(flimits, 255, block_width);
+
+  // Initialize pixels in the input:
+  //   block pixels to value 1,
+  //   border pixels to value 10.
+  (void)vpx_memset(src_image, 10, input_size);
+  uint8_t *pixel_ptr = src_image_ptr;
+  for (int i = 0; i < block_height; ++i) {
+    for (int j = 0; j < block_width; ++j) {
+      pixel_ptr[j] = 1;
+    }
+    pixel_ptr += input_stride;
+  }
+
+  // Initialize pixels in the output to 99.
+  (void)vpx_memset(dst_image, 99, output_size);
+
+  REGISTER_STATE_CHECK(GetParam()(src_image_ptr, dst_image_ptr, input_stride,
+                                  output_stride, block_width, flimits, 16));
+
+  static const uint8_t expected_data[block_height] = {
+    4, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 4
+  };
+
+  pixel_ptr = dst_image_ptr;
+  for (int i = 0; i < block_height; ++i) {
+    for (int j = 0; j < block_width; ++j) {
+      EXPECT_EQ(expected_data[i], pixel_ptr[j])
+          << "Vp8PostProcessingFilterTest failed with invalid filter output";
+    }
+    pixel_ptr += output_stride;
+  }
+
+  vpx_free(src_image);
+  vpx_free(dst_image);
+  vpx_free(flimits);
+};
+
+INSTANTIATE_TEST_CASE_P(C, Vp8PostProcessingFilterTest,
+    ::testing::Values(vp8_post_proc_down_and_across_mb_row_c));
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, Vp8PostProcessingFilterTest,
+    ::testing::Values(vp8_post_proc_down_and_across_mb_row_sse2));
+#endif
+
+}  // namespace

test/register_state_check.h

@@ -0,0 +1,95 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+#ifndef LIBVPX_TEST_REGISTER_STATE_CHECK_H_
+#define LIBVPX_TEST_REGISTER_STATE_CHECK_H_
+
+#ifdef _WIN64
+
+#define _WIN32_LEAN_AND_MEAN
+#include <windows.h>
+#include <winnt.h>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace testing {
+namespace internal {
+
+inline bool operator==(const M128A& lhs, const M128A& rhs) {
+  return (lhs.Low == rhs.Low && lhs.High == rhs.High);
+}
+
+}  // namespace internal
+}  // namespace testing
+
+namespace libvpx_test {
+
+// Compares the state of xmm[6-15] at construction with their state at
+// destruction. These registers should be preserved by the callee on
+// Windows x64.
+// Usage:
+// {
+//   RegisterStateCheck reg_check;
+//   FunctionToVerify();
+// }
+class RegisterStateCheck {
+ public:
+  RegisterStateCheck() { initialized_ = StoreRegisters(&pre_context_); }
+  ~RegisterStateCheck() { EXPECT_TRUE(Check()); }
+
+ private:
+  static bool StoreRegisters(CONTEXT* const context) {
+    const HANDLE this_thread = GetCurrentThread();
+    EXPECT_TRUE(this_thread != NULL);
+    context->ContextFlags = CONTEXT_FLOATING_POINT;
+    const bool context_saved = GetThreadContext(this_thread, context) == TRUE;
+    EXPECT_TRUE(context_saved) << "GetLastError: " << GetLastError();
+    return context_saved;
+  }
+
+  // Compares the register state. Returns true if the states match.
+  bool Check() const {
+    if (!initialized_) return false;
+    CONTEXT post_context;
+    if (!StoreRegisters(&post_context)) return false;
+
+    const M128A* xmm_pre = &pre_context_.Xmm6;
+    const M128A* xmm_post = &post_context.Xmm6;
+    for (int i = 6; i <= 15; ++i) {
+      EXPECT_EQ(*xmm_pre, *xmm_post) << "xmm" << i << " has been modified!";
+      ++xmm_pre;
+      ++xmm_post;
+    }
+    return !testing::Test::HasNonfatalFailure();
+  }
+
+  bool initialized_;
+  CONTEXT pre_context_;
+};
+
+#define REGISTER_STATE_CHECK(statement) do { \
+  libvpx_test::RegisterStateCheck reg_check; \
+  statement;                               \
+} while (false)
+
+}  // namespace libvpx_test
+
+#else  // !_WIN64
+
+namespace libvpx_test {
+
+class RegisterStateCheck {};
+#define REGISTER_STATE_CHECK(statement) statement
+
+}  // namespace libvpx_test
+
+#endif  // _WIN64
+
+#endif  // LIBVPX_TEST_REGISTER_STATE_CHECK_H_

test/resize_test.cc

@@ -0,0 +1,104 @@
+/*
+ *  Copyright (c) 2012 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 <climits>
+#include <vector>
+#include "test/encode_test_driver.h"
+#include "test/video_source.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+namespace {
+
+const unsigned int kInitialWidth = 320;
+const unsigned int kInitialHeight = 240;
+
+unsigned int ScaleForFrameNumber(unsigned int frame, unsigned int val) {
+  if (frame < 10)
+    return val;
+  if (frame < 20)
+    return val / 2;
+  if (frame < 30)
+    return val * 2 / 3;
+  if (frame < 40)
+    return val / 4;
+  if (frame < 50)
+    return val * 7 / 8;
+  return val;
+}
+
+class ResizingVideoSource : public ::libvpx_test::DummyVideoSource {
+ public:
+  ResizingVideoSource() {
+    SetSize(kInitialWidth, kInitialHeight);
+    limit_ = 60;
+  }
+
+ protected:
+  virtual void Next() {
+    ++frame_;
+    SetSize(ScaleForFrameNumber(frame_, kInitialWidth),
+            ScaleForFrameNumber(frame_, kInitialHeight));
+    FillFrame();
+  }
+};
+
+class ResizeTest : public ::libvpx_test::EncoderTest,
+  public ::testing::TestWithParam<enum libvpx_test::TestMode> {
+ protected:
+  struct FrameInfo {
+    FrameInfo(vpx_codec_pts_t _pts, unsigned int _w, unsigned int _h)
+        : pts(_pts), w(_w), h(_h) {}
+
+    vpx_codec_pts_t pts;
+    unsigned int    w;
+    unsigned int    h;
+  };
+
+  virtual void SetUp() {
+    InitializeConfig();
+    SetMode(GetParam());
+  }
+
+  virtual bool Continue() const {
+    return !HasFatalFailure() && !abort_;
+  }
+
+  virtual void FramePktHook(const vpx_codec_cx_pkt_t *pkt) {
+    if (pkt->data.frame.flags & VPX_FRAME_IS_KEY) {
+      const unsigned char *buf =
+          reinterpret_cast<const unsigned char *>(pkt->data.frame.buf);
+      const unsigned int w = (buf[6] | (buf[7] << 8)) & 0x3fff;
+      const unsigned int h = (buf[8] | (buf[9] << 8)) & 0x3fff;
+
+      frame_info_list_.push_back(FrameInfo(pkt->data.frame.pts, w, h));
+    }
+  }
+
+  std::vector< FrameInfo > frame_info_list_;
+};
+
+TEST_P(ResizeTest, TestExternalResizeWorks) {
+  ResizingVideoSource video;
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+
+  for (std::vector<FrameInfo>::iterator info = frame_info_list_.begin();
+       info != frame_info_list_.end(); ++info) {
+    const vpx_codec_pts_t pts = info->pts;
+    const unsigned int expected_w = ScaleForFrameNumber(pts, kInitialWidth);
+    const unsigned int expected_h = ScaleForFrameNumber(pts, kInitialHeight);
+
+    EXPECT_EQ(expected_w, info->w)
+        << "Frame " << pts << "had unexpected width";
+    EXPECT_EQ(expected_h, info->h)
+        << "Frame " << pts << "had unexpected height";
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(OnePass, ResizeTest, ONE_PASS_TEST_MODES);
+}  // namespace

test/sad_test.cc

@@ -0,0 +1,253 @@
+/*
+ *  Copyright (c) 2012 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 <string.h>
+#include <limits.h>
+#include <stdio.h>
+
+extern "C" {
+#include "./vpx_config.h"
+#include "./vpx_rtcd.h"
+#include "vp8/common/blockd.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+#include "test/acm_random.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+
+typedef unsigned int (*sad_m_by_n_fn_t)(const unsigned char *source_ptr,
+                                        int source_stride,
+                                        const unsigned char *reference_ptr,
+                                        int reference_stride,
+                                        unsigned int max_sad);
+
+using libvpx_test::ACMRandom;
+
+namespace {
+class SADTest : public PARAMS(int, int, sad_m_by_n_fn_t) {
+ public:
+  static void SetUpTestCase() {
+    source_data_ = reinterpret_cast<uint8_t*>(
+        vpx_memalign(kDataAlignment, kDataBufferSize));
+    reference_data_ = reinterpret_cast<uint8_t*>(
+        vpx_memalign(kDataAlignment, kDataBufferSize));
+  }
+
+  static void TearDownTestCase() {
+    vpx_free(source_data_);
+    source_data_ = NULL;
+    vpx_free(reference_data_);
+    reference_data_ = NULL;
+  }
+
+ protected:
+  static const int kDataAlignment = 16;
+  static const int kDataBufferSize = 16 * 32;
+
+  virtual void SetUp() {
+    sad_fn_ = GET_PARAM(2);
+    height_ = GET_PARAM(1);
+    width_ = GET_PARAM(0);
+    source_stride_ = width_ * 2;
+    reference_stride_ = width_ * 2;
+    rnd_.Reset(ACMRandom::DeterministicSeed());
+  }
+
+  sad_m_by_n_fn_t sad_fn_;
+  virtual unsigned int SAD(unsigned int max_sad) {
+    unsigned int ret;
+    REGISTER_STATE_CHECK(ret = sad_fn_(source_data_, source_stride_,
+                                       reference_data_, reference_stride_,
+                                       max_sad));
+    return ret;
+  }
+
+  // Sum of Absolute Differences. Given two blocks, calculate the absolute
+  // difference between two pixels in the same relative location; accumulate.
+  unsigned int ReferenceSAD(unsigned int max_sad) {
+    unsigned int sad = 0;
+
+    for (int h = 0; h < height_; ++h) {
+      for (int w = 0; w < width_; ++w) {
+        sad += abs(source_data_[h * source_stride_ + w]
+               - reference_data_[h * reference_stride_ + w]);
+      }
+      if (sad > max_sad) {
+        break;
+      }
+    }
+    return sad;
+  }
+
+  void FillConstant(uint8_t *data, int stride, uint8_t fill_constant) {
+    for (int h = 0; h < height_; ++h) {
+      for (int w = 0; w < width_; ++w) {
+        data[h * stride + w] = fill_constant;
+      }
+    }
+  }
+
+  void FillRandom(uint8_t *data, int stride) {
+    for (int h = 0; h < height_; ++h) {
+      for (int w = 0; w < width_; ++w) {
+        data[h * stride + w] = rnd_.Rand8();
+      }
+    }
+  }
+
+  void CheckSad(unsigned int max_sad) {
+    unsigned int reference_sad, exp_sad;
+
+    reference_sad = ReferenceSAD(max_sad);
+    exp_sad = SAD(max_sad);
+
+    if (reference_sad <= max_sad) {
+      ASSERT_EQ(exp_sad, reference_sad);
+    } else {
+      // Alternative implementations are not required to check max_sad
+      ASSERT_GE(exp_sad, reference_sad);
+    }
+  }
+
+  // Handle blocks up to 16x16 with stride up to 32
+  int height_, width_;
+  static uint8_t* source_data_;
+  int source_stride_;
+  static uint8_t* reference_data_;
+  int reference_stride_;
+
+  ACMRandom rnd_;
+};
+
+uint8_t* SADTest::source_data_ = NULL;
+uint8_t* SADTest::reference_data_ = NULL;
+
+TEST_P(SADTest, MaxRef) {
+  FillConstant(source_data_, source_stride_, 0);
+  FillConstant(reference_data_, reference_stride_, 255);
+  CheckSad(UINT_MAX);
+}
+
+TEST_P(SADTest, MaxSrc) {
+  FillConstant(source_data_, source_stride_, 255);
+  FillConstant(reference_data_, reference_stride_, 0);
+  CheckSad(UINT_MAX);
+}
+
+TEST_P(SADTest, ShortRef) {
+  int tmp_stride = reference_stride_;
+  reference_stride_ >>= 1;
+  FillRandom(source_data_, source_stride_);
+  FillRandom(reference_data_, reference_stride_);
+  CheckSad(UINT_MAX);
+  reference_stride_ = tmp_stride;
+}
+
+TEST_P(SADTest, UnalignedRef) {
+  // The reference frame, but not the source frame, may be unaligned for
+  // certain types of searches.
+  int tmp_stride = reference_stride_;
+  reference_stride_ -= 1;
+  FillRandom(source_data_, source_stride_);
+  FillRandom(reference_data_, reference_stride_);
+  CheckSad(UINT_MAX);
+  reference_stride_ = tmp_stride;
+}
+
+TEST_P(SADTest, ShortSrc) {
+  int tmp_stride = source_stride_;
+  source_stride_ >>= 1;
+  FillRandom(source_data_, source_stride_);
+  FillRandom(reference_data_, reference_stride_);
+  CheckSad(UINT_MAX);
+  source_stride_ = tmp_stride;
+}
+
+TEST_P(SADTest, MaxSAD) {
+  // Verify that, when max_sad is set, the implementation does not return a
+  // value lower than the reference.
+  FillConstant(source_data_, source_stride_, 255);
+  FillConstant(reference_data_, reference_stride_, 0);
+  CheckSad(128);
+}
+
+using std::tr1::make_tuple;
+
+const sad_m_by_n_fn_t sad_16x16_c = vp8_sad16x16_c;
+const sad_m_by_n_fn_t sad_8x16_c = vp8_sad8x16_c;
+const sad_m_by_n_fn_t sad_16x8_c = vp8_sad16x8_c;
+const sad_m_by_n_fn_t sad_8x8_c = vp8_sad8x8_c;
+const sad_m_by_n_fn_t sad_4x4_c = vp8_sad4x4_c;
+INSTANTIATE_TEST_CASE_P(C, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_c),
+                        make_tuple(8, 16, sad_8x16_c),
+                        make_tuple(16, 8, sad_16x8_c),
+                        make_tuple(8, 8, sad_8x8_c),
+                        make_tuple(4, 4, sad_4x4_c)));
+
+// ARM tests
+#if HAVE_MEDIA
+const sad_m_by_n_fn_t sad_16x16_armv6 = vp8_sad16x16_armv6;
+INSTANTIATE_TEST_CASE_P(MEDIA, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_armv6)));
+
+#endif
+#if HAVE_NEON
+const sad_m_by_n_fn_t sad_16x16_neon = vp8_sad16x16_neon;
+const sad_m_by_n_fn_t sad_8x16_neon = vp8_sad8x16_neon;
+const sad_m_by_n_fn_t sad_16x8_neon = vp8_sad16x8_neon;
+const sad_m_by_n_fn_t sad_8x8_neon = vp8_sad8x8_neon;
+const sad_m_by_n_fn_t sad_4x4_neon = vp8_sad4x4_neon;
+INSTANTIATE_TEST_CASE_P(NEON, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_neon),
+                        make_tuple(8, 16, sad_8x16_neon),
+                        make_tuple(16, 8, sad_16x8_neon),
+                        make_tuple(8, 8, sad_8x8_neon),
+                        make_tuple(4, 4, sad_4x4_neon)));
+#endif
+
+// X86 tests
+#if HAVE_MMX
+const sad_m_by_n_fn_t sad_16x16_mmx = vp8_sad16x16_mmx;
+const sad_m_by_n_fn_t sad_8x16_mmx = vp8_sad8x16_mmx;
+const sad_m_by_n_fn_t sad_16x8_mmx = vp8_sad16x8_mmx;
+const sad_m_by_n_fn_t sad_8x8_mmx = vp8_sad8x8_mmx;
+const sad_m_by_n_fn_t sad_4x4_mmx = vp8_sad4x4_mmx;
+INSTANTIATE_TEST_CASE_P(MMX, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_mmx),
+                        make_tuple(8, 16, sad_8x16_mmx),
+                        make_tuple(16, 8, sad_16x8_mmx),
+                        make_tuple(8, 8, sad_8x8_mmx),
+                        make_tuple(4, 4, sad_4x4_mmx)));
+#endif
+#if HAVE_SSE2
+const sad_m_by_n_fn_t sad_16x16_wmt = vp8_sad16x16_wmt;
+const sad_m_by_n_fn_t sad_8x16_wmt = vp8_sad8x16_wmt;
+const sad_m_by_n_fn_t sad_16x8_wmt = vp8_sad16x8_wmt;
+const sad_m_by_n_fn_t sad_8x8_wmt = vp8_sad8x8_wmt;
+const sad_m_by_n_fn_t sad_4x4_wmt = vp8_sad4x4_wmt;
+INSTANTIATE_TEST_CASE_P(SSE2, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_wmt),
+                        make_tuple(8, 16, sad_8x16_wmt),
+                        make_tuple(16, 8, sad_16x8_wmt),
+                        make_tuple(8, 8, sad_8x8_wmt),
+                        make_tuple(4, 4, sad_4x4_wmt)));
+#endif
+#if HAVE_SSSE3
+const sad_m_by_n_fn_t sad_16x16_sse3 = vp8_sad16x16_sse3;
+INSTANTIATE_TEST_CASE_P(SSE3, SADTest, ::testing::Values(
+                        make_tuple(16, 16, sad_16x16_sse3)));
+#endif
+
+}  // namespace

test/set_roi.cc

@@ -0,0 +1,182 @@
+/*
+ *  Copyright (c) 2012 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 <math.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "vpx/vpx_integer.h"
+#include "vpx_mem/vpx_mem.h"
+extern "C" {
+#include "vp8/encoder/onyx_int.h"
+}
+
+namespace {
+
+TEST(Vp8RoiMapTest, ParameterCheck) {
+  int delta_q[MAX_MB_SEGMENTS] = { -2, -25, 0, 31 };
+  int delta_lf[MAX_MB_SEGMENTS] = { -2, -25, 0, 31 };
+  unsigned int threshold[MAX_MB_SEGMENTS] = { 0, 100, 200, 300 };
+
+  const int internalq_trans[] = {
+    0,   1,  2,  3,  4,  5,  7,  8,
+    9,  10, 12, 13, 15, 17, 18, 19,
+    20,  21, 23, 24, 25, 26, 27, 28,
+    29,  30, 31, 33, 35, 37, 39, 41,
+    43,  45, 47, 49, 51, 53, 55, 57,
+    59,  61, 64, 67, 70, 73, 76, 79,
+    82,  85, 88, 91, 94, 97, 100, 103,
+    106, 109, 112, 115, 118, 121, 124, 127,
+  };
+
+  // Initialize elements of cpi with valid defaults.
+  VP8_COMP cpi;
+  cpi.mb.e_mbd.mb_segement_abs_delta = SEGMENT_DELTADATA;
+  cpi.cyclic_refresh_mode_enabled = 0;
+  cpi.mb.e_mbd.segmentation_enabled = 0;
+  cpi.mb.e_mbd.update_mb_segmentation_map = 0;
+  cpi.mb.e_mbd.update_mb_segmentation_data = 0;
+  cpi.common.mb_rows = 240 >> 4;
+  cpi.common.mb_cols = 320 >> 4;
+  const int mbs = (cpi.common.mb_rows * cpi.common.mb_cols);
+  vpx_memset(cpi.segment_feature_data, 0, sizeof(cpi.segment_feature_data));
+
+  // Segment map
+  cpi.segmentation_map = reinterpret_cast<unsigned char *>(vpx_calloc(mbs, 1));
+
+  // Allocate memory for the source memory map.
+  unsigned char *roi_map =
+    reinterpret_cast<unsigned char *>(vpx_calloc(mbs, 1));
+  vpx_memset(&roi_map[mbs >> 2], 1, (mbs >> 2));
+  vpx_memset(&roi_map[mbs >> 1], 2, (mbs >> 2));
+  vpx_memset(&roi_map[mbs -(mbs >> 2)], 3, (mbs >> 2));
+
+  // Do a test call with valid parameters.
+  int roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows,
+                                  cpi.common.mb_cols, delta_q, delta_lf,
+                                  threshold);
+  EXPECT_EQ(0, roi_retval)
+        << "vp8_set_roimap roi failed with default test parameters";
+
+  // Check that the values in the cpi structure get set as expected.
+  if (roi_retval == 0) {
+    // Check that the segment map got set.
+    const int mapcompare = memcmp(roi_map, cpi.segmentation_map, mbs);
+    EXPECT_EQ(0, mapcompare) << "segment map error";
+
+    // Check the q deltas (note the need to translate into
+    // the interanl range of 0-127.
+    for (int i = 0; i < MAX_MB_SEGMENTS; ++i) {
+      const int transq = internalq_trans[abs(delta_q[i])];
+      if (abs(cpi.segment_feature_data[MB_LVL_ALT_Q][i]) != transq) {
+          EXPECT_EQ(transq, cpi.segment_feature_data[MB_LVL_ALT_Q][i])
+                    << "segment delta_q  error";
+          break;
+      }
+    }
+
+    // Check the loop filter deltas
+    for (int i = 0; i < MAX_MB_SEGMENTS; ++i) {
+      if (cpi.segment_feature_data[MB_LVL_ALT_LF][i] != delta_lf[i]) {
+        EXPECT_EQ(delta_lf[i], cpi.segment_feature_data[MB_LVL_ALT_LF][i])
+                  << "segment delta_lf error";
+        break;
+      }
+    }
+
+    // Check the breakout thresholds
+    for (int i = 0; i < MAX_MB_SEGMENTS; ++i) {
+      unsigned int breakout =
+        static_cast<unsigned int>(cpi.segment_encode_breakout[i]);
+
+      if (threshold[i] != breakout) {
+        EXPECT_EQ(threshold[i], breakout)
+                  << "breakout threshold error";
+        break;
+      }
+    }
+
+    // Segmentation, and segmentation update flages should be set.
+    EXPECT_EQ(1, cpi.mb.e_mbd.segmentation_enabled)
+              << "segmentation_enabled error";
+    EXPECT_EQ(1, cpi.mb.e_mbd.update_mb_segmentation_map)
+              << "update_mb_segmentation_map error";
+    EXPECT_EQ(1, cpi.mb.e_mbd.update_mb_segmentation_data)
+              << "update_mb_segmentation_data error";
+
+
+    // Try a range of delta q and lf parameters (some legal, some not)
+    for (int i = 0; i < 1000; ++i) {
+      int rand_deltas[4];
+      int deltas_valid;
+      rand_deltas[0] = (rand() % 160) - 80;
+      rand_deltas[1] = (rand() % 160) - 80;
+      rand_deltas[2] = (rand() % 160) - 80;
+      rand_deltas[3] = (rand() % 160) - 80;
+
+      deltas_valid = ((abs(rand_deltas[0]) <= 63) &&
+                      (abs(rand_deltas[1]) <= 63) &&
+                      (abs(rand_deltas[2]) <= 63) &&
+                      (abs(rand_deltas[3]) <= 63)) ? 0 : -1;
+
+      // Test with random delta q values.
+      roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows,
+                                  cpi.common.mb_cols, rand_deltas,
+                                  delta_lf, threshold);
+      EXPECT_EQ(deltas_valid, roi_retval) << "dq range check error";
+
+      // One delta_q error shown at a time
+      if (deltas_valid != roi_retval)
+        break;
+
+      // Test with random loop filter values.
+      roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows,
+                                  cpi.common.mb_cols, delta_q,
+                                  rand_deltas, threshold);
+      EXPECT_EQ(deltas_valid, roi_retval) << "dlf range check error";
+
+      // One delta loop filter error shown at a time
+      if (deltas_valid != roi_retval)
+        break;
+    }
+
+    // Test that we report and error if cyclic refresh is enabled.
+    cpi.cyclic_refresh_mode_enabled = 1;
+    roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows,
+                                cpi.common.mb_cols, delta_q,
+                                delta_lf, threshold);
+    EXPECT_EQ(-1, roi_retval) << "cyclic refresh check error";
+    cpi.cyclic_refresh_mode_enabled = 0;
+
+    // Test invalid number of rows or colums.
+    roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows + 1,
+                                cpi.common.mb_cols, delta_q,
+                                delta_lf, threshold);
+    EXPECT_EQ(-1, roi_retval) << "MB rows bounds check error";
+
+    roi_retval = vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows,
+                                cpi.common.mb_cols - 1, delta_q,
+                                delta_lf, threshold);
+    EXPECT_EQ(-1, roi_retval) << "MB cols bounds check error";
+  }
+
+  // Free allocated memory
+  if (cpi.segmentation_map)
+    vpx_free(cpi.segmentation_map);
+  if (roi_map)
+    vpx_free(roi_map);
+};
+
+}  // namespace

test/sixtap_predict_test.cc

@@ -0,0 +1,224 @@
+/*
+*  Copyright (c) 2012 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 <math.h>
+#include <stdlib.h>
+#include <string.h>
+#include "test/acm_random.h"
+#include "test/register_state_check.h"
+#include "test/util.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+extern "C" {
+#include "./vpx_config.h"
+#include "./vpx_rtcd.h"
+#include "vpx/vpx_integer.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+namespace {
+
+typedef void (*sixtap_predict_fn_t)(uint8_t *src_ptr,
+                                    int  src_pixels_per_line,
+                                    int  xoffset,
+                                    int  yoffset,
+                                    uint8_t *dst_ptr,
+                                    int  dst_pitch);
+
+class SixtapPredictTest : public PARAMS(int, int, sixtap_predict_fn_t) {
+ public:
+  static void SetUpTestCase() {
+    src_ = reinterpret_cast<uint8_t*>(vpx_memalign(kDataAlignment, kSrcSize));
+    dst_ = reinterpret_cast<uint8_t*>(vpx_memalign(kDataAlignment, kDstSize));
+    dst_c_ = reinterpret_cast<uint8_t*>(vpx_memalign(kDataAlignment, kDstSize));
+  }
+
+  static void TearDownTestCase() {
+    vpx_free(src_);
+    src_ = NULL;
+    vpx_free(dst_);
+    dst_ = NULL;
+    vpx_free(dst_c_);
+    dst_c_ = NULL;
+  }
+
+ protected:
+  // Make test arrays big enough for 16x16 functions. Six-tap filters
+  // need 5 extra pixels outside of the macroblock.
+  static const int kSrcStride = 21;
+  static const int kDstStride = 16;
+  static const int kDataAlignment = 16;
+  static const int kSrcSize = kSrcStride * kSrcStride + 1;
+  static const int kDstSize = kDstStride * kDstStride;
+
+  virtual void SetUp() {
+    width_ = GET_PARAM(0);
+    height_ = GET_PARAM(1);
+    sixtap_predict_ = GET_PARAM(2);
+    memset(src_, 0, sizeof(src_));
+    memset(dst_, 0, sizeof(dst_));
+    memset(dst_c_, 0, sizeof(dst_c_));
+  }
+
+  int width_;
+  int height_;
+  sixtap_predict_fn_t sixtap_predict_;
+  // The src stores the macroblock we will filter on, and makes it 1 byte larger
+  // in order to test unaligned access. The result is stored in dst and dst_c(c
+  // reference code result).
+  static uint8_t* src_;
+  static uint8_t* dst_;
+  static uint8_t* dst_c_;
+};
+
+uint8_t* SixtapPredictTest::src_ = NULL;
+uint8_t* SixtapPredictTest::dst_ = NULL;
+uint8_t* SixtapPredictTest::dst_c_ = NULL;
+
+TEST_P(SixtapPredictTest, TestWithPresetData) {
+  // Test input
+  static const uint8_t test_data[kSrcSize] = {
+    216, 184, 4, 191, 82, 92, 41, 0, 1, 226, 236, 172, 20, 182, 42, 226, 177,
+    79, 94, 77, 179, 203, 206, 198, 22, 192, 19, 75, 17, 192, 44, 233, 120,
+    48, 168, 203, 141, 210, 203, 143, 180, 184, 59, 201, 110, 102, 171, 32,
+    182, 10, 109, 105, 213, 60, 47, 236, 253, 67, 55, 14, 3, 99, 247, 124,
+    148, 159, 71, 34, 114, 19, 177, 38, 203, 237, 239, 58, 83, 155, 91, 10,
+    166, 201, 115, 124, 5, 163, 104, 2, 231, 160, 16, 234, 4, 8, 103, 153,
+    167, 174, 187, 26, 193, 109, 64, 141, 90, 48, 200, 174, 204, 36, 184,
+    114, 237, 43, 238, 242, 207, 86, 245, 182, 247, 6, 161, 251, 14, 8, 148,
+    182, 182, 79, 208, 120, 188, 17, 6, 23, 65, 206, 197, 13, 242, 126, 128,
+    224, 170, 110, 211, 121, 197, 200, 47, 188, 207, 208, 184, 221, 216, 76,
+    148, 143, 156, 100, 8, 89, 117, 14, 112, 183, 221, 54, 197, 208, 180, 69,
+    176, 94, 180, 131, 215, 121, 76, 7, 54, 28, 216, 238, 249, 176, 58, 142,
+    64, 215, 242, 72, 49, 104, 87, 161, 32, 52, 216, 230, 4, 141, 44, 181,
+    235, 224, 57, 195, 89, 134, 203, 144, 162, 163, 126, 156, 84, 185, 42,
+    148, 145, 29, 221, 194, 134, 52, 100, 166, 105, 60, 140, 110, 201, 184,
+    35, 181, 153, 93, 121, 243, 227, 68, 131, 134, 232, 2, 35, 60, 187, 77,
+    209, 76, 106, 174, 15, 241, 227, 115, 151, 77, 175, 36, 187, 121, 221,
+    223, 47, 118, 61, 168, 105, 32, 237, 236, 167, 213, 238, 202, 17, 170,
+    24, 226, 247, 131, 145, 6, 116, 117, 121, 11, 194, 41, 48, 126, 162, 13,
+    93, 209, 131, 154, 122, 237, 187, 103, 217, 99, 60, 200, 45, 78, 115, 69,
+    49, 106, 200, 194, 112, 60, 56, 234, 72, 251, 19, 120, 121, 182, 134, 215,
+    135, 10, 114, 2, 247, 46, 105, 209, 145, 165, 153, 191, 243, 12, 5, 36,
+    119, 206, 231, 231, 11, 32, 209, 83, 27, 229, 204, 149, 155, 83, 109, 35,
+    93, 223, 37, 84, 14, 142, 37, 160, 52, 191, 96, 40, 204, 101, 77, 67, 52,
+    53, 43, 63, 85, 253, 147, 113, 226, 96, 6, 125, 179, 115, 161, 17, 83,
+    198, 101, 98, 85, 139, 3, 137, 75, 99, 178, 23, 201, 255, 91, 253, 52,
+    134, 60, 138, 131, 208, 251, 101, 48, 2, 227, 228, 118, 132, 245, 202,
+    75, 91, 44, 160, 231, 47, 41, 50, 147, 220, 74, 92, 219, 165, 89, 16
+  };
+
+  // Expected result
+  static const uint8_t expected_dst[kDstSize] = {
+    117, 102, 74, 135, 42, 98, 175, 206, 70, 73, 222, 197, 50, 24, 39, 49, 38,
+    105, 90, 47, 169, 40, 171, 215, 200, 73, 109, 141, 53, 85, 177, 164, 79,
+    208, 124, 89, 212, 18, 81, 145, 151, 164, 217, 153, 91, 154, 102, 102,
+    159, 75, 164, 152, 136, 51, 213, 219, 186, 116, 193, 224, 186, 36, 231,
+    208, 84, 211, 155, 167, 35, 59, 42, 76, 216, 149, 73, 201, 78, 149, 184,
+    100, 96, 196, 189, 198, 188, 235, 195, 117, 129, 120, 129, 49, 25, 133,
+    113, 69, 221, 114, 70, 143, 99, 157, 108, 189, 140, 78, 6, 55, 65, 240,
+    255, 245, 184, 72, 90, 100, 116, 131, 39, 60, 234, 167, 33, 160, 88, 185,
+    200, 157, 159, 176, 127, 151, 138, 102, 168, 106, 170, 86, 82, 219, 189,
+    76, 33, 115, 197, 106, 96, 198, 136, 97, 141, 237, 151, 98, 137, 191,
+    185, 2, 57, 95, 142, 91, 255, 185, 97, 137, 76, 162, 94, 173, 131, 193,
+    161, 81, 106, 72, 135, 222, 234, 137, 66, 137, 106, 243, 210, 147, 95,
+    15, 137, 110, 85, 66, 16, 96, 167, 147, 150, 173, 203, 140, 118, 196,
+    84, 147, 160, 19, 95, 101, 123, 74, 132, 202, 82, 166, 12, 131, 166,
+    189, 170, 159, 85, 79, 66, 57, 152, 132, 203, 194, 0, 1, 56, 146, 180,
+    224, 156, 28, 83, 181, 79, 76, 80, 46, 160, 175, 59, 106, 43, 87, 75,
+    136, 85, 189, 46, 71, 200, 90
+  };
+
+  uint8_t *src = const_cast<uint8_t*>(test_data);
+
+  REGISTER_STATE_CHECK(sixtap_predict_(&src[kSrcStride * 2 + 2 + 1], kSrcStride,
+                                       2, 2, dst_, kDstStride));
+
+  for (int i = 0; i < height_; ++i)
+    for (int j = 0; j < width_; ++j)
+      ASSERT_EQ(expected_dst[i * kDstStride + j], dst_[i * kDstStride + j])
+          << "i==" << (i * width_ + j);
+}
+
+using libvpx_test::ACMRandom;
+
+TEST_P(SixtapPredictTest, TestWithRandomData) {
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  for (int i = 0; i < kSrcSize; ++i)
+    src_[i] = rnd.Rand8();
+
+  // Run tests for all possible offsets.
+  for (int xoffset = 0; xoffset < 8; ++xoffset) {
+    for (int yoffset = 0; yoffset < 8; ++yoffset) {
+      // Call c reference function.
+      // Move start point to next pixel to test if the function reads
+      // unaligned data correctly.
+      vp8_sixtap_predict16x16_c(&src_[kSrcStride * 2 + 2 + 1], kSrcStride,
+                                xoffset, yoffset, dst_c_, kDstStride);
+
+      // Run test.
+      REGISTER_STATE_CHECK(
+          sixtap_predict_(&src_[kSrcStride * 2 + 2 + 1], kSrcStride,
+                          xoffset, yoffset, dst_, kDstStride));
+
+      for (int i = 0; i < height_; ++i)
+        for (int j = 0; j < width_; ++j)
+          ASSERT_EQ(dst_c_[i * kDstStride + j], dst_[i * kDstStride + j])
+              << "i==" << (i * width_ + j);
+    }
+  }
+}
+
+using std::tr1::make_tuple;
+
+const sixtap_predict_fn_t sixtap_16x16_c = vp8_sixtap_predict16x16_c;
+const sixtap_predict_fn_t sixtap_8x8_c = vp8_sixtap_predict8x8_c;
+const sixtap_predict_fn_t sixtap_8x4_c = vp8_sixtap_predict8x4_c;
+const sixtap_predict_fn_t sixtap_4x4_c = vp8_sixtap_predict4x4_c;
+INSTANTIATE_TEST_CASE_P(
+    C, SixtapPredictTest, ::testing::Values(
+        make_tuple(16, 16, sixtap_16x16_c),
+        make_tuple(8, 8, sixtap_8x8_c),
+        make_tuple(8, 4, sixtap_8x4_c),
+        make_tuple(4, 4, sixtap_4x4_c)));
+#if HAVE_MMX
+const sixtap_predict_fn_t sixtap_16x16_mmx = vp8_sixtap_predict16x16_mmx;
+const sixtap_predict_fn_t sixtap_8x8_mmx = vp8_sixtap_predict8x8_mmx;
+const sixtap_predict_fn_t sixtap_8x4_mmx = vp8_sixtap_predict8x4_mmx;
+const sixtap_predict_fn_t sixtap_4x4_mmx = vp8_sixtap_predict4x4_mmx;
+INSTANTIATE_TEST_CASE_P(
+    MMX, SixtapPredictTest, ::testing::Values(
+        make_tuple(16, 16, sixtap_16x16_mmx),
+        make_tuple(8, 8, sixtap_8x8_mmx),
+        make_tuple(8, 4, sixtap_8x4_mmx),
+        make_tuple(4, 4, sixtap_4x4_mmx)));
+#endif
+#if HAVE_SSE2
+const sixtap_predict_fn_t sixtap_16x16_sse2 = vp8_sixtap_predict16x16_sse2;
+const sixtap_predict_fn_t sixtap_8x8_sse2 = vp8_sixtap_predict8x8_sse2;
+const sixtap_predict_fn_t sixtap_8x4_sse2 = vp8_sixtap_predict8x4_sse2;
+INSTANTIATE_TEST_CASE_P(
+    SSE2, SixtapPredictTest, ::testing::Values(
+        make_tuple(16, 16, sixtap_16x16_sse2),
+        make_tuple(8, 8, sixtap_8x8_sse2),
+        make_tuple(8, 4, sixtap_8x4_sse2)));
+#endif
+#if HAVE_SSSE3
+const sixtap_predict_fn_t sixtap_16x16_ssse3 = vp8_sixtap_predict16x16_ssse3;
+const sixtap_predict_fn_t sixtap_8x8_ssse3 = vp8_sixtap_predict8x8_ssse3;
+const sixtap_predict_fn_t sixtap_8x4_ssse3 = vp8_sixtap_predict8x4_ssse3;
+const sixtap_predict_fn_t sixtap_4x4_ssse3 = vp8_sixtap_predict4x4_ssse3;
+INSTANTIATE_TEST_CASE_P(
+    SSSE3, SixtapPredictTest, ::testing::Values(
+        make_tuple(16, 16, sixtap_16x16_ssse3),
+        make_tuple(8, 8, sixtap_8x8_ssse3),
+        make_tuple(8, 4, sixtap_8x4_ssse3),
+        make_tuple(4, 4, sixtap_4x4_ssse3)));
+#endif
+}  // namespace

test/subtract_test.cc

@@ -0,0 +1,114 @@
+/*
+ *  Copyright (c) 2012 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 "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/acm_random.h"
+#include "test/register_state_check.h"
+extern "C" {
+#include "vpx_config.h"
+#include "vpx_rtcd.h"
+#include "vp8/common/blockd.h"
+#include "vp8/encoder/block.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+typedef void (*subtract_b_fn_t)(BLOCK *be, BLOCKD *bd, int pitch);
+
+namespace {
+
+class SubtractBlockTest : public ::testing::TestWithParam<subtract_b_fn_t> {};
+
+using libvpx_test::ACMRandom;
+
+TEST_P(SubtractBlockTest, SimpleSubtract) {
+  ACMRandom rnd(ACMRandom::DeterministicSeed());
+  BLOCK be;
+  BLOCKD bd;
+  // in libvpx, this stride is always 16
+  const int kDiffPredStride = 16;
+  const int kSrcStride[] = {32, 16, 8, 4, 0};
+  const int kBlockWidth = 4;
+  const int kBlockHeight = 4;
+
+  // Allocate... align to 16 for mmx/sse tests
+  uint8_t *source = reinterpret_cast<uint8_t*>(
+      vpx_memalign(16, kBlockHeight * kSrcStride[0] * sizeof(*source)));
+  be.src_diff = reinterpret_cast<int16_t*>(
+      vpx_memalign(16, kBlockHeight * kDiffPredStride * sizeof(*be.src_diff)));
+  bd.predictor = reinterpret_cast<unsigned char*>(
+      vpx_memalign(16, kBlockHeight * kDiffPredStride * sizeof(*bd.predictor)));
+
+  for(int i = 0; kSrcStride[i] > 0; ++i) {
+    // start at block0
+    be.src = 0;
+    be.base_src = &source;
+    be.src_stride = kSrcStride[i];
+
+    // set difference
+    int16_t *src_diff = be.src_diff;
+    for (int r = 0; r < kBlockHeight; ++r) {
+      for (int c = 0; c < kBlockWidth; ++c) {
+        src_diff[c] = 0xa5a5;
+      }
+      src_diff += kDiffPredStride;
+    }
+
+    // set destination
+    uint8_t *base_src = *be.base_src;
+    for (int r = 0; r < kBlockHeight; ++r) {
+      for (int c = 0; c < kBlockWidth; ++c) {
+        base_src[c] = rnd.Rand8();
+      }
+      base_src += be.src_stride;
+    }
+
+    // set predictor
+    uint8_t *predictor = bd.predictor;
+    for (int r = 0; r < kBlockHeight; ++r) {
+      for (int c = 0; c < kBlockWidth; ++c) {
+        predictor[c] = rnd.Rand8();
+      }
+      predictor += kDiffPredStride;
+    }
+
+    REGISTER_STATE_CHECK(GetParam()(&be, &bd, kDiffPredStride));
+
+    base_src = *be.base_src;
+    src_diff = be.src_diff;
+    predictor = bd.predictor;
+    for (int r = 0; r < kBlockHeight; ++r) {
+      for (int c = 0; c < kBlockWidth; ++c) {
+        EXPECT_EQ(base_src[c], (src_diff[c] + predictor[c])) << "r = " << r
+                                                             << ", c = " << c;
+      }
+      src_diff += kDiffPredStride;
+      predictor += kDiffPredStride;
+      base_src += be.src_stride;
+    }
+  }
+  vpx_free(be.src_diff);
+  vpx_free(source);
+  vpx_free(bd.predictor);
+}
+
+INSTANTIATE_TEST_CASE_P(C, SubtractBlockTest,
+                        ::testing::Values(vp8_subtract_b_c));
+
+#if HAVE_MMX
+INSTANTIATE_TEST_CASE_P(MMX, SubtractBlockTest,
+                        ::testing::Values(vp8_subtract_b_mmx));
+#endif
+
+#if HAVE_SSE2
+INSTANTIATE_TEST_CASE_P(SSE2, SubtractBlockTest,
+                        ::testing::Values(vp8_subtract_b_sse2));
+#endif
+
+}  // namespace

test/test-data.sha1

@@ -0,0 +1,123 @@
+d5dfb0151c9051f8c85999255645d7a23916d3c0  hantro_collage_w352h288.yuv
+5184c46ddca8b1fadd16742e8500115bc8f749da  vp80-00-comprehensive-001.ivf
+65bf1bbbced81b97bd030f376d1b7f61a224793f  vp80-00-comprehensive-002.ivf
+906b4c1e99eb734504c504b3f1ad8052137ce672  vp80-00-comprehensive-003.ivf
+ec144b1af53af895db78355785650b96dd3f0ade  vp80-00-comprehensive-004.ivf
+afc7091785c62f1c121c4554a2830c30704587d9  vp80-00-comprehensive-005.ivf
+42ea9d55c818145d06a9b633b8e85c6a6164fd3e  vp80-00-comprehensive-006.ivf
+e5b3a73ab79fe024c14309d653d6bed92902ee3b  vp80-00-comprehensive-007.ivf
+f3c50a58875930adfb84525c0ef59d7e4c08540c  vp80-00-comprehensive-008.ivf
+4b2841fdb83db51ae322096ae468bbb9dc2c8362  vp80-00-comprehensive-009.ivf
+efbff736e3a91ab6a98c5bc2dce65d645944c7b1  vp80-00-comprehensive-010.ivf
+6b315102cae008d22a3d2c231be92cb704a222f8  vp80-00-comprehensive-011.ivf
+f3214a4fea14c2d5ec689936c1613f274c859ee8  vp80-00-comprehensive-012.ivf
+e4094e96d308c8a35b74c480a43d853c5294cd34  vp80-00-comprehensive-013.ivf
+5b0adfaf60a69e0aaf3ec021a39d0a68fc0e1b5a  vp80-00-comprehensive-014.ivf
+e8467688ddf26b5000664f904faf0d70506aa653  vp80-00-comprehensive-015.ivf
+aab55582337dfd2a39ff54fb2576a91910d49337  vp80-00-comprehensive-016.ivf
+1ba24724f80203c9bae4f1d0f99d534721980016  vp80-00-comprehensive-017.ivf
+143a15512b46f436280ddb4d0e6411eb4af434f2  vp80-00-comprehensive-018.ivf
+c5baeaf5714fdfb3a8bc960a8e33ac438e83b16b  vp80-01-intra-1400.ivf
+f383955229afe3408453e316d11553d923ca60d5  vp80-01-intra-1411.ivf
+84e1f4343f174c9f3c83f834bac3196fb325bf2c  vp80-01-intra-1416.ivf
+fb6e712a47dd57a28a3727d2ae2c97a8b7c7ca51  vp80-01-intra-1417.ivf
+71ea772d3e9d315b8cbecf41207b8a237c34853b  vp80-02-inter-1402.ivf
+d85dbc4271525dcd128c503f936fe69091d1f8d0  vp80-02-inter-1412.ivf
+d4e5d3ad56511867d025f93724d090f92ba6ec3d  vp80-02-inter-1418.ivf
+91791cbcc37c60f35dbd8090bacb54e5ec6dd4fa  vp80-02-inter-1424.ivf
+17fbfe2fea70f6e2f3fa6ca4efaae6c0b03b5f02  vp80-03-segmentation-01.ivf
+3c3600dbbcde08e20d54c66fe3b7eadd4f09bdbb  vp80-03-segmentation-02.ivf
+c156778d5340967d4b369c490848076e92f1f875  vp80-03-segmentation-03.ivf
+d25dcff6c60e87a1af70945b8911b6b4998533b0  vp80-03-segmentation-04.ivf
+362baba2ce454c9db21218f35e81c27a5ed0b730  vp80-03-segmentation-1401.ivf
+d223ae7ee748ce07e74c4679bfd219e84aa9f4b0  vp80-03-segmentation-1403.ivf
+033adf7f3a13836a3f1cffcb87c1972900f2b5c6  vp80-03-segmentation-1407.ivf
+4d51dfbf9f3e2c590ec99d1d6f59dd731d04375f  vp80-03-segmentation-1408.ivf
+f37a62b197c2600d75e0ccfbb31b60efdedac251  vp80-03-segmentation-1409.ivf
+eb25bd7bfba5b2f6935018a930f42d123b1e7fcd  vp80-03-segmentation-1410.ivf
+b9d5c436663a30c27cfff84b53a002e501258843  vp80-03-segmentation-1413.ivf
+6da92b9d1a180cc3a8afe348ab12258f5a37be1a  vp80-03-segmentation-1414.ivf
+a4f5842602886bd669f115f93d8a35c035cb0948  vp80-03-segmentation-1415.ivf
+f295dceb8ef278b77251b3f9df8aee22e161d547  vp80-03-segmentation-1425.ivf
+198dbf9f36f733200e432664cc8c5752d59779de  vp80-03-segmentation-1426.ivf
+7704804e32f5de976803929934a7fafe101ac7b0  vp80-03-segmentation-1427.ivf
+831ccd862ea95ca025d2f3bd8b88678752f5416d  vp80-03-segmentation-1432.ivf
+b3c11978529289f9109f2766fcaba3ebc40e11ef  vp80-03-segmentation-1435.ivf
+a835a731f5520ebfc1002c40121264d0020559ac  vp80-03-segmentation-1436.ivf
+1d1732942f773bb2a5775fcb9689b1579ce28eab  vp80-03-segmentation-1437.ivf
+db04799adfe089dfdf74dbd43cc05ede7161f99e  vp80-03-segmentation-1441.ivf
+7caf39b3f20cfd52b998210878062e52a5edf1e6  vp80-03-segmentation-1442.ivf
+3607f6bb4ee106c38fa1ea370dc4ff8b8cde2261  vp80-04-partitions-1404.ivf
+93cc323b6b6867f1b12dd48773424549c6960a6b  vp80-04-partitions-1405.ivf
+047eedb14b865bdac8a3538e63801054e0295e9c  vp80-04-partitions-1406.ivf
+0f1233bd2bc33f56ce5e495dbd455d122339f384  vp80-05-sharpness-1428.ivf
+51767fc136488a9535c2a4c38067c542ee2048df  vp80-05-sharpness-1429.ivf
+9805aa107672de25d6fb8c35e20d06deca5efe18  vp80-05-sharpness-1430.ivf
+61db6b965f9c27aebe71b85bf2d5877e58e4bbdf  vp80-05-sharpness-1431.ivf
+10420d266290d2923555f84af38eeb96edbd3ae8  vp80-05-sharpness-1433.ivf
+3ed24f9a80cddfdf75824ba95cdb4ff9286cb443  vp80-05-sharpness-1434.ivf
+c87599cbecd72d4cd4f7ace3313b7a6bc6eb8163  vp80-05-sharpness-1438.ivf
+aff51d865c2621b60510459244ea83e958e4baed  vp80-05-sharpness-1439.ivf
+da386e72b19b5485a6af199c5eb60ef25e510dd1  vp80-05-sharpness-1440.ivf
+6759a095203d96ccd267ce09b1b050b8cc4c2f1f  vp80-05-sharpness-1443.ivf
+db55ec7fd02c864ba996ff060b25b1e08611330b  vp80-00-comprehensive-001.ivf.md5
+29db0ad011cba1e45f856d5623cd38dac3e3bf19  vp80-00-comprehensive-002.ivf.md5
+e84f258f69e173e7d68f8f8c037a0a3766902182  vp80-00-comprehensive-003.ivf.md5
+eb7912eaf69559a16fd82bc3f5fb1524cf4a4466  vp80-00-comprehensive-004.ivf.md5
+4206f71c94894bd5b5b376f6c09b3817dbc65206  vp80-00-comprehensive-005.ivf.md5
+4f89b356f6f2fecb928f330a10f804f00f5325f5  vp80-00-comprehensive-006.ivf.md5
+2813236a32964dd8007e17648bcf035a20fcda6c  vp80-00-comprehensive-007.ivf.md5
+10746c72098f872803c900e17c5680e451f5f498  vp80-00-comprehensive-008.ivf.md5
+39a23d0692ce64421a7bb7cdf6ccec5928d37fff  vp80-00-comprehensive-009.ivf.md5
+f6e3de8931a0cc659bda8fbc14050346955e72d4  vp80-00-comprehensive-010.ivf.md5
+101683ec195b6e944f7cd1e468fc8921439363e6  vp80-00-comprehensive-011.ivf.md5
+1f592751ce46d8688998fa0fa4fbdcda0fd4058c  vp80-00-comprehensive-012.ivf.md5
+6066176f90ca790251e795fca1a5797d59999841  vp80-00-comprehensive-013.ivf.md5
+2656da94ba93691f23edc4d60b3a09e2be46c217  vp80-00-comprehensive-014.ivf.md5
+c6e0d5f5d61460c8ac8edfa4e701f10312c03133  vp80-00-comprehensive-015.ivf.md5
+ee60fee501d8493e34e8d6a1fe315b51ed09b24a  vp80-00-comprehensive-016.ivf.md5
+9f1914ceffcad4546c0a29de3ef591d8bea304dc  vp80-00-comprehensive-017.ivf.md5
+e0305178fe288a9fd8082b39e2d03181edb19054  vp80-00-comprehensive-018.ivf.md5
+612494da2fa799cc9d76dcdd835ae6c7cb2e5c05  vp80-01-intra-1400.ivf.md5
+48ea06097ac8269c5e8c2131d3d0639f431fcf0e  vp80-01-intra-1411.ivf.md5
+6e2ab4e7677ad0ba868083ca6bc387ee922b400c  vp80-01-intra-1416.ivf.md5
+eca0a90348959ce3854142f8d8641b13050e8349  vp80-01-intra-1417.ivf.md5
+920feea203145d5c2258a91c4e6991934a79a99e  vp80-02-inter-1402.ivf.md5
+f71d97909fe2b3dd65be7e1f56c72237f0cef200  vp80-02-inter-1412.ivf.md5
+e911254569a30bbb2a237ff8b79f69ed9da0672d  vp80-02-inter-1418.ivf.md5
+58c789c50c9bb9cc90580bed291164a0939d28ba  vp80-02-inter-1424.ivf.md5
+ff3e2f441327b9c20a0b37c524e0f5a48a36de7b  vp80-03-segmentation-01.ivf.md5
+0791f417f076a542ae66fbc3426ab4d94cbd6c75  vp80-03-segmentation-02.ivf.md5
+722e50f1a6a91c34302d68681faffc1c26d1cc57  vp80-03-segmentation-03.ivf.md5
+c701f1885bcfb27fb8e70cc65606b289172ef889  vp80-03-segmentation-04.ivf.md5
+f79bc9ec189a2b4807632a3d0c5bf04a178b5300  vp80-03-segmentation-1401.ivf.md5
+b9aa4c74c0219b639811c44760d0b24cd8bb436a  vp80-03-segmentation-1403.ivf.md5
+70d5a2207ca1891bcaebd5cf6dd88ce8d57b4334  vp80-03-segmentation-1407.ivf.md5
+265f962ee781531f9a93b9309461316fd32b2a1d  vp80-03-segmentation-1408.ivf.md5
+0c4ecbbd6dc042d30e626d951b65f460dd6cd563  vp80-03-segmentation-1409.ivf.md5
+cf779af36a937f06570a0fca9db64ba133451dee  vp80-03-segmentation-1410.ivf.md5
+0e6c5036d51ab078842f133934926c598a9cff02  vp80-03-segmentation-1413.ivf.md5
+eb3930aaf229116c80d507516c34759c3f6cdf69  vp80-03-segmentation-1414.ivf.md5
+123d6c0f72ee87911c4ae7538e87b7d163b22d6c  vp80-03-segmentation-1415.ivf.md5
+e70551d1a38920e097a5d8782390b79ecaeb7505  vp80-03-segmentation-1425.ivf.md5
+44e8f4117e46dbb302b2cfd81171cc1a1846e431  vp80-03-segmentation-1426.ivf.md5
+52636e54aee5f95bbace37021bd67de5db767e9a  vp80-03-segmentation-1427.ivf.md5
+b1ad3eff20215c28e295b15ef3636ed926d59cba  vp80-03-segmentation-1432.ivf.md5
+24c22a552fa28a90e5978f67f57181cc2d7546d7  vp80-03-segmentation-1435.ivf.md5
+96c49c390abfced18a7a8c9b9ea10af778e10edb  vp80-03-segmentation-1436.ivf.md5
+f95eb6214571434f1f73ab7833b9ccdf47588020  vp80-03-segmentation-1437.ivf.md5
+1c0700ca27c9b0090a7747a4b0b4dc21d1843181  vp80-03-segmentation-1441.ivf.md5
+81d4f23ca32667ee958bae579c8f5e97ba72eb97  vp80-03-segmentation-1442.ivf.md5
+272efcef07a3a30fbca51bfd566063d8258ec0be  vp80-04-partitions-1404.ivf.md5
+66ed219ab812ac801b256d35cf495d193d4cf478  vp80-04-partitions-1405.ivf.md5
+36083f37f56f502bd60ec5e07502ee9e6b8699b0  vp80-04-partitions-1406.ivf.md5
+6ca909bf168a64c09415626294665dc1be3d1973  vp80-05-sharpness-1428.ivf.md5
+1667d2ee2334e5fdea8a8a866f4ccf3cf76f033a  vp80-05-sharpness-1429.ivf.md5
+71bcbe5357d36a19df5b07fbe3e27bffa8893f0a  vp80-05-sharpness-1430.ivf.md5
+89a09b1dffce2d55770a89e58d9925c70ef79bf8  vp80-05-sharpness-1431.ivf.md5
+08444a18b4e6ba3450c0796dd728d48c399a2dc9  vp80-05-sharpness-1433.ivf.md5
+6d6223719a90c13e848aa2a8a6642098cdb5977a  vp80-05-sharpness-1434.ivf.md5
+41d70bb5fa45bc88da1604a0af466930b8dd77b5  vp80-05-sharpness-1438.ivf.md5
+086c56378df81b6cee264d7540a7b8f2b405c7a4  vp80-05-sharpness-1439.ivf.md5
+d32dc2c4165eb266ea4c23c14a45459b363def32  vp80-05-sharpness-1440.ivf.md5
+8c69dc3d8e563f56ffab5ad1e400d9e689dd23df  vp80-05-sharpness-1443.ivf.md5

test/test.mk

@@ -0,0 +1,179 @@
+LIBVPX_TEST_SRCS-yes += acm_random.h
+LIBVPX_TEST_SRCS-yes += register_state_check.h
+LIBVPX_TEST_SRCS-yes += test.mk
+LIBVPX_TEST_SRCS-yes += test_libvpx.cc
+LIBVPX_TEST_SRCS-yes += util.h
+LIBVPX_TEST_SRCS-yes += video_source.h
+
+##
+## BLACK BOX TESTS
+##
+## Black box tests only use the public API.
+##
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += altref_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += config_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += cq_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += datarate_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += encode_test_driver.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += encode_test_driver.h
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += error_resilience_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += i420_video_source.h
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += keyframe_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += resize_test.cc
+
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += ../md5_utils.h ../md5_utils.c
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += decode_test_driver.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += decode_test_driver.h
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += ivf_video_source.h
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_DECODER) += test_vector_test.cc
+##
+## WHITE BOX TESTS
+##
+## Whitebox tests invoke functions not exposed via the public API. Certain
+## shared library builds don't make these functions accessible.
+##
+ifeq ($(CONFIG_SHARED),)
+
+# These tests require both the encoder and decoder to be built.
+ifeq ($(CONFIG_VP8_ENCODER)$(CONFIG_VP8_DECODER),yesyes)
+LIBVPX_TEST_SRCS-yes                   += boolcoder_test.cc
+endif
+
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += fdct4x4_test.cc
+LIBVPX_TEST_SRCS-yes                   += idctllm_test.cc
+LIBVPX_TEST_SRCS-yes                   += intrapred_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_POSTPROC)    += pp_filter_test.cc
+LIBVPX_TEST_SRCS-yes                   += sad_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += set_roi.cc
+LIBVPX_TEST_SRCS-yes                   += sixtap_predict_test.cc
+LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += subtract_test.cc
+
+endif
+
+
+##
+## TEST DATA
+##
+LIBVPX_TEST_DATA-$(CONFIG_VP8_ENCODER) += hantro_collage_w352h288.yuv
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-002.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-003.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-004.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-005.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-006.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-007.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-008.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-009.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-010.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-011.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-012.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-013.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-014.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-015.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-016.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-017.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-018.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1400.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1411.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1416.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-01-intra-1417.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1402.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1412.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1418.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-02-inter-1424.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1401.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1403.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1407.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1408.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1409.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1410.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1413.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1414.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1415.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1425.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1426.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1427.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1432.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1435.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1436.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1437.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1441.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-1442.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-01.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-02.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-03.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-03-segmentation-04.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1404.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1405.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-04-partitions-1406.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1428.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1429.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1430.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1431.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1433.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1434.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1438.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1439.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1440.ivf.md5
+LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-05-sharpness-1443.ivf.md5

test/test_libvpx.cc

@@ -0,0 +1,45 @@
+/*
+ *  Copyright (c) 2012 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 <string>
+#include "vpx_config.h"
+#if ARCH_X86 || ARCH_X86_64
+extern "C" {
+#include "vpx_ports/x86.h"
+}
+#endif
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+static void append_gtest_filter(const char *str) {
+  std::string filter = ::testing::FLAGS_gtest_filter;
+  filter += str;
+  ::testing::FLAGS_gtest_filter = filter;
+}
+
+int main(int argc, char **argv) {
+  ::testing::InitGoogleTest(&argc, argv);
+
+#if ARCH_X86 || ARCH_X86_64
+  const int simd_caps = x86_simd_caps();
+  if(!(simd_caps & HAS_MMX))
+    append_gtest_filter(":-MMX/*");
+  if(!(simd_caps & HAS_SSE))
+    append_gtest_filter(":-SSE/*");
+  if(!(simd_caps & HAS_SSE2))
+    append_gtest_filter(":-SSE2/*");
+  if(!(simd_caps & HAS_SSE3))
+    append_gtest_filter(":-SSE3/*");
+  if(!(simd_caps & HAS_SSSE3))
+    append_gtest_filter(":-SSSE3/*");
+  if(!(simd_caps & HAS_SSE4_1))
+    append_gtest_filter(":-SSE4_1/*");
+#endif
+
+  return RUN_ALL_TESTS();
+}

test/test_vector_test.cc

@@ -0,0 +1,144 @@
+/*
+ Copyright (c) 2012 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 <cstdio>
+#include <cstdlib>
+#include <string>
+#include "third_party/googletest/src/include/gtest/gtest.h"
+#include "test/decode_test_driver.h"
+#include "test/ivf_video_source.h"
+extern "C" {
+#include "./md5_utils.h"
+#include "vpx_mem/vpx_mem.h"
+}
+
+#if defined(_MSC_VER)
+#define snprintf sprintf_s
+#endif
+
+namespace {
+// There are 61 test vectors in total.
+const char *kTestVectors[] = {
+  "vp80-00-comprehensive-001.ivf",
+  "vp80-00-comprehensive-002.ivf", "vp80-00-comprehensive-003.ivf",
+  "vp80-00-comprehensive-004.ivf", "vp80-00-comprehensive-005.ivf",
+  "vp80-00-comprehensive-006.ivf", "vp80-00-comprehensive-007.ivf",
+  "vp80-00-comprehensive-008.ivf", "vp80-00-comprehensive-009.ivf",
+  "vp80-00-comprehensive-010.ivf", "vp80-00-comprehensive-011.ivf",
+  "vp80-00-comprehensive-012.ivf", "vp80-00-comprehensive-013.ivf",
+  "vp80-00-comprehensive-014.ivf", "vp80-00-comprehensive-015.ivf",
+  "vp80-00-comprehensive-016.ivf", "vp80-00-comprehensive-017.ivf",
+  "vp80-00-comprehensive-018.ivf", "vp80-01-intra-1400.ivf",
+  "vp80-01-intra-1411.ivf", "vp80-01-intra-1416.ivf",
+  "vp80-01-intra-1417.ivf", "vp80-02-inter-1402.ivf",
+  "vp80-02-inter-1412.ivf", "vp80-02-inter-1418.ivf",
+  "vp80-02-inter-1424.ivf", "vp80-03-segmentation-01.ivf",
+  "vp80-03-segmentation-02.ivf", "vp80-03-segmentation-03.ivf",
+  "vp80-03-segmentation-04.ivf", "vp80-03-segmentation-1401.ivf",
+  "vp80-03-segmentation-1403.ivf", "vp80-03-segmentation-1407.ivf",
+  "vp80-03-segmentation-1408.ivf", "vp80-03-segmentation-1409.ivf",
+  "vp80-03-segmentation-1410.ivf", "vp80-03-segmentation-1413.ivf",
+  "vp80-03-segmentation-1414.ivf", "vp80-03-segmentation-1415.ivf",
+  "vp80-03-segmentation-1425.ivf", "vp80-03-segmentation-1426.ivf",
+  "vp80-03-segmentation-1427.ivf", "vp80-03-segmentation-1432.ivf",
+  "vp80-03-segmentation-1435.ivf", "vp80-03-segmentation-1436.ivf",
+  "vp80-03-segmentation-1437.ivf", "vp80-03-segmentation-1441.ivf",
+  "vp80-03-segmentation-1442.ivf", "vp80-04-partitions-1404.ivf",
+  "vp80-04-partitions-1405.ivf", "vp80-04-partitions-1406.ivf",
+  "vp80-05-sharpness-1428.ivf", "vp80-05-sharpness-1429.ivf",
+  "vp80-05-sharpness-1430.ivf", "vp80-05-sharpness-1431.ivf",
+  "vp80-05-sharpness-1433.ivf", "vp80-05-sharpness-1434.ivf",
+  "vp80-05-sharpness-1438.ivf", "vp80-05-sharpness-1439.ivf",
+  "vp80-05-sharpness-1440.ivf", "vp80-05-sharpness-1443.ivf"
+};
+
+class TestVectorTest : public libvpx_test::DecoderTest,
+    public ::testing::TestWithParam<const char*> {
+ protected:
+  TestVectorTest() : md5_file_(NULL) {}
+
+  virtual ~TestVectorTest() {
+    if (md5_file_)
+      fclose(md5_file_);
+  }
+
+  void OpenMD5File(const std::string& md5_file_name_) {
+    md5_file_ = libvpx_test::OpenTestDataFile(md5_file_name_);
+    ASSERT_TRUE(md5_file_) << "Md5 file open failed. Filename: "
+        << md5_file_name_;
+  }
+
+  virtual void DecompressedFrameHook(const vpx_image_t& img,
+                                     const unsigned int frame_number) {
+    char expected_md5[33];
+    char junk[128];
+
+    // Read correct md5 checksums.
+    const int res = fscanf(md5_file_, "%s  %s", expected_md5, junk);
+    ASSERT_NE(res, EOF) << "Read md5 data failed";
+    expected_md5[32] = '\0';
+
+    MD5Context md5;
+    MD5Init(&md5);
+
+    // Compute and update md5 for each raw in decompressed data.
+    for (int plane = 0; plane < 3; ++plane) {
+      uint8_t *buf = img.planes[plane];
+
+      for (unsigned int y = 0; y < (plane ? (img.d_h + 1) >> 1 : img.d_h);
+           ++y) {
+        MD5Update(&md5, buf, (plane ? (img.d_w + 1) >> 1 : img.d_w));
+        buf += img.stride[plane];
+      }
+    }
+
+    uint8_t md5_sum[16];
+    MD5Final(md5_sum, &md5);
+
+    char actual_md5[33];
+    // Convert to get the actual md5.
+    for (int i = 0; i < 16; i++) {
+      snprintf(&actual_md5[i * 2], sizeof(actual_md5) - i * 2, "%02x",
+               md5_sum[i]);
+    }
+    actual_md5[32] = '\0';
+
+    // Check md5 match.
+    ASSERT_STREQ(expected_md5, actual_md5)
+        << "Md5 checksums don't match: frame number = " << frame_number;
+  }
+
+ private:
+  FILE *md5_file_;
+};
+
+// This test runs through the whole set of test vectors, and decodes them.
+// The md5 checksums are computed for each frame in the video file. If md5
+// checksums match the correct md5 data, then the test is passed. Otherwise,
+// the test failed.
+TEST_P(TestVectorTest, MD5Match) {
+  const std::string filename = GetParam();
+  // Open compressed video file.
+  libvpx_test::IVFVideoSource video(filename);
+
+  video.Init();
+
+  // Construct md5 file name.
+  const std::string md5_filename = filename + ".md5";
+  OpenMD5File(md5_filename);
+
+  // Decode frame, and check the md5 matching.
+  ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
+}
+
+INSTANTIATE_TEST_CASE_P(TestVectorSequence, TestVectorTest,
+                        ::testing::ValuesIn(kTestVectors));
+
+}  // namespace

test/util.h

@@ -0,0 +1,18 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+
+#ifndef TEST_UTIL_H_
+#define TEST_UTIL_H_
+
+// Macros
+#define PARAMS(...) ::testing::TestWithParam< std::tr1::tuple< __VA_ARGS__ > >
+#define GET_PARAM(k) std::tr1::get< k >(GetParam())
+
+#endif  // TEST_UTIL_H_

test/video_source.h

@@ -0,0 +1,175 @@
+/*
+ *  Copyright (c) 2012 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.
+ */
+#ifndef TEST_VIDEO_SOURCE_H_
+#define TEST_VIDEO_SOURCE_H_
+
+#include <cstdio>
+#include <cstdlib>
+#include <string>
+#include "test/acm_random.h"
+#include "vpx/vpx_encoder.h"
+
+namespace libvpx_test {
+
+static FILE *OpenTestDataFile(const std::string& file_name) {
+  std::string path_to_source = file_name;
+  const char *kDataPath = getenv("LIBVPX_TEST_DATA_PATH");
+
+  if (kDataPath) {
+    path_to_source = kDataPath;
+    path_to_source += "/";
+    path_to_source += file_name;
+  }
+
+  return fopen(path_to_source.c_str(), "rb");
+}
+
+// Abstract base class for test video sources, which provide a stream of
+// vpx_image_t images with associated timestamps and duration.
+class VideoSource {
+ public:
+  virtual ~VideoSource() {}
+
+  // Prepare the stream for reading, rewind/open as necessary.
+  virtual void Begin() = 0;
+
+  // Advance the cursor to the next frame
+  virtual void Next() = 0;
+
+  // Get the current video frame, or NULL on End-Of-Stream.
+  virtual vpx_image_t *img() const = 0;
+
+  // Get the presentation timestamp of the current frame.
+  virtual vpx_codec_pts_t pts() const = 0;
+
+  // Get the current frame's duration
+  virtual unsigned long duration() const = 0;
+
+  // Get the timebase for the stream
+  virtual vpx_rational_t timebase() const = 0;
+
+  // Get the current frame counter, starting at 0.
+  virtual unsigned int frame() const = 0;
+
+  // Get the current file limit.
+  virtual unsigned int limit() const = 0;
+};
+
+
+class DummyVideoSource : public VideoSource {
+ public:
+  DummyVideoSource() : img_(NULL), limit_(100), width_(0), height_(0) {
+    SetSize(80, 64);
+  }
+
+  virtual ~DummyVideoSource() { vpx_img_free(img_); }
+
+  virtual void Begin() {
+    frame_ = 0;
+    FillFrame();
+  }
+
+  virtual void Next() {
+    ++frame_;
+    FillFrame();
+  }
+
+  virtual vpx_image_t *img() const {
+    return (frame_ < limit_) ? img_ : NULL;
+  }
+
+  // Models a stream where Timebase = 1/FPS, so pts == frame.
+  virtual vpx_codec_pts_t pts() const { return frame_; }
+
+  virtual unsigned long duration() const { return 1; }
+
+  virtual vpx_rational_t timebase() const {
+    const vpx_rational_t t = {1, 30};
+    return t;
+  }
+
+  virtual unsigned int frame() const { return frame_; }
+
+  virtual unsigned int limit() const { return limit_; }
+
+  void SetSize(unsigned int width, unsigned int height) {
+    if (width != width_ || height != height_) {
+      vpx_img_free(img_);
+      raw_sz_ = ((width + 31)&~31) * height * 3 / 2;
+      img_ = vpx_img_alloc(NULL, VPX_IMG_FMT_VPXI420, width, height, 32);
+      width_ = width;
+      height_ = height;
+    }
+  }
+
+ protected:
+  virtual void FillFrame() { memset(img_->img_data, 0, raw_sz_); }
+
+  vpx_image_t *img_;
+  size_t       raw_sz_;
+  unsigned int limit_;
+  unsigned int frame_;
+  unsigned int width_;
+  unsigned int height_;
+};
+
+
+class RandomVideoSource : public DummyVideoSource {
+ public:
+  RandomVideoSource(int seed = ACMRandom::DeterministicSeed())
+      : rnd_(seed),
+        seed_(seed) { }
+
+ protected:
+  // Reset the RNG to get a matching stream for the second pass
+  virtual void Begin() {
+    frame_ = 0;
+    rnd_.Reset(seed_);
+    FillFrame();
+  }
+
+  // 15 frames of noise, followed by 15 static frames. Reset to 0 rather
+  // than holding previous frames to encourage keyframes to be thrown.
+  virtual void FillFrame() {
+    if (frame_ % 30 < 15)
+      for (size_t i = 0; i < raw_sz_; ++i)
+        img_->img_data[i] = rnd_.Rand8();
+    else
+      memset(img_->img_data, 0, raw_sz_);
+  }
+
+  ACMRandom rnd_;
+  int seed_;
+};
+
+// Abstract base class for test video sources, which provide a stream of
+// decompressed images to the decoder.
+class CompressedVideoSource {
+ public:
+  virtual ~CompressedVideoSource() {}
+
+  virtual void Init() = 0;
+
+  // Prepare the stream for reading, rewind/open as necessary.
+  virtual void Begin() = 0;
+
+  // Advance the cursor to the next frame
+  virtual void Next() = 0;
+
+  virtual const uint8_t *cxdata() const = 0;
+
+  virtual const unsigned int frame_size() const = 0;
+
+  virtual const unsigned int frame_number() const = 0;
+};
+
+}  // namespace libvpx_test
+
+#endif  // TEST_VIDEO_SOURCE_H_

third_party/libyuv/source/scale.c

@@ -60,7 +60,7 @@ void SetUseReferenceImpl(int use) {
 
 #if defined(__ARM_NEON__) && !defined(YUV_DISABLE_ASM)
 #define HAS_SCALEROWDOWN2_NEON
-void ScaleRowDown2_NEON(const uint8* src_ptr, int /* src_stride */,
+void ScaleRowDown2_NEON(const uint8* src_ptr, int  src_stride,
                         uint8* dst, int dst_width) {
   asm volatile (
     "1:                                        \n"
@@ -102,7 +102,7 @@ void ScaleRowDown2Int_NEON(const uint8* src_ptr, int src_stride,
 }
 
 #define HAS_SCALEROWDOWN4_NEON
-static void ScaleRowDown4_NEON(const uint8* src_ptr, int /* src_stride */,
+static void ScaleRowDown4_NEON(const uint8* src_ptr, int src_stride,
                                uint8* dst_ptr, int dst_width) {
   asm volatile (
     "1:                                        \n"
@@ -160,7 +160,7 @@ static void ScaleRowDown4Int_NEON(const uint8* src_ptr, int src_stride,
 // Down scale from 4 to 3 pixels.  Use the neon multilane read/write
 //  to load up the every 4th pixel into a 4 different registers.
 // Point samples 32 pixels to 24 pixels.
-static void ScaleRowDown34_NEON(const uint8* src_ptr, int /* src_stride */,
+static void ScaleRowDown34_NEON(const uint8* src_ptr, int src_stride,
                                 uint8* dst_ptr, int dst_width) {
   asm volatile (
     "1:                                        \n"
@@ -284,7 +284,7 @@ const unsigned short mult38_div9[8] __attribute__ ((aligned(16))) =
     65536 / 18, 65536 / 18, 65536 / 18, 65536 / 18 };
 
 // 32 -> 12
-static void ScaleRowDown38_NEON(const uint8* src_ptr, int,
+static void ScaleRowDown38_NEON(const uint8* src_ptr, int src_stride,
                                 uint8* dst_ptr, int dst_width) {
   asm volatile (
     "vld1.u8      {q3}, [%3]                   \n"

tools/ftfy.sh

@@ -34,7 +34,7 @@ vpx_style() {
          --align-pointer=name \
          --indent-preprocessor --convert-tabs --indent-labels \
          --suffix=none --quiet "$@"
-  sed -i 's/[[:space:]]\{1,\},/,/g' "$@"
+  sed -i "" 's/[[:space:]]\{1,\},/,/g' "$@"
 }
 
 

vp8/common/alloccommon.c

@@ -17,23 +17,6 @@
 #include "entropymode.h"
 #include "systemdependent.h"
 
-
-extern  void vp8_init_scan_order_mask();
-
-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));
-    }
-}
-
 void vp8_de_alloc_frame_buffers(VP8_COMMON *oci)
 {
     int i;
@@ -45,16 +28,20 @@ void vp8_de_alloc_frame_buffers(VP8_COMMON *oci)
     vp8_yv12_de_alloc_frame_buffer(&oci->post_proc_buffer);
     if (oci->post_proc_buffer_int_used)
         vp8_yv12_de_alloc_frame_buffer(&oci->post_proc_buffer_int);
+
+    vpx_free(oci->pp_limits_buffer);
+    oci->pp_limits_buffer = NULL;
 #endif
 
     vpx_free(oci->above_context);
     vpx_free(oci->mip);
+#if CONFIG_ERROR_CONCEALMENT
     vpx_free(oci->prev_mip);
+    oci->prev_mip = NULL;
+#endif
 
-    oci->above_context = 0;
-    oci->mip = 0;
-    oci->prev_mip = 0;
-
+    oci->above_context = NULL;
+    oci->mip = NULL;
 }
 
 int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
@@ -76,10 +63,7 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
         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;
-        }
+            goto allocation_fail;
     }
 
     oci->new_fb_idx = 0;
@@ -93,22 +77,7 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
     oci->fb_idx_ref_cnt[3] = 1;
 
     if (vp8_yv12_alloc_frame_buffer(&oci->temp_scale_frame,   width, 16, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return 1;
-    }
-
-#if CONFIG_POSTPROC
-    if (vp8_yv12_alloc_frame_buffer(&oci->post_proc_buffer, width, height, VP8BORDERINPIXELS) < 0)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return 1;
-    }
-
-    oci->post_proc_buffer_int_used = 0;
-    vpx_memset(&oci->postproc_state, 0, sizeof(oci->postproc_state));
-    vpx_memset((&oci->post_proc_buffer)->buffer_alloc,128,(&oci->post_proc_buffer)->frame_size);
-#endif
+        goto allocation_fail;
 
     oci->mb_rows = height >> 4;
     oci->mb_cols = width >> 4;
@@ -117,44 +86,43 @@ int vp8_alloc_frame_buffers(VP8_COMMON *oci, int width, int height)
     oci->mip = vpx_calloc((oci->mb_cols + 1) * (oci->mb_rows + 1), sizeof(MODE_INFO));
 
     if (!oci->mip)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return 1;
-    }
+        goto allocation_fail;
 
     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
+    /* Allocation of previous mode info will be done in vp8_decode_frame()
+     * as it is a decoder only data */
 
     oci->above_context = vpx_calloc(sizeof(ENTROPY_CONTEXT_PLANES) * oci->mb_cols, 1);
 
     if (!oci->above_context)
-    {
-        vp8_de_alloc_frame_buffers(oci);
-        return 1;
-    }
+        goto allocation_fail;
 
-    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);
+#if CONFIG_POSTPROC
+    if (vp8_yv12_alloc_frame_buffer(&oci->post_proc_buffer, width, height, VP8BORDERINPIXELS) < 0)
+        goto allocation_fail;
+
+    oci->post_proc_buffer_int_used = 0;
+    vpx_memset(&oci->postproc_state, 0, sizeof(oci->postproc_state));
+    vpx_memset(oci->post_proc_buffer.buffer_alloc, 128,
+               oci->post_proc_buffer.frame_size);
+
+    /* Allocate buffer to store post-processing filter coefficients.
+     *
+     * Note: Round up mb_cols to support SIMD reads
+     */
+    oci->pp_limits_buffer = vpx_memalign(16, 24 * ((oci->mb_cols + 1) & ~1));
+    if (!oci->pp_limits_buffer)
+        goto allocation_fail;
 #endif
 
     return 0;
+
+allocation_fail:
+    vp8_de_alloc_frame_buffers(oci);
+    return 1;
 }
+
 void vp8_setup_version(VP8_COMMON *cm)
 {
     switch (cm->version)

vp8/common/arm/armv6/intra4x4_predict_v6.asm

@@ -18,15 +18,23 @@
     AREA ||.text||, CODE, READONLY, ALIGN=2
 
 
-;void vp8_intra4x4_predict(unsigned char *src, int src_stride, int b_mode,
-;                          unsigned char *dst, int dst_stride)
+;void vp8_intra4x4_predict_armv6(unsigned char *Above, unsigned char *yleft,
+;                                B_PREDICTION_MODE left_stride, int b_mode,
+;                                unsigned char *dst, int dst_stride,
+;                                unsigned char top_left)
 
+; r0: *Above
+; r1: *yleft
+; r2: left_stride
+; r3: b_mode
+; sp + #40: dst
+; sp + #44: dst_stride
+; sp + #48: top_left
 |vp8_intra4x4_predict_armv6| PROC
     push        {r4-r12, lr}
 
-
-    cmp         r2, #10
-    addlt       pc, pc, r2, lsl #2       ; position independent switch
+    cmp         r3, #10
+    addlt       pc, pc, r3, lsl #2       ; position independent switch
     pop         {r4-r12, pc}             ; default
     b           b_dc_pred
     b           b_tm_pred
@@ -41,13 +49,13 @@
 
 b_dc_pred
     ; load values
-    ldr         r8, [r0, -r1]            ; Above
-    ldrb        r4, [r0, #-1]!           ; Left[0]
+    ldr         r8, [r0]                 ; Above
+    ldrb        r4, [r1], r2             ; Left[0]
     mov         r9, #0
-    ldrb        r5, [r0, r1]             ; Left[1]
-    ldrb        r6, [r0, r1, lsl #1]!    ; Left[2]
+    ldrb        r5, [r1], r2             ; Left[1]
+    ldrb        r6, [r1], r2             ; Left[2]
     usad8       r12, r8, r9
-    ldrb        r7, [r0, r1]             ; Left[3]
+    ldrb        r7, [r1]                 ; Left[3]
 
     ; calculate dc
     add         r4, r4, r5
@@ -55,31 +63,30 @@ b_dc_pred
     add         r4, r4, r7
     add         r4, r4, r12
     add         r4, r4, #4
-    ldr         r0, [sp, #40]           ; load stride
+    ldr         r0, [sp, #44]           ; dst_stride
     mov         r12, r4, asr #3         ; (expected_dc + 4) >> 3
 
     add         r12, r12, r12, lsl #8
-    add         r3, r3, r0
+    ldr         r3, [sp, #40]           ; dst
     add         r12, r12, r12, lsl #16
 
     ; store values
-    str         r12, [r3, -r0]
+    str         r12, [r3], r0
+    str         r12, [r3], r0
+    str         r12, [r3], r0
     str         r12, [r3]
-    str         r12, [r3, r0]
-    str         r12, [r3, r0, lsl #1]
 
     pop        {r4-r12, pc}
 
 b_tm_pred
-    sub         r10, r0, #1             ; Left
-    ldr         r8, [r0, -r1]           ; Above
-    ldrb        r9, [r10, -r1]          ; top_left
-    ldrb        r4, [r0, #-1]!          ; Left[0]
-    ldrb        r5, [r10, r1]!          ; Left[1]
-    ldrb        r6, [r0, r1, lsl #1]    ; Left[2]
-    ldrb        r7, [r10, r1, lsl #1]   ; Left[3]
-    ldr         r0, [sp, #40]           ; load stride
-
+    ldr         r8, [r0]                ; Above
+    ldrb        r9, [sp, #48]           ; top_left
+    ldrb        r4, [r1], r2            ; Left[0]
+    ldrb        r5, [r1], r2            ; Left[1]
+    ldrb        r6, [r1], r2            ; Left[2]
+    ldrb        r7, [r1]                ; Left[3]
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     add         r9, r9, r9, lsl #16     ; [tl|tl]
     uxtb16      r10, r8                 ; a[2|0]
@@ -126,25 +133,26 @@ b_tm_pred
     str         r12, [r3], r0
 
     add         r12, r4, r5, lsl #8     ; [3|2|1|0]
-    str         r12, [r3], r0
+    str         r12, [r3]
 
     pop        {r4-r12, pc}
 
 b_ve_pred
-    ldr         r8, [r0, -r1]!          ; a[3|2|1|0]
+    ldr         r8, [r0]                ; a[3|2|1|0]
     ldr         r11, c00FF00FF
-    ldrb        r9, [r0, #-1]           ; top_left
+    ldrb        r9, [sp, #48]           ; top_left
     ldrb        r10, [r0, #4]           ; a[4]
 
     ldr         r0, c00020002
 
     uxtb16      r4, r8                  ; a[2|0]
     uxtb16      r5, r8, ror #8          ; a[3|1]
-    ldr         r2, [sp, #40]           ; stride
+    ldr         r2, [sp, #44]           ; dst_stride
     pkhbt       r9, r9, r5, lsl #16     ; a[1|-1]
 
     add         r9, r9, r4, lsl #1      ;[a[1]+2*a[2]       | tl+2*a[0]       ]
     uxtab16     r9, r9, r5              ;[a[1]+2*a[2]+a[3]  | tl+2*a[0]+a[1]  ]
+    ldr         r3, [sp, #40]           ; dst
     uxtab16     r9, r9, r0              ;[a[1]+2*a[2]+a[3]+2| tl+2*a[0]+a[1]+2]
 
     add         r0, r0, r10, lsl #16    ;[a[4]+2            |                 2]
@@ -154,25 +162,23 @@ b_ve_pred
 
     and         r9, r11, r9, asr #2
     and         r4, r11, r4, asr #2
-    add         r3, r3, r2              ; dst + dst_stride
     add         r9, r9, r4, lsl #8
 
     ; store values
-    str         r9, [r3, -r2]
+    str         r9, [r3], r2
+    str         r9, [r3], r2
+    str         r9, [r3], r2
     str         r9, [r3]
-    str         r9, [r3, r2]
-    str         r9, [r3, r2, lsl #1]
 
     pop        {r4-r12, pc}
 
 
 b_he_pred
-    sub         r10, r0, #1             ; Left
-    ldrb        r4, [r0, #-1]!          ; Left[0]
-    ldrb        r8, [r10, -r1]          ; top_left
-    ldrb        r5, [r10, r1]!          ; Left[1]
-    ldrb        r6, [r0, r1, lsl #1]    ; Left[2]
-    ldrb        r7, [r10, r1, lsl #1]   ; Left[3]
+    ldrb        r4, [r1], r2            ; Left[0]
+    ldrb        r8, [sp, #48]           ; top_left
+    ldrb        r5, [r1], r2            ; Left[1]
+    ldrb        r6, [r1], r2            ; Left[2]
+    ldrb        r7, [r1]                ; Left[3]
 
     add         r8, r8, r4              ; tl   + l[0]
     add         r9, r4, r5              ; l[0] + l[1]
@@ -197,7 +203,8 @@ b_he_pred
     pkhtb       r10, r10, r10, asr #16  ; l[-|2|-|2]
     pkhtb       r11, r11, r11, asr #16  ; l[-|3|-|3]
 
-    ldr         r0, [sp, #40]           ; stride
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     add         r8, r8, r8, lsl #8      ; l[0|0|0|0]
     add         r9, r9, r9, lsl #8      ; l[1|1|1|1]
@@ -206,16 +213,16 @@ b_he_pred
 
     ; store values
     str         r8, [r3], r0
-    str         r9, [r3]
-    str         r10, [r3, r0]
-    str         r11, [r3, r0, lsl #1]
+    str         r9, [r3], r0
+    str         r10, [r3], r0
+    str         r11, [r3]
 
     pop        {r4-r12, pc}
 
 b_ld_pred
-    ldr         r4, [r0, -r1]!          ; Above
+    ldr         r4, [r0]                ; Above[0-3]
     ldr         r12, c00020002
-    ldr         r5, [r0, #4]
+    ldr         r5, [r0, #4]            ; Above[4-7]
     ldr         lr,  c00FF00FF
 
     uxtb16      r6, r4                  ; a[2|0]
@@ -225,7 +232,6 @@ b_ld_pred
     pkhtb       r10, r6, r8             ; a[2|4]
     pkhtb       r11, r7, r9             ; a[3|5]
 
-
     add         r4, r6, r7, lsl #1      ; [a2+2*a3      |      a0+2*a1]
     add         r4, r4, r10, ror #16    ; [a2+2*a3+a4   |   a0+2*a1+a2]
     uxtab16     r4, r4, r12             ; [a2+2*a3+a4+2 | a0+2*a1+a2+2]
@@ -244,7 +250,8 @@ b_ld_pred
     add         r7, r7, r9, asr #16     ; [                 a5+2*a6+a7]
     uxtah       r7, r7, r12             ; [               a5+2*a6+a7+2]
 
-    ldr         r0, [sp, #40]           ; stride
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     ; scale down
     and         r4, lr, r4, asr #2
@@ -266,18 +273,17 @@ b_ld_pred
     mov         r6, r6, lsr #16
     mov         r11, r10, lsr #8
     add         r11, r11, r6, lsl #24   ; [6|5|4|3]
-    str         r11, [r3], r0
+    str         r11, [r3]
 
     pop        {r4-r12, pc}
 
 b_rd_pred
-    sub         r12, r0, r1             ; Above = src - src_stride
-    ldrb        r7, [r0, #-1]!          ; l[0] = pp[3]
-    ldr         lr, [r12]               ; Above = pp[8|7|6|5]
-    ldrb        r8, [r12, #-1]!         ; tl   = pp[4]
-    ldrb        r6, [r12, r1, lsl #1]   ; l[1] = pp[2]
-    ldrb        r5, [r0, r1, lsl #1]    ; l[2] = pp[1]
-    ldrb        r4, [r12, r1, lsl #2]   ; l[3] = pp[0]
+    ldrb        r7, [r1], r2            ; l[0] = pp[3]
+    ldr         lr, [r0]                ; Above = pp[8|7|6|5]
+    ldrb        r8, [sp, #48]           ; tl   = pp[4]
+    ldrb        r6, [r1], r2            ; l[1] = pp[2]
+    ldrb        r5, [r1], r2            ; l[2] = pp[1]
+    ldrb        r4, [r1], r2            ; l[3] = pp[0]
 
 
     uxtb16      r9, lr                  ; p[7|5]
@@ -307,7 +313,8 @@ b_rd_pred
     add         r7, r7, r10             ; [p6+2*p7+p8   |   p4+2*p5+p6]
     uxtab16     r7, r7, r12             ; [p6+2*p7+p8+2 | p4+2*p5+p6+2]
 
-    ldr         r0, [sp, #40]           ; stride
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     ; scale down
     and         r7, lr, r7, asr #2
@@ -328,18 +335,17 @@ b_rd_pred
 
     mov         r11, r10, lsl #8        ; [3|2|1|-]
     uxtab       r11, r11, r4            ; [3|2|1|0]
-    str         r11, [r3], r0
+    str         r11, [r3]
 
     pop        {r4-r12, pc}
 
 b_vr_pred
-    sub         r12, r0, r1             ; Above = src - src_stride
-    ldrb        r7, [r0, #-1]!          ; l[0] = pp[3]
-    ldr         lr, [r12]               ; Above = pp[8|7|6|5]
-    ldrb        r8, [r12, #-1]!         ; tl   = pp[4]
-    ldrb        r6, [r12, r1, lsl #1]   ; l[1] = pp[2]
-    ldrb        r5, [r0, r1, lsl #1]    ; l[2] = pp[1]
-    ldrb        r4, [r12, r1, lsl #2]   ; l[3] = pp[0]
+    ldrb        r7, [r1], r2            ; l[0] = pp[3]
+    ldr         lr, [r0]                ; Above = pp[8|7|6|5]
+    ldrb        r8, [sp, #48]           ; tl   = pp[4]
+    ldrb        r6, [r1], r2            ; l[1] = pp[2]
+    ldrb        r5, [r1], r2            ; l[2] = pp[1]
+    ldrb        r4, [r1]                ; l[3] = pp[0]
 
     add         r5, r5, r7, lsl #16     ; p[3|1]
     add         r6, r6, r8, lsl #16     ; p[4|2]
@@ -376,7 +382,8 @@ b_vr_pred
     add         r8, r8, r10             ; [p6+2*p7+p8   |   p4+2*p5+p6]
     uxtab16     r8, r8, r12             ; [p6+2*p7+p8+2 | p4+2*p5+p6+2]
 
-    ldr         r0, [sp, #40]           ; stride
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     ; scale down
     and         r5, lr, r5, asr #2      ; [B|A]
@@ -397,14 +404,14 @@ b_vr_pred
     pkhtb       r10, r7, r5, asr #16    ; [-|H|-|B]
     str         r2, [r3], r0
     add         r12, r12, r10, lsl #8   ; [H|D|B|A]
-    str         r12, [r3], r0
+    str         r12, [r3]
 
     pop        {r4-r12, pc}
 
 b_vl_pred
-    ldr         r4, [r0, -r1]!          ; [3|2|1|0]
+    ldr         r4, [r0]                ; [3|2|1|0] = Above[0-3]
     ldr         r12, c00020002
-    ldr         r5, [r0, #4]            ; [7|6|5|4]
+    ldr         r5, [r0, #4]            ; [7|6|5|4] = Above[4-7]
     ldr         lr,  c00FF00FF
     ldr         r2,  c00010001
 
@@ -441,7 +448,8 @@ b_vl_pred
     add         r9, r9, r11             ; [p5+2*p6+p7   |   p3+2*p4+p5]
     uxtab16     r9, r9, r12             ; [p5+2*p6+p7+2 | p3+2*p4+p5+2]
 
-    ldr         r0, [sp, #40]           ; stride
+    ldr         r0, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     ; scale down
     and         r5, lr, r5, asr #2      ; [D|C]
@@ -449,7 +457,6 @@ b_vl_pred
     and         r8, lr, r8, asr #2      ; [I|D]
     and         r9, lr, r9, asr #2      ; [J|H]
 
-
     add         r10, r4, r6, lsl #8     ; [F|B|E|A]
     str         r10, [r3], r0
 
@@ -463,18 +470,17 @@ b_vl_pred
     str         r12, [r3], r0
 
     add         r10, r7, r10, lsl #8    ; [J|H|D|G]
-    str         r10, [r3], r0
+    str         r10, [r3]
 
     pop        {r4-r12, pc}
 
 b_hd_pred
-    sub         r12, r0, r1             ; Above = src - src_stride
-    ldrb        r7, [r0, #-1]!          ; l[0] = pp[3]
-    ldr         lr, [r12]               ; Above = pp[8|7|6|5]
-    ldrb        r8, [r12, #-1]!         ; tl   = pp[4]
-    ldrb        r6, [r0, r1]            ; l[1] = pp[2]
-    ldrb        r5, [r0, r1, lsl #1]    ; l[2] = pp[1]
-    ldrb        r4, [r12, r1, lsl #2]   ; l[3] = pp[0]
+    ldrb        r7, [r1], r2            ; l[0] = pp[3]
+    ldr         lr, [r0]                ; Above = pp[8|7|6|5]
+    ldrb        r8, [sp, #48]           ; tl   = pp[4]
+    ldrb        r6, [r1], r2            ; l[1] = pp[2]
+    ldrb        r5, [r1], r2            ; l[2] = pp[1]
+    ldrb        r4, [r1]                ; l[3] = pp[0]
 
     uxtb16      r9, lr                  ; p[7|5]
     uxtb16      r10, lr, ror #8         ; p[8|6]
@@ -492,7 +498,6 @@ b_hd_pred
     pkhtb       r1, r9, r10             ; p[7|6]
     pkhbt       r10, r8, r10, lsl #16   ; p[6|5]
 
-
     uadd16      r11, r4, r5             ; [p1+p2        |        p0+p1]
     uhadd16     r11, r11, r2            ; [(p1+p2+1)>>1 | (p0+p1+1)>>1]
                                         ; [B|A]
@@ -518,7 +523,8 @@ b_hd_pred
     and         r5, lr, r5, asr #2      ; [H|G]
     and         r6, lr, r6, asr #2      ; [J|I]
 
-    ldr         lr, [sp, #40]           ; stride
+    ldr         lr, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
 
     pkhtb       r2, r0, r6              ; [-|F|-|I]
     pkhtb       r12, r6, r5, asr #16    ; [-|J|-|H]
@@ -527,7 +533,6 @@ b_hd_pred
     mov         r12, r12, ror #24       ; [J|I|H|F]
     str         r12, [r3], lr
 
-
     mov         r7, r11, asr #16        ; [-|-|-|B]
     str         r2, [r3], lr
     add         r7, r7, r0, lsl #16     ; [-|E|-|B]
@@ -536,21 +541,20 @@ b_hd_pred
     str         r7, [r3], lr
 
     add         r5, r11, r4, lsl #8     ; [D|B|C|A]
-    str         r5, [r3], lr
+    str         r5, [r3]
 
     pop        {r4-r12, pc}
 
 
 
 b_hu_pred
-    ldrb        r4, [r0, #-1]!          ; Left[0]
+    ldrb        r4, [r1], r2            ; Left[0]
     ldr         r12, c00020002
-    ldrb        r5, [r0, r1]!           ; Left[1]
+    ldrb        r5, [r1], r2            ; Left[1]
     ldr         lr,  c00FF00FF
-    ldrb        r6, [r0, r1]!           ; Left[2]
+    ldrb        r6, [r1], r2            ; Left[2]
     ldr         r2,  c00010001
-    ldrb        r7, [r0, r1]            ; Left[3]
-
+    ldrb        r7, [r1]                ; Left[3]
 
     add         r4, r4, r5, lsl #16     ; [1|0]
     add         r5, r5, r6, lsl #16     ; [2|1]
@@ -563,7 +567,8 @@ b_hu_pred
     add         r4, r4, r5, lsl #1      ; [p1+2*p2      |      p0+2*p1]
     add         r4, r4, r9              ; [p1+2*p2+p3   |   p0+2*p1+p2]
     uxtab16     r4, r4, r12             ; [p1+2*p2+p3+2 | p0+2*p1+p2+2]
-    ldr         r2, [sp, #40]           ; stride
+    ldr         r2, [sp, #44]           ; dst_stride
+    ldr         r3, [sp, #40]           ; dst
     and         r4, lr, r4, asr #2      ; [D|C]
 
     add         r10, r6, r7             ; [p2+p3]
@@ -587,9 +592,9 @@ b_hu_pred
 
     add         r10, r11, lsl #8        ; [-|-|F|E]
     add         r10, r10, r9, lsl #16   ; [G|G|F|E]
-    str         r10, [r3]
+    str         r10, [r3], r2
 
-    str         r7, [r3, r2]
+    str         r7, [r3]
 
     pop        {r4-r12, pc}
 

vp8/common/blockd.h

@@ -161,22 +161,32 @@ typedef struct
     uint8_t segment_id;                  /* Which set of segmentation parameters should be used for this MB */
 } MB_MODE_INFO;
 
-typedef struct
+typedef struct modeinfo
 {
     MB_MODE_INFO mbmi;
     union b_mode_info bmi[16];
 } MODE_INFO;
 
 #if CONFIG_MULTI_RES_ENCODING
-/* The information needed to be stored for higher-resolution encoder */
+/* The mb-level information needed to be stored for higher-resolution encoder */
 typedef struct
 {
     MB_PREDICTION_MODE mode;
     MV_REFERENCE_FRAME ref_frame;
     int_mv mv;
-    //union b_mode_info bmi[16];
-    int dissim;    // dissimilarity level of the macroblock
-} LOWER_RES_INFO;
+    int dissim;    /* dissimilarity level of the macroblock */
+} LOWER_RES_MB_INFO;
+
+/* The frame-level information needed to be stored for higher-resolution
+ *  encoder */
+typedef struct
+{
+    FRAME_TYPE frame_type;
+    int is_frame_dropped;
+    /* The frame number of each reference frames */
+    unsigned int low_res_ref_frames[MAX_REF_FRAMES];
+    LOWER_RES_MB_INFO *mb_info;
+} LOWER_RES_FRAME_INFO;
 #endif
 
 typedef struct blockd

vp8/common/entropy.c

@@ -101,7 +101,7 @@ const vp8_tree_index vp8_coef_tree[ 22] =     /* corresponding _CONTEXT_NODEs */
 /* vp8_coef_encodings generated with:
     vp8_tokens_from_tree(vp8_coef_encodings, vp8_coef_tree);
 */
-const vp8_token vp8_coef_encodings[MAX_ENTROPY_TOKENS] =
+vp8_token vp8_coef_encodings[MAX_ENTROPY_TOKENS] =
 {
     {2, 2},
     {6, 3},

vp8/common/entropymode.c

@@ -160,9 +160,7 @@ const vp8_tree_index vp8_small_mvtree [14] =
 void vp8_init_mbmode_probs(VP8_COMMON *x)
 {
     vpx_memcpy(x->fc.ymode_prob, vp8_ymode_prob, sizeof(vp8_ymode_prob));
-    vpx_memcpy(x->kf_ymode_prob, vp8_kf_ymode_prob, sizeof(vp8_kf_ymode_prob));
     vpx_memcpy(x->fc.uv_mode_prob, vp8_uv_mode_prob, sizeof(vp8_uv_mode_prob));
-    vpx_memcpy(x->kf_uv_mode_prob, vp8_kf_uv_mode_prob, sizeof(vp8_kf_uv_mode_prob));
     vpx_memcpy(x->fc.sub_mv_ref_prob, sub_mv_ref_prob, sizeof(sub_mv_ref_prob));
 }
 
@@ -171,7 +169,3 @@ void vp8_default_bmode_probs(vp8_prob p [VP8_BINTRAMODES-1])
     vpx_memcpy(p, vp8_bmode_prob, sizeof(vp8_bmode_prob));
 }
 
-void vp8_kf_default_bmode_probs(vp8_prob p [VP8_BINTRAMODES] [VP8_BINTRAMODES] [VP8_BINTRAMODES-1])
-{
-    vpx_memcpy(p, vp8_kf_bmode_prob, sizeof(vp8_kf_bmode_prob));
-}

vp8/common/entropymode.h

@@ -24,11 +24,11 @@ typedef enum
     SUBMVREF_LEFT_ABOVE_ZED
 } sumvfref_t;
 
-typedef const int vp8_mbsplit[16];
+typedef int vp8_mbsplit[16];
 
 #define VP8_NUMMBSPLITS 4
 
-extern vp8_mbsplit vp8_mbsplits [VP8_NUMMBSPLITS];
+extern const vp8_mbsplit vp8_mbsplits [VP8_NUMMBSPLITS];
 
 extern const int vp8_mbsplit_count [VP8_NUMMBSPLITS];    /* # of subsets */
 
@@ -67,9 +67,14 @@ extern const vp8_tree_index vp8_small_mvtree[];
 
 extern const struct vp8_token_struct vp8_small_mvencodings[8];
 
-void vp8_init_mbmode_probs(VP8_COMMON *x);
+/* Key frame default mode probs */
+extern const vp8_prob vp8_kf_bmode_prob[VP8_BINTRAMODES][VP8_BINTRAMODES]
+[VP8_BINTRAMODES-1];
+extern const vp8_prob vp8_kf_uv_mode_prob[VP8_UV_MODES-1];
+extern const vp8_prob vp8_kf_ymode_prob[VP8_YMODES-1];
 
-void   vp8_default_bmode_probs(vp8_prob dest [VP8_BINTRAMODES-1]);
+void vp8_init_mbmode_probs(VP8_COMMON *x);
+void vp8_default_bmode_probs(vp8_prob dest [VP8_BINTRAMODES-1]);
 void vp8_kf_default_bmode_probs(vp8_prob dest [VP8_BINTRAMODES] [VP8_BINTRAMODES] [VP8_BINTRAMODES-1]);
 
 #endif

vp8/common/extend.c

@@ -116,7 +116,7 @@ void vp8_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
     int src_uv_offset = ((srcy * src->uv_stride) >> 1) + (srcx >> 1);
     int dst_uv_offset = ((srcy * dst->uv_stride) >> 1) + (srcx >> 1);
 
-    // If the side is not touching the bounder then don't extend.
+    /* If the side is not touching the bounder then don't extend. */
     if (srcy)
       et = 0;
     if (srcx)
@@ -157,7 +157,10 @@ void vp8_copy_and_extend_frame_with_rect(YV12_BUFFER_CONFIG *src,
 
 
 /* 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)
+void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf,
+                       unsigned char *YPtr,
+                       unsigned char *UPtr,
+                       unsigned char *VPtr)
 {
     int i;
 

vp8/common/filter.h

@@ -19,4 +19,4 @@
 extern const short vp8_bilinear_filters[8][2];
 extern const short vp8_sub_pel_filters[8][6];
 
-#endif //FILTER_H
+#endif

vp8/common/idctllm_test.cc

@@ -1,31 +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.
- */
-
-
- extern "C" {
-    void vp8_short_idct4x4llm_c(short *input, unsigned char *pred_ptr,
-                            int pred_stride, unsigned char *dst_ptr,
-                            int dst_stride);
-}
-
-#include "vpx_config.h"
-#include "idctllm_test.h"
-namespace
-{
-
-INSTANTIATE_TEST_CASE_P(C, IDCTTest,
-                        ::testing::Values(vp8_short_idct4x4llm_c));
-
-} // namespace
-
-int main(int argc, char **argv) {
-  ::testing::InitGoogleTest(&argc, argv);
-  return RUN_ALL_TESTS();
-}

vp8/common/loopfilter.c

@@ -196,6 +196,114 @@ void vp8_loop_filter_frame_init(VP8_COMMON *cm,
     }
 }
 
+
+void vp8_loop_filter_row_normal(VP8_COMMON *cm, MODE_INFO *mode_info_context,
+                         int mb_row, int post_ystride, int post_uvstride,
+                         unsigned char *y_ptr, unsigned char *u_ptr,
+                         unsigned char *v_ptr)
+{
+    int mb_col;
+    int filter_level;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+    FRAME_TYPE frame_type = cm->frame_type;
+
+    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+    {
+        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                        mode_info_context->mbmi.mode != SPLITMV &&
+                        mode_info_context->mbmi.mb_skip_coeff);
+
+        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)
+        {
+            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 (mb_col > 0)
+                vp8_loop_filter_mbv
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            if (!skip_lf)
+                vp8_loop_filter_bv
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            /* don't apply across umv border */
+            if (mb_row > 0)
+                vp8_loop_filter_mbh
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            if (!skip_lf)
+                vp8_loop_filter_bh
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+        }
+
+        y_ptr += 16;
+        u_ptr += 8;
+        v_ptr += 8;
+
+        mode_info_context++;     /* step to next MB */
+    }
+
+}
+
+void vp8_loop_filter_row_simple(VP8_COMMON *cm, MODE_INFO *mode_info_context,
+                         int mb_row, int post_ystride, int post_uvstride,
+                         unsigned char *y_ptr, unsigned char *u_ptr,
+                         unsigned char *v_ptr)
+{
+    int mb_col;
+    int filter_level;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+
+    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+    {
+        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                        mode_info_context->mbmi.mode != SPLITMV &&
+                        mode_info_context->mbmi.mb_skip_coeff);
+
+        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)
+                vp8_loop_filter_simple_mbv
+                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
+
+            if (!skip_lf)
+                vp8_loop_filter_simple_bv
+                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
+
+            /* don't apply across umv border */
+            if (mb_row > 0)
+                vp8_loop_filter_simple_mbh
+                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
+
+            if (!skip_lf)
+                vp8_loop_filter_simple_bh
+                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
+        }
+
+        y_ptr += 16;
+        u_ptr += 8;
+        v_ptr += 8;
+
+        mode_info_context++;     /* step to next MB */
+    }
+
+}
 void vp8_loop_filter_frame(VP8_COMMON *cm,
                            MACROBLOCKD *mbd,
                            int frame_type)

vp8/common/loopfilter.h

@@ -69,6 +69,7 @@ typedef void loop_filter_uvfunction
 /* assorted loopfilter functions which get used elsewhere */
 struct VP8Common;
 struct macroblockd;
+struct modeinfo;
 
 void vp8_loop_filter_init(struct VP8Common *cm);
 
@@ -90,4 +91,15 @@ void vp8_loop_filter_frame_yonly(struct VP8Common *cm,
 void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
                                       int sharpness_lvl);
 
+void vp8_loop_filter_row_normal(struct VP8Common *cm,
+                                struct modeinfo *mode_info_context,
+                                int mb_row, int post_ystride, int post_uvstride,
+                                unsigned char *y_ptr, unsigned char *u_ptr,
+                                unsigned char *v_ptr);
+
+void vp8_loop_filter_row_simple(struct VP8Common *cm,
+                                struct modeinfo *mode_info_context,
+                                int mb_row, int post_ystride, int post_uvstride,
+                                unsigned char *y_ptr, unsigned char *u_ptr,
+                                unsigned char *v_ptr);
 #endif

vp8/common/mfqe.c

@@ -160,9 +160,9 @@ static void multiframe_quality_enhance_block
         vsad = (vp8_variance8x8(v, uv_stride, vd, uvd_stride, &sse));
         vsad = (sse + 32)>>6;
 #else
-        sad = (vp8_sad16x16(y, y_stride, yd, yd_stride, INT_MAX)+128)>>8;
-        usad = (vp8_sad8x8(u, uv_stride, ud, uvd_stride, INT_MAX)+32)>>6;
-        vsad = (vp8_sad8x8(v, uv_stride, vd, uvd_stride, INT_MAX)+32)>>6;
+        sad = (vp8_sad16x16(y, y_stride, yd, yd_stride, UINT_MAX) + 128) >> 8;
+        usad = (vp8_sad8x8(u, uv_stride, ud, uvd_stride, UINT_MAX) + 32) >> 6;
+        vsad = (vp8_sad8x8(v, uv_stride, vd, uvd_stride, UINT_MAX)+ 32) >> 6;
 #endif
     }
     else /* if (blksize == 8) */
@@ -177,16 +177,16 @@ static void multiframe_quality_enhance_block
         vsad = (vp8_variance4x4(v, uv_stride, vd, uvd_stride, &sse));
         vsad = (sse + 8)>>4;
 #else
-        sad = (vp8_sad8x8(y, y_stride, yd, yd_stride, INT_MAX)+32)>>6;
-        usad = (vp8_sad4x4(u, uv_stride, ud, uvd_stride, INT_MAX)+8)>>4;
-        vsad = (vp8_sad4x4(v, uv_stride, vd, uvd_stride, INT_MAX)+8)>>4;
+        sad = (vp8_sad8x8(y, y_stride, yd, yd_stride, UINT_MAX) + 32) >> 6;
+        usad = (vp8_sad4x4(u, uv_stride, ud, uvd_stride, UINT_MAX) + 8) >> 4;
+        vsad = (vp8_sad4x4(v, uv_stride, vd, uvd_stride, UINT_MAX) + 8) >> 4;
 #endif
     }
 
     actrisk = (actd > act * 5);
 
-    /* thr = qdiff/8 + log2(act) + log4(qprev) */
-    thr = (qdiff >> 3);
+    /* thr = qdiff/16 + log2(act) + log4(qprev) */
+    thr = (qdiff >> 4);
     while (actd >>= 1) thr++;
     while (qprev >>= 2) thr++;
 

vp8/common/mips/dspr2/dequantize_dspr2.c

@@ -0,0 +1,33 @@
+/*
+ *  Copyright (c) 2012 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_config.h"
+#include "vpx_rtcd.h"
+#include "vpx_mem/vpx_mem.h"
+
+#if HAVE_DSPR2
+void vp8_dequant_idct_add_dspr2(short *input, short *dq,
+                                unsigned char *dest, int stride)
+{
+    int i;
+
+    for (i = 0; i < 16; i++)
+    {
+        input[i] = dq[i] * input[i];
+    }
+
+    vp8_short_idct4x4llm_dspr2(input, dest, stride, dest, stride);
+
+    vpx_memset(input, 0, 32);
+
+}
+
+#endif

vp8/common/mips/dspr2/idct_blk_dspr2.c

@@ -0,0 +1,88 @@
+/*
+ *  Copyright (c) 2012 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_config.h"
+#include "vpx_rtcd.h"
+
+#if HAVE_DSPR2
+
+void vp8_dequant_idct_add_y_block_dspr2
+(short *q, short *dq,
+ unsigned char *dst, int stride, char *eobs)
+{
+    int i, j;
+
+    for (i = 0; i < 4; i++)
+    {
+        for (j = 0; j < 4; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_dspr2(q, dq, dst, stride);
+            else
+            {
+                vp8_dc_only_idct_add_dspr2(q[0]*dq[0], dst, stride, dst, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q   += 16;
+            dst += 4;
+        }
+
+        dst += 4 * stride - 16;
+    }
+}
+
+void vp8_dequant_idct_add_uv_block_dspr2
+(short *q, short *dq,
+ unsigned char *dstu, unsigned char *dstv, int stride, char *eobs)
+{
+    int i, j;
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 2; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_dspr2(q, dq, dstu, stride);
+            else
+            {
+                vp8_dc_only_idct_add_dspr2(q[0]*dq[0], dstu, stride, dstu, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q    += 16;
+            dstu += 4;
+        }
+
+        dstu += 4 * stride - 8;
+    }
+
+    for (i = 0; i < 2; i++)
+    {
+        for (j = 0; j < 2; j++)
+        {
+            if (*eobs++ > 1)
+                vp8_dequant_idct_add_dspr2(q, dq, dstv, stride);
+            else
+            {
+                vp8_dc_only_idct_add_dspr2(q[0]*dq[0], dstv, stride, dstv, stride);
+                ((int *)q)[0] = 0;
+            }
+
+            q    += 16;
+            dstv += 4;
+        }
+
+        dstv += 4 * stride - 8;
+    }
+}
+
+#endif
+

vp8/common/mips/dspr2/idctllm_dspr2.c

@@ -0,0 +1,369 @@
+/*
+ *  Copyright (c) 2012 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_rtcd.h"
+
+#if HAVE_DSPR2
+#define CROP_WIDTH 256
+
+/******************************************************************************
+ * Notes:
+ *
+ * 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)
+ * Since 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).
+ ****************************************************************************/
+extern unsigned char ff_cropTbl[256 + 2 * CROP_WIDTH];
+static const int cospi8sqrt2minus1 = 20091;
+static const int sinpi8sqrt2      = 35468;
+
+inline void prefetch_load_short(short *src)
+{
+    __asm__ __volatile__ (
+        "pref   0,  0(%[src])   \n\t"
+        :
+        : [src] "r" (src)
+    );
+}
+
+void vp8_short_idct4x4llm_dspr2(short *input, unsigned char *pred_ptr,
+                                int pred_stride, unsigned char *dst_ptr,
+                                int dst_stride)
+{
+    int r, c;
+    int a1, b1, c1, d1;
+    short output[16];
+    short *ip = input;
+    short *op = output;
+    int temp1, temp2;
+    int shortpitch = 4;
+
+    int c2, d2;
+    int temp3, temp4;
+    unsigned char *cm = ff_cropTbl + CROP_WIDTH;
+
+    /* prepare data for load */
+    prefetch_load_short(ip + 8);
+
+    /* first loop is unrolled */
+    a1 = ip[0] + ip[8];
+    b1 = ip[0] - ip[8];
+
+    temp1 = (ip[4] * sinpi8sqrt2) >> 16;
+    temp2 = ip[12] + ((ip[12] * cospi8sqrt2minus1) >> 16);
+    c1 = temp1 - temp2;
+
+    temp1 = ip[4] + ((ip[4] * cospi8sqrt2minus1) >> 16);
+    temp2 = (ip[12] * sinpi8sqrt2) >> 16;
+    d1 = temp1 + temp2;
+
+    temp3 = (ip[5] * sinpi8sqrt2) >> 16;
+    temp4 = ip[13] + ((ip[13] * cospi8sqrt2minus1) >> 16);
+    c2 = temp3 - temp4;
+
+    temp3 = ip[5] + ((ip[5] * cospi8sqrt2minus1) >> 16);
+    temp4 = (ip[13] * sinpi8sqrt2) >> 16;
+    d2 = temp3 + temp4;
+
+    op[0] = a1 + d1;
+    op[12] = a1 - d1;
+    op[4] = b1 + c1;
+    op[8] = b1 - c1;
+
+    a1 = ip[1] + ip[9];
+    b1 = ip[1] - ip[9];
+
+    op[1] = a1 + d2;
+    op[13] = a1 - d2;
+    op[5] = b1 + c2;
+    op[9] = b1 - c2;
+
+    a1 = ip[2] + ip[10];
+    b1 = ip[2] - ip[10];
+
+    temp1 = (ip[6] * sinpi8sqrt2) >> 16;
+    temp2 = ip[14] + ((ip[14] * cospi8sqrt2minus1) >> 16);
+    c1 = temp1 - temp2;
+
+    temp1 = ip[6] + ((ip[6] * cospi8sqrt2minus1) >> 16);
+    temp2 = (ip[14] * sinpi8sqrt2) >> 16;
+    d1 = temp1 + temp2;
+
+    temp3 = (ip[7] * sinpi8sqrt2) >> 16;
+    temp4 = ip[15] + ((ip[15] * cospi8sqrt2minus1) >> 16);
+    c2 = temp3 - temp4;
+
+    temp3 = ip[7] + ((ip[7] * cospi8sqrt2minus1) >> 16);
+    temp4 = (ip[15] * sinpi8sqrt2) >> 16;
+    d2 = temp3 + temp4;
+
+    op[2] = a1 + d1;
+    op[14] = a1 - d1;
+    op[6] = b1 + c1;
+    op[10] = b1 - c1;
+
+    a1 = ip[3] + ip[11];
+    b1 = ip[3] - ip[11];
+
+    op[3] = a1 + d2;
+    op[15] = a1 - d2;
+    op[7] = b1 + c2;
+    op[11] = b1 - c2;
+
+    ip = output;
+
+    /* prepare data for load */
+    prefetch_load_short(ip + shortpitch);
+
+    /* second loop is unrolled */
+    a1 = ip[0] + ip[2];
+    b1 = ip[0] - ip[2];
+
+    temp1 = (ip[1] * sinpi8sqrt2) >> 16;
+    temp2 = ip[3] + ((ip[3] * cospi8sqrt2minus1) >> 16);
+    c1 = temp1 - temp2;
+
+    temp1 = ip[1] + ((ip[1] * cospi8sqrt2minus1) >> 16);
+    temp2 = (ip[3] * sinpi8sqrt2) >> 16;
+    d1 = temp1 + temp2;
+
+    temp3 = (ip[5] * sinpi8sqrt2) >> 16;
+    temp4 = ip[7] + ((ip[7] * cospi8sqrt2minus1) >> 16);
+    c2 = temp3 - temp4;
+
+    temp3 = ip[5] + ((ip[5] * cospi8sqrt2minus1) >> 16);
+    temp4 = (ip[7] * sinpi8sqrt2) >> 16;
+    d2 = temp3 + temp4;
+
+    op[0] = (a1 + d1 + 4) >> 3;
+    op[3] = (a1 - d1 + 4) >> 3;
+    op[1] = (b1 + c1 + 4) >> 3;
+    op[2] = (b1 - c1 + 4) >> 3;
+
+    a1 = ip[4] + ip[6];
+    b1 = ip[4] - ip[6];
+
+    op[4] = (a1 + d2 + 4) >> 3;
+    op[7] = (a1 - d2 + 4) >> 3;
+    op[5] = (b1 + c2 + 4) >> 3;
+    op[6] = (b1 - c2 + 4) >> 3;
+
+    a1 = ip[8] + ip[10];
+    b1 = ip[8] - ip[10];
+
+    temp1 = (ip[9] * sinpi8sqrt2) >> 16;
+    temp2 = ip[11] + ((ip[11] * cospi8sqrt2minus1) >> 16);
+    c1 = temp1 - temp2;
+
+    temp1 = ip[9] + ((ip[9] * cospi8sqrt2minus1) >> 16);
+    temp2 = (ip[11] * sinpi8sqrt2) >> 16;
+    d1 = temp1 + temp2;
+
+    temp3 = (ip[13] * sinpi8sqrt2) >> 16;
+    temp4 = ip[15] + ((ip[15] * cospi8sqrt2minus1) >> 16);
+    c2 = temp3 - temp4;
+
+    temp3 = ip[13] + ((ip[13] * cospi8sqrt2minus1) >> 16);
+    temp4 = (ip[15] * sinpi8sqrt2) >> 16;
+    d2 = temp3 + temp4;
+
+    op[8] = (a1 + d1 + 4) >> 3;
+    op[11] = (a1 - d1 + 4) >> 3;
+    op[9] = (b1 + c1 + 4) >> 3;
+    op[10] = (b1 - c1 + 4) >> 3;
+
+    a1 = ip[12] + ip[14];
+    b1 = ip[12] - ip[14];
+
+    op[12] = (a1 + d2 + 4) >> 3;
+    op[15] = (a1 - d2 + 4) >> 3;
+    op[13] = (b1 + c2 + 4) >> 3;
+    op[14] = (b1 - c2 + 4) >> 3;
+
+    ip = output;
+
+    for (r = 0; r < 4; r++)
+    {
+        for (c = 0; c < 4; c++)
+        {
+            short a = ip[c] + pred_ptr[c] ;
+            dst_ptr[c] = cm[a] ;
+        }
+
+        ip += 4;
+        dst_ptr += dst_stride;
+        pred_ptr += pred_stride;
+    }
+}
+
+void vp8_dc_only_idct_add_dspr2(short input_dc, unsigned char *pred_ptr, int pred_stride, unsigned char *dst_ptr, int dst_stride)
+{
+    int a1;
+    int i, absa1;
+    int t2, vector_a1, vector_a;
+
+    /* a1 = ((input_dc + 4) >> 3); */
+    __asm__ __volatile__ (
+        "addi  %[a1], %[input_dc], 4   \n\t"
+        "sra   %[a1], %[a1],       3   \n\t"
+        : [a1] "=r" (a1)
+        : [input_dc] "r" (input_dc)
+    );
+
+    if (a1 < 0)
+    {
+        /* use quad-byte
+         * input and output memory are four byte aligned
+         */
+        __asm__ __volatile__ (
+            "abs        %[absa1],     %[a1]         \n\t"
+            "replv.qb   %[vector_a1], %[absa1]      \n\t"
+            : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1)
+            : [a1] "r" (a1)
+        );
+
+        /* use (a1 - predptr[c]) instead a1 + predptr[c] */
+        for (i = 4; i--;)
+        {
+            __asm__ __volatile__ (
+                "lw             %[t2],       0(%[pred_ptr])                     \n\t"
+                "add            %[pred_ptr], %[pred_ptr],    %[pred_stride]     \n\t"
+                "subu_s.qb      %[vector_a], %[t2],          %[vector_a1]       \n\t"
+                "sw             %[vector_a], 0(%[dst_ptr])                      \n\t"
+                "add            %[dst_ptr],  %[dst_ptr],     %[dst_stride]      \n\t"
+                : [t2] "=&r" (t2), [vector_a] "=&r" (vector_a),
+                  [dst_ptr] "+&r" (dst_ptr), [pred_ptr] "+&r" (pred_ptr)
+                : [dst_stride] "r" (dst_stride), [pred_stride] "r" (pred_stride), [vector_a1] "r" (vector_a1)
+            );
+        }
+    }
+    else
+    {
+        /* use quad-byte
+         * input and output memory are four byte aligned
+         */
+        __asm__ __volatile__ (
+            "replv.qb       %[vector_a1], %[a1]     \n\t"
+            : [vector_a1] "=r" (vector_a1)
+            : [a1] "r" (a1)
+        );
+
+        for (i = 4; i--;)
+        {
+            __asm__ __volatile__ (
+                "lw             %[t2],       0(%[pred_ptr])                 \n\t"
+                "add            %[pred_ptr], %[pred_ptr],    %[pred_stride] \n\t"
+                "addu_s.qb      %[vector_a], %[vector_a1],   %[t2]          \n\t"
+                "sw             %[vector_a], 0(%[dst_ptr])                  \n\t"
+                "add            %[dst_ptr],  %[dst_ptr],     %[dst_stride]  \n\t"
+                : [t2] "=&r" (t2), [vector_a] "=&r" (vector_a),
+                  [dst_ptr] "+&r" (dst_ptr), [pred_ptr] "+&r" (pred_ptr)
+                : [dst_stride] "r" (dst_stride), [pred_stride] "r" (pred_stride), [vector_a1] "r" (vector_a1)
+            );
+        }
+    }
+
+}
+
+void vp8_short_inv_walsh4x4_dspr2(short *input, short *mb_dqcoeff)
+{
+    short output[16];
+    int i;
+    int a1, b1, c1, d1;
+    int a2, b2, c2, d2;
+    short *ip = input;
+    short *op = output;
+
+    prefetch_load_short(ip);
+
+    for (i = 4; i--;)
+    {
+        a1 = ip[0] + ip[12];
+        b1 = ip[4] + ip[8];
+        c1 = ip[4] - ip[8];
+        d1 = ip[0] - ip[12];
+
+        op[0] = a1 + b1;
+        op[4] = c1 + d1;
+        op[8] = a1 - b1;
+        op[12] = d1 - c1;
+
+        ip++;
+        op++;
+    }
+
+    ip = output;
+    op = output;
+
+    prefetch_load_short(ip);
+
+    for (i = 4; i--;)
+    {
+        a1 = ip[0] + ip[3] + 3;
+        b1 = ip[1] + ip[2];
+        c1 = ip[1] - ip[2];
+        d1 = ip[0] - ip[3] + 3;
+
+        a2 = a1 + b1;
+        b2 = d1 + c1;
+        c2 = a1 - b1;
+        d2 = d1 - c1;
+
+        op[0] = a2 >> 3;
+        op[1] = b2 >> 3;
+        op[2] = c2 >> 3;
+        op[3] = d2 >> 3;
+
+        ip += 4;
+        op += 4;
+    }
+
+    for (i = 0; i < 16; i++)
+    {
+        mb_dqcoeff[i * 16] = output[i];
+    }
+}
+
+void vp8_short_inv_walsh4x4_1_dspr2(short *input, short *mb_dqcoeff)
+{
+    int a1;
+
+    a1 = ((input[0] + 3) >> 3);
+
+    __asm__ __volatile__ (
+        "sh             %[a1], 0(%[mb_dqcoeff])                    \n\t"
+        "sh             %[a1], 32(%[mb_dqcoeff])                   \n\t"
+        "sh             %[a1], 64(%[mb_dqcoeff])                   \n\t"
+        "sh             %[a1], 96(%[mb_dqcoeff])                   \n\t"
+        "sh             %[a1], 128(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 160(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 192(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 224(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 256(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 288(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 320(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 352(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 384(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 416(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 448(%[mb_dqcoeff])                  \n\t"
+        "sh             %[a1], 480(%[mb_dqcoeff])                  \n\t"
+
+        :
+        : [a1] "r" (a1), [mb_dqcoeff] "r" (mb_dqcoeff)
+    );
+}
+
+#endif

vp8/common/mips/dspr2/reconinter_dspr2.c

@@ -0,0 +1,121 @@
+/*
+ *  Copyright (c) 2012 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_config.h"
+#include "vpx_rtcd.h"
+#include "vpx/vpx_integer.h"
+
+#if HAVE_DSPR2
+inline void prefetch_load_int(unsigned char *src)
+{
+    __asm__ __volatile__ (
+        "pref   0,  0(%[src])   \n\t"
+        :
+        : [src] "r" (src)
+    );
+}
+
+
+__inline void vp8_copy_mem16x16_dspr2(
+    unsigned char *RESTRICT src,
+    int src_stride,
+    unsigned char *RESTRICT dst,
+    int dst_stride)
+{
+    int r;
+    unsigned int a0, a1, a2, a3;
+
+    for (r = 16; r--;)
+    {
+        /* load src data in cache memory */
+        prefetch_load_int(src + src_stride);
+
+        /* use unaligned memory load and store */
+        __asm__ __volatile__ (
+            "ulw    %[a0], 0(%[src])            \n\t"
+            "ulw    %[a1], 4(%[src])            \n\t"
+            "ulw    %[a2], 8(%[src])            \n\t"
+            "ulw    %[a3], 12(%[src])           \n\t"
+            "sw     %[a0], 0(%[dst])            \n\t"
+            "sw     %[a1], 4(%[dst])            \n\t"
+            "sw     %[a2], 8(%[dst])            \n\t"
+            "sw     %[a3], 12(%[dst])           \n\t"
+            : [a0] "=&r" (a0), [a1] "=&r" (a1),
+              [a2] "=&r" (a2), [a3] "=&r" (a3)
+            : [src] "r" (src), [dst] "r" (dst)
+        );
+
+        src += src_stride;
+        dst += dst_stride;
+    }
+}
+
+
+__inline void vp8_copy_mem8x8_dspr2(
+    unsigned char *RESTRICT src,
+    int src_stride,
+    unsigned char *RESTRICT dst,
+    int dst_stride)
+{
+    int r;
+    unsigned int a0, a1;
+
+    /* load src data in cache memory */
+    prefetch_load_int(src + src_stride);
+
+    for (r = 8; r--;)
+    {
+        /* use unaligned memory load and store */
+        __asm__ __volatile__ (
+            "ulw    %[a0], 0(%[src])            \n\t"
+            "ulw    %[a1], 4(%[src])            \n\t"
+            "sw     %[a0], 0(%[dst])            \n\t"
+            "sw     %[a1], 4(%[dst])            \n\t"
+            : [a0] "=&r" (a0), [a1] "=&r" (a1)
+            : [src] "r" (src), [dst] "r" (dst)
+        );
+
+        src += src_stride;
+        dst += dst_stride;
+    }
+}
+
+
+__inline void vp8_copy_mem8x4_dspr2(
+    unsigned char *RESTRICT src,
+    int src_stride,
+    unsigned char *RESTRICT dst,
+    int dst_stride)
+{
+    int r;
+    unsigned int a0, a1;
+
+    /* load src data in cache memory */
+    prefetch_load_int(src + src_stride);
+
+    for (r = 4; r--;)
+    {
+        /* use unaligned memory load and store */
+        __asm__ __volatile__ (
+            "ulw    %[a0], 0(%[src])            \n\t"
+            "ulw    %[a1], 4(%[src])            \n\t"
+            "sw     %[a0], 0(%[dst])            \n\t"
+            "sw     %[a1], 4(%[dst])            \n\t"
+           : [a0] "=&r" (a0), [a1] "=&r" (a1)
+           : [src] "r" (src), [dst] "r" (dst)
+        );
+
+        src += src_stride;
+        dst += dst_stride;
+    }
+}
+
+#endif

vp8/common/onyx.h

@@ -37,14 +37,6 @@ extern "C"
 
     } VPX_SCALING;
 
-    typedef enum
-    {
-        VP8_LAST_FLAG = 1,
-        VP8_GOLD_FLAG = 2,
-        VP8_ALT_FLAG = 4
-    } VP8_REFFRAME;
-
-
     typedef enum
     {
         USAGE_STREAM_FROM_SERVER    = 0x0,
@@ -102,83 +94,101 @@ extern "C"
 
     typedef struct
     {
-        int Version;            // 4 versions of bitstream defined 0 best quality/slowest decode, 3 lowest quality/fastest decode
-        int Width;              // width of data passed to the compressor
-        int Height;             // height of data passed to the compressor
+        /* 4 versions of bitstream defined:
+         *   0 best quality/slowest decode, 3 lowest quality/fastest decode
+         */
+        int Version;
+        int Width;
+        int Height;
         struct vpx_rational  timebase;
-        unsigned int target_bandwidth;    // bandwidth to be used in kilobits per second
+        unsigned int target_bandwidth;    /* kilobits per second */
 
-        int noise_sensitivity;   // parameter used for applying pre processing blur: recommendation 0
-        int Sharpness;          // parameter used for sharpening output: recommendation 0:
+        /* parameter used for applying pre processing blur: recommendation 0 */
+        int noise_sensitivity;
+
+        /* parameter used for sharpening output: recommendation 0: */
+        int Sharpness;
         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
-        //    a television signal or feed from a live camera). ( speed setting controls how fast )
-        //(1)=Good Quality Fast Encoding. The encoder balances quality with the amount of time it takes to
-        //    encode the output. ( speed setting controls how fast )
-        //(2)=One Pass - Best Quality. The encoder places priority on the quality of the output over encoding
-        //    speed. The output is compressed at the highest possible quality. This option takes the longest
-        //    amount of time to encode. ( speed setting ignored )
-        //(3)=Two Pass - First Pass. The encoder generates a file of statistics for use in the second encoding
-        //    pass. ( speed setting controls how fast )
-        //(4)=Two Pass - Second Pass. The encoder uses the statistics that were generated in the first encoding
-        //    pass to create the compressed output. ( speed setting controls how fast )
-        //(5)=Two Pass - Second Pass Best.  The encoder uses the statistics that were generated in the first
-        //    encoding pass to create the compressed output using the highest possible quality, and taking a
-        //    longer amount of time to encode.. ( speed setting ignored )
-        int Mode;               //
+        /* mode ->
+         *(0)=Realtime/Live Encoding. This mode is optimized for realtim
+         *    encoding (for example, capturing a television signal or feed
+         *    from a live camera). ( speed setting controls how fast )
+         *(1)=Good Quality Fast Encoding. The encoder balances quality with
+         *    the amount of time it takes to encode the output. ( speed
+         *    setting controls how fast )
+         *(2)=One Pass - Best Quality. The encoder places priority on the
+         *    quality of the output over encoding speed. The output is
+         *    compressed at the highest possible quality. This option takes
+         *    the longest amount of time to encode. ( speed setting ignored
+         *    )
+         *(3)=Two Pass - First Pass. The encoder generates a file of
+         *    statistics for use in the second encoding pass. ( speed
+         *    setting controls how fast )
+         *(4)=Two Pass - Second Pass. The encoder uses the statistics that
+         *    were generated in the first encoding pass to create the
+         *    compressed output. ( speed setting controls how fast )
+         *(5)=Two Pass - Second Pass Best.  The encoder uses the statistics
+         *    that were generated in the first encoding pass to create the
+         *    compressed output using the highest possible quality, and
+         *    taking a longer amount of time to encode.. ( speed setting
+         *    ignored )
+         */
+        int Mode;
 
-        // Key Framing Operations
-        int auto_key;            // automatically detect cut scenes and set the keyframes
-        int key_freq;            // maximum distance to key frame.
+        /* Key Framing Operations */
+        int auto_key;       /* automatically detect cut scenes */
+        int key_freq;       /* maximum distance to key frame. */
 
-        int allow_lag;           // allow lagged compression (if 0 lagin frames is ignored)
-        int lag_in_frames;        // how many frames lag before we start encoding
+        /* lagged compression (if allow_lag == 0 lag_in_frames is ignored) */
+        int allow_lag;
+        int lag_in_frames; /* how many frames lag before we start encoding */
 
-        //----------------------------------------------------------------
-        // DATARATE CONTROL OPTIONS
+        /*
+         * DATARATE CONTROL OPTIONS
+         */
 
-        int end_usage; // vbr or cbr
+        int end_usage; /* vbr or cbr */
 
-        // buffer targeting aggressiveness
+        /* buffer targeting aggressiveness */
         int under_shoot_pct;
         int over_shoot_pct;
 
-        // buffering parameters
-        int64_t starting_buffer_level;  // in bytes
+        /* buffering parameters */
+        int64_t starting_buffer_level;
         int64_t optimal_buffer_level;
         int64_t maximum_buffer_size;
 
-        int64_t starting_buffer_level_in_ms;  // in milli-seconds
+        int64_t starting_buffer_level_in_ms;
         int64_t optimal_buffer_level_in_ms;
         int64_t maximum_buffer_size_in_ms;
 
-        // controlling quality
+        /* controlling quality */
         int fixed_q;
         int worst_allowed_q;
         int best_allowed_q;
         int cq_level;
 
-        // allow internal resizing ( currently disabled in the build !!!!!)
+        /* allow internal resizing */
         int allow_spatial_resampling;
         int resample_down_water_mark;
         int resample_up_water_mark;
 
-        // allow internal frame rate alterations
+        /* allow internal frame rate alterations */
         int allow_df;
         int drop_frames_water_mark;
 
-        // two pass datarate control
-        int two_pass_vbrbias;        // two pass datarate control tweaks
+        /* two pass datarate control */
+        int two_pass_vbrbias;
         int two_pass_vbrmin_section;
         int two_pass_vbrmax_section;
-        // END DATARATE CONTROL OPTIONS
-        //----------------------------------------------------------------
 
+        /*
+         * END DATARATE CONTROL OPTIONS
+         */
 
-        // these parameters aren't to be used in final build don't use!!!
+        /* these parameters aren't to be used in final build don't use!!! */
         int play_alternate;
         int alt_freq;
         int alt_q;
@@ -186,26 +196,28 @@ extern "C"
         int gold_q;
 
 
-        int multi_threaded;   // how many threads to run the encoder on
-        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 multi_threaded;   /* how many threads to run the encoder on */
+        int token_partitions; /* how many token partitions to create */
 
-        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.
+        /* early breakout threshold: for video conf recommend 800 */
+        int encode_breakout;
+
+        /* 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.
+         */
+        unsigned int error_resilient_mode;
 
         int arnr_max_frames;
-        int arnr_strength ;
-        int arnr_type     ;
+        int arnr_strength;
+        int arnr_type;
 
-        struct vpx_fixed_buf         two_pass_stats_in;
+        struct vpx_fixed_buf        two_pass_stats_in;
         struct vpx_codec_pkt_list  *output_pkt_list;
 
         vp8e_tuning tuning;
 
-        // Temporal scaling parameters
+        /* Temporal scaling parameters */
         unsigned int number_of_layers;
         unsigned int target_bitrate[VPX_TS_MAX_PERIODICITY];
         unsigned int rate_decimator[VPX_TS_MAX_PERIODICITY];
@@ -236,16 +248,14 @@ extern "C"
     void vp8_init_config(struct VP8_COMP* onyx, VP8_CONFIG *oxcf);
     void vp8_change_config(struct VP8_COMP* onyx, VP8_CONFIG *oxcf);
 
-// 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(struct VP8_COMP* comp, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, int64_t time_stamp, int64_t end_time_stamp);
     int vp8_get_compressed_data(struct VP8_COMP* comp, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, unsigned char *dest_end, int64_t *time_stamp, int64_t *time_end, int flush);
     int vp8_get_preview_raw_frame(struct VP8_COMP* comp, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t *flags);
 
     int vp8_use_as_reference(struct VP8_COMP* comp, int ref_frame_flags);
     int vp8_update_reference(struct VP8_COMP* comp, int ref_frame_flags);
-    int vp8_get_reference(struct VP8_COMP* comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
-    int vp8_set_reference(struct VP8_COMP* comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    int vp8_get_reference(struct VP8_COMP* comp, enum vpx_ref_frame_type ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    int vp8_set_reference(struct VP8_COMP* comp, enum vpx_ref_frame_type ref_frame_flag, YV12_BUFFER_CONFIG *sd);
     int vp8_update_entropy(struct VP8_COMP* comp, int update);
     int vp8_set_roimap(struct VP8_COMP* comp, unsigned char *map, unsigned int rows, unsigned int cols, int delta_q[4], int delta_lf[4], unsigned int threshold[4]);
     int vp8_set_active_map(struct VP8_COMP* comp, unsigned char *map, unsigned int rows, unsigned int cols);

vp8/common/onyxc_int.h

@@ -42,7 +42,6 @@ typedef struct frame_contexts
     vp8_prob sub_mv_ref_prob [VP8_SUBMVREFS-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;
 
 typedef enum
@@ -59,12 +58,6 @@ typedef enum
     RECON_CLAMP_NOTREQUIRED     = 1
 } CLAMP_TYPE;
 
-typedef enum
-{
-    SIXTAP   = 0,
-    BILINEAR = 1
-} INTERPOLATIONFILTERTYPE;
-
 typedef struct VP8Common
 
 {
@@ -94,6 +87,7 @@ typedef struct VP8Common
     YV12_BUFFER_CONFIG post_proc_buffer;
     YV12_BUFFER_CONFIG post_proc_buffer_int;
     int post_proc_buffer_int_used;
+    unsigned char *pp_limits_buffer;   /* post-processing filter coefficients */
 #endif
 
     FRAME_TYPE last_frame_type;  /* Save last frame's frame type for motion search. */
@@ -114,7 +108,6 @@ typedef struct VP8Common
     int full_pixel;
 
     int base_qindex;
-    int last_kf_gf_q;  /* Q used on the last GF or KF */
 
     int y1dc_delta_q;
     int y2dc_delta_q;
@@ -130,11 +123,11 @@ typedef struct VP8Common
 
     MODE_INFO *mip; /* Base of allocated array */
     MODE_INFO *mi;  /* Corresponds to upper left visible macroblock */
+#if CONFIG_ERROR_CONCEALMENT
     MODE_INFO *prev_mip; /* MODE_INFO array 'mip' from last decoded frame */
     MODE_INFO *prev_mi;  /* 'mi' from last frame (points into prev_mip) */
+#endif
 
-
-    INTERPOLATIONFILTERTYPE mcomp_filter_type;
     LOOPFILTERTYPE filter_type;
 
     loop_filter_info_n lf_info;
@@ -158,14 +151,6 @@ typedef struct VP8Common
     ENTROPY_CONTEXT_PLANES *above_context;   /* row of context for each plane */
     ENTROPY_CONTEXT_PLANES left_context;  /* (up to) 4 contexts "" */
 
-
-    /* keyframe block modes are predicted by their above, left neighbors */
-
-    vp8_prob kf_bmode_prob [VP8_BINTRAMODES] [VP8_BINTRAMODES] [VP8_BINTRAMODES-1];
-    vp8_prob kf_ymode_prob [VP8_YMODES-1];  /* keyframe "" */
-    vp8_prob kf_uv_mode_prob [VP8_UV_MODES-1];
-
-
     FRAME_CONTEXT lfc; /* last frame entropy */
     FRAME_CONTEXT fc;  /* this frame entropy */
 

vp8/common/onyxd.h

@@ -22,6 +22,7 @@ extern "C"
 #include "ppflags.h"
 #include "vpx_ports/mem.h"
 #include "vpx/vpx_codec.h"
+#include "vpx/vp8.h"
 
     struct VP8D_COMP;
 
@@ -35,12 +36,6 @@ extern "C"
         int     error_concealment;
         int     input_fragments;
     } VP8D_CONFIG;
-    typedef enum
-    {
-        VP8_LAST_FLAG = 1,
-        VP8_GOLD_FLAG = 2,
-        VP8_ALT_FLAG = 4
-    } VP8_REFFRAME;
 
     typedef enum
     {
@@ -53,11 +48,13 @@ extern "C"
 
     int vp8dx_get_setting(struct VP8D_COMP* comp, VP8D_SETTING oxst);
 
-    int vp8dx_receive_compressed_data(struct VP8D_COMP* comp, unsigned long size, const unsigned char *dest, int64_t time_stamp);
+    int vp8dx_receive_compressed_data(struct VP8D_COMP* comp,
+                                      size_t size, const uint8_t *dest,
+                                      int64_t time_stamp);
     int vp8dx_get_raw_frame(struct VP8D_COMP* comp, YV12_BUFFER_CONFIG *sd, int64_t *time_stamp, int64_t *time_end_stamp, vp8_ppflags_t *flags);
 
-    vpx_codec_err_t vp8dx_get_reference(struct VP8D_COMP* comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
-    vpx_codec_err_t vp8dx_set_reference(struct VP8D_COMP* comp, VP8_REFFRAME ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_get_reference(struct VP8D_COMP* comp, enum vpx_ref_frame_type ref_frame_flag, YV12_BUFFER_CONFIG *sd);
+    vpx_codec_err_t vp8dx_set_reference(struct VP8D_COMP* comp, enum vpx_ref_frame_type ref_frame_flag, YV12_BUFFER_CONFIG *sd);
 
     struct VP8D_COMP* vp8dx_create_decompressor(VP8D_CONFIG *oxcf);
 

vp8/common/postproc.c

@@ -127,27 +127,24 @@ extern void vp8_blit_text(const char *msg, unsigned char *address, const int pit
 extern void vp8_blit_line(int x0, int x1, int y0, int y1, unsigned char *image, const int pitch);
 /***********************************************************************************************************
  */
-void vp8_post_proc_down_and_across_c
+void vp8_post_proc_down_and_across_mb_row_c
 (
     unsigned char *src_ptr,
     unsigned char *dst_ptr,
     int src_pixels_per_line,
     int dst_pixels_per_line,
-    int rows,
     int cols,
-    int flimit
+    unsigned char *f,
+    int size
 )
 {
     unsigned char *p_src, *p_dst;
     int row;
     int col;
-    int i;
-    int v;
-    int pitch = src_pixels_per_line;
-    unsigned char d[8];
-    (void)dst_pixels_per_line;
+    unsigned char v;
+    unsigned char d[4];
 
-    for (row = 0; row < rows; row++)
+    for (row = 0; row < size; row++)
     {
         /* post_proc_down for one row */
         p_src = src_ptr;
@@ -155,20 +152,23 @@ void vp8_post_proc_down_and_across_c
 
         for (col = 0; col < cols; col++)
         {
+            unsigned char p_above2 = p_src[col - 2 * src_pixels_per_line];
+            unsigned char p_above1 = p_src[col - src_pixels_per_line];
+            unsigned char p_below1 = p_src[col + src_pixels_per_line];
+            unsigned char p_below2 = p_src[col + 2 * src_pixels_per_line];
 
-            int kernel = 4;
-            int v = p_src[col];
+            v = p_src[col];
 
-            for (i = -2; i <= 2; i++)
+            if ((abs(v - p_above2) < f[col]) && (abs(v - p_above1) < f[col])
+                && (abs(v - p_below1) < f[col]) && (abs(v - p_below2) < f[col]))
             {
-                if (abs(v - p_src[col+i*pitch]) > flimit)
-                    goto down_skip_convolve;
-
-                kernel += kernel5[2+i] * p_src[col+i*pitch];
+                unsigned char k1, k2, k3;
+                k1 = (p_above2 + p_above1 + 1) >> 1;
+                k2 = (p_below2 + p_below1 + 1) >> 1;
+                k3 = (k1 + k2 + 1) >> 1;
+                v = (k3 + v + 1) >> 1;
             }
 
-            v = (kernel >> 3);
-        down_skip_convolve:
             p_dst[col] = v;
         }
 
@@ -176,45 +176,38 @@ void vp8_post_proc_down_and_across_c
         p_src = dst_ptr;
         p_dst = dst_ptr;
 
-        for (i = -8; i<0; i++)
-          p_src[i]=p_src[0];
-
-        for (i = cols; i<cols+8; i++)
-          p_src[i]=p_src[cols-1];
-
-        for (i = 0; i < 8; i++)
-            d[i] = p_src[i];
+        p_src[-2] = p_src[-1] = p_src[0];
+        p_src[cols] = p_src[cols + 1] = p_src[cols - 1];
 
         for (col = 0; col < cols; col++)
         {
-            int kernel = 4;
             v = p_src[col];
 
-            d[col&7] = v;
-
-            for (i = -2; i <= 2; i++)
+            if ((abs(v - p_src[col - 2]) < f[col])
+                && (abs(v - p_src[col - 1]) < f[col])
+                && (abs(v - p_src[col + 1]) < f[col])
+                && (abs(v - p_src[col + 2]) < f[col]))
             {
-                if (abs(v - p_src[col+i]) > flimit)
-                    goto across_skip_convolve;
-
-                kernel += kernel5[2+i] * p_src[col+i];
+                unsigned char k1, k2, k3;
+                k1 = (p_src[col - 2] + p_src[col - 1] + 1) >> 1;
+                k2 = (p_src[col + 2] + p_src[col + 1] + 1) >> 1;
+                k3 = (k1 + k2 + 1) >> 1;
+                v = (k3 + v + 1) >> 1;
             }
 
-            d[col&7] = (kernel >> 3);
-        across_skip_convolve:
+            d[col & 3] = v;
 
             if (col >= 2)
-                p_dst[col-2] = d[(col-2)&7];
+                p_dst[col - 2] = d[(col - 2) & 3];
         }
 
         /* handle the last two pixels */
-        p_dst[col-2] = d[(col-2)&7];
-        p_dst[col-1] = d[(col-1)&7];
-
+        p_dst[col - 2] = d[(col - 2) & 3];
+        p_dst[col - 1] = d[(col - 1) & 3];
 
         /* next row */
-        src_ptr += pitch;
-        dst_ptr += pitch;
+        src_ptr += src_pixels_per_line;
+        dst_ptr += dst_pixels_per_line;
     }
 }
 
@@ -240,8 +233,9 @@ void vp8_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows, int co
         for (i = -8; i<0; i++)
           s[i]=s[0];
 
-        // 17 avoids valgrind warning - we buffer values in c in d
-        // and only write them when we've read 8 ahead...
+        /* 17 avoids valgrind warning - we buffer values in c in d
+         * and only write them when we've read 8 ahead...
+         */
         for (i = cols; i<cols+17; i++)
           s[i]=s[cols-1];
 
@@ -275,9 +269,6 @@ void vp8_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows, int co
 }
 
 
-
-
-
 void vp8_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols, int flimit)
 {
     int r, c, i;
@@ -294,8 +285,9 @@ void vp8_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols, i
         for (i = -8; i < 0; i++)
           s[i*pitch]=s[0];
 
-        // 17 avoids valgrind warning - we buffer values in c in d
-        // and only write them when we've read 8 ahead...
+        /* 17 avoids valgrind warning - we buffer values in c in d
+         * and only write them when we've read 8 ahead...
+         */
         for (i = rows; i < rows+17; i++)
           s[i*pitch]=s[(rows-1)*pitch];
 
@@ -322,28 +314,17 @@ void vp8_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols, i
     }
 }
 
-
-static void vp8_deblock_and_de_macro_block(YV12_BUFFER_CONFIG         *source,
-        YV12_BUFFER_CONFIG         *post,
-        int                         q,
-        int                         low_var_thresh,
-        int                         flag)
+static void vp8_de_mblock(YV12_BUFFER_CONFIG         *post,
+                          int                         q)
 {
-    double level = 6.0e-05 * q * q * q - .0067 * q * q + .306 * q + .0065;
-    int ppl = (int)(level + .5);
-    (void) low_var_thresh;
-    (void) flag;
-
-    vp8_post_proc_down_and_across(source->y_buffer, post->y_buffer, source->y_stride,  post->y_stride, source->y_height, source->y_width,  ppl);
-    vp8_mbpost_proc_across_ip(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
-    vp8_mbpost_proc_down(post->y_buffer, post->y_stride, post->y_height, post->y_width, q2mbl(q));
-
-    vp8_post_proc_down_and_across(source->u_buffer, post->u_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
-    vp8_post_proc_down_and_across(source->v_buffer, post->v_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
-
+    vp8_mbpost_proc_across_ip(post->y_buffer, post->y_stride, post->y_height,
+                              post->y_width, q2mbl(q));
+    vp8_mbpost_proc_down(post->y_buffer, post->y_stride, post->y_height,
+                         post->y_width, q2mbl(q));
 }
 
-void vp8_deblock(YV12_BUFFER_CONFIG         *source,
+void vp8_deblock(VP8_COMMON                 *cm,
+                 YV12_BUFFER_CONFIG         *source,
                  YV12_BUFFER_CONFIG         *post,
                  int                         q,
                  int                         low_var_thresh,
@@ -351,16 +332,64 @@ void vp8_deblock(YV12_BUFFER_CONFIG         *source,
 {
     double level = 6.0e-05 * q * q * q - .0067 * q * q + .306 * q + .0065;
     int ppl = (int)(level + .5);
+
+    const MODE_INFO *mode_info_context = cm->mi;
+    int mbr, mbc;
+
+    /* The pixel thresholds are adjusted according to if or not the macroblock
+     * is a skipped block.  */
+    unsigned char *ylimits = cm->pp_limits_buffer;
+    unsigned char *uvlimits = cm->pp_limits_buffer + 16 * cm->mb_cols;
     (void) low_var_thresh;
     (void) flag;
 
-    vp8_post_proc_down_and_across(source->y_buffer, post->y_buffer, source->y_stride,  post->y_stride, source->y_height, source->y_width,   ppl);
-    vp8_post_proc_down_and_across(source->u_buffer, post->u_buffer, source->uv_stride, post->uv_stride,  source->uv_height, source->uv_width, ppl);
-    vp8_post_proc_down_and_across(source->v_buffer, post->v_buffer, source->uv_stride, post->uv_stride, source->uv_height, source->uv_width, ppl);
+    if (ppl > 0)
+    {
+        for (mbr = 0; mbr < cm->mb_rows; mbr++)
+        {
+            unsigned char *ylptr = ylimits;
+            unsigned char *uvlptr = uvlimits;
+            for (mbc = 0; mbc < cm->mb_cols; mbc++)
+            {
+                unsigned char mb_ppl;
+
+                if (mode_info_context->mbmi.mb_skip_coeff)
+                    mb_ppl = (unsigned char)ppl >> 1;
+                else
+                    mb_ppl = (unsigned char)ppl;
+
+                vpx_memset(ylptr, mb_ppl, 16);
+                vpx_memset(uvlptr, mb_ppl, 8);
+
+                ylptr += 16;
+                uvlptr += 8;
+                mode_info_context++;
+            }
+            mode_info_context++;
+
+            vp8_post_proc_down_and_across_mb_row(
+                source->y_buffer + 16 * mbr * source->y_stride,
+                post->y_buffer + 16 * mbr * post->y_stride, source->y_stride,
+                post->y_stride, source->y_width, ylimits, 16);
+
+            vp8_post_proc_down_and_across_mb_row(
+                source->u_buffer + 8 * mbr * source->uv_stride,
+                post->u_buffer + 8 * mbr * post->uv_stride, source->uv_stride,
+                post->uv_stride, source->uv_width, uvlimits, 8);
+            vp8_post_proc_down_and_across_mb_row(
+                source->v_buffer + 8 * mbr * source->uv_stride,
+                post->v_buffer + 8 * mbr * post->uv_stride, source->uv_stride,
+                post->uv_stride, source->uv_width, uvlimits, 8);
+        }
+    } else
+    {
+        vp8_yv12_copy_frame(source, post);
+    }
 }
 
 #if !(CONFIG_TEMPORAL_DENOISING)
-void vp8_de_noise(YV12_BUFFER_CONFIG         *source,
+void vp8_de_noise(VP8_COMMON                 *cm,
+                  YV12_BUFFER_CONFIG         *source,
                   YV12_BUFFER_CONFIG         *post,
                   int                         q,
                   int                         low_var_thresh,
@@ -368,33 +397,33 @@ void vp8_de_noise(YV12_BUFFER_CONFIG         *source,
 {
     double level = 6.0e-05 * q * q * q - .0067 * q * q + .306 * q + .0065;
     int ppl = (int)(level + .5);
+    int mb_rows = source->y_width >> 4;
+    int mb_cols = source->y_height >> 4;
+    unsigned char *limits = cm->pp_limits_buffer;;
+    int mbr, mbc;
     (void) post;
     (void) low_var_thresh;
     (void) flag;
 
-    vp8_post_proc_down_and_across(
-        source->y_buffer + 2 * source->y_stride + 2,
-        source->y_buffer + 2 * source->y_stride + 2,
-        source->y_stride,
-        source->y_stride,
-        source->y_height - 4,
-        source->y_width - 4,
-        ppl);
-    vp8_post_proc_down_and_across(
-        source->u_buffer + 2 * source->uv_stride + 2,
-        source->u_buffer + 2 * source->uv_stride + 2,
-        source->uv_stride,
-        source->uv_stride,
-        source->uv_height - 4,
-        source->uv_width - 4, ppl);
-    vp8_post_proc_down_and_across(
-        source->v_buffer + 2 * source->uv_stride + 2,
-        source->v_buffer + 2 * source->uv_stride + 2,
-        source->uv_stride,
-        source->uv_stride,
-        source->uv_height - 4,
-        source->uv_width - 4, ppl);
+    vpx_memset(limits, (unsigned char)ppl, 16 * mb_cols);
 
+    /* TODO: The original code don't filter the 2 outer rows and columns. */
+    for (mbr = 0; mbr < mb_rows; mbr++)
+    {
+        vp8_post_proc_down_and_across_mb_row(
+            source->y_buffer + 16 * mbr * source->y_stride,
+            source->y_buffer + 16 * mbr * source->y_stride,
+            source->y_stride, source->y_stride, source->y_width, limits, 16);
+
+        vp8_post_proc_down_and_across_mb_row(
+            source->u_buffer + 8 * mbr * source->uv_stride,
+            source->u_buffer + 8 * mbr * source->uv_stride,
+            source->uv_stride, source->uv_stride, source->uv_width, limits, 8);
+        vp8_post_proc_down_and_across_mb_row(
+            source->v_buffer + 8 * mbr * source->uv_stride,
+            source->v_buffer + 8 * mbr * source->uv_stride,
+            source->uv_stride, source->uv_stride, source->uv_width, limits, 8);
+    }
 }
 #endif
 
@@ -441,7 +470,7 @@ static void fillrd(struct postproc_state *state, int q, int a)
 
         }
 
-        for (next = next; next < 256; next++)
+        for (; next < 256; next++)
             char_dist[next] = 0;
 
     }
@@ -731,21 +760,21 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
 
             oci->post_proc_buffer_int_used = 1;
 
-            // insure that postproc is set to all 0's so that post proc
-            // doesn't pull random data in from edge
+            /* insure that postproc is set to all 0's so that post proc
+             * doesn't pull random data in from edge
+             */
             vpx_memset((&oci->post_proc_buffer_int)->buffer_alloc,128,(&oci->post_proc_buffer)->frame_size);
 
         }
     }
 
-#if ARCH_X86||ARCH_X86_64
-    vpx_reset_mmx_state();
-#endif
+    vp8_clear_system_state();
 
     if ((flags & VP8D_MFQE) &&
          oci->postproc_state.last_frame_valid &&
          oci->current_video_frame >= 2 &&
-         oci->base_qindex - oci->postproc_state.last_base_qindex >= 10)
+         oci->postproc_state.last_base_qindex < 60 &&
+         oci->base_qindex - oci->postproc_state.last_base_qindex >= 20)
     {
         vp8_multiframe_quality_enhance(oci);
         if (((flags & VP8D_DEBLOCK) || (flags & VP8D_DEMACROBLOCK)) &&
@@ -754,12 +783,14 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
             vp8_yv12_copy_frame(&oci->post_proc_buffer, &oci->post_proc_buffer_int);
             if (flags & VP8D_DEMACROBLOCK)
             {
-                vp8_deblock_and_de_macro_block(&oci->post_proc_buffer_int, &oci->post_proc_buffer,
+                vp8_deblock(oci, &oci->post_proc_buffer_int, &oci->post_proc_buffer,
                                                q + (deblock_level - 5) * 10, 1, 0);
+                vp8_de_mblock(&oci->post_proc_buffer,
+                              q + (deblock_level - 5) * 10);
             }
             else if (flags & VP8D_DEBLOCK)
             {
-                vp8_deblock(&oci->post_proc_buffer_int, &oci->post_proc_buffer,
+                vp8_deblock(oci, &oci->post_proc_buffer_int, &oci->post_proc_buffer,
                             q, 1, 0);
             }
         }
@@ -768,13 +799,15 @@ int vp8_post_proc_frame(VP8_COMMON *oci, YV12_BUFFER_CONFIG *dest, vp8_ppflags_t
     }
     else if (flags & VP8D_DEMACROBLOCK)
     {
-        vp8_deblock_and_de_macro_block(oci->frame_to_show, &oci->post_proc_buffer,
-                                       q + (deblock_level - 5) * 10, 1, 0);
+        vp8_deblock(oci, oci->frame_to_show, &oci->post_proc_buffer,
+                                     q + (deblock_level - 5) * 10, 1, 0);
+        vp8_de_mblock(&oci->post_proc_buffer, q + (deblock_level - 5) * 10);
+
         oci->postproc_state.last_base_qindex = oci->base_qindex;
     }
     else if (flags & VP8D_DEBLOCK)
     {
-        vp8_deblock(oci->frame_to_show, &oci->post_proc_buffer,
+        vp8_deblock(oci, oci->frame_to_show, &oci->post_proc_buffer,
                     q, 1, 0);
         oci->postproc_state.last_base_qindex = oci->base_qindex;
     }

vp8/common/postproc.h

@@ -30,13 +30,15 @@ int vp8_post_proc_frame(struct VP8Common *oci, YV12_BUFFER_CONFIG *dest,
                         vp8_ppflags_t *flags);
 
 
-void vp8_de_noise(YV12_BUFFER_CONFIG         *source,
+void vp8_de_noise(struct VP8Common           *oci,
+                  YV12_BUFFER_CONFIG         *source,
                   YV12_BUFFER_CONFIG         *post,
                   int                         q,
                   int                         low_var_thresh,
                   int                         flag);
 
-void vp8_deblock(YV12_BUFFER_CONFIG         *source,
+void vp8_deblock(struct VP8Common           *oci,
+                 YV12_BUFFER_CONFIG         *source,
                  YV12_BUFFER_CONFIG         *post,
                  int                         q,
                  int                         low_var_thresh,

vp8/common/ppc/systemdependent.c

@@ -19,14 +19,14 @@ void (*vp8_short_idct4x4)(short *input, short *output, int pitch);
 void (*vp8_short_idct4x4_1)(short *input, short *output, int pitch);
 void (*vp8_dc_only_idct)(short input_dc, short *output, int pitch);
 
-extern void (*vp8_post_proc_down_and_across)(
+extern void (*vp8_post_proc_down_and_across_mb_row)(
     unsigned char *src_ptr,
     unsigned char *dst_ptr,
     int src_pixels_per_line,
     int dst_pixels_per_line,
-    int rows,
     int cols,
-    int flimit
+    unsigned char *f,
+    int size
 );
 
 extern void (*vp8_mbpost_proc_down)(unsigned char *dst, int pitch, int rows, int cols, int flimit);
@@ -34,15 +34,15 @@ extern void vp8_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int
 extern void (*vp8_mbpost_proc_across_ip)(unsigned char *src, int pitch, int rows, int cols, int flimit);
 extern void vp8_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows, int cols, int flimit);
 
-extern void vp8_post_proc_down_and_across_c
+extern void vp8_post_proc_down_and_across_mb_row_c
 (
     unsigned char *src_ptr,
     unsigned char *dst_ptr,
     int src_pixels_per_line,
     int dst_pixels_per_line,
-    int rows,
     int cols,
-    int flimit
+    unsigned char *f,
+    int size
 );
 void vp8_plane_add_noise_c(unsigned char *Start, unsigned int Width, unsigned int Height, int Pitch, int q, int a);
 
@@ -158,7 +158,7 @@ void vp8_machine_specific_config(void)
     vp8_lf_mbhsimple                     = loop_filter_mbhs_ppc;
     vp8_lf_bhsimple                      = loop_filter_bhs_ppc;
 
-    vp8_post_proc_down_and_across           = vp8_post_proc_down_and_across_c;
+    vp8_post_proc_down_and_across_mb_row = vp8_post_proc_down_and_across_mb_row_c;
     vp8_mbpost_proc_down                  = vp8_mbpost_proc_down_c;
     vp8_mbpost_proc_across_ip              = vp8_mbpost_proc_across_ip_c;
     vp8_plane_add_noise                   = vp8_plane_add_noise_c;

vp8/common/quant_common.c

@@ -109,7 +109,10 @@ int vp8_ac2quant(int QIndex, int Delta)
     else if (QIndex < 0)
         QIndex = 0;
 
-    retval = (ac_qlookup[ QIndex ] * 155) / 100;
+    /* For all x in [0..284], x*155/100 is bitwise equal to (x*101581) >> 16.
+     * The smallest precision for that is '(x*6349) >> 12' but 16 is a good
+     * word size. */
+    retval = (ac_qlookup[ QIndex ] * 101581) >> 16;
 
     if (retval < 8)
         retval = 8;

vp8/common/reconintra4x4.c

@@ -13,11 +13,11 @@
 #include "vpx_rtcd.h"
 #include "blockd.h"
 
-void vp8_intra4x4_predict_d_c(unsigned char *Above,
-                              unsigned char *yleft, int left_stride,
-                              int b_mode,
-                              unsigned char *dst, int dst_stride,
-                              unsigned char top_left)
+void vp8_intra4x4_predict_c(unsigned char *Above,
+                            unsigned char *yleft, int left_stride,
+                            B_PREDICTION_MODE b_mode,
+                            unsigned char *dst, int dst_stride,
+                            unsigned char top_left)
 {
     int i, r, c;
 
@@ -290,19 +290,8 @@ void vp8_intra4x4_predict_d_c(unsigned char *Above,
     }
     break;
 
+    default:
+    break;
 
     }
 }
-
-void vp8_intra4x4_predict_c(unsigned char *src, int src_stride,
-                            int b_mode,
-                            unsigned char *dst, int dst_stride)
-{
-    unsigned char *Above = src - src_stride;
-
-    vp8_intra4x4_predict_d_c(Above,
-                             src - 1, src_stride,
-                             b_mode,
-                             dst, dst_stride,
-                             Above[-1]);
-}

vp8/common/rtcd.c

@@ -11,28 +11,45 @@
 #define RTCD_C
 #include "vpx_rtcd.h"
 
-#if CONFIG_MULTITHREAD && HAVE_PTHREAD_H
-#include <pthread.h>
-static void once(void (*func)(void))
-{
-    static pthread_once_t lock = PTHREAD_ONCE_INIT;
-    pthread_once(&lock, func);
-}
-
-
-#elif CONFIG_MULTITHREAD && defined(_WIN32)
+#if CONFIG_MULTITHREAD && defined(_WIN32)
 #include <windows.h>
+#include <stdlib.h>
 static void once(void (*func)(void))
 {
-    /* Using a static initializer here rather than InitializeCriticalSection()
-     * since there's no race-free context in which to execute it. Protecting
-     * it with an atomic op like InterlockedCompareExchangePointer introduces
-     * an x86 dependency, and InitOnceExecuteOnce requires Vista.
-     */
-    static CRITICAL_SECTION lock = {(void *)-1, -1, 0, 0, 0, 0};
+    static CRITICAL_SECTION *lock;
+    static LONG waiters;
     static int done;
+    void *lock_ptr = &lock;
 
-    EnterCriticalSection(&lock);
+    /* If the initialization is complete, return early. This isn't just an
+     * optimization, it prevents races on the destruction of the global
+     * lock.
+     */
+    if(done)
+        return;
+
+    InterlockedIncrement(&waiters);
+
+    /* Get a lock. We create one and try to make it the one-true-lock,
+     * throwing it away if we lost the race.
+     */
+
+    {
+        /* Scope to protect access to new_lock */
+        CRITICAL_SECTION *new_lock = malloc(sizeof(CRITICAL_SECTION));
+        InitializeCriticalSection(new_lock);
+        if (InterlockedCompareExchangePointer(lock_ptr, new_lock, NULL) != NULL)
+        {
+            DeleteCriticalSection(new_lock);
+            free(new_lock);
+        }
+    }
+
+    /* At this point, we have a lock that can be synchronized on. We don't
+     * care which thread actually performed the allocation.
+     */
+
+    EnterCriticalSection(lock);
 
     if (!done)
     {
@@ -40,7 +57,26 @@ static void once(void (*func)(void))
         done = 1;
     }
 
-    LeaveCriticalSection(&lock);
+    LeaveCriticalSection(lock);
+
+    /* Last one out should free resources. The destructed objects are
+     * protected by checking if(done) above.
+     */
+    if(!InterlockedDecrement(&waiters))
+    {
+        DeleteCriticalSection(lock);
+        free(lock);
+        lock = NULL;
+    }
+}
+
+
+#elif CONFIG_MULTITHREAD && HAVE_PTHREAD_H
+#include <pthread.h>
+static void once(void (*func)(void))
+{
+    static pthread_once_t lock = PTHREAD_ONCE_INIT;
+    pthread_once(&lock, func);
 }
 
 

vp8/common/rtcd_defs.sh

@@ -1,5 +1,7 @@
 common_forward_decls() {
 cat <<EOF
+#include "vp8/common/blockd.h"
+
 struct blockd;
 struct macroblockd;
 struct loop_filter_info;
@@ -22,35 +24,42 @@ specialize vp8_dequantize_b mmx media neon
 vp8_dequantize_b_media=vp8_dequantize_b_v6
 
 prototype void vp8_dequant_idct_add "short *input, short *dq, unsigned char *output, int stride"
-specialize vp8_dequant_idct_add mmx media neon
+specialize vp8_dequant_idct_add mmx media neon dspr2
 vp8_dequant_idct_add_media=vp8_dequant_idct_add_v6
+vp8_dequant_idct_add_dspr2=vp8_dequant_idct_add_dspr2
 
 prototype void vp8_dequant_idct_add_y_block "short *q, short *dq, unsigned char *dst, int stride, char *eobs"
-specialize vp8_dequant_idct_add_y_block mmx sse2 media neon
+specialize vp8_dequant_idct_add_y_block mmx sse2 media neon dspr2
 vp8_dequant_idct_add_y_block_media=vp8_dequant_idct_add_y_block_v6
+vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2
 
 prototype void vp8_dequant_idct_add_uv_block "short *q, short *dq, unsigned char *dst_u, unsigned char *dst_v, int stride, char *eobs"
-specialize vp8_dequant_idct_add_uv_block mmx sse2 media neon
+specialize vp8_dequant_idct_add_uv_block mmx sse2 media neon dspr2
 vp8_dequant_idct_add_uv_block_media=vp8_dequant_idct_add_uv_block_v6
+vp8_dequant_idct_add_y_block_dspr2=vp8_dequant_idct_add_y_block_dspr2
 
 #
 # Loopfilter
 #
 prototype void vp8_loop_filter_mbv "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
-specialize vp8_loop_filter_mbv mmx sse2 media neon
+specialize vp8_loop_filter_mbv mmx sse2 media neon dspr2
 vp8_loop_filter_mbv_media=vp8_loop_filter_mbv_armv6
+vp8_loop_filter_mbv_dspr2=vp8_loop_filter_mbv_dspr2
 
 prototype void vp8_loop_filter_bv "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
-specialize vp8_loop_filter_bv mmx sse2 media neon
+specialize vp8_loop_filter_bv mmx sse2 media neon dspr2
 vp8_loop_filter_bv_media=vp8_loop_filter_bv_armv6
+vp8_loop_filter_bv_dspr2=vp8_loop_filter_bv_dspr2
 
 prototype void vp8_loop_filter_mbh "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
-specialize vp8_loop_filter_mbh mmx sse2 media neon
+specialize vp8_loop_filter_mbh mmx sse2 media neon dspr2
 vp8_loop_filter_mbh_media=vp8_loop_filter_mbh_armv6
+vp8_loop_filter_mbh_dspr2=vp8_loop_filter_mbh_dspr2
 
 prototype void vp8_loop_filter_bh "unsigned char *y, unsigned char *u, unsigned char *v, int ystride, int uv_stride, struct loop_filter_info *lfi"
-specialize vp8_loop_filter_bh mmx sse2 media neon
+specialize vp8_loop_filter_bh mmx sse2 media neon dspr2
 vp8_loop_filter_bh_media=vp8_loop_filter_bh_armv6
+vp8_loop_filter_bh_dspr2=vp8_loop_filter_bh_dspr2
 
 
 prototype void vp8_loop_filter_simple_mbv "unsigned char *y, int ystride, const unsigned char *blimit"
@@ -90,37 +99,45 @@ vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon
 #
 #idct16
 prototype void vp8_short_idct4x4llm "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride"
-specialize vp8_short_idct4x4llm mmx media neon
+specialize vp8_short_idct4x4llm mmx media neon dspr2
 vp8_short_idct4x4llm_media=vp8_short_idct4x4llm_v6_dual
+vp8_short_idct4x4llm_dspr2=vp8_short_idct4x4llm_dspr2
 
 #iwalsh1
 prototype void vp8_short_inv_walsh4x4_1 "short *input, short *output"
+specialize vp8_short_inv_walsh4x4_1 dspr2
+vp8_short_inv_walsh4x4_1_dspr2=vp8_short_inv_walsh4x4_1_dspr2
 # no asm yet
 
 #iwalsh16
 prototype void vp8_short_inv_walsh4x4 "short *input, short *output"
-specialize vp8_short_inv_walsh4x4 mmx sse2 media neon
+specialize vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2
 vp8_short_inv_walsh4x4_media=vp8_short_inv_walsh4x4_v6
+vp8_short_inv_walsh4x4_dspr2=vp8_short_inv_walsh4x4_dspr2
 
 #idct1_scalar_add
 prototype void vp8_dc_only_idct_add "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride"
-specialize vp8_dc_only_idct_add	mmx media neon
+specialize vp8_dc_only_idct_add	mmx media neon dspr2
 vp8_dc_only_idct_add_media=vp8_dc_only_idct_add_v6
+vp8_dc_only_idct_add_dspr2=vp8_dc_only_idct_add_dspr2
 
 #
 # RECON
 #
 prototype void vp8_copy_mem16x16 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
-specialize vp8_copy_mem16x16 mmx sse2 media neon
+specialize vp8_copy_mem16x16 mmx sse2 media neon dspr2
 vp8_copy_mem16x16_media=vp8_copy_mem16x16_v6
+vp8_copy_mem16x16_dspr2=vp8_copy_mem16x16_dspr2
 
 prototype void vp8_copy_mem8x8 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
-specialize vp8_copy_mem8x8 mmx media neon
+specialize vp8_copy_mem8x8 mmx media neon dspr2
 vp8_copy_mem8x8_media=vp8_copy_mem8x8_v6
+vp8_copy_mem8x8_dspr2=vp8_copy_mem8x8_dspr2
 
 prototype void vp8_copy_mem8x4 "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch"
-specialize vp8_copy_mem8x4 mmx media neon
+specialize vp8_copy_mem8x4 mmx media neon dspr2
 vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6
+vp8_copy_mem8x4_dspr2=vp8_copy_mem8x4_dspr2
 
 prototype void vp8_build_intra_predictors_mby_s "struct macroblockd *x, unsigned char * yabove_row, unsigned char * yleft, int left_stride, unsigned char * ypred_ptr, int y_stride"
 specialize vp8_build_intra_predictors_mby_s sse2 ssse3
@@ -129,8 +146,7 @@ specialize vp8_build_intra_predictors_mby_s sse2 ssse3
 prototype void vp8_build_intra_predictors_mbuv_s "struct macroblockd *x, unsigned char * uabove_row, unsigned char * vabove_row,  unsigned char *uleft, unsigned char *vleft, int left_stride, unsigned char * upred_ptr, unsigned char * vpred_ptr, int pred_stride"
 specialize vp8_build_intra_predictors_mbuv_s sse2 ssse3
 
-prototype void vp8_intra4x4_predict_d "unsigned char *above, unsigned char *left, int left_stride, int b_mode, unsigned char *dst, int dst_stride, unsigned char top_left"
-prototype void vp8_intra4x4_predict "unsigned char *src, int src_stride, int b_mode, unsigned char *dst, int dst_stride"
+prototype void vp8_intra4x4_predict "unsigned char *Above, unsigned char *yleft, int left_stride, B_PREDICTION_MODE b_mode, unsigned char *dst, int dst_stride, unsigned char top_left"
 specialize vp8_intra4x4_predict media
 vp8_intra4x4_predict_media=vp8_intra4x4_predict_armv6
 
@@ -146,9 +162,8 @@ if [ "$CONFIG_POSTPROC" = "yes" ]; then
     specialize vp8_mbpost_proc_across_ip sse2
     vp8_mbpost_proc_across_ip_sse2=vp8_mbpost_proc_across_ip_xmm
 
-    prototype void vp8_post_proc_down_and_across "unsigned char *src, unsigned char *dst, int src_pitch, int dst_pitch, int rows, int cols, int flimit"
-    specialize vp8_post_proc_down_and_across mmx sse2
-    vp8_post_proc_down_and_across_sse2=vp8_post_proc_down_and_across_xmm
+    prototype void vp8_post_proc_down_and_across_mb_row "unsigned char *src, unsigned char *dst, int src_pitch, int dst_pitch, int cols, unsigned char *flimits, int size"
+    specialize vp8_post_proc_down_and_across_mb_row sse2
 
     prototype void vp8_plane_add_noise "unsigned char *s, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int w, unsigned int h, int pitch"
     specialize vp8_plane_add_noise mmx sse2
@@ -177,20 +192,24 @@ fi
 # Subpixel
 #
 prototype void vp8_sixtap_predict16x16 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_sixtap_predict16x16 mmx sse2 ssse3 media neon
+specialize vp8_sixtap_predict16x16 mmx sse2 ssse3 media neon dspr2
 vp8_sixtap_predict16x16_media=vp8_sixtap_predict16x16_armv6
+vp8_sixtap_predict16x16_dspr2=vp8_sixtap_predict16x16_dspr2
 
 prototype void vp8_sixtap_predict8x8 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_sixtap_predict8x8 mmx sse2 ssse3 media neon
+specialize vp8_sixtap_predict8x8 mmx sse2 ssse3 media neon dspr2
 vp8_sixtap_predict8x8_media=vp8_sixtap_predict8x8_armv6
+vp8_sixtap_predict8x8_dspr2=vp8_sixtap_predict8x8_dspr2
 
 prototype void vp8_sixtap_predict8x4 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_sixtap_predict8x4 mmx sse2 ssse3 media neon
+specialize vp8_sixtap_predict8x4 mmx sse2 ssse3 media neon dspr2
 vp8_sixtap_predict8x4_media=vp8_sixtap_predict8x4_armv6
+vp8_sixtap_predict8x4_dspr2=vp8_sixtap_predict8x4_dspr2
 
 prototype void vp8_sixtap_predict4x4 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_sixtap_predict4x4 mmx ssse3 media neon
+specialize vp8_sixtap_predict4x4 mmx ssse3 media neon dspr2
 vp8_sixtap_predict4x4_media=vp8_sixtap_predict4x4_armv6
+vp8_sixtap_predict4x4_dspr2=vp8_sixtap_predict4x4_dspr2
 
 prototype void vp8_bilinear_predict16x16 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
 specialize vp8_bilinear_predict16x16 mmx sse2 ssse3 media neon
@@ -276,23 +295,23 @@ vp8_variance_halfpixvar16x16_hv_media=vp8_variance_halfpixvar16x16_hv_armv6
 #
 # Single block SAD
 #
-prototype unsigned int vp8_sad4x4 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
+prototype unsigned int vp8_sad4x4 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
 specialize vp8_sad4x4 mmx sse2 neon
 vp8_sad4x4_sse2=vp8_sad4x4_wmt
 
-prototype unsigned int vp8_sad8x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
+prototype unsigned int vp8_sad8x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
 specialize vp8_sad8x8 mmx sse2 neon
 vp8_sad8x8_sse2=vp8_sad8x8_wmt
 
-prototype unsigned int vp8_sad8x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
+prototype unsigned int vp8_sad8x16 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
 specialize vp8_sad8x16 mmx sse2 neon
 vp8_sad8x16_sse2=vp8_sad8x16_wmt
 
-prototype unsigned int vp8_sad16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
+prototype unsigned int vp8_sad16x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
 specialize vp8_sad16x8 mmx sse2 neon
 vp8_sad16x8_sse2=vp8_sad16x8_wmt
 
-prototype unsigned int vp8_sad16x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int max_sad"
+prototype unsigned int vp8_sad16x16 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad"
 specialize vp8_sad16x16 mmx sse2 sse3 media neon
 vp8_sad16x16_sse2=vp8_sad16x16_wmt
 vp8_sad16x16_media=vp8_sad16x16_armv6
@@ -300,59 +319,59 @@ vp8_sad16x16_media=vp8_sad16x16_armv6
 #
 # Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
 #
-prototype void vp8_sad4x4x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad4x4x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad4x4x3 sse3
 
-prototype void vp8_sad8x8x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad8x8x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad8x8x3 sse3
 
-prototype void vp8_sad8x16x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad8x16x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad8x16x3 sse3
 
-prototype void vp8_sad16x8x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad16x8x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad16x8x3 sse3 ssse3
 
-prototype void vp8_sad16x16x3 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad16x16x3 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad16x16x3 sse3 ssse3
 
 # Note the only difference in the following prototypes is that they return into
 # an array of short
-prototype void vp8_sad4x4x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
+prototype void vp8_sad4x4x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
 specialize vp8_sad4x4x8 sse4_1
 vp8_sad4x4x8_sse4_1=vp8_sad4x4x8_sse4
 
-prototype void vp8_sad8x8x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
+prototype void vp8_sad8x8x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
 specialize vp8_sad8x8x8 sse4_1
 vp8_sad8x8x8_sse4_1=vp8_sad8x8x8_sse4
 
-prototype void vp8_sad8x16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
+prototype void vp8_sad8x16x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
 specialize vp8_sad8x16x8 sse4_1
 vp8_sad8x16x8_sse4_1=vp8_sad8x16x8_sse4
 
-prototype void vp8_sad16x8x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
+prototype void vp8_sad16x8x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
 specialize vp8_sad16x8x8 sse4_1
 vp8_sad16x8x8_sse4_1=vp8_sad16x8x8_sse4
 
-prototype void vp8_sad16x16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
+prototype void vp8_sad16x16x8 "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array"
 specialize vp8_sad16x16x8 sse4_1
 vp8_sad16x16x8_sse4_1=vp8_sad16x16x8_sse4
 
 #
 # Multi-block SAD, comparing a reference to N independent blocks
 #
-prototype void vp8_sad4x4x4d "const unsigned char *src_ptr, int source_stride, unsigned char *ref_ptr[4], int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad4x4x4d "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad4x4x4d sse3
 
-prototype void vp8_sad8x8x4d "const unsigned char *src_ptr, int source_stride, unsigned char *ref_ptr[4], int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad8x8x4d "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad8x8x4d sse3
 
-prototype void vp8_sad8x16x4d "const unsigned char *src_ptr, int source_stride, unsigned char *ref_ptr[4], int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad8x16x4d "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad8x16x4d sse3
 
-prototype void vp8_sad16x8x4d "const unsigned char *src_ptr, int source_stride, unsigned char *ref_ptr[4], int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad16x8x4d "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad16x8x4d sse3
 
-prototype void vp8_sad16x16x4d "const unsigned char *src_ptr, int source_stride, unsigned char *ref_ptr[4], int  ref_stride, unsigned int *sad_array"
+prototype void vp8_sad16x16x4d "const unsigned char *src_ptr, int src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
 specialize vp8_sad16x16x4d sse3
 
 #

vp8/common/sad_c.c

@@ -9,21 +9,15 @@
  */
 
 
+#include <limits.h>
 #include <stdlib.h>
 #include "vpx_config.h"
 #include "vpx/vpx_integer.h"
 
-static
-unsigned int sad_mx_n_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad,
-    int  m,
-    int  n)
+static unsigned int sad_mx_n_c(const unsigned char *src_ptr, int src_stride,
+                               const unsigned char *ref_ptr, int ref_stride,
+                               unsigned int max_sad, int m, int n)
 {
-
     int r, c;
     unsigned int sad = 0;
 
@@ -48,298 +42,211 @@ unsigned int sad_mx_n_c(
  * implementations of these functions are not required to check it.
  */
 
-unsigned int vp8_sad16x16_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad)
+unsigned int vp8_sad16x16_c(const unsigned char *src_ptr, int src_stride,
+                            const unsigned char *ref_ptr, int ref_stride,
+                            unsigned int max_sad)
 {
-
     return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 16);
 }
 
-
-unsigned int vp8_sad8x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad)
+unsigned int vp8_sad8x8_c(const unsigned char *src_ptr, int src_stride,
+                          const unsigned char *ref_ptr, int ref_stride,
+                          unsigned int max_sad)
 {
-
     return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 8);
 }
 
-
-unsigned int vp8_sad16x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad)
+unsigned int vp8_sad16x8_c(const unsigned char *src_ptr, int src_stride,
+                           const unsigned char *ref_ptr, int ref_stride,
+                           unsigned int max_sad)
 {
-
     return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 8);
 
 }
 
-
-unsigned int vp8_sad8x16_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad)
+unsigned int vp8_sad8x16_c(const unsigned char *src_ptr, int src_stride,
+                           const unsigned char *ref_ptr, int ref_stride,
+                           unsigned int max_sad)
 {
-
     return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 16);
 }
 
-
-unsigned int vp8_sad4x4_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    int  max_sad)
+unsigned int vp8_sad4x4_c(const unsigned char *src_ptr, int src_stride,
+                          const unsigned char *ref_ptr, int ref_stride,
+                          unsigned int max_sad)
 {
-
     return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 4, 4);
 }
 
-void vp8_sad16x16x3_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad16x16x3_c(const unsigned char *src_ptr, int src_stride,
+                      const unsigned char *ref_ptr, int ref_stride,
+                      unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
 }
 
-void vp8_sad16x16x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned short *sad_array
-)
+void vp8_sad16x16x8_c(const unsigned char *src_ptr, int src_stride,
+                      const unsigned char *ref_ptr, int ref_stride,
+                      unsigned short *sad_array)
 {
-    sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-    sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
-    sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
-    sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
-    sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
-    sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+    sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+    sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+    sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+    sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+    sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+    sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
 }
 
-void vp8_sad16x8x3_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad16x8x3_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char *ref_ptr, int ref_stride,
+                     unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
 }
 
-void vp8_sad16x8x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned short *sad_array
-)
+void vp8_sad16x8x8_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char *ref_ptr, int ref_stride,
+                     unsigned short *sad_array)
 {
-    sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-    sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
-    sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
-    sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
-    sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
-    sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+    sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+    sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+    sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+    sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+    sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+    sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x8x3_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad8x8x3_c(const unsigned char *src_ptr, int src_stride,
+                    const unsigned char *ref_ptr, int ref_stride,
+                    unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x8x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned short *sad_array
-)
+void vp8_sad8x8x8_c(const unsigned char *src_ptr, int src_stride,
+                    const unsigned char *ref_ptr, int ref_stride,
+                    unsigned short *sad_array)
 {
-    sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-    sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
-    sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
-    sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
-    sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
-    sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+    sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+    sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+    sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+    sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+    sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+    sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x16x3_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad8x16x3_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char *ref_ptr, int ref_stride,
+                     unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x16x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned short *sad_array
-)
+void vp8_sad8x16x8_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char *ref_ptr, int ref_stride,
+                     unsigned short *sad_array)
 {
-    sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-    sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
-    sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
-    sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
-    sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
-    sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+    sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+    sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+    sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+    sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+    sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+    sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
 }
 
-void vp8_sad4x4x3_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad4x4x3_c(const unsigned char *src_ptr, int src_stride,
+                    const unsigned char *ref_ptr, int ref_stride,
+                    unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
 }
 
-void vp8_sad4x4x8_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    const unsigned char *ref_ptr,
-    int  ref_stride,
-    unsigned short *sad_array
-)
+void vp8_sad4x4x8_c(const unsigned char *src_ptr, int src_stride,
+                    const unsigned char *ref_ptr, int ref_stride,
+                    unsigned short *sad_array)
 {
-    sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr  , ref_stride, 0x7fffffff);
-    sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
-    sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-    sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
-    sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
-    sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
-    sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
-    sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+    sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 0, ref_stride, UINT_MAX);
+    sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, UINT_MAX);
+    sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, UINT_MAX);
+    sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3, ref_stride, UINT_MAX);
+    sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, UINT_MAX);
+    sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, UINT_MAX);
+    sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6, ref_stride, UINT_MAX);
+    sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, UINT_MAX);
 }
 
-void vp8_sad16x16x4d_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *ref_ptr[],
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad16x16x4d_c(const unsigned char *src_ptr, int src_stride,
+                       const unsigned char * const ref_ptr[], int ref_stride,
+                       unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
-    sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+    sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
 }
 
-void vp8_sad16x8x4d_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *ref_ptr[],
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad16x8x4d_c(const unsigned char *src_ptr, int src_stride,
+                      const unsigned char * const ref_ptr[], int ref_stride,
+                      unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
-    sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+    sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x8x4d_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *ref_ptr[],
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad8x8x4d_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char * const ref_ptr[], int ref_stride,
+                     unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
-    sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+    sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
 }
 
-void vp8_sad8x16x4d_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *ref_ptr[],
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad8x16x4d_c(const unsigned char *src_ptr, int src_stride,
+                      const unsigned char * const ref_ptr[], int ref_stride,
+                      unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
-    sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+    sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
 }
 
-void vp8_sad4x4x4d_c(
-    const unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *ref_ptr[],
-    int  ref_stride,
-    unsigned int *sad_array
-)
+void vp8_sad4x4x4d_c(const unsigned char *src_ptr, int src_stride,
+                     const unsigned char * const ref_ptr[], int  ref_stride,
+                     unsigned int *sad_array)
 {
-    sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
-    sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
-    sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
-    sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+    sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride, UINT_MAX);
+    sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride, UINT_MAX);
+    sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride, UINT_MAX);
+    sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride, UINT_MAX);
 }
 
 /* Copy 2 macroblocks to a buffer */
-void vp8_copy32xn_c(
-    unsigned char *src_ptr,
-    int  src_stride,
-    unsigned char *dst_ptr,
-    int  dst_stride,
-    int height)
+void vp8_copy32xn_c(unsigned char *src_ptr, int src_stride,
+                    unsigned char *dst_ptr, int dst_stride,
+                    int height)
 {
     int r;
 

vp8/common/setupintrarecon.c

@@ -30,3 +30,10 @@ void vp8_setup_intra_recon(YV12_BUFFER_CONFIG *ybf)
         ybf->v_buffer[ybf->uv_stride *i - 1] = (unsigned char) 129;
 
 }
+
+void vp8_setup_intra_recon_top_line(YV12_BUFFER_CONFIG *ybf)
+{
+    vpx_memset(ybf->y_buffer - 1 - ybf->y_stride, 127, ybf->y_width + 5);
+    vpx_memset(ybf->u_buffer - 1 - ybf->uv_stride, 127, ybf->uv_width + 5);
+    vpx_memset(ybf->v_buffer - 1 - ybf->uv_stride, 127, ybf->uv_width + 5);
+}

vp8/common/setupintrarecon.h

@@ -11,3 +11,23 @@
 
 #include "vpx_scale/yv12config.h"
 extern void vp8_setup_intra_recon(YV12_BUFFER_CONFIG *ybf);
+extern void vp8_setup_intra_recon_top_line(YV12_BUFFER_CONFIG *ybf);
+
+static
+void setup_intra_recon_left(unsigned char *y_buffer,
+                            unsigned char *u_buffer,
+                            unsigned char *v_buffer,
+                            int y_stride,
+                            int uv_stride)
+{
+    int i;
+
+    for (i = 0; i < 16; i++)
+        y_buffer[y_stride *i] = (unsigned char) 129;
+
+    for (i = 0; i < 8; i++)
+        u_buffer[uv_stride *i] = (unsigned char) 129;
+
+    for (i = 0; i < 8; i++)
+        v_buffer[uv_stride *i] = (unsigned char) 129;
+}

vp8/common/variance.h

@@ -12,14 +12,14 @@
 #ifndef VARIANCE_H
 #define VARIANCE_H
 
-typedef unsigned int(*vp8_sad_fn_t)
-    (
+#include "vpx_config.h"
+
+typedef unsigned int(*vp8_sad_fn_t)(
     const unsigned char *src_ptr,
     int source_stride,
     const unsigned char *ref_ptr,
     int ref_stride,
-    int max_sad
-    );
+    unsigned int max_sad);
 
 typedef void (*vp8_copy32xn_fn_t)(
     const unsigned char *src_ptr,
@@ -48,7 +48,7 @@ typedef void (*vp8_sad_multi_d_fn_t)
     (
      const unsigned char *src_ptr,
      int source_stride,
-     unsigned char *ref_ptr[4],
+     const unsigned char * const ref_ptr[],
      int  ref_stride,
      unsigned int *sad_array
     );

vp8/common/variance_c.c

@@ -205,14 +205,14 @@ static void var_filter_block2d_bil_first_pass
     {
         for (j = 0; j < output_width; j++)
         {
-            // Apply bilinear filter
+            /* Apply bilinear filter */
             output_ptr[j] = (((int)src_ptr[0]          * vp8_filter[0]) +
                              ((int)src_ptr[pixel_step] * vp8_filter[1]) +
                              (VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
             src_ptr++;
         }
 
-        // Next row...
+        /* Next row... */
         src_ptr    += src_pixels_per_line - output_width;
         output_ptr += output_width;
     }
@@ -264,15 +264,15 @@ static void var_filter_block2d_bil_second_pass
     {
         for (j = 0; j < output_width; j++)
         {
-            // Apply filter
-            Temp = ((int)src_ptr[0]         * vp8_filter[0]) +
+            /* Apply filter */
+            Temp = ((int)src_ptr[0]          * vp8_filter[0]) +
                    ((int)src_ptr[pixel_step] * vp8_filter[1]) +
                    (VP8_FILTER_WEIGHT / 2);
             output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
             src_ptr++;
         }
 
-        // Next row...
+        /* Next row... */
         src_ptr    += src_pixels_per_line - output_width;
         output_ptr += output_width;
     }
@@ -292,15 +292,15 @@ unsigned int vp8_sub_pixel_variance4x4_c
 {
     unsigned char  temp2[20*16];
     const short *HFilter, *VFilter;
-    unsigned short FData3[5*4]; // Temp data bufffer used in filtering
+    unsigned short FData3[5*4]; /* Temp data bufffer used in filtering */
 
     HFilter = vp8_bilinear_filters[xoffset];
     VFilter = vp8_bilinear_filters[yoffset];
 
-    // First filter 1d Horizontal
+    /* First filter 1d Horizontal */
     var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 5, 4, HFilter);
 
-    // Now filter Verticaly
+    /* Now filter Verticaly */
     var_filter_block2d_bil_second_pass(FData3, temp2, 4,  4,  4,  4, VFilter);
 
     return vp8_variance4x4_c(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
@@ -318,7 +318,7 @@ unsigned int vp8_sub_pixel_variance8x8_c
     unsigned int *sse
 )
 {
-    unsigned short FData3[9*8]; // Temp data bufffer used in filtering
+    unsigned short FData3[9*8]; /* Temp data bufffer used in filtering */
     unsigned char  temp2[20*16];
     const short *HFilter, *VFilter;
 
@@ -342,7 +342,7 @@ unsigned int vp8_sub_pixel_variance16x16_c
     unsigned int *sse
 )
 {
-    unsigned short FData3[17*16];   // Temp data bufffer used in filtering
+    unsigned short FData3[17*16];   /* Temp data bufffer used in filtering */
     unsigned char  temp2[20*16];
     const short *HFilter, *VFilter;
 
@@ -418,7 +418,7 @@ unsigned int vp8_sub_pixel_variance16x8_c
     unsigned int *sse
 )
 {
-    unsigned short FData3[16*9];    // Temp data bufffer used in filtering
+    unsigned short FData3[16*9];    /* Temp data bufffer used in filtering */
     unsigned char  temp2[20*16];
     const short *HFilter, *VFilter;
 
@@ -442,7 +442,7 @@ unsigned int vp8_sub_pixel_variance8x16_c
     unsigned int *sse
 )
 {
-    unsigned short FData3[9*16];    // Temp data bufffer used in filtering
+    unsigned short FData3[9*16];    /* Temp data bufffer used in filtering */
     unsigned char  temp2[20*16];
     const short *HFilter, *VFilter;
 

vp8/common/x86/idctllm_mmx_test.cc

@@ -1,31 +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.
- */
-
-
- extern "C" {
-    void vp8_short_idct4x4llm_mmx(short *input, unsigned char *pred_ptr,
-                            int pred_stride, unsigned char *dst_ptr,
-                            int dst_stride);
-}
-
-#include "vp8/common/idctllm_test.h"
-
-namespace
-{
-
-INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
-                        ::testing::Values(vp8_short_idct4x4llm_mmx));
-
-} // namespace
-
-int main(int argc, char **argv) {
-  ::testing::InitGoogleTest(&argc, argv);
-  return RUN_ALL_TESTS();
-}

vp8/common/x86/loopfilter_block_sse2.asm

@@ -150,6 +150,7 @@ sym(vp8_loop_filter_bh_y_sse2):
 
     push    rbp
     mov     rbp, rsp
+    SAVE_XMM 11
     push    r12
     push    r13
     mov     thresh, arg(4)
@@ -258,6 +259,7 @@ LF_FILTER xmm0, xmm1, xmm3, xmm8, xmm4, xmm2
 %ifidn __OUTPUT_FORMAT__,x64
     pop    r13
     pop    r12
+    RESTORE_XMM
     pop    rbp
 %endif
 

vp8/common/x86/postproc_mmx.asm

@@ -14,271 +14,6 @@
 %define VP8_FILTER_WEIGHT 128
 %define VP8_FILTER_SHIFT  7
 
-;void vp8_post_proc_down_and_across_mmx
-;(
-;    unsigned char *src_ptr,
-;    unsigned char *dst_ptr,
-;    int src_pixels_per_line,
-;    int dst_pixels_per_line,
-;    int rows,
-;    int cols,
-;    int flimit
-;)
-global sym(vp8_post_proc_down_and_across_mmx) PRIVATE
-sym(vp8_post_proc_down_and_across_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-%if ABI_IS_32BIT=1 && CONFIG_PIC=1
-    ; move the global rd onto the stack, since we don't have enough registers
-    ; to do PIC addressing
-    movq        mm0, [GLOBAL(rd)]
-    sub         rsp, 8
-    movq        [rsp], mm0
-%define RD [rsp]
-%else
-%define RD [GLOBAL(rd)]
-%endif
-
-        push        rbx
-        lea         rbx, [GLOBAL(Blur)]
-        movd        mm2, dword ptr arg(6) ;flimit
-        punpcklwd   mm2, mm2
-        punpckldq   mm2, mm2
-
-        mov         rsi,        arg(0) ;src_ptr
-        mov         rdi,        arg(1) ;dst_ptr
-
-        movsxd      rcx, DWORD PTR arg(4) ;rows
-        movsxd      rax, DWORD PTR arg(2) ;src_pixels_per_line ; destination pitch?
-        pxor        mm0, mm0              ; mm0 = 00000000
-
-.nextrow:
-
-        xor         rdx,        rdx       ; clear out rdx for use as loop counter
-.nextcol:
-
-        pxor        mm7, mm7              ; mm7 = 00000000
-        movq        mm6, [rbx + 32 ]      ; mm6 = kernel 2 taps
-        movq        mm3, [rsi]            ; mm4 = r0 p0..p7
-        punpcklbw   mm3, mm0              ; mm3 = p0..p3
-        movq        mm1, mm3              ; mm1 = p0..p3
-        pmullw      mm3, mm6              ; mm3 *= kernel 2 modifiers
-
-        movq        mm6, [rbx + 48]       ; mm6 = kernel 3 taps
-        movq        mm5, [rsi + rax]      ; mm4 = r1 p0..p7
-        punpcklbw   mm5, mm0              ; mm5 = r1 p0..p3
-        pmullw      mm6, mm5              ; mm6 *= p0..p3 * kernel 3 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm6
-
-        ; thresholding
-        movq        mm7, mm1              ; mm7 = r0 p0..p3
-        psubusw     mm7, mm5              ; mm7 = r0 p0..p3 - r1 p0..p3
-        psubusw     mm5, mm1              ; mm5 = r1 p0..p3 - r0 p0..p3
-        paddusw     mm7, mm5              ; mm7 = abs(r0 p0..p3 - r1 p0..p3)
-        pcmpgtw     mm7, mm2
-
-        movq        mm6, [rbx + 64 ]      ; mm6 = kernel 4 modifiers
-        movq        mm5, [rsi + 2*rax]    ; mm4 = r2 p0..p7
-        punpcklbw   mm5, mm0              ; mm5 = r2 p0..p3
-        pmullw      mm6, mm5              ; mm5 *= kernel 4 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = r0 p0..p3
-        psubusw     mm6, mm5              ; mm6 = r0 p0..p3 - r2 p0..p3
-        psubusw     mm5, mm1              ; mm5 = r2 p0..p3 - r2 p0..p3
-        paddusw     mm6, mm5              ; mm6 = abs(r0 p0..p3 - r2 p0..p3)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-
-        neg         rax
-        movq        mm6, [rbx ]           ; kernel 0 taps
-        movq        mm5, [rsi+2*rax]      ; mm4 = r-2 p0..p7
-        punpcklbw   mm5, mm0              ; mm5 = r-2 p0..p3
-        pmullw      mm6, mm5              ; mm5 *= kernel 0 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = r0 p0..p3
-        psubusw     mm6, mm5              ; mm6 = p0..p3 - r-2 p0..p3
-        psubusw     mm5, mm1              ; mm5 = r-2 p0..p3 - p0..p3
-        paddusw     mm6, mm5              ; mm6 = abs(r0 p0..p3 - r-2 p0..p3)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-        movq        mm6, [rbx + 16]       ; kernel 1 taps
-        movq        mm4, [rsi+rax]        ; mm4 = r-1 p0..p7
-        punpcklbw   mm4, mm0              ; mm4 = r-1 p0..p3
-        pmullw      mm6, mm4              ; mm4 *= kernel 1 modifiers.
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = r0 p0..p3
-        psubusw     mm6, mm4              ; mm6 = p0..p3 - r-2 p0..p3
-        psubusw     mm4, mm1              ; mm5 = r-1 p0..p3 - p0..p3
-        paddusw     mm6, mm4              ; mm6 = abs(r0 p0..p3 - r-1 p0..p3)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-
-        paddusw     mm3, RD               ; mm3 += round value
-        psraw       mm3, VP8_FILTER_SHIFT     ; mm3 /= 128
-
-        pand        mm1, mm7              ; mm1 select vals > thresh from source
-        pandn       mm7, mm3              ; mm7 select vals < thresh from blurred result
-        paddusw     mm1, mm7              ; combination
-
-        packuswb    mm1, mm0              ; pack to bytes
-
-        movd        [rdi], mm1            ;
-        neg         rax                   ; pitch is positive
-
-
-        add         rsi, 4
-        add         rdi, 4
-        add         rdx, 4
-
-        cmp         edx, dword ptr arg(5) ;cols
-        jl          .nextcol
-        ; done with the all cols, start the across filtering in place
-        sub         rsi, rdx
-        sub         rdi, rdx
-
-        ; dup the first byte into the left border 8 times
-        movq        mm1,   [rdi]
-        punpcklbw   mm1,   mm1
-        punpcklwd   mm1,   mm1
-        punpckldq   mm1,   mm1
-
-        mov         rdx,    -8
-        movq        [rdi+rdx], mm1
-
-        ; dup the last byte into the right border
-        movsxd      rdx,    dword arg(5)
-        movq        mm1,   [rdi + rdx + -1]
-        punpcklbw   mm1,   mm1
-        punpcklwd   mm1,   mm1
-        punpckldq   mm1,   mm1
-        movq        [rdi+rdx], mm1
-
-
-        push        rax
-        xor         rdx,    rdx
-        mov         rax,    [rdi-4];
-
-.acrossnextcol:
-        pxor        mm7, mm7              ; mm7 = 00000000
-        movq        mm6, [rbx + 32 ]      ;
-        movq        mm4, [rdi+rdx]        ; mm4 = p0..p7
-        movq        mm3, mm4              ; mm3 = p0..p7
-        punpcklbw   mm3, mm0              ; mm3 = p0..p3
-        movq        mm1, mm3              ; mm1 = p0..p3
-        pmullw      mm3, mm6              ; mm3 *= kernel 2 modifiers
-
-        movq        mm6, [rbx + 48]
-        psrlq       mm4, 8                ; mm4 = p1..p7
-        movq        mm5, mm4              ; mm5 = p1..p7
-        punpcklbw   mm5, mm0              ; mm5 = p1..p4
-        pmullw      mm6, mm5              ; mm6 *= p1..p4 * kernel 3 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm6
-
-        ; thresholding
-        movq        mm7, mm1              ; mm7 = p0..p3
-        psubusw     mm7, mm5              ; mm7 = p0..p3 - p1..p4
-        psubusw     mm5, mm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     mm7, mm5              ; mm7 = abs(p0..p3 - p1..p4)
-        pcmpgtw     mm7, mm2
-
-        movq        mm6, [rbx + 64 ]
-        psrlq       mm4, 8                ; mm4 = p2..p7
-        movq        mm5, mm4              ; mm5 = p2..p7
-        punpcklbw   mm5, mm0              ; mm5 = p2..p5
-        pmullw      mm6, mm5              ; mm5 *= kernel 4 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = p0..p3
-        psubusw     mm6, mm5              ; mm6 = p0..p3 - p1..p4
-        psubusw     mm5, mm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     mm6, mm5              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-
-        movq        mm6, [rbx ]
-        movq        mm4, [rdi+rdx-2]      ; mm4 = p-2..p5
-        movq        mm5, mm4              ; mm5 = p-2..p5
-        punpcklbw   mm5, mm0              ; mm5 = p-2..p1
-        pmullw      mm6, mm5              ; mm5 *= kernel 0 modifiers
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = p0..p3
-        psubusw     mm6, mm5              ; mm6 = p0..p3 - p1..p4
-        psubusw     mm5, mm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     mm6, mm5              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-        movq        mm6, [rbx + 16]
-        psrlq       mm4, 8                ; mm4 = p-1..p5
-        punpcklbw   mm4, mm0              ; mm4 = p-1..p2
-        pmullw      mm6, mm4              ; mm4 *= kernel 1 modifiers.
-        paddusw     mm3, mm6              ; mm3 += mm5
-
-        ; thresholding
-        movq        mm6, mm1              ; mm6 = p0..p3
-        psubusw     mm6, mm4              ; mm6 = p0..p3 - p1..p4
-        psubusw     mm4, mm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     mm6, mm4              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     mm6, mm2
-        por         mm7, mm6              ; accumulate thresholds
-
-        paddusw     mm3, RD               ; mm3 += round value
-        psraw       mm3, VP8_FILTER_SHIFT     ; mm3 /= 128
-
-        pand        mm1, mm7              ; mm1 select vals > thresh from source
-        pandn       mm7, mm3              ; mm7 select vals < thresh from blurred result
-        paddusw     mm1, mm7              ; combination
-
-        packuswb    mm1, mm0              ; pack to bytes
-        mov         DWORD PTR [rdi+rdx-4],  eax   ; store previous four bytes
-        movd        eax,    mm1
-
-        add         rdx, 4
-        cmp         edx, dword ptr arg(5) ;cols
-        jl          .acrossnextcol;
-
-        mov         DWORD PTR [rdi+rdx-4],  eax
-        pop         rax
-
-        ; done with this rwo
-        add         rsi,rax               ; next line
-        movsxd      rax, dword ptr arg(3) ;dst_pixels_per_line ; destination pitch?
-        add         rdi,rax               ; next destination
-        movsxd      rax, dword ptr arg(2) ;src_pixels_per_line ; destination pitch?
-
-        dec         rcx                   ; decrement count
-        jnz         .nextrow               ; next row
-        pop         rbx
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-%undef RD
-
-
 ;void vp8_mbpost_proc_down_mmx(unsigned char *dst,
 ;                             int pitch, int rows, int cols,int flimit)
 extern sym(vp8_rv)

vp8/common/x86/postproc_sse2.asm

@@ -11,146 +11,159 @@
 
 %include "vpx_ports/x86_abi_support.asm"
 
-;void vp8_post_proc_down_and_across_xmm
+;macro in deblock functions
+%macro FIRST_2_ROWS 0
+        movdqa      xmm4,       xmm0
+        movdqa      xmm6,       xmm0
+        movdqa      xmm5,       xmm1
+        pavgb       xmm5,       xmm3
+
+        ;calculate absolute value
+        psubusb     xmm4,       xmm1
+        psubusb     xmm1,       xmm0
+        psubusb     xmm6,       xmm3
+        psubusb     xmm3,       xmm0
+        paddusb     xmm4,       xmm1
+        paddusb     xmm6,       xmm3
+
+        ;get threshold
+        movdqa      xmm2,       flimit
+        pxor        xmm1,       xmm1
+        movdqa      xmm7,       xmm2
+
+        ;get mask
+        psubusb     xmm2,       xmm4
+        psubusb     xmm7,       xmm6
+        pcmpeqb     xmm2,       xmm1
+        pcmpeqb     xmm7,       xmm1
+        por         xmm7,       xmm2
+%endmacro
+
+%macro SECOND_2_ROWS 0
+        movdqa      xmm6,       xmm0
+        movdqa      xmm4,       xmm0
+        movdqa      xmm2,       xmm1
+        pavgb       xmm1,       xmm3
+
+        ;calculate absolute value
+        psubusb     xmm6,       xmm2
+        psubusb     xmm2,       xmm0
+        psubusb     xmm4,       xmm3
+        psubusb     xmm3,       xmm0
+        paddusb     xmm6,       xmm2
+        paddusb     xmm4,       xmm3
+
+        pavgb       xmm5,       xmm1
+
+        ;get threshold
+        movdqa      xmm2,       flimit
+        pxor        xmm1,       xmm1
+        movdqa      xmm3,       xmm2
+
+        ;get mask
+        psubusb     xmm2,       xmm6
+        psubusb     xmm3,       xmm4
+        pcmpeqb     xmm2,       xmm1
+        pcmpeqb     xmm3,       xmm1
+
+        por         xmm7,       xmm2
+        por         xmm7,       xmm3
+
+        pavgb       xmm5,       xmm0
+
+        ;decide if or not to use filtered value
+        pand        xmm0,       xmm7
+        pandn       xmm7,       xmm5
+        paddusb     xmm0,       xmm7
+%endmacro
+
+%macro UPDATE_FLIMIT 0
+        movdqa      xmm2,       XMMWORD PTR [rbx]
+        movdqa      [rsp],      xmm2
+        add         rbx,        16
+%endmacro
+
+;void vp8_post_proc_down_and_across_mb_row_sse2
 ;(
 ;    unsigned char *src_ptr,
 ;    unsigned char *dst_ptr,
 ;    int src_pixels_per_line,
 ;    int dst_pixels_per_line,
-;    int rows,
 ;    int cols,
-;    int flimit
+;    int *flimits,
+;    int size
 ;)
-global sym(vp8_post_proc_down_and_across_xmm) PRIVATE
-sym(vp8_post_proc_down_and_across_xmm):
+global sym(vp8_post_proc_down_and_across_mb_row_sse2) PRIVATE
+sym(vp8_post_proc_down_and_across_mb_row_sse2):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
     SAVE_XMM 7
-    GET_GOT     rbx
+    push        rbx
     push        rsi
     push        rdi
     ; end prolog
-
-%if ABI_IS_32BIT=1 && CONFIG_PIC=1
     ALIGN_STACK 16, rax
-    ; move the global rd onto the stack, since we don't have enough registers
-    ; to do PIC addressing
-    movdqa      xmm0, [GLOBAL(rd42)]
     sub         rsp, 16
-    movdqa      [rsp], xmm0
-%define RD42 [rsp]
-%else
-%define RD42 [GLOBAL(rd42)]
-%endif
 
+        ; put flimit on stack
+        mov         rbx,        arg(5)           ;flimits ptr
+        UPDATE_FLIMIT
 
-        movd        xmm2,       dword ptr arg(6) ;flimit
-        punpcklwd   xmm2,       xmm2
-        punpckldq   xmm2,       xmm2
-        punpcklqdq  xmm2,       xmm2
+%define flimit [rsp]
 
-        mov         rsi,        arg(0) ;src_ptr
-        mov         rdi,        arg(1) ;dst_ptr
-
-        movsxd      rcx,        DWORD PTR arg(4) ;rows
-        movsxd      rax,        DWORD PTR arg(2) ;src_pixels_per_line ; destination pitch?
-        pxor        xmm0,       xmm0              ; mm0 = 00000000
+        mov         rsi,        arg(0)           ;src_ptr
+        mov         rdi,        arg(1)           ;dst_ptr
 
+        movsxd      rax,        DWORD PTR arg(2) ;src_pixels_per_line
+        movsxd      rcx,        DWORD PTR arg(6) ;rows in a macroblock
 .nextrow:
-
-        xor         rdx,        rdx       ; clear out rdx for use as loop counter
+        xor         rdx,        rdx              ;col
 .nextcol:
-        movq        xmm3,       QWORD PTR [rsi]         ; mm4 = r0 p0..p7
-        punpcklbw   xmm3,       xmm0                    ; mm3 = p0..p3
-        movdqa      xmm1,       xmm3                    ; mm1 = p0..p3
-        psllw       xmm3,       2                       ;
-
-        movq        xmm5,       QWORD PTR [rsi + rax]   ; mm4 = r1 p0..p7
-        punpcklbw   xmm5,       xmm0                    ; mm5 = r1 p0..p3
-        paddusw     xmm3,       xmm5                    ; mm3 += mm6
-
-        ; thresholding
-        movdqa      xmm7,       xmm1                    ; mm7 = r0 p0..p3
-        psubusw     xmm7,       xmm5                    ; mm7 = r0 p0..p3 - r1 p0..p3
-        psubusw     xmm5,       xmm1                    ; mm5 = r1 p0..p3 - r0 p0..p3
-        paddusw     xmm7,       xmm5                    ; mm7 = abs(r0 p0..p3 - r1 p0..p3)
-        pcmpgtw     xmm7,       xmm2
-
-        movq        xmm5,       QWORD PTR [rsi + 2*rax] ; mm4 = r2 p0..p7
-        punpcklbw   xmm5,       xmm0                    ; mm5 = r2 p0..p3
-        paddusw     xmm3,       xmm5                    ; mm3 += mm5
-
-        ; thresholding
-        movdqa      xmm6,       xmm1                    ; mm6 = r0 p0..p3
-        psubusw     xmm6,       xmm5                    ; mm6 = r0 p0..p3 - r2 p0..p3
-        psubusw     xmm5,       xmm1                    ; mm5 = r2 p0..p3 - r2 p0..p3
-        paddusw     xmm6,       xmm5                    ; mm6 = abs(r0 p0..p3 - r2 p0..p3)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6                    ; accumulate thresholds
+        ;load current and next 2 rows
+        movdqu      xmm0,       XMMWORD PTR [rsi]
+        movdqu      xmm1,       XMMWORD PTR [rsi + rax]
+        movdqu      xmm3,       XMMWORD PTR [rsi + 2*rax]
 
+        FIRST_2_ROWS
 
+        ;load above 2 rows
         neg         rax
-        movq        xmm5,       QWORD PTR [rsi+2*rax]   ; mm4 = r-2 p0..p7
-        punpcklbw   xmm5,       xmm0                    ; mm5 = r-2 p0..p3
-        paddusw     xmm3,       xmm5                    ; mm3 += mm5
+        movdqu      xmm1,       XMMWORD PTR [rsi + 2*rax]
+        movdqu      xmm3,       XMMWORD PTR [rsi + rax]
 
-        ; thresholding
-        movdqa      xmm6,       xmm1                    ; mm6 = r0 p0..p3
-        psubusw     xmm6,       xmm5                    ; mm6 = p0..p3 - r-2 p0..p3
-        psubusw     xmm5,       xmm1                    ; mm5 = r-2 p0..p3 - p0..p3
-        paddusw     xmm6,       xmm5                    ; mm6 = abs(r0 p0..p3 - r-2 p0..p3)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6                    ; accumulate thresholds
+        SECOND_2_ROWS
 
-        movq        xmm4,       QWORD PTR [rsi+rax]     ; mm4 = r-1 p0..p7
-        punpcklbw   xmm4,       xmm0                    ; mm4 = r-1 p0..p3
-        paddusw     xmm3,       xmm4                    ; mm3 += mm5
+        movdqu      XMMWORD PTR [rdi], xmm0
 
-        ; thresholding
-        movdqa      xmm6,       xmm1                    ; mm6 = r0 p0..p3
-        psubusw     xmm6,       xmm4                    ; mm6 = p0..p3 - r-2 p0..p3
-        psubusw     xmm4,       xmm1                    ; mm5 = r-1 p0..p3 - p0..p3
-        paddusw     xmm6,       xmm4                    ; mm6 = abs(r0 p0..p3 - r-1 p0..p3)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6                    ; accumulate thresholds
+        neg         rax                          ; positive stride
+        add         rsi,        16
+        add         rdi,        16
 
+        add         rdx,        16
+        cmp         edx,        dword arg(4)     ;cols
+        jge         .downdone
+        UPDATE_FLIMIT
+        jmp         .nextcol
 
-        paddusw     xmm3,       RD42                    ; mm3 += round value
-        psraw       xmm3,       3                       ; mm3 /= 8
-
-        pand        xmm1,       xmm7                    ; mm1 select vals > thresh from source
-        pandn       xmm7,       xmm3                    ; mm7 select vals < thresh from blurred result
-        paddusw     xmm1,       xmm7                    ; combination
-
-        packuswb    xmm1,       xmm0                    ; pack to bytes
-        movq        QWORD PTR [rdi], xmm1             ;
-
-        neg         rax                   ; pitch is positive
-        add         rsi,        8
-        add         rdi,        8
-
-        add         rdx,        8
-        cmp         edx,        dword arg(5) ;cols
-
-        jl          .nextcol
-
+.downdone:
         ; done with the all cols, start the across filtering in place
         sub         rsi,        rdx
         sub         rdi,        rdx
 
+        mov         rbx,        arg(5) ; flimits
+        UPDATE_FLIMIT
 
         ; dup the first byte into the left border 8 times
         movq        mm1,   [rdi]
         punpcklbw   mm1,   mm1
         punpcklwd   mm1,   mm1
         punpckldq   mm1,   mm1
-
         mov         rdx,    -8
         movq        [rdi+rdx], mm1
 
         ; dup the last byte into the right border
-        movsxd      rdx,    dword arg(5)
+        movsxd      rdx,    dword arg(4)
         movq        mm1,   [rdi + rdx + -1]
         punpcklbw   mm1,   mm1
         punpcklwd   mm1,   mm1
@@ -158,113 +171,64 @@ sym(vp8_post_proc_down_and_across_xmm):
         movq        [rdi+rdx], mm1
 
         xor         rdx,        rdx
-        movq        mm0,        QWORD PTR [rdi-8];
+        movq        mm0,        QWORD PTR [rdi-16];
+        movq        mm1,        QWORD PTR [rdi-8];
 
 .acrossnextcol:
-        movq        xmm7,       QWORD PTR [rdi +rdx -2]
-        movd        xmm4,       DWORD PTR [rdi +rdx +6]
+        movdqu      xmm0,       XMMWORD PTR [rdi + rdx]
+        movdqu      xmm1,       XMMWORD PTR [rdi + rdx -2]
+        movdqu      xmm3,       XMMWORD PTR [rdi + rdx -1]
 
-        pslldq      xmm4,       8
-        por         xmm4,       xmm7
+        FIRST_2_ROWS
 
-        movdqa      xmm3,       xmm4
-        psrldq      xmm3,       2
-        punpcklbw   xmm3,       xmm0              ; mm3 = p0..p3
-        movdqa      xmm1,       xmm3              ; mm1 = p0..p3
-        psllw       xmm3,       2
+        movdqu      xmm1,       XMMWORD PTR [rdi + rdx +1]
+        movdqu      xmm3,       XMMWORD PTR [rdi + rdx +2]
 
+        SECOND_2_ROWS
 
-        movdqa      xmm5,       xmm4
-        psrldq      xmm5,       3
-        punpcklbw   xmm5,       xmm0              ; mm5 = p1..p4
-        paddusw     xmm3,       xmm5              ; mm3 += mm6
+        movq        QWORD PTR [rdi+rdx-16], mm0  ; store previous 8 bytes
+        movq        QWORD PTR [rdi+rdx-8], mm1   ; store previous 8 bytes
+        movdq2q     mm0,        xmm0
+        psrldq      xmm0,       8
+        movdq2q     mm1,        xmm0
 
-        ; thresholding
-        movdqa      xmm7,       xmm1              ; mm7 = p0..p3
-        psubusw     xmm7,       xmm5              ; mm7 = p0..p3 - p1..p4
-        psubusw     xmm5,       xmm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     xmm7,       xmm5              ; mm7 = abs(p0..p3 - p1..p4)
-        pcmpgtw     xmm7,       xmm2
+        add         rdx,        16
+        cmp         edx,        dword arg(4)     ;cols
+        jge         .acrossdone
+        UPDATE_FLIMIT
+        jmp         .acrossnextcol
 
-        movdqa      xmm5,       xmm4
-        psrldq      xmm5,       4
-        punpcklbw   xmm5,       xmm0              ; mm5 = p2..p5
-        paddusw     xmm3,       xmm5              ; mm3 += mm5
-
-        ; thresholding
-        movdqa      xmm6,       xmm1              ; mm6 = p0..p3
-        psubusw     xmm6,       xmm5              ; mm6 = p0..p3 - p1..p4
-        psubusw     xmm5,       xmm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     xmm6,       xmm5              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6              ; accumulate thresholds
-
-
-        movdqa      xmm5,       xmm4              ; mm5 = p-2..p5
-        punpcklbw   xmm5,       xmm0              ; mm5 = p-2..p1
-        paddusw     xmm3,       xmm5              ; mm3 += mm5
-
-        ; thresholding
-        movdqa      xmm6,       xmm1              ; mm6 = p0..p3
-        psubusw     xmm6,       xmm5              ; mm6 = p0..p3 - p1..p4
-        psubusw     xmm5,       xmm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     xmm6,       xmm5              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6              ; accumulate thresholds
-
-        psrldq      xmm4,       1                   ; mm4 = p-1..p5
-        punpcklbw   xmm4,       xmm0              ; mm4 = p-1..p2
-        paddusw     xmm3,       xmm4              ; mm3 += mm5
-
-        ; thresholding
-        movdqa      xmm6,       xmm1              ; mm6 = p0..p3
-        psubusw     xmm6,       xmm4              ; mm6 = p0..p3 - p1..p4
-        psubusw     xmm4,       xmm1              ; mm5 = p1..p4 - p0..p3
-        paddusw     xmm6,       xmm4              ; mm6 = abs(p0..p3 - p1..p4)
-        pcmpgtw     xmm6,       xmm2
-        por         xmm7,       xmm6              ; accumulate thresholds
-
-        paddusw     xmm3,       RD42              ; mm3 += round value
-        psraw       xmm3,       3                 ; mm3 /= 8
-
-        pand        xmm1,       xmm7              ; mm1 select vals > thresh from source
-        pandn       xmm7,       xmm3              ; mm7 select vals < thresh from blurred result
-        paddusw     xmm1,       xmm7              ; combination
-
-        packuswb    xmm1,       xmm0              ; pack to bytes
-        movq        QWORD PTR [rdi+rdx-8],  mm0   ; store previous four bytes
-        movdq2q     mm0,        xmm1
-
-        add         rdx,        8
-        cmp         edx,        dword arg(5) ;cols
-        jl          .acrossnextcol;
-
-        ; last 8 pixels
-        movq        QWORD PTR [rdi+rdx-8],  mm0
+.acrossdone
+        ; last 16 pixels
+        movq        QWORD PTR [rdi+rdx-16], mm0
 
+        cmp         edx,        dword arg(4)
+        jne         .throw_last_8
+        movq        QWORD PTR [rdi+rdx-8], mm1
+.throw_last_8:
         ; done with this rwo
-        add         rsi,rax               ; next line
-        mov         eax, dword arg(3) ;dst_pixels_per_line ; destination pitch?
-        add         rdi,rax               ; next destination
-        mov         eax, dword arg(2) ;src_pixels_per_line ; destination pitch?
+        add         rsi,rax                      ;next src line
+        mov         eax, dword arg(3)            ;dst_pixels_per_line
+        add         rdi,rax                      ;next destination
+        mov         eax, dword arg(2)            ;src_pixels_per_line
 
-        dec         rcx                   ; decrement count
-        jnz         .nextrow              ; next row
+        mov         rbx,        arg(5)           ;flimits
+        UPDATE_FLIMIT
 
-%if ABI_IS_32BIT=1 && CONFIG_PIC=1
-    add rsp,16
+        dec         rcx                          ;decrement count
+        jnz         .nextrow                     ;next row
+
+    add rsp, 16
     pop rsp
-%endif
     ; begin epilog
     pop rdi
     pop rsi
-    RESTORE_GOT
+    pop rbx
     RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
-%undef RD42
-
+%undef flimit
 
 ;void vp8_mbpost_proc_down_xmm(unsigned char *dst,
 ;                            int pitch, int rows, int cols,int flimit)
@@ -753,7 +717,5 @@ sym(vp8_plane_add_noise_wmt):
 
 SECTION_RODATA
 align 16
-rd42:
-    times 8 dw 0x04
 four8s:
     times 4 dd 8

vp8/common/x86/postproc_x86.c

@@ -18,4 +18,7 @@ extern int rand(void)
 {
   return __rand();
 }
+#else
+/* ISO C forbids an empty translation unit. */
+int vp8_unused;
 #endif

vp8/common/x86/recon_sse2.asm

@@ -890,6 +890,7 @@ sym(vp8_intra_pred_y_tm_%1):
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
+    SAVE_XMM 7
     push        rsi
     push        rdi
     GET_GOT     rbx
@@ -957,6 +958,7 @@ vp8_intra_pred_y_tm_%1_loop:
     RESTORE_GOT
     pop         rdi
     pop         rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret

vp8/common/x86/sad_sse2.asm

@@ -115,7 +115,7 @@ sym(vp8_sad8x16_wmt):
 
         movq            rax,        mm7
         cmp             eax,        arg(4)
-        jg              .x8x16sad_wmt_early_exit
+        ja              .x8x16sad_wmt_early_exit
 
         movq            mm0,        QWORD PTR [rsi]
         movq            mm1,        QWORD PTR [rdi]
@@ -176,7 +176,7 @@ sym(vp8_sad8x8_wmt):
 
         movq            rax,        mm7
         cmp             eax,        arg(4)
-        jg              .x8x8sad_wmt_early_exit
+        ja              .x8x8sad_wmt_early_exit
 
         movq            mm0,        QWORD PTR [rsi]
         movq            mm1,        QWORD PTR [rdi]
@@ -285,7 +285,7 @@ sym(vp8_sad16x8_wmt):
 
         movq            rax,        mm7
         cmp             eax,        arg(4)
-        jg              .x16x8sad_wmt_early_exit
+        ja              .x16x8sad_wmt_early_exit
 
         movq            mm0,        QWORD PTR [rsi]
         movq            mm2,        QWORD PTR [rsi+8]

vp8/common/x86/subpixel_ssse3.asm

@@ -352,6 +352,7 @@ sym(vp8_filter_block1d4_h6_ssse3):
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret

vp8/common/x86/variance_sse2.c

@@ -332,8 +332,9 @@ unsigned int vp8_sub_pixel_variance16x16_wmt
     unsigned int xxsum0, xxsum1;
 
 
-    // note we could avoid these if statements if the calling function
-    // just called the appropriate functions inside.
+    /* note we could avoid these if statements if the calling function
+     * just called the appropriate functions inside.
+     */
     if (xoffset == 4 && yoffset == 0)
     {
         vp8_half_horiz_variance16x_h_sse2(

vp8/common/x86/variance_ssse3.c

@@ -79,8 +79,9 @@ unsigned int vp8_sub_pixel_variance16x16_ssse3
     int xsum0;
     unsigned int xxsum0;
 
-    // note we could avoid these if statements if the calling function
-    // just called the appropriate functions inside.
+    /* note we could avoid these if statements if the calling function
+     * just called the appropriate functions inside.
+     */
     if (xoffset == 4 && yoffset == 0)
     {
         vp8_half_horiz_variance16x_h_sse2(

vp8/common/x86/vp8_asm_stubs.c

@@ -438,19 +438,35 @@ void vp8_sixtap_predict16x16_ssse3
     {
         if (yoffset)
         {
-            vp8_filter_block1d16_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 16, 21, xoffset);
-            vp8_filter_block1d16_v6_ssse3(FData2 , 16, dst_ptr, dst_pitch, 16, yoffset);
+            vp8_filter_block1d16_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                          src_pixels_per_line, FData2,
+                                          16, 21, xoffset);
+            vp8_filter_block1d16_v6_ssse3(FData2 , 16, dst_ptr, dst_pitch,
+                                          16, yoffset);
         }
         else
         {
             /* First-pass only */
-            vp8_filter_block1d16_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 16, xoffset);
+            vp8_filter_block1d16_h6_ssse3(src_ptr, src_pixels_per_line,
+                                          dst_ptr, dst_pitch, 16, xoffset);
         }
     }
     else
     {
-        /* Second-pass only */
-        vp8_filter_block1d16_v6_ssse3(src_ptr - (2 * src_pixels_per_line) , src_pixels_per_line, dst_ptr, dst_pitch, 16, yoffset);
+        if (yoffset)
+        {
+            /* Second-pass only */
+            vp8_filter_block1d16_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                          src_pixels_per_line,
+                                          dst_ptr, dst_pitch, 16, yoffset);
+        }
+        else
+        {
+            /* ssse3 second-pass only function couldn't handle (xoffset==0 &&
+             * yoffset==0) case correctly. Add copy function here to guarantee
+             * six-tap function handles all possible offsets. */
+            vp8_copy_mem16x16(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch);
+        }
     }
 }
 
@@ -470,18 +486,34 @@ void vp8_sixtap_predict8x8_ssse3
     {
         if (yoffset)
         {
-            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 8, 13, xoffset);
-            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch, 8, yoffset);
+            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                         src_pixels_per_line, FData2,
+                                         8, 13, xoffset);
+            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch,
+                                         8, yoffset);
         }
         else
         {
-            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 8, xoffset);
+            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
+                                         dst_ptr, dst_pitch, 8, xoffset);
         }
     }
     else
     {
-        /* Second-pass only */
-        vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 8, yoffset);
+        if (yoffset)
+        {
+            /* Second-pass only */
+            vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                         src_pixels_per_line,
+                                         dst_ptr, dst_pitch, 8, yoffset);
+        }
+        else
+        {
+            /* ssse3 second-pass only function couldn't handle (xoffset==0 &&
+             * yoffset==0) case correctly. Add copy function here to guarantee
+             * six-tap function handles all possible offsets. */
+            vp8_copy_mem8x8(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch);
+        }
     }
 }
 
@@ -502,19 +534,35 @@ void vp8_sixtap_predict8x4_ssse3
     {
         if (yoffset)
         {
-            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 8, 9, xoffset);
-            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch, 4, yoffset);
+            vp8_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                         src_pixels_per_line, FData2,
+                                         8, 9, xoffset);
+            vp8_filter_block1d8_v6_ssse3(FData2, 8, dst_ptr, dst_pitch,
+                                         4, yoffset);
         }
         else
         {
             /* First-pass only */
-            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 4, xoffset);
+            vp8_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
+                                         dst_ptr, dst_pitch, 4, xoffset);
         }
     }
     else
     {
-        /* Second-pass only */
-        vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 4, yoffset);
+        if (yoffset)
+        {
+            /* Second-pass only */
+            vp8_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                         src_pixels_per_line,
+                                         dst_ptr, dst_pitch, 4, yoffset);
+        }
+        else
+        {
+            /* ssse3 second-pass only function couldn't handle (xoffset==0 &&
+             * yoffset==0) case correctly. Add copy function here to guarantee
+             * six-tap function handles all possible offsets. */
+            vp8_copy_mem8x4(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch);
+        }
     }
 }
 
@@ -534,19 +582,48 @@ void vp8_sixtap_predict4x4_ssse3
   {
       if (yoffset)
       {
-          vp8_filter_block1d4_h6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, FData2, 4, 9, xoffset);
-          vp8_filter_block1d4_v6_ssse3(FData2, 4, dst_ptr, dst_pitch, 4, yoffset);
+          vp8_filter_block1d4_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                       src_pixels_per_line,
+                                       FData2, 4, 9, xoffset);
+          vp8_filter_block1d4_v6_ssse3(FData2, 4, dst_ptr, dst_pitch,
+                                       4, yoffset);
       }
       else
       {
-          vp8_filter_block1d4_h6_ssse3(src_ptr, src_pixels_per_line, dst_ptr, dst_pitch, 4, xoffset);
+          vp8_filter_block1d4_h6_ssse3(src_ptr, src_pixels_per_line,
+                                       dst_ptr, dst_pitch, 4, xoffset);
       }
   }
   else
   {
-      vp8_filter_block1d4_v6_ssse3(src_ptr - (2 * src_pixels_per_line), src_pixels_per_line, dst_ptr, dst_pitch, 4, yoffset);
-  }
+      if (yoffset)
+      {
+          vp8_filter_block1d4_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
+                                       src_pixels_per_line,
+                                       dst_ptr, dst_pitch, 4, yoffset);
+      }
+      else
+      {
+        /* ssse3 second-pass only function couldn't handle (xoffset==0 &&
+          * yoffset==0) case correctly. Add copy function here to guarantee
+          * six-tap function handles all possible offsets. */
+          int r;
 
+          for (r = 0; r < 4; r++)
+          {
+  #if !(CONFIG_FAST_UNALIGNED)
+            dst_ptr[0]  = src_ptr[0];
+            dst_ptr[1]  = src_ptr[1];
+            dst_ptr[2]  = src_ptr[2];
+            dst_ptr[3]  = src_ptr[3];
+  #else
+              *(uint32_t *)dst_ptr = *(uint32_t *)src_ptr ;
+  #endif
+              dst_ptr     += dst_pitch;
+              src_ptr     += src_pixels_per_line;
+          }
+      }
+  }
 }
 
 #endif

vp8/decoder/dboolhuff.h

@@ -55,7 +55,7 @@ void vp8dx_bool_decoder_fill(BOOL_DECODER *br);
         int loop_end, x; \
         size_t bits_left = ((_bufend)-(_bufptr))*CHAR_BIT; \
         \
-        x = shift + CHAR_BIT - bits_left; \
+        x = (int)(shift + CHAR_BIT - bits_left); \
         loop_end = 0; \
         if(x >= 0) \
         { \

vp8/decoder/decodemv.c

@@ -48,11 +48,11 @@ static MB_PREDICTION_MODE read_uv_mode(vp8_reader *bc, const vp8_prob *p)
 
 static void read_kf_modes(VP8D_COMP *pbi, MODE_INFO *mi)
 {
-    vp8_reader *const bc = & pbi->bc;
+    vp8_reader *const bc = & pbi->mbc[8];
     const int mis = pbi->common.mode_info_stride;
 
     mi->mbmi.ref_frame = INTRA_FRAME;
-    mi->mbmi.mode = read_kf_ymode(bc, pbi->common.kf_ymode_prob);
+    mi->mbmi.mode = read_kf_ymode(bc, vp8_kf_ymode_prob);
 
     if (mi->mbmi.mode == B_PRED)
     {
@@ -65,12 +65,12 @@ static void read_kf_modes(VP8D_COMP *pbi, MODE_INFO *mi)
             const B_PREDICTION_MODE L = left_block_mode(mi, i);
 
             mi->bmi[i].as_mode =
-                read_bmode(bc, pbi->common.kf_bmode_prob [A] [L]);
+                read_bmode(bc, vp8_kf_bmode_prob [A] [L]);
         }
         while (++i < 16);
     }
 
-    mi->mbmi.uv_mode = read_uv_mode(bc, pbi->common.kf_uv_mode_prob);
+    mi->mbmi.uv_mode = read_uv_mode(bc, vp8_kf_uv_mode_prob);
 }
 
 static int read_mvcomponent(vp8_reader *r, const MV_CONTEXT *mvc)
@@ -150,7 +150,7 @@ static const unsigned char mbsplit_fill_offset[4][16] = {
 
 static void mb_mode_mv_init(VP8D_COMP *pbi)
 {
-    vp8_reader *const bc = & pbi->bc;
+    vp8_reader *const bc = & pbi->mbc[8];
     MV_CONTEXT *const mvc = pbi->common.fc.mvc;
 
 #if CONFIG_ERROR_CONCEALMENT
@@ -159,6 +159,9 @@ static void mb_mode_mv_init(VP8D_COMP *pbi)
      * outside the frame. */
     pbi->mvs_corrupt_from_mb = UINT_MAX;
 #endif
+    /* Read the mb_no_coeff_skip flag */
+    pbi->common.mb_no_coeff_skip = (int)vp8_read_bit(bc);
+
     pbi->prob_skip_false = 0;
     if (pbi->common.mb_no_coeff_skip)
         pbi->prob_skip_false = (vp8_prob)vp8_read_literal(bc, 8);
@@ -293,26 +296,24 @@ static void decode_split_mv(vp8_reader *const bc, MODE_INFO *mi,
                     blockmv.as_mv.row += best_mv.as_mv.row;
                     blockmv.as_mv.col = read_mvcomponent(bc, &mvc[1]) << 1;
                     blockmv.as_mv.col += best_mv.as_mv.col;
-
-                    mbmi->need_to_clamp_mvs |= vp8_check_mv_bounds(&blockmv,
-                                                              mb_to_left_edge,
-                                                              mb_to_right_edge,
-                                                              mb_to_top_edge,
-                                                              mb_to_bottom_edge);
                 }
             }
             else
             {
                 blockmv.as_int = abovemv.as_int;
-                mbmi->need_to_clamp_mvs |= above_mb->mbmi.need_to_clamp_mvs;
             }
         }
         else
         {
             blockmv.as_int = leftmv.as_int;
-            mbmi->need_to_clamp_mvs |= left_mb->mbmi.need_to_clamp_mvs;
         }
 
+        mbmi->need_to_clamp_mvs |= vp8_check_mv_bounds(&blockmv,
+                                                  mb_to_left_edge,
+                                                  mb_to_right_edge,
+                                                  mb_to_top_edge,
+                                                  mb_to_bottom_edge);
+
         {
             /* Fill (uniform) modes, mvs of jth subset.
              Must do it here because ensuing subsets can
@@ -337,7 +338,7 @@ static void decode_split_mv(vp8_reader *const bc, MODE_INFO *mi,
 
 static void read_mb_modes_mv(VP8D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi)
 {
-    vp8_reader *const bc = & pbi->bc;
+    vp8_reader *const bc = & pbi->mbc[8];
     mbmi->ref_frame = (MV_REFERENCE_FRAME) vp8_read(bc, pbi->prob_intra);
     if (mbmi->ref_frame)    /* inter MB */
     {
@@ -595,14 +596,14 @@ static void decode_mb_mode_mvs(VP8D_COMP *pbi, MODE_INFO *mi,
      * By default on a key frame reset all MBs to segment 0
      */
     if (pbi->mb.update_mb_segmentation_map)
-        read_mb_features(&pbi->bc, &mi->mbmi, &pbi->mb);
+        read_mb_features(&pbi->mbc[8], &mi->mbmi, &pbi->mb);
     else if(pbi->common.frame_type == KEY_FRAME)
         mi->mbmi.segment_id = 0;
 
     /* Read the macroblock coeff skip flag if this feature is in use,
      * else default to 0 */
     if (pbi->common.mb_no_coeff_skip)
-        mi->mbmi.mb_skip_coeff = vp8_read(&pbi->bc, pbi->prob_skip_false);
+        mi->mbmi.mb_skip_coeff = vp8_read(&pbi->mbc[8], pbi->prob_skip_false);
     else
         mi->mbmi.mb_skip_coeff = 0;
 
@@ -644,7 +645,8 @@ void vp8_decode_mode_mvs(VP8D_COMP *pbi)
 #if CONFIG_ERROR_CONCEALMENT
             /* look for corruption. set mvs_corrupt_from_mb to the current
              * mb_num if the frame is corrupt from this macroblock. */
-            if (vp8dx_bool_error(&pbi->bc) && mb_num < pbi->mvs_corrupt_from_mb)
+            if (vp8dx_bool_error(&pbi->mbc[8]) && mb_num <
+                (int)pbi->mvs_corrupt_from_mb)
             {
                 pbi->mvs_corrupt_from_mb = mb_num;
                 /* no need to continue since the partition is corrupt from