diff --git a/build/make/rtcd.pl b/build/make/rtcd.pl
new file mode 100755
index 0000000..18ee80d
--- /dev/null
+++ b/build/make/rtcd.pl
@@ -0,0 +1,414 @@
+#!/usr/bin/env perl
+
+no strict 'refs';
+use warnings;
+use Getopt::Long;
+Getopt::Long::Configure("auto_help");
+
+my %ALL_FUNCS = ();
+my @ALL_ARCHS;
+my @ALL_FORWARD_DECLS;
+my @REQUIRES;
+
+my %opts = ();
+my %disabled = ();
+my %required = ();
+
+my @argv;
+foreach (@ARGV) {
+  $disabled{$1} = 1, next if /--disable-(.*)/;
+  $required{$1} = 1, next if /--require-(.*)/;
+  push @argv, $_;
+}
+
+# NB: use GetOptions() instead of GetOptionsFromArray() for compatibility.
+@ARGV = @argv;
+GetOptions(
+  \%opts,
+  'arch=s',
+  'sym=s',
+  'config=s',
+);
+
+foreach my $opt (qw/arch config/) {
+  if (!defined($opts{$opt})) {
+    warn "--$opt is required!\n";
+    Getopt::Long::HelpMessage('-exit' => 1);
+  }
+}
+
+foreach my $defs_file (@ARGV) {
+  if (!-f $defs_file) {
+    warn "$defs_file: $!\n";
+    Getopt::Long::HelpMessage('-exit' => 1);
+  }
+}
+
+open CONFIG_FILE, $opts{config} or
+  die "Error opening config file '$opts{config}': $!\n";
+
+my %config = ();
+while (<CONFIG_FILE>) {
+  next if !/^CONFIG_/;
+  chomp;
+  my @pair = split /=/;
+  $config{$pair[0]} = $pair[1];
+}
+close CONFIG_FILE;
+
+#
+# Routines for the RTCD DSL to call
+#
+sub vpx_config($) {
+  return (defined $config{$_[0]}) ? $config{$_[0]} : "";
+}
+
+sub specialize {
+  my $fn=$_[0];
+  shift;
+  foreach my $opt (@_) {
+    eval "\$${fn}_${opt}=${fn}_${opt}";
+  }
+}
+
+sub add_proto {
+  my $fn = splice(@_, -2, 1);
+  $ALL_FUNCS{$fn} = \@_;
+  specialize $fn, "c";
+}
+
+sub require {
+  foreach my $fn (keys %ALL_FUNCS) {
+    foreach my $opt (@_) {
+      my $ofn = eval "\$${fn}_${opt}";
+      next if !$ofn;
+
+      # if we already have a default, then we can disable it, as we know
+      # we can do better.
+      my $best = eval "\$${fn}_default";
+      if ($best) {
+        my $best_ofn = eval "\$${best}";
+        if ($best_ofn && "$best_ofn" ne "$ofn") {
+          eval "\$${best}_link = 'false'";
+        }
+      }
+      eval "\$${fn}_default=${fn}_${opt}";
+      eval "\$${fn}_${opt}_link='true'";
+    }
+  }
+}
+
+sub forward_decls {
+  push @ALL_FORWARD_DECLS, @_;
+}
+
+#
+# Include the user's directives
+#
+foreach my $f (@ARGV) {
+  open FILE, "<", $f or die "cannot open $f: $!\n";
+  my $contents = join('', <FILE>);
+  close FILE;
+  eval $contents or warn "eval failed: $@\n";
+}
+
+#
+# Process the directives according to the command line
+#
+sub process_forward_decls() {
+  foreach (@ALL_FORWARD_DECLS) {
+    $_->();
+  }
+}
+
+sub determine_indirection {
+  vpx_config("CONFIG_RUNTIME_CPU_DETECT") eq "yes" or &require(@ALL_ARCHS);
+  foreach my $fn (keys %ALL_FUNCS) {
+    my $n = "";
+    my @val = @{$ALL_FUNCS{$fn}};
+    my $args = pop @val;
+    my $rtyp = "@val";
+    my $dfn = eval "\$${fn}_default";
+    $dfn = eval "\$${dfn}";
+    foreach my $opt (@_) {
+      my $ofn = eval "\$${fn}_${opt}";
+      next if !$ofn;
+      my $link = eval "\$${fn}_${opt}_link";
+      next if $link && $link eq "false";
+      $n .= "x";
+    }
+    if ($n eq "x") {
+      eval "\$${fn}_indirect = 'false'";
+    } else {
+      eval "\$${fn}_indirect = 'true'";
+    }
+  }
+}
+
+sub declare_function_pointers {
+  foreach my $fn (sort keys %ALL_FUNCS) {
+    my @val = @{$ALL_FUNCS{$fn}};
+    my $args = pop @val;
+    my $rtyp = "@val";
+    my $dfn = eval "\$${fn}_default";
+    $dfn = eval "\$${dfn}";
+    foreach my $opt (@_) {
+      my $ofn = eval "\$${fn}_${opt}";
+      next if !$ofn;
+      print "$rtyp ${ofn}($args);\n";
+    }
+    if (eval "\$${fn}_indirect" eq "false") {
+      print "#define ${fn} ${dfn}\n";
+    } else {
+      print "RTCD_EXTERN $rtyp (*${fn})($args);\n";
+    }
+    print "\n";
+  }
+}
+
+sub set_function_pointers {
+  foreach my $fn (sort keys %ALL_FUNCS) {
+    my @val = @{$ALL_FUNCS{$fn}};
+    my $args = pop @val;
+    my $rtyp = "@val";
+    my $dfn = eval "\$${fn}_default";
+    $dfn = eval "\$${dfn}";
+    if (eval "\$${fn}_indirect" eq "true") {
+      print "    $fn = $dfn;\n";
+      foreach my $opt (@_) {
+        my $ofn = eval "\$${fn}_${opt}";
+        next if !$ofn;
+        next if "$ofn" eq "$dfn";
+        my $link = eval "\$${fn}_${opt}_link";
+        next if $link && $link eq "false";
+        my $cond = eval "\$have_${opt}";
+        print "    if (${cond}) $fn = $ofn;\n"
+      }
+    }
+  }
+}
+
+sub filter {
+  my @filtered;
+  foreach (@_) { push @filtered, $_ unless $disabled{$_}; }
+  return @filtered;
+}
+
+#
+# Helper functions for generating the arch specific RTCD files
+#
+sub common_top() {
+  my $include_guard = uc($opts{sym})."_H_";
+  print <<EOF;
+#ifndef ${include_guard}
+#define ${include_guard}
+
+#ifdef RTCD_C
+#define RTCD_EXTERN
+#else
+#define RTCD_EXTERN extern
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+EOF
+
+process_forward_decls();
+print "\n";
+declare_function_pointers("c", @ALL_ARCHS);
+
+print <<EOF;
+void $opts{sym}(void);
+
+EOF
+}
+
+sub common_bottom() {
+  print <<EOF;
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif
+EOF
+}
+
+sub x86() {
+  determine_indirection("c", @ALL_ARCHS);
+
+  # Assign the helper variable for each enabled extension
+  foreach my $opt (@ALL_ARCHS) {
+    my $opt_uc = uc $opt;
+    eval "\$have_${opt}=\"flags & HAS_${opt_uc}\"";
+  }
+
+  common_top;
+  print <<EOF;
+#ifdef RTCD_C
+#include "vpx_ports/x86.h"
+static void setup_rtcd_internal(void)
+{
+    int flags = x86_simd_caps();
+
+    (void)flags;
+
+EOF
+
+  set_function_pointers("c", @ALL_ARCHS);
+
+  print <<EOF;
+}
+#endif
+EOF
+  common_bottom;
+}
+
+sub arm() {
+  determine_indirection("c", @ALL_ARCHS);
+
+  # Assign the helper variable for each enabled extension
+  foreach my $opt (@ALL_ARCHS) {
+    my $opt_uc = uc $opt;
+    eval "\$have_${opt}=\"flags & HAS_${opt_uc}\"";
+  }
+
+  common_top;
+  print <<EOF;
+#include "vpx_config.h"
+
+#ifdef RTCD_C
+#include "vpx_ports/arm.h"
+static void setup_rtcd_internal(void)
+{
+    int flags = arm_cpu_caps();
+
+    (void)flags;
+
+EOF
+
+  set_function_pointers("c", @ALL_ARCHS);
+
+  print <<EOF;
+}
+#endif
+EOF
+  common_bottom;
+}
+
+sub mips() {
+  determine_indirection("c", @ALL_ARCHS);
+  common_top;
+
+  print <<EOF;
+#include "vpx_config.h"
+
+#ifdef RTCD_C
+static void setup_rtcd_internal(void)
+{
+EOF
+
+  set_function_pointers("c", @ALL_ARCHS);
+
+  print <<EOF;
+#if HAVE_DSPR2
+#if CONFIG_VP8
+void dsputil_static_init();
+dsputil_static_init();
+#endif
+#if CONFIG_VP9
+void vp9_dsputil_static_init();
+vp9_dsputil_static_init();
+#endif
+#endif
+}
+#endif
+EOF
+  common_bottom;
+}
+
+sub unoptimized() {
+  determine_indirection "c";
+  common_top;
+  print <<EOF;
+#include "vpx_config.h"
+
+#ifdef RTCD_C
+static void setup_rtcd_internal(void)
+{
+EOF
+
+  set_function_pointers "c";
+
+  print <<EOF;
+}
+#endif
+EOF
+  common_bottom;
+}
+
+#
+# Main Driver
+#
+
+&require("c");
+if ($opts{arch} eq 'x86') {
+  @ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2/);
+  x86;
+} elsif ($opts{arch} eq 'x86_64') {
+  @ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2/);
+  @REQUIRES = filter(keys %required ? keys %required : qw/mmx sse sse2/);
+  &require(@REQUIRES);
+  x86;
+} elsif ($opts{arch} eq 'mips32') {
+  @ALL_ARCHS = filter(qw/mips32/);
+  open CONFIG_FILE, $opts{config} or
+    die "Error opening config file '$opts{config}': $!\n";
+  while (<CONFIG_FILE>) {
+    if (/HAVE_DSPR2=yes/) {
+      @ALL_ARCHS = filter(qw/mips32 dspr2/);
+      last;
+    }
+  }
+  close CONFIG_FILE;
+  mips;
+} elsif ($opts{arch} eq 'armv5te') {
+  @ALL_ARCHS = filter(qw/edsp/);
+  arm;
+} elsif ($opts{arch} eq 'armv6') {
+  @ALL_ARCHS = filter(qw/edsp media/);
+  arm;
+} elsif ($opts{arch} eq 'armv7') {
+  @ALL_ARCHS = filter(qw/edsp media neon/);
+  arm;
+} else {
+  unoptimized;
+}
+
+__END__
+
+=head1 NAME
+
+rtcd -
+
+=head1 SYNOPSIS
+
+Usage: rtcd.pl [options] FILE
+
+See 'perldoc rtcd.pl' for more details.
+
+=head1 DESCRIPTION
+
+Reads the Run Time CPU Detections definitions from FILE and generates a
+C header file on stdout.
+
+=head1 OPTIONS
+
+Options:
+  --arch=ARCH       Architecture to generate defs for (required)
+  --disable-EXT     Disable support for EXT extensions
+  --require-EXT     Require support for EXT extensions
+  --sym=SYMBOL      Unique symbol to use for RTCD initialization function
+  --config=FILE     File with CONFIG_FOO=yes lines to parse
diff --git a/build/make/rtcd.sh b/build/make/rtcd.sh
deleted file mode 100755
index 93c9adc..0000000
--- a/build/make/rtcd.sh
+++ /dev/null
@@ -1,373 +0,0 @@
-#!/bin/sh
-self=$0
-
-usage() {
-  cat <<EOF >&2
-Usage: $self [options] FILE
-
-Reads the Run Time CPU Detections definitions from FILE and generates a
-C header file on stdout.
-
-Options:
-  --arch=ARCH   Architecture to generate defs for (required)
-  --disable-EXT Disable support for EXT extensions
-  --require-EXT Require support for EXT extensions
-  --sym=SYMBOL  Unique symbol to use for RTCD initialization function
-  --config=FILE File with CONFIG_FOO=yes lines to parse
-EOF
-  exit 1
-}
-
-die() {
-  echo "$@" >&2
-  exit 1
-}
-
-die_argument_required() {
-  die "Option $opt requires argument"
-}
-
-for opt; do
-  optval="${opt#*=}"
-  case "$opt" in
-    --arch) die_argument_required;;
-    --arch=*) arch=${optval};;
-    --disable-*) eval "disable_${opt#--disable-}=true";;
-    --require-*) REQUIRES="${REQUIRES}${opt#--require-} ";;
-    --sym) die_argument_required;;
-    --sym=*) symbol=${optval};;
-    --config=*) config_file=${optval};;
-    -h|--help)
-      usage
-      ;;
-    -*)
-      die "Unrecognized option: ${opt%%=*}"
-      ;;
-    *)
-      defs_file="$defs_file $opt"
-      ;;
-  esac
-  shift
-done
-for f in $defs_file; do [ -f "$f" ] || usage; done
-[ -n "$arch" ] || usage
-
-# Import the configuration
-[ -f "$config_file" ] && eval $(grep CONFIG_ "$config_file")
-
-#
-# Routines for the RTCD DSL to call
-#
-prototype() {
-  rtyp=""
-  case "$1" in
-    unsigned) rtyp="$1 "; shift;;
-  esac
-  rtyp="${rtyp}$1"
-  fn="$2"
-  args="$3"
-
-  eval "${2}_rtyp='$rtyp'"
-  eval "${2}_args='$3'"
-  ALL_FUNCS="$ALL_FUNCS $fn"
-  specialize $fn c
-}
-
-specialize() {
-  fn="$1"
-  shift
-  for opt in "$@"; do
-    eval "${fn}_${opt}=${fn}_${opt}"
-  done
-}
-
-require() {
-  for fn in $ALL_FUNCS; do
-    for opt in "$@"; do
-      ofn=$(eval "echo \$${fn}_${opt}")
-      [ -z "$ofn" ] && continue
-
-      # if we already have a default, then we can disable it, as we know
-      # we can do better.
-      best=$(eval "echo \$${fn}_default")
-      best_ofn=$(eval "echo \$${best}")
-      [ -n "$best" ] && [ "$best_ofn" != "$ofn" ] && eval "${best}_link=false"
-      eval "${fn}_default=${fn}_${opt}"
-      eval "${fn}_${opt}_link=true"
-    done
-  done
-}
-
-forward_decls() {
-  ALL_FORWARD_DECLS="$ALL_FORWARD_DECLS $1"
-}
-
-#
-# Include the user's directives
-#
-for f in $defs_file; do
-  . $f
-done
-
-#
-# Process the directives according to the command line
-#
-process_forward_decls() {
-  for fn in $ALL_FORWARD_DECLS; do
-    eval $fn
-  done
-}
-
-determine_indirection() {
-  [ "$CONFIG_RUNTIME_CPU_DETECT" = "yes" ] || require $ALL_ARCHS
-  for fn in $ALL_FUNCS; do
-    n=""
-    rtyp="$(eval "echo \$${fn}_rtyp")"
-    args="$(eval "echo \"\$${fn}_args\"")"
-    dfn="$(eval "echo \$${fn}_default")"
-    dfn=$(eval "echo \$${dfn}")
-    for opt in "$@"; do
-      ofn=$(eval "echo \$${fn}_${opt}")
-      [ -z "$ofn" ] && continue
-      link=$(eval "echo \$${fn}_${opt}_link")
-      [ "$link" = "false" ] && continue
-      n="${n}x"
-    done
-    if [ "$n" = "x" ]; then
-      eval "${fn}_indirect=false"
-    else
-      eval "${fn}_indirect=true"
-    fi
-  done
-}
-
-declare_function_pointers() {
-  for fn in $ALL_FUNCS; do
-    rtyp="$(eval "echo \$${fn}_rtyp")"
-    args="$(eval "echo \"\$${fn}_args\"")"
-    dfn="$(eval "echo \$${fn}_default")"
-    dfn=$(eval "echo \$${dfn}")
-    for opt in "$@"; do
-      ofn=$(eval "echo \$${fn}_${opt}")
-      [ -z "$ofn" ] && continue
-      echo "$rtyp ${ofn}($args);"
-    done
-    if [ "$(eval "echo \$${fn}_indirect")" = "false" ]; then
-      echo "#define ${fn} ${dfn}"
-    else
-      echo "RTCD_EXTERN $rtyp (*${fn})($args);"
-    fi
-    echo
-  done
-}
-
-set_function_pointers() {
-  for fn in $ALL_FUNCS; do
-    n=""
-    rtyp="$(eval "echo \$${fn}_rtyp")"
-    args="$(eval "echo \"\$${fn}_args\"")"
-    dfn="$(eval "echo \$${fn}_default")"
-    dfn=$(eval "echo \$${dfn}")
-    if $(eval "echo \$${fn}_indirect"); then
-      echo "    $fn = $dfn;"
-      for opt in "$@"; do
-        ofn=$(eval "echo \$${fn}_${opt}")
-        [ -z "$ofn" ] && continue
-        [ "$ofn" = "$dfn" ] && continue;
-        link=$(eval "echo \$${fn}_${opt}_link")
-        [ "$link" = "false" ] && continue
-        cond="$(eval "echo \$have_${opt}")"
-        echo "    if (${cond}) $fn = $ofn;"
-      done
-    fi
-    echo
-  done
-}
-
-filter() {
-  filtered=""
-  for opt in "$@"; do
-    [ -z $(eval "echo \$disable_${opt}") ] && filtered="$filtered $opt"
-  done
-  echo $filtered
-}
-
-#
-# Helper functions for generating the arch specific RTCD files
-#
-common_top() {
-  outfile_basename=$(basename ${symbol:-rtcd})
-  include_guard=$(echo $outfile_basename | tr '[a-z]' '[A-Z]' | \
-    tr -c '[A-Z0-9]' _)H_
-  cat <<EOF
-#ifndef ${include_guard}
-#define ${include_guard}
-
-#ifdef RTCD_C
-#define RTCD_EXTERN
-#else
-#define RTCD_EXTERN extern
-#endif
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-$(process_forward_decls)
-
-$(declare_function_pointers c $ALL_ARCHS)
-
-void ${symbol:-rtcd}(void);
-EOF
-}
-
-common_bottom() {
-  cat <<EOF
-
-#ifdef __cplusplus
-}  // extern "C"
-#endif
-
-#endif
-EOF
-}
-
-x86() {
-  determine_indirection c $ALL_ARCHS
-
-  # Assign the helper variable for each enabled extension
-  for opt in $ALL_ARCHS; do
-    uc=$(echo $opt | tr '[a-z]' '[A-Z]')
-    eval "have_${opt}=\"flags & HAS_${uc}\""
-  done
-
-  cat <<EOF
-$(common_top)
-
-#ifdef RTCD_C
-#include "vpx_ports/x86.h"
-static void setup_rtcd_internal(void)
-{
-    int flags = x86_simd_caps();
-
-    (void)flags;
-
-$(set_function_pointers c $ALL_ARCHS)
-}
-#endif
-$(common_bottom)
-EOF
-}
-
-arm() {
-  determine_indirection c $ALL_ARCHS
-
-  # Assign the helper variable for each enabled extension
-  for opt in $ALL_ARCHS; do
-    uc=$(echo $opt | tr '[a-z]' '[A-Z]')
-    eval "have_${opt}=\"flags & HAS_${uc}\""
-  done
-
-  cat <<EOF
-$(common_top)
-#include "vpx_config.h"
-
-#ifdef RTCD_C
-#include "vpx_ports/arm.h"
-static void setup_rtcd_internal(void)
-{
-    int flags = arm_cpu_caps();
-
-    (void)flags;
-
-$(set_function_pointers c $ALL_ARCHS)
-}
-#endif
-$(common_bottom)
-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 $ALL_ARCHS)
-#if HAVE_DSPR2
-#if CONFIG_VP8
-void dsputil_static_init();
-dsputil_static_init();
-#endif
-#if CONFIG_VP9
-void vp9_dsputil_static_init();
-vp9_dsputil_static_init();
-#endif
-#endif
-}
-#endif
-$(common_bottom)
-EOF
-}
-
-unoptimized() {
-  determine_indirection c
-  cat <<EOF
-$(common_top)
-#include "vpx_config.h"
-
-#ifdef RTCD_C
-static void setup_rtcd_internal(void)
-{
-$(set_function_pointers c)
-}
-#endif
-$(common_bottom)
-EOF
-
-}
-#
-# Main Driver
-#
-ALL_FUNCS=$(export LC_ALL=C; echo $ALL_FUNCS | tr ' ' '\n' | sort |tr '\n' ' ')
-require c
-case $arch in
-  x86)
-    ALL_ARCHS=$(filter mmx sse sse2 sse3 ssse3 sse4_1 avx avx2)
-    x86
-    ;;
-  x86_64)
-    ALL_ARCHS=$(filter mmx sse sse2 sse3 ssse3 sse4_1 avx avx2)
-    REQUIRES=${REQUIRES:-mmx sse sse2}
-    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
-    ;;
-  armv6)
-    ALL_ARCHS=$(filter edsp media)
-    arm
-    ;;
-  armv7)
-    ALL_ARCHS=$(filter edsp media neon)
-    arm
-    ;;
-  *)
-    unoptimized
-    ;;
-esac
diff --git a/configure b/configure
index 9f5a435..f37615f 100755
--- a/configure
+++ b/configure
@@ -160,6 +160,10 @@
     [ -f ${source_path}/${t}.mk ] && enable_feature ${t}
 done
 
+if ! perl --version >/dev/null; then
+    die "Perl is required to build"
+fi
+
 # check installed doxygen version
 doxy_version=$(doxygen --version 2>/dev/null)
 doxy_major=${doxy_version%%.*}
diff --git a/libs.mk b/libs.mk
index 302d2af..a5c4b76 100644
--- a/libs.mk
+++ b/libs.mk
@@ -49,7 +49,7 @@
 define rtcd_h_template
 $$(BUILD_PFX)$(1).h: $$(SRC_PATH_BARE)/$(2)
 	@echo "    [CREATE] $$@"
-	$$(qexec)$$(SRC_PATH_BARE)/build/make/rtcd.sh --arch=$$(TGT_ISA) \
+	$$(qexec)$$(SRC_PATH_BARE)/build/make/rtcd.pl --arch=$$(TGT_ISA) \
           --sym=$(1) \
           --config=$$(CONFIG_DIR)$$(target)$$(if $$(FAT_ARCHS),,-$$(TOOLCHAIN)).mk \
           $$(RTCD_OPTIONS) $$^ > $$@
@@ -162,7 +162,7 @@
 endif
 
 CODEC_SRCS-$(BUILD_LIBVPX) += build/make/version.sh
-CODEC_SRCS-$(BUILD_LIBVPX) += build/make/rtcd.sh
+CODEC_SRCS-$(BUILD_LIBVPX) += build/make/rtcd.pl
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/emmintrin_compat.h
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/mem_ops.h
 CODEC_SRCS-$(BUILD_LIBVPX) += vpx_ports/mem_ops_aligned.h
@@ -236,6 +236,13 @@
             --out=$@ $^
 CLEAN-OBJS += vpx.def
 
+# Assembly files that are included, but don't define symbols themselves.
+# Filtered out to avoid Visual Studio build warnings.
+ASM_INCLUDES := \
+    third_party/x86inc/x86inc.asm \
+    vpx_config.asm \
+    vpx_ports/x86_abi_support.asm \
+
 vpx.$(VCPROJ_SFX): $(CODEC_SRCS) vpx.def obj_int_extract.$(VCPROJ_SFX)
 	@echo "    [CREATE] $@"
 	$(qexec)$(GEN_VCPROJ) \
@@ -246,7 +253,8 @@
             --proj-guid=DCE19DAF-69AC-46DB-B14A-39F0FAA5DB74 \
             --module-def=vpx.def \
             --ver=$(CONFIG_VS_VERSION) \
-            --out=$@ $(CFLAGS) $^ \
+            --out=$@ $(CFLAGS) \
+            $(filter-out $(addprefix %, $(ASM_INCLUDES)), $^) \
             --src-path-bare="$(SRC_PATH_BARE)" \
 
 PROJECTS-$(BUILD_LIBVPX) += vpx.$(VCPROJ_SFX)
diff --git a/test/vp8_boolcoder_test.cc b/test/vp8_boolcoder_test.cc
index 7c6c601..9cd1987 100644
--- a/test/vp8_boolcoder_test.cc
+++ b/test/vp8_boolcoder_test.cc
@@ -35,14 +35,14 @@
   0x89, 0x9a, 0xab, 0xbc, 0xcd, 0xde, 0xef, 0xf0
 };
 
-void encrypt_buffer(uint8_t *buffer, int size) {
-  for (int i = 0; i < size; ++i) {
+void encrypt_buffer(uint8_t *buffer, size_t size) {
+  for (size_t i = 0; i < size; ++i) {
     buffer[i] ^= secret_key[i & 15];
   }
 }
 
 void test_decrypt_cb(void *decrypt_state, const uint8_t *input,
-                           uint8_t *output, int count) {
+                     uint8_t *output, int count) {
   const size_t offset = input - reinterpret_cast<uint8_t*>(decrypt_state);
   for (int i = 0; i < count; i++) {
     output[i] = input[i] ^ secret_key[(offset + i) & 15];
diff --git a/test/vp8_decrypt_test.cc b/test/vp8_decrypt_test.cc
index b092509..1b5b083 100644
--- a/test/vp8_decrypt_test.cc
+++ b/test/vp8_decrypt_test.cc
@@ -26,9 +26,9 @@
   0x89, 0x9a, 0xab, 0xbc, 0xcd, 0xde, 0xef, 0xf0
 };
 
-void encrypt_buffer(const uint8_t *src, uint8_t *dst,
-                    int size, int offset = 0) {
-  for (int i = 0; i < size; ++i) {
+void encrypt_buffer(const uint8_t *src, uint8_t *dst, size_t size,
+                    ptrdiff_t offset) {
+  for (size_t i = 0; i < size; ++i) {
     dst[i] = src[i] ^ test_key[(offset + i) & 15];
   }
 }
@@ -61,7 +61,7 @@
 
 #if CONFIG_DECRYPT
   std::vector<uint8_t> encrypted(video.frame_size());
-  encrypt_buffer(video.cxdata(), &encrypted[0], video.frame_size());
+  encrypt_buffer(video.cxdata(), &encrypted[0], video.frame_size(), 0);
   vp8_decrypt_init di = { test_decrypt_cb, &encrypted[0] };
   decoder.Control(VP8D_SET_DECRYPTOR, &di);
 #endif  // CONFIG_DECRYPT
diff --git a/third_party/nestegg/README.webm b/third_party/nestegg/README.webm
index 7860a7c..8e3760b 100644
--- a/third_party/nestegg/README.webm
+++ b/third_party/nestegg/README.webm
@@ -18,3 +18,7 @@
 - fix track_number uint64->uint32 warnings
 - fix track_scale double->uint64 warning
 - nestegg_packet_track: fix uint64->uint32 warning
+- ne_read_(string|binary|block): normalize size_t usage
+- ne_parse: normalize size_t usage
+- quiet read related uint64->size_t warnings
+- ne_buffer_read: quiet uint64->size_t warning
diff --git a/third_party/nestegg/src/nestegg.c b/third_party/nestegg/src/nestegg.c
index 35ce9f1..c7e2b02 100644
--- a/third_party/nestegg/src/nestegg.c
+++ b/third_party/nestegg/src/nestegg.c
@@ -694,14 +694,15 @@
 {
   char * str;
   int r;
+  const size_t alloc_size = (size_t)length + 1;
 
   if (length == 0 || length > LIMIT_STRING)
     return -1;
-  str = ne_pool_alloc(length + 1, ctx->alloc_pool);
-  r = ne_io_read(ctx->io, (unsigned char *) str, length);
+  str = ne_pool_alloc(alloc_size, ctx->alloc_pool);
+  r = ne_io_read(ctx->io, (unsigned char *) str, alloc_size - 1);
   if (r != 1)
     return r;
-  str[length] = '\0';
+  str[alloc_size - 1] = '\0';
   *val = str;
   return 1;
 }
@@ -711,9 +712,9 @@
 {
   if (length == 0 || length > LIMIT_BINARY)
     return -1;
-  val->data = ne_pool_alloc(length, ctx->alloc_pool);
-  val->length = length;
-  return ne_io_read(ctx->io, val->data, length);
+  val->length = (size_t)length;
+  val->data = ne_pool_alloc(val->length, ctx->alloc_pool);
+  return ne_io_read(ctx->io, val->data, val->length);
 }
 
 static int
@@ -1043,7 +1044,7 @@
           ne_read_single_master(ctx, element);
         continue;
       } else {
-        r = ne_read_simple(ctx, element, size);
+        r = ne_read_simple(ctx, element, (size_t)size);
         if (r < 0)
           break;
       }
@@ -1062,7 +1063,7 @@
 
       if (id != ID_VOID && id != ID_CRC32)
         ctx->log(ctx, NESTEGG_LOG_DEBUG, "unknown element %llx", id);
-      r = ne_io_read_skip(ctx->io, size);
+      r = ne_io_read_skip(ctx->io, (size_t)size);
       if (r != 1)
         break;
     }
@@ -1151,7 +1152,8 @@
   r = ne_read_vint(io, &lace, &length);
   if (r != 1)
     return r;
-  *read += length;
+  assert(length <= 8);
+  *read += (size_t)length;
 
   sizes[i] = lace;
   sum = sizes[i];
@@ -1163,7 +1165,8 @@
     r = ne_read_svint(io, &slace, &length);
     if (r != 1)
       return r;
-    *read += length;
+    assert(length <= 8);
+    *read += (size_t)length;
     sizes[i] = sizes[i - 1] + slace;
     sum += sizes[i];
     i += 1;
@@ -1263,7 +1266,8 @@
   if (track_number == 0 || (unsigned int)track_number != track_number)
     return -1;
 
-  consumed += length;
+  assert(length <= 8);
+  consumed += (size_t)length;
 
   r = ne_read_int(ctx->io, &timecode, 2);
   if (r != 1)
@@ -1307,7 +1311,7 @@
   case LACING_XIPH:
     if (frames == 1)
       return -1;
-    r = ne_read_xiph_lacing(ctx->io, block_size, &consumed, frames, frame_sizes);
+    r = ne_read_xiph_lacing(ctx->io, (size_t)block_size, &consumed, frames, frame_sizes);
     if (r != 1)
       return r;
     break;
@@ -1320,7 +1324,7 @@
   case LACING_EBML:
     if (frames == 1)
       return -1;
-    r = ne_read_ebml_lacing(ctx->io, block_size, &consumed, frames, frame_sizes);
+    r = ne_read_ebml_lacing(ctx->io, (size_t)block_size, &consumed, frames, frame_sizes);
     if (r != 1)
       return r;
     break;
@@ -1365,9 +1369,9 @@
       return -1;
     }
     f = ne_alloc(sizeof(*f));
-    f->data = ne_alloc(frame_sizes[i]);
-    f->length = frame_sizes[i];
-    r = ne_io_read(ctx->io, f->data, frame_sizes[i]);
+    f->length = (size_t)frame_sizes[i];
+    f->data = ne_alloc(f->length);
+    r = ne_io_read(ctx->io, f->data, f->length);
     if (r != 1) {
       free(f->data);
       free(f);
@@ -1406,7 +1410,8 @@
   if (!element)
     return 1;
 
-  r = ne_read_simple(ctx, element, size);
+  assert((size_t)size == size);
+  r = ne_read_simple(ctx, element, (size_t)size);
   if (r != 1)
     return r;
   storage = (struct ebml_type *) (ctx->ancestor->data + element->offset);
@@ -1600,7 +1605,7 @@
   struct sniff_buffer * sb = user_data;
 
   int rv = 1;
-  size_t available = sb->length - sb->offset;
+  size_t available = sb->length - (size_t)sb->offset;
 
   if (available < length)
     return 0;
@@ -2074,7 +2079,7 @@
         p += sizes[i];
       }
       *data = p;
-      *length = sizes[item];
+      *length = (size_t)sizes[item];
   } else {
     *data = codec_private.data;
     *length = codec_private.length;
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
new file mode 100644
index 0000000..130d965
--- /dev/null
+++ b/vp8/common/rtcd_defs.pl
@@ -0,0 +1,541 @@
+sub vp8_common_forward_decls() {
+print <<EOF
+/*
+ * VP8
+ */
+
+struct blockd;
+struct macroblockd;
+struct loop_filter_info;
+
+/* Encoder forward decls */
+struct block;
+struct macroblock;
+struct variance_vtable;
+union int_mv;
+struct yv12_buffer_config;
+EOF
+}
+forward_decls qw/vp8_common_forward_decls/;
+
+#
+# system state
+#
+add_proto qw/void vp8_clear_system_state/, "";
+specialize qw/vp8_clear_system_state mmx/;
+$vp8_clear_system_state_mmx=vpx_reset_mmx_state;
+
+#
+# Dequant
+#
+add_proto qw/void vp8_dequantize_b/, "struct blockd*, short *dqc";
+specialize qw/vp8_dequantize_b mmx media neon/;
+$vp8_dequantize_b_media=vp8_dequantize_b_v6;
+
+add_proto qw/void vp8_dequant_idct_add/, "short *input, short *dq, unsigned char *output, int stride";
+specialize qw/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;
+
+add_proto qw/void vp8_dequant_idct_add_y_block/, "short *q, short *dq, unsigned char *dst, int stride, char *eobs";
+specialize qw/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;
+
+add_proto qw/void vp8_dequant_idct_add_uv_block/, "short *q, short *dq, unsigned char *dst_u, unsigned char *dst_v, int stride, char *eobs";
+specialize qw/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
+#
+add_proto qw/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 qw/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;
+
+add_proto qw/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 qw/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;
+
+add_proto qw/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 qw/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;
+
+add_proto qw/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 qw/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;
+
+
+add_proto qw/void vp8_loop_filter_simple_mbv/, "unsigned char *y, int ystride, const unsigned char *blimit";
+specialize qw/vp8_loop_filter_simple_mbv mmx sse2 media neon/;
+$vp8_loop_filter_simple_mbv_c=vp8_loop_filter_simple_vertical_edge_c;
+$vp8_loop_filter_simple_mbv_mmx=vp8_loop_filter_simple_vertical_edge_mmx;
+$vp8_loop_filter_simple_mbv_sse2=vp8_loop_filter_simple_vertical_edge_sse2;
+$vp8_loop_filter_simple_mbv_media=vp8_loop_filter_simple_vertical_edge_armv6;
+$vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon;
+
+add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit";
+specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon/;
+$vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c;
+$vp8_loop_filter_simple_mbh_mmx=vp8_loop_filter_simple_horizontal_edge_mmx;
+$vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2;
+$vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6;
+$vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon;
+
+add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit";
+specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon/;
+$vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c;
+$vp8_loop_filter_simple_bv_mmx=vp8_loop_filter_bvs_mmx;
+$vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2;
+$vp8_loop_filter_simple_bv_media=vp8_loop_filter_bvs_armv6;
+$vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon;
+
+add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit";
+specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon/;
+$vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c;
+$vp8_loop_filter_simple_bh_mmx=vp8_loop_filter_bhs_mmx;
+$vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2;
+$vp8_loop_filter_simple_bh_media=vp8_loop_filter_bhs_armv6;
+$vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon;
+
+#
+# IDCT
+#
+#idct16
+add_proto qw/void vp8_short_idct4x4llm/, "short *input, unsigned char *pred, int pitch, unsigned char *dst, int dst_stride";
+specialize qw/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
+add_proto qw/void vp8_short_inv_walsh4x4_1/, "short *input, short *output";
+specialize qw/vp8_short_inv_walsh4x4_1 dspr2/;
+$vp8_short_inv_walsh4x4_1_dspr2=vp8_short_inv_walsh4x4_1_dspr2;
+# no asm yet
+
+#iwalsh16
+add_proto qw/void vp8_short_inv_walsh4x4/, "short *input, short *output";
+specialize qw/vp8_short_inv_walsh4x4 mmx sse2 media neon dspr2/;
+$vp8_short_inv_walsh4x4_media=vp8_short_inv_walsh4x4_v6;
+$vp8_short_inv_walsh4x4_dspr2=vp8_short_inv_walsh4x4_dspr2;
+
+#idct1_scalar_add
+add_proto qw/void vp8_dc_only_idct_add/, "short input, unsigned char *pred, int pred_stride, unsigned char *dst, int dst_stride";
+specialize qw/vp8_dc_only_idct_add	mmx media neon dspr2/;
+$vp8_dc_only_idct_add_media=vp8_dc_only_idct_add_v6;
+$vp8_dc_only_idct_add_dspr2=vp8_dc_only_idct_add_dspr2;
+
+#
+# RECON
+#
+add_proto qw/void vp8_copy_mem16x16/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_copy_mem16x16 mmx sse2 media neon dspr2/;
+$vp8_copy_mem16x16_media=vp8_copy_mem16x16_v6;
+$vp8_copy_mem16x16_dspr2=vp8_copy_mem16x16_dspr2;
+
+add_proto qw/void vp8_copy_mem8x8/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_copy_mem8x8 mmx media neon dspr2/;
+$vp8_copy_mem8x8_media=vp8_copy_mem8x8_v6;
+$vp8_copy_mem8x8_dspr2=vp8_copy_mem8x8_dspr2;
+
+add_proto qw/void vp8_copy_mem8x4/, "unsigned char *src, int src_pitch, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_copy_mem8x4 mmx media neon dspr2/;
+$vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6;
+$vp8_copy_mem8x4_dspr2=vp8_copy_mem8x4_dspr2;
+
+add_proto qw/void vp8_build_intra_predictors_mby_s/, "struct macroblockd *x, unsigned char * yabove_row, unsigned char * yleft, int left_stride, unsigned char * ypred_ptr, int y_stride";
+specialize qw/vp8_build_intra_predictors_mby_s sse2 ssse3/;
+#TODO: fix assembly for neon
+
+add_proto qw/void vp8_build_intra_predictors_mbuv_s/, "struct macroblockd *x, unsigned char * uabove_row, unsigned char * vabove_row,  unsigned char *uleft, unsigned char *vleft, int left_stride, unsigned char * upred_ptr, unsigned char * vpred_ptr, int pred_stride";
+specialize qw/vp8_build_intra_predictors_mbuv_s sse2 ssse3/;
+
+add_proto qw/void vp8_intra4x4_predict/, "unsigned char *Above, unsigned char *yleft, int left_stride, int b_mode, unsigned char *dst, int dst_stride, unsigned char top_left";
+specialize qw/vp8_intra4x4_predict media/;
+$vp8_intra4x4_predict_media=vp8_intra4x4_predict_armv6;
+
+#
+# Postproc
+#
+if (vpx_config("CONFIG_POSTPROC") eq "yes") {
+    add_proto qw/void vp8_mbpost_proc_down/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
+    specialize qw/vp8_mbpost_proc_down mmx sse2/;
+    $vp8_mbpost_proc_down_sse2=vp8_mbpost_proc_down_xmm;
+
+    add_proto qw/void vp8_mbpost_proc_across_ip/, "unsigned char *dst, int pitch, int rows, int cols,int flimit";
+    specialize qw/vp8_mbpost_proc_across_ip sse2/;
+    $vp8_mbpost_proc_across_ip_sse2=vp8_mbpost_proc_across_ip_xmm;
+
+    add_proto qw/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 qw/vp8_post_proc_down_and_across_mb_row sse2/;
+
+    add_proto qw/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 qw/vp8_plane_add_noise mmx sse2/;
+    $vp8_plane_add_noise_sse2=vp8_plane_add_noise_wmt;
+
+    add_proto qw/void vp8_blend_mb_inner/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
+    # no asm yet
+
+    add_proto qw/void vp8_blend_mb_outer/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
+    # no asm yet
+
+    add_proto qw/void vp8_blend_b/, "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride";
+    # no asm yet
+
+    add_proto qw/void vp8_filter_by_weight16x16/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
+    specialize qw/vp8_filter_by_weight16x16 sse2/;
+
+    add_proto qw/void vp8_filter_by_weight8x8/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
+    specialize qw/vp8_filter_by_weight8x8 sse2/;
+
+    add_proto qw/void vp8_filter_by_weight4x4/, "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight";
+    # no asm yet
+}
+
+#
+# Subpixel
+#
+add_proto qw/void vp8_sixtap_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_sixtap_predict16x16 mmx sse2 ssse3 media neon dspr2/;
+$vp8_sixtap_predict16x16_media=vp8_sixtap_predict16x16_armv6;
+$vp8_sixtap_predict16x16_dspr2=vp8_sixtap_predict16x16_dspr2;
+
+add_proto qw/void vp8_sixtap_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_sixtap_predict8x8 mmx sse2 ssse3 media neon dspr2/;
+$vp8_sixtap_predict8x8_media=vp8_sixtap_predict8x8_armv6;
+$vp8_sixtap_predict8x8_dspr2=vp8_sixtap_predict8x8_dspr2;
+
+add_proto qw/void vp8_sixtap_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_sixtap_predict8x4 mmx sse2 ssse3 media neon dspr2/;
+$vp8_sixtap_predict8x4_media=vp8_sixtap_predict8x4_armv6;
+$vp8_sixtap_predict8x4_dspr2=vp8_sixtap_predict8x4_dspr2;
+
+add_proto qw/void vp8_sixtap_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_sixtap_predict4x4 mmx ssse3 media neon dspr2/;
+$vp8_sixtap_predict4x4_media=vp8_sixtap_predict4x4_armv6;
+$vp8_sixtap_predict4x4_dspr2=vp8_sixtap_predict4x4_dspr2;
+
+add_proto qw/void vp8_bilinear_predict16x16/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_bilinear_predict16x16 mmx sse2 ssse3 media neon/;
+$vp8_bilinear_predict16x16_media=vp8_bilinear_predict16x16_armv6;
+
+add_proto qw/void vp8_bilinear_predict8x8/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_bilinear_predict8x8 mmx sse2 ssse3 media neon/;
+$vp8_bilinear_predict8x8_media=vp8_bilinear_predict8x8_armv6;
+
+add_proto qw/void vp8_bilinear_predict8x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_bilinear_predict8x4 mmx media neon/;
+$vp8_bilinear_predict8x4_media=vp8_bilinear_predict8x4_armv6;
+
+add_proto qw/void vp8_bilinear_predict4x4/, "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch";
+specialize qw/vp8_bilinear_predict4x4 mmx media neon/;
+$vp8_bilinear_predict4x4_media=vp8_bilinear_predict4x4_armv6;
+
+#
+# Whole-pixel Variance
+#
+add_proto qw/unsigned int vp8_variance4x4/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance4x4 mmx sse2/;
+$vp8_variance4x4_sse2=vp8_variance4x4_wmt;
+
+add_proto qw/unsigned int vp8_variance8x8/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance8x8 mmx sse2 media neon/;
+$vp8_variance8x8_sse2=vp8_variance8x8_wmt;
+$vp8_variance8x8_media=vp8_variance8x8_armv6;
+
+add_proto qw/unsigned int vp8_variance8x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance8x16 mmx sse2 neon/;
+$vp8_variance8x16_sse2=vp8_variance8x16_wmt;
+
+add_proto qw/unsigned int vp8_variance16x8/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance16x8 mmx sse2 neon/;
+$vp8_variance16x8_sse2=vp8_variance16x8_wmt;
+
+add_proto qw/unsigned int vp8_variance16x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance16x16 mmx sse2 media neon/;
+$vp8_variance16x16_sse2=vp8_variance16x16_wmt;
+$vp8_variance16x16_media=vp8_variance16x16_armv6;
+
+#
+# Sub-pixel Variance
+#
+add_proto qw/unsigned int vp8_sub_pixel_variance4x4/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_variance4x4 mmx sse2/;
+$vp8_sub_pixel_variance4x4_sse2=vp8_sub_pixel_variance4x4_wmt;
+
+add_proto qw/unsigned int vp8_sub_pixel_variance8x8/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_variance8x8 mmx sse2 media neon/;
+$vp8_sub_pixel_variance8x8_sse2=vp8_sub_pixel_variance8x8_wmt;
+$vp8_sub_pixel_variance8x8_media=vp8_sub_pixel_variance8x8_armv6;
+
+add_proto qw/unsigned int vp8_sub_pixel_variance8x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_variance8x16 mmx sse2/;
+$vp8_sub_pixel_variance8x16_sse2=vp8_sub_pixel_variance8x16_wmt;
+
+add_proto qw/unsigned int vp8_sub_pixel_variance16x8/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_variance16x8 mmx sse2 ssse3/;
+$vp8_sub_pixel_variance16x8_sse2=vp8_sub_pixel_variance16x8_wmt;
+
+add_proto qw/unsigned int vp8_sub_pixel_variance16x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_variance16x16 mmx sse2 ssse3 media neon/;
+$vp8_sub_pixel_variance16x16_sse2=vp8_sub_pixel_variance16x16_wmt;
+$vp8_sub_pixel_variance16x16_media=vp8_sub_pixel_variance16x16_armv6;
+
+add_proto qw/unsigned int vp8_variance_halfpixvar16x16_h/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance_halfpixvar16x16_h mmx sse2 media neon/;
+$vp8_variance_halfpixvar16x16_h_sse2=vp8_variance_halfpixvar16x16_h_wmt;
+$vp8_variance_halfpixvar16x16_h_media=vp8_variance_halfpixvar16x16_h_armv6;
+
+add_proto qw/unsigned int vp8_variance_halfpixvar16x16_v/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance_halfpixvar16x16_v mmx sse2 media neon/;
+$vp8_variance_halfpixvar16x16_v_sse2=vp8_variance_halfpixvar16x16_v_wmt;
+$vp8_variance_halfpixvar16x16_v_media=vp8_variance_halfpixvar16x16_v_armv6;
+
+add_proto qw/unsigned int vp8_variance_halfpixvar16x16_hv/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_variance_halfpixvar16x16_hv mmx sse2 media neon/;
+$vp8_variance_halfpixvar16x16_hv_sse2=vp8_variance_halfpixvar16x16_hv_wmt;
+$vp8_variance_halfpixvar16x16_hv_media=vp8_variance_halfpixvar16x16_hv_armv6;
+
+#
+# Single block SAD
+#
+add_proto qw/unsigned int vp8_sad4x4/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp8_sad4x4 mmx sse2 neon/;
+$vp8_sad4x4_sse2=vp8_sad4x4_wmt;
+
+add_proto qw/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 qw/vp8_sad8x8 mmx sse2 neon/;
+$vp8_sad8x8_sse2=vp8_sad8x8_wmt;
+
+add_proto qw/unsigned int vp8_sad8x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp8_sad8x16 mmx sse2 neon/;
+$vp8_sad8x16_sse2=vp8_sad8x16_wmt;
+
+add_proto qw/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 qw/vp8_sad16x8 mmx sse2 neon/;
+$vp8_sad16x8_sse2=vp8_sad16x8_wmt;
+
+add_proto qw/unsigned int vp8_sad16x16/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp8_sad16x16 mmx sse2 sse3 media neon/;
+$vp8_sad16x16_sse2=vp8_sad16x16_wmt;
+$vp8_sad16x16_media=vp8_sad16x16_armv6;
+
+#
+# Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
+#
+add_proto qw/void vp8_sad4x4x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp8_sad4x4x3 sse3/;
+
+add_proto qw/void vp8_sad8x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp8_sad8x8x3 sse3/;
+
+add_proto qw/void vp8_sad8x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp8_sad8x16x3 sse3/;
+
+add_proto qw/void vp8_sad16x8x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp8_sad16x8x3 sse3 ssse3/;
+
+add_proto qw/void vp8_sad16x16x3/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp8_sad16x16x3 sse3 ssse3/;
+
+# Note the only difference in the following prototypes is that they return into
+# an array of short
+add_proto qw/void vp8_sad4x4x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
+specialize qw/vp8_sad4x4x8 sse4_1/;
+$vp8_sad4x4x8_sse4_1=vp8_sad4x4x8_sse4;
+
+add_proto qw/void vp8_sad8x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
+specialize qw/vp8_sad8x8x8 sse4_1/;
+$vp8_sad8x8x8_sse4_1=vp8_sad8x8x8_sse4;
+
+add_proto qw/void vp8_sad8x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
+specialize qw/vp8_sad8x16x8 sse4_1/;
+$vp8_sad8x16x8_sse4_1=vp8_sad8x16x8_sse4;
+
+add_proto qw/void vp8_sad16x8x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
+specialize qw/vp8_sad16x8x8 sse4_1/;
+$vp8_sad16x8x8_sse4_1=vp8_sad16x8x8_sse4;
+
+add_proto qw/void vp8_sad16x16x8/, "const unsigned char *src_ptr, int src_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned short *sad_array";
+specialize qw/vp8_sad16x16x8 sse4_1/;
+$vp8_sad16x16x8_sse4_1=vp8_sad16x16x8_sse4;
+
+#
+# Multi-block SAD, comparing a reference to N independent blocks
+#
+add_proto qw/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 qw/vp8_sad4x4x4d sse3/;
+
+add_proto qw/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 qw/vp8_sad8x8x4d sse3/;
+
+add_proto qw/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 qw/vp8_sad8x16x4d sse3/;
+
+add_proto qw/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 qw/vp8_sad16x8x4d sse3/;
+
+add_proto qw/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 qw/vp8_sad16x16x4d sse3/;
+
+#
+# Encoder functions below this point.
+#
+if (vpx_config("CONFIG_VP8_ENCODER") eq "yes") {
+
+#
+# Sum of squares (vector)
+#
+add_proto qw/unsigned int vp8_get_mb_ss/, "const short *";
+specialize qw/vp8_get_mb_ss mmx sse2/;
+
+#
+# SSE (Sum Squared Error)
+#
+add_proto qw/unsigned int vp8_sub_pixel_mse16x16/, "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse";
+specialize qw/vp8_sub_pixel_mse16x16 mmx sse2/;
+$vp8_sub_pixel_mse16x16_sse2=vp8_sub_pixel_mse16x16_wmt;
+
+add_proto qw/unsigned int vp8_mse16x16/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse";
+specialize qw/vp8_mse16x16 mmx sse2 media neon/;
+$vp8_mse16x16_sse2=vp8_mse16x16_wmt;
+$vp8_mse16x16_media=vp8_mse16x16_armv6;
+
+add_proto qw/unsigned int vp8_get4x4sse_cs/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride";
+specialize qw/vp8_get4x4sse_cs mmx neon/;
+
+#
+# Block copy
+#
+if ($opts{arch} =~ /x86/) {
+    add_proto qw/void vp8_copy32xn/, "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int n";
+    specialize qw/vp8_copy32xn sse2 sse3/;
+}
+
+#
+# Structured Similarity (SSIM)
+#
+if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") {
+    $opts{arch} eq "x86_64" and $sse2_on_x86_64 = "sse2";
+
+    add_proto qw/void vp8_ssim_parms_8x8/, "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
+    specialize qw/vp8_ssim_parms_8x8/, "$sse2_on_x86_64";
+
+    add_proto qw/void vp8_ssim_parms_16x16/, "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
+    specialize qw/vp8_ssim_parms_16x16/, "$sse2_on_x86_64";
+}
+
+#
+# Forward DCT
+#
+add_proto qw/void vp8_short_fdct4x4/, "short *input, short *output, int pitch";
+specialize qw/vp8_short_fdct4x4 mmx sse2 media neon/;
+$vp8_short_fdct4x4_media=vp8_short_fdct4x4_armv6;
+
+add_proto qw/void vp8_short_fdct8x4/, "short *input, short *output, int pitch";
+specialize qw/vp8_short_fdct8x4 mmx sse2 media neon/;
+$vp8_short_fdct8x4_media=vp8_short_fdct8x4_armv6;
+
+add_proto qw/void vp8_short_walsh4x4/, "short *input, short *output, int pitch";
+specialize qw/vp8_short_walsh4x4 sse2 media neon/;
+$vp8_short_walsh4x4_media=vp8_short_walsh4x4_armv6;
+
+#
+# Quantizer
+#
+add_proto qw/void vp8_regular_quantize_b/, "struct block *, struct blockd *";
+specialize qw/vp8_regular_quantize_b sse2/;
+# TODO(johann) Update sse4 implementation and re-enable
+#$vp8_regular_quantize_b_sse4_1=vp8_regular_quantize_b_sse4;
+
+add_proto qw/void vp8_fast_quantize_b/, "struct block *, struct blockd *";
+specialize qw/vp8_fast_quantize_b sse2 ssse3 media neon/;
+$vp8_fast_quantize_b_media=vp8_fast_quantize_b_armv6;
+
+add_proto qw/void vp8_regular_quantize_b_pair/, "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2";
+# no asm yet
+
+add_proto qw/void vp8_fast_quantize_b_pair/, "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2";
+specialize qw/vp8_fast_quantize_b_pair neon/;
+
+add_proto qw/void vp8_quantize_mb/, "struct macroblock *";
+specialize qw/vp8_quantize_mb neon/;
+
+add_proto qw/void vp8_quantize_mby/, "struct macroblock *";
+specialize qw/vp8_quantize_mby neon/;
+
+add_proto qw/void vp8_quantize_mbuv/, "struct macroblock *";
+specialize qw/vp8_quantize_mbuv neon/;
+
+#
+# Block subtraction
+#
+add_proto qw/int vp8_block_error/, "short *coeff, short *dqcoeff";
+specialize qw/vp8_block_error mmx sse2/;
+$vp8_block_error_sse2=vp8_block_error_xmm;
+
+add_proto qw/int vp8_mbblock_error/, "struct macroblock *mb, int dc";
+specialize qw/vp8_mbblock_error mmx sse2/;
+$vp8_mbblock_error_sse2=vp8_mbblock_error_xmm;
+
+add_proto qw/int vp8_mbuverror/, "struct macroblock *mb";
+specialize qw/vp8_mbuverror mmx sse2/;
+$vp8_mbuverror_sse2=vp8_mbuverror_xmm;
+
+add_proto qw/void vp8_subtract_b/, "struct block *be, struct blockd *bd, int pitch";
+specialize qw/vp8_subtract_b mmx sse2 media neon/;
+$vp8_subtract_b_media=vp8_subtract_b_armv6;
+
+add_proto qw/void vp8_subtract_mby/, "short *diff, unsigned char *src, int src_stride, unsigned char *pred, int pred_stride";
+specialize qw/vp8_subtract_mby mmx sse2 media neon/;
+$vp8_subtract_mby_media=vp8_subtract_mby_armv6;
+
+add_proto qw/void vp8_subtract_mbuv/, "short *diff, unsigned char *usrc, unsigned char *vsrc, int src_stride, unsigned char *upred, unsigned char *vpred, int pred_stride";
+specialize qw/vp8_subtract_mbuv mmx sse2 media neon/;
+$vp8_subtract_mbuv_media=vp8_subtract_mbuv_armv6;
+
+#
+# Motion search
+#
+add_proto qw/int vp8_full_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
+specialize qw/vp8_full_search_sad sse3 sse4_1/;
+$vp8_full_search_sad_sse3=vp8_full_search_sadx3;
+$vp8_full_search_sad_sse4_1=vp8_full_search_sadx8;
+
+add_proto qw/int vp8_refining_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
+specialize qw/vp8_refining_search_sad sse3/;
+$vp8_refining_search_sad_sse3=vp8_refining_search_sadx4;
+
+add_proto qw/int vp8_diamond_search_sad/, "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, union int_mv *best_mv, int search_param, int sad_per_bit, int *num00, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv";
+$vp8_diamond_search_sad_sse3=vp8_diamond_search_sadx4;
+
+#
+# Alt-ref Noise Reduction (ARNR)
+#
+if (vpx_config("CONFIG_REALTIME_ONLY") ne "yes") {
+    add_proto qw/void vp8_temporal_filter_apply/, "unsigned char *frame1, unsigned int stride, unsigned char *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, unsigned short *count";
+    specialize qw/vp8_temporal_filter_apply sse2/;
+}
+
+#
+# Pick Loopfilter
+#
+add_proto qw/void vp8_yv12_copy_partial_frame/, "struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc";
+specialize qw/vp8_yv12_copy_partial_frame neon/;
+
+#
+# Denoiser filter
+#
+if (vpx_config("CONFIG_TEMPORAL_DENOISING") eq "yes") {
+    add_proto qw/int vp8_denoiser_filter/, "struct yv12_buffer_config* mc_running_avg, struct yv12_buffer_config* running_avg, struct macroblock* signal, unsigned int motion_magnitude2, int y_offset, int uv_offset";
+    specialize qw/vp8_denoiser_filter sse2 neon/;
+}
+
+# End of encoder only functions
+}
+1;
diff --git a/vp8/common/rtcd_defs.sh b/vp8/common/rtcd_defs.sh
deleted file mode 100644
index 28e6754..0000000
--- a/vp8/common/rtcd_defs.sh
+++ /dev/null
@@ -1,542 +0,0 @@
-vp8_common_forward_decls() {
-cat <<EOF
-/*
- * VP8
- */
-
-struct blockd;
-struct macroblockd;
-struct loop_filter_info;
-
-/* Encoder forward decls */
-struct block;
-struct macroblock;
-struct variance_vtable;
-union int_mv;
-struct yv12_buffer_config;
-EOF
-}
-forward_decls vp8_common_forward_decls
-
-#
-# system state
-#
-prototype void vp8_clear_system_state ""
-specialize vp8_clear_system_state mmx
-vp8_clear_system_state_mmx=vpx_reset_mmx_state
-
-#
-# Dequant
-#
-prototype void vp8_dequantize_b "struct blockd*, short *dqc"
-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 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 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 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 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 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 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 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"
-specialize vp8_loop_filter_simple_mbv mmx sse2 media neon
-vp8_loop_filter_simple_mbv_c=vp8_loop_filter_simple_vertical_edge_c
-vp8_loop_filter_simple_mbv_mmx=vp8_loop_filter_simple_vertical_edge_mmx
-vp8_loop_filter_simple_mbv_sse2=vp8_loop_filter_simple_vertical_edge_sse2
-vp8_loop_filter_simple_mbv_media=vp8_loop_filter_simple_vertical_edge_armv6
-vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon
-
-prototype void vp8_loop_filter_simple_mbh "unsigned char *y, int ystride, const unsigned char *blimit"
-specialize vp8_loop_filter_simple_mbh mmx sse2 media neon
-vp8_loop_filter_simple_mbh_c=vp8_loop_filter_simple_horizontal_edge_c
-vp8_loop_filter_simple_mbh_mmx=vp8_loop_filter_simple_horizontal_edge_mmx
-vp8_loop_filter_simple_mbh_sse2=vp8_loop_filter_simple_horizontal_edge_sse2
-vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6
-vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon
-
-prototype void vp8_loop_filter_simple_bv "unsigned char *y, int ystride, const unsigned char *blimit"
-specialize vp8_loop_filter_simple_bv mmx sse2 media neon
-vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c
-vp8_loop_filter_simple_bv_mmx=vp8_loop_filter_bvs_mmx
-vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2
-vp8_loop_filter_simple_bv_media=vp8_loop_filter_bvs_armv6
-vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon
-
-prototype void vp8_loop_filter_simple_bh "unsigned char *y, int ystride, const unsigned char *blimit"
-specialize vp8_loop_filter_simple_bh mmx sse2 media neon
-vp8_loop_filter_simple_bh_c=vp8_loop_filter_bhs_c
-vp8_loop_filter_simple_bh_mmx=vp8_loop_filter_bhs_mmx
-vp8_loop_filter_simple_bh_sse2=vp8_loop_filter_bhs_sse2
-vp8_loop_filter_simple_bh_media=vp8_loop_filter_bhs_armv6
-vp8_loop_filter_simple_bh_neon=vp8_loop_filter_bhs_neon
-
-#
-# IDCT
-#
-#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 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 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 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 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 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 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
-#TODO: fix assembly for neon
-
-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 "unsigned char *Above, unsigned char *yleft, int left_stride, int 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
-
-#
-# Postproc
-#
-if [ "$CONFIG_POSTPROC" = "yes" ]; then
-    prototype void vp8_mbpost_proc_down "unsigned char *dst, int pitch, int rows, int cols,int flimit"
-    specialize vp8_mbpost_proc_down mmx sse2
-    vp8_mbpost_proc_down_sse2=vp8_mbpost_proc_down_xmm
-
-    prototype void vp8_mbpost_proc_across_ip "unsigned char *dst, int pitch, int rows, int cols,int flimit"
-    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_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
-    vp8_plane_add_noise_sse2=vp8_plane_add_noise_wmt
-
-    prototype void vp8_blend_mb_inner "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
-    # no asm yet
-
-    prototype void vp8_blend_mb_outer "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
-    # no asm yet
-
-    prototype void vp8_blend_b "unsigned char *y, unsigned char *u, unsigned char *v, int y1, int u1, int v1, int alpha, int stride"
-    # no asm yet
-
-    prototype void vp8_filter_by_weight16x16 "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"
-    specialize vp8_filter_by_weight16x16 sse2
-
-    prototype void vp8_filter_by_weight8x8 "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"
-    specialize vp8_filter_by_weight8x8 sse2
-
-    prototype void vp8_filter_by_weight4x4 "unsigned char *src, int src_stride, unsigned char *dst, int dst_stride, int src_weight"
-    # no asm yet
-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 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 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 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 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
-vp8_bilinear_predict16x16_media=vp8_bilinear_predict16x16_armv6
-
-prototype void vp8_bilinear_predict8x8 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_bilinear_predict8x8 mmx sse2 ssse3 media neon
-vp8_bilinear_predict8x8_media=vp8_bilinear_predict8x8_armv6
-
-prototype void vp8_bilinear_predict8x4 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_bilinear_predict8x4 mmx media neon
-vp8_bilinear_predict8x4_media=vp8_bilinear_predict8x4_armv6
-
-prototype void vp8_bilinear_predict4x4 "unsigned char *src, int src_pitch, int xofst, int yofst, unsigned char *dst, int dst_pitch"
-specialize vp8_bilinear_predict4x4 mmx media neon
-vp8_bilinear_predict4x4_media=vp8_bilinear_predict4x4_armv6
-
-#
-# Whole-pixel Variance
-#
-prototype unsigned int vp8_variance4x4 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance4x4 mmx sse2
-vp8_variance4x4_sse2=vp8_variance4x4_wmt
-
-prototype unsigned int vp8_variance8x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance8x8 mmx sse2 media neon
-vp8_variance8x8_sse2=vp8_variance8x8_wmt
-vp8_variance8x8_media=vp8_variance8x8_armv6
-
-prototype unsigned int vp8_variance8x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance8x16 mmx sse2 neon
-vp8_variance8x16_sse2=vp8_variance8x16_wmt
-
-prototype unsigned int vp8_variance16x8 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance16x8 mmx sse2 neon
-vp8_variance16x8_sse2=vp8_variance16x8_wmt
-
-prototype unsigned int vp8_variance16x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance16x16 mmx sse2 media neon
-vp8_variance16x16_sse2=vp8_variance16x16_wmt
-vp8_variance16x16_media=vp8_variance16x16_armv6
-
-#
-# Sub-pixel Variance
-#
-prototype unsigned int vp8_sub_pixel_variance4x4 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_variance4x4 mmx sse2
-vp8_sub_pixel_variance4x4_sse2=vp8_sub_pixel_variance4x4_wmt
-
-prototype unsigned int vp8_sub_pixel_variance8x8 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_variance8x8 mmx sse2 media neon
-vp8_sub_pixel_variance8x8_sse2=vp8_sub_pixel_variance8x8_wmt
-vp8_sub_pixel_variance8x8_media=vp8_sub_pixel_variance8x8_armv6
-
-prototype unsigned int vp8_sub_pixel_variance8x16 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_variance8x16 mmx sse2
-vp8_sub_pixel_variance8x16_sse2=vp8_sub_pixel_variance8x16_wmt
-
-prototype unsigned int vp8_sub_pixel_variance16x8 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_variance16x8 mmx sse2 ssse3
-vp8_sub_pixel_variance16x8_sse2=vp8_sub_pixel_variance16x8_wmt
-
-prototype unsigned int vp8_sub_pixel_variance16x16 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_variance16x16 mmx sse2 ssse3 media neon
-vp8_sub_pixel_variance16x16_sse2=vp8_sub_pixel_variance16x16_wmt
-vp8_sub_pixel_variance16x16_media=vp8_sub_pixel_variance16x16_armv6
-
-prototype unsigned int vp8_variance_halfpixvar16x16_h "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance_halfpixvar16x16_h mmx sse2 media neon
-vp8_variance_halfpixvar16x16_h_sse2=vp8_variance_halfpixvar16x16_h_wmt
-vp8_variance_halfpixvar16x16_h_media=vp8_variance_halfpixvar16x16_h_armv6
-
-prototype unsigned int vp8_variance_halfpixvar16x16_v "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance_halfpixvar16x16_v mmx sse2 media neon
-vp8_variance_halfpixvar16x16_v_sse2=vp8_variance_halfpixvar16x16_v_wmt
-vp8_variance_halfpixvar16x16_v_media=vp8_variance_halfpixvar16x16_v_armv6
-
-prototype unsigned int vp8_variance_halfpixvar16x16_hv "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_variance_halfpixvar16x16_hv mmx sse2 media neon
-vp8_variance_halfpixvar16x16_hv_sse2=vp8_variance_halfpixvar16x16_hv_wmt
-vp8_variance_halfpixvar16x16_hv_media=vp8_variance_halfpixvar16x16_hv_armv6
-
-#
-# Single block 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 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 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 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 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
-
-#
-# Multi-block SAD, comparing a reference to N blocks 1 pixel apart horizontally
-#
-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 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 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 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 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 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 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 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 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 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 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 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 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 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 src_stride, const unsigned char * const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp8_sad16x16x4d sse3
-
-#
-# Encoder functions below this point.
-#
-if [ "$CONFIG_VP8_ENCODER" = "yes" ]; then
-
-#
-# Sum of squares (vector)
-#
-prototype unsigned int vp8_get_mb_ss "const short *"
-specialize vp8_get_mb_ss mmx sse2
-
-#
-# SSE (Sum Squared Error)
-#
-prototype unsigned int vp8_sub_pixel_mse16x16 "const unsigned char  *src_ptr, int  source_stride, int  xoffset, int  yoffset, const unsigned char *ref_ptr, int Refstride, unsigned int *sse"
-specialize vp8_sub_pixel_mse16x16 mmx sse2
-vp8_sub_pixel_mse16x16_sse2=vp8_sub_pixel_mse16x16_wmt
-
-prototype unsigned int vp8_mse16x16 "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride, unsigned int *sse"
-specialize vp8_mse16x16 mmx sse2 media neon
-vp8_mse16x16_sse2=vp8_mse16x16_wmt
-vp8_mse16x16_media=vp8_mse16x16_armv6
-
-prototype unsigned int vp8_get4x4sse_cs "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int  ref_stride"
-specialize vp8_get4x4sse_cs mmx neon
-
-#
-# Block copy
-#
-case $arch in
-    x86*)
-    prototype void vp8_copy32xn "const unsigned char *src_ptr, int source_stride, const unsigned char *ref_ptr, int ref_stride, int n"
-    specialize vp8_copy32xn sse2 sse3
-    ;;
-esac
-
-#
-# Structured Similarity (SSIM)
-#
-if [ "$CONFIG_INTERNAL_STATS" = "yes" ]; then
-    [ $arch = "x86_64" ] && sse2_on_x86_64=sse2
-
-    prototype void vp8_ssim_parms_8x8 "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp8_ssim_parms_8x8 $sse2_on_x86_64
-
-    prototype void vp8_ssim_parms_16x16 "unsigned char *s, int sp, unsigned char *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp8_ssim_parms_16x16 $sse2_on_x86_64
-fi
-
-#
-# Forward DCT
-#
-prototype void vp8_short_fdct4x4 "short *input, short *output, int pitch"
-specialize vp8_short_fdct4x4 mmx sse2 media neon
-vp8_short_fdct4x4_media=vp8_short_fdct4x4_armv6
-
-prototype void vp8_short_fdct8x4 "short *input, short *output, int pitch"
-specialize vp8_short_fdct8x4 mmx sse2 media neon
-vp8_short_fdct8x4_media=vp8_short_fdct8x4_armv6
-
-prototype void vp8_short_walsh4x4 "short *input, short *output, int pitch"
-specialize vp8_short_walsh4x4 sse2 media neon
-vp8_short_walsh4x4_media=vp8_short_walsh4x4_armv6
-
-#
-# Quantizer
-#
-prototype void vp8_regular_quantize_b "struct block *, struct blockd *"
-specialize vp8_regular_quantize_b sse2 #sse4_1
-# TODO(johann) Update sse4 implementation and re-enable
-#vp8_regular_quantize_b_sse4_1=vp8_regular_quantize_b_sse4
-
-prototype void vp8_fast_quantize_b "struct block *, struct blockd *"
-specialize vp8_fast_quantize_b sse2 ssse3 media neon
-vp8_fast_quantize_b_media=vp8_fast_quantize_b_armv6
-
-prototype void vp8_regular_quantize_b_pair "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2"
-# no asm yet
-
-prototype void vp8_fast_quantize_b_pair "struct block *b1, struct block *b2, struct blockd *d1, struct blockd *d2"
-specialize vp8_fast_quantize_b_pair neon
-
-prototype void vp8_quantize_mb "struct macroblock *"
-specialize vp8_quantize_mb neon
-
-prototype void vp8_quantize_mby "struct macroblock *"
-specialize vp8_quantize_mby neon
-
-prototype void vp8_quantize_mbuv "struct macroblock *"
-specialize vp8_quantize_mbuv neon
-
-#
-# Block subtraction
-#
-prototype int vp8_block_error "short *coeff, short *dqcoeff"
-specialize vp8_block_error mmx sse2
-vp8_block_error_sse2=vp8_block_error_xmm
-
-prototype int vp8_mbblock_error "struct macroblock *mb, int dc"
-specialize vp8_mbblock_error mmx sse2
-vp8_mbblock_error_sse2=vp8_mbblock_error_xmm
-
-prototype int vp8_mbuverror "struct macroblock *mb"
-specialize vp8_mbuverror mmx sse2
-vp8_mbuverror_sse2=vp8_mbuverror_xmm
-
-prototype void vp8_subtract_b "struct block *be, struct blockd *bd, int pitch"
-specialize vp8_subtract_b mmx sse2 media neon
-vp8_subtract_b_media=vp8_subtract_b_armv6
-
-prototype void vp8_subtract_mby "short *diff, unsigned char *src, int src_stride, unsigned char *pred, int pred_stride"
-specialize vp8_subtract_mby mmx sse2 media neon
-vp8_subtract_mby_media=vp8_subtract_mby_armv6
-
-prototype void vp8_subtract_mbuv "short *diff, unsigned char *usrc, unsigned char *vsrc, int src_stride, unsigned char *upred, unsigned char *vpred, int pred_stride"
-specialize vp8_subtract_mbuv mmx sse2 media neon
-vp8_subtract_mbuv_media=vp8_subtract_mbuv_armv6
-
-#
-# Motion search
-#
-prototype int vp8_full_search_sad "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv"
-specialize vp8_full_search_sad sse3 sse4_1
-vp8_full_search_sad_sse3=vp8_full_search_sadx3
-vp8_full_search_sad_sse4_1=vp8_full_search_sadx8
-
-prototype int vp8_refining_search_sad "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, int sad_per_bit, int distance, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv"
-specialize vp8_refining_search_sad sse3
-vp8_refining_search_sad_sse3=vp8_refining_search_sadx4
-
-prototype int vp8_diamond_search_sad "struct macroblock *x, struct block *b, struct blockd *d, union int_mv *ref_mv, union int_mv *best_mv, int search_param, int sad_per_bit, int *num00, struct variance_vtable *fn_ptr, int *mvcost[2], union int_mv *center_mv"
-vp8_diamond_search_sad_sse3=vp8_diamond_search_sadx4
-
-#
-# Alt-ref Noise Reduction (ARNR)
-#
-if [ "$CONFIG_REALTIME_ONLY" != "yes" ]; then
-    prototype void vp8_temporal_filter_apply "unsigned char *frame1, unsigned int stride, unsigned char *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, unsigned short *count"
-    specialize vp8_temporal_filter_apply sse2
-fi
-
-#
-# Pick Loopfilter
-#
-prototype void vp8_yv12_copy_partial_frame "struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc"
-specialize vp8_yv12_copy_partial_frame neon
-
-#
-# Denoiser filter
-#
-if [ "$CONFIG_TEMPORAL_DENOISING" = "yes" ]; then
-    prototype int vp8_denoiser_filter "struct yv12_buffer_config* mc_running_avg, struct yv12_buffer_config* running_avg, struct macroblock* signal, unsigned int motion_magnitude2, int y_offset, int uv_offset"
-    specialize vp8_denoiser_filter sse2 neon
-fi
-
-# End of encoder only functions
-fi
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index ac91d7a..dfb54a5 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -47,7 +47,7 @@
 VP8_COMMON_SRCS-yes += common/reconinter.h
 VP8_COMMON_SRCS-yes += common/reconintra4x4.h
 VP8_COMMON_SRCS-yes += common/rtcd.c
-VP8_COMMON_SRCS-yes += common/rtcd_defs.sh
+VP8_COMMON_SRCS-yes += common/rtcd_defs.pl
 VP8_COMMON_SRCS-yes += common/setupintrarecon.h
 VP8_COMMON_SRCS-yes += common/swapyv12buffer.h
 VP8_COMMON_SRCS-yes += common/systemdependent.h
@@ -189,4 +189,4 @@
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dequantizeb_neon.c
 
 
-$(eval $(call rtcd_h_template,vp8_rtcd,vp8/common/rtcd_defs.sh))
+$(eval $(call rtcd_h_template,vp8_rtcd,vp8/common/rtcd_defs.pl))
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
new file mode 100644
index 0000000..e4cd9d4
--- /dev/null
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -0,0 +1,778 @@
+sub vp9_common_forward_decls() {
+print <<EOF
+/*
+ * VP9
+ */
+
+#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_enums.h"
+
+struct macroblockd;
+
+/* Encoder forward decls */
+struct macroblock;
+struct vp9_variance_vtable;
+
+#define DEC_MVCOSTS int *mvjcost, int *mvcost[2]
+struct mv;
+union int_mv;
+struct yv12_buffer_config;
+EOF
+}
+forward_decls qw/vp9_common_forward_decls/;
+
+# x86inc.asm doesn't work if pic is enabled on 32 bit platforms so no assembly.
+if (vpx_config("CONFIG_USE_X86INC") eq "yes") {
+  $mmx_x86inc = 'mmx';
+  $sse_x86inc = 'sse';
+  $sse2_x86inc = 'sse2';
+  $ssse3_x86inc = 'ssse3';
+  $avx_x86inc = 'avx';
+  $avx2_x86inc = 'avx2';
+} else {
+  $mmx_x86inc = $sse_x86inc = $sse2_x86inc = $ssse3_x86inc =
+  $avx_x86inc = $avx2_x86inc = '';
+}
+
+# this variable is for functions that are 64 bit only.
+if ($opts{arch} eq "x86_64") {
+  $mmx_x86_64 = 'mmx';
+  $sse2_x86_64 = 'sse2';
+  $ssse3_x86_64 = 'ssse3';
+  $avx_x86_64 = 'avx';
+  $avx2_x86_64 = 'avx2';
+} else {
+  $mmx_x86_64 = $sse2_x86_64 = $ssse3_x86_64 =
+  $avx_x86_64 = $avx2_x86_64 = '';
+}
+
+#
+# RECON
+#
+add_proto qw/void vp9_d207_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d207_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d45_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d45_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d63_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d63_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_h_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_h_predictor_4x4 neon dspr2/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d117_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d117_predictor_4x4/;
+
+add_proto qw/void vp9_d135_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d135_predictor_4x4/;
+
+add_proto qw/void vp9_d153_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d153_predictor_4x4/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_v_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_v_predictor_4x4 neon/, "$sse_x86inc";
+
+add_proto qw/void vp9_tm_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_tm_predictor_4x4 neon dspr2/, "$sse_x86inc";
+
+add_proto qw/void vp9_dc_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_predictor_4x4 dspr2/, "$sse_x86inc";
+
+add_proto qw/void vp9_dc_top_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_top_predictor_4x4/;
+
+add_proto qw/void vp9_dc_left_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_left_predictor_4x4/;
+
+add_proto qw/void vp9_dc_128_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_128_predictor_4x4/;
+
+add_proto qw/void vp9_d207_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d207_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d45_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d45_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d63_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d63_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_h_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_h_predictor_8x8 neon dspr2/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d117_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d117_predictor_8x8/;
+
+add_proto qw/void vp9_d135_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d135_predictor_8x8/;
+
+add_proto qw/void vp9_d153_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d153_predictor_8x8/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_v_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_v_predictor_8x8 neon/, "$sse_x86inc";
+
+add_proto qw/void vp9_tm_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_tm_predictor_8x8 neon dspr2/, "$sse2_x86inc";
+
+add_proto qw/void vp9_dc_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_predictor_8x8 dspr2/, "$sse_x86inc";
+
+add_proto qw/void vp9_dc_top_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_top_predictor_8x8/;
+
+add_proto qw/void vp9_dc_left_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_left_predictor_8x8/;
+
+add_proto qw/void vp9_dc_128_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_128_predictor_8x8/;
+
+add_proto qw/void vp9_d207_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d207_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d45_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d45_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d63_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d63_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_h_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_h_predictor_16x16 neon dspr2/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d117_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d117_predictor_16x16/;
+
+add_proto qw/void vp9_d135_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d135_predictor_16x16/;
+
+add_proto qw/void vp9_d153_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d153_predictor_16x16/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_v_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_v_predictor_16x16 neon/, "$sse2_x86inc";
+
+add_proto qw/void vp9_tm_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_tm_predictor_16x16 neon/, "$sse2_x86inc";
+
+add_proto qw/void vp9_dc_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_predictor_16x16 dspr2/, "$sse2_x86inc";
+
+add_proto qw/void vp9_dc_top_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_top_predictor_16x16/;
+
+add_proto qw/void vp9_dc_left_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_left_predictor_16x16/;
+
+add_proto qw/void vp9_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_128_predictor_16x16/;
+
+add_proto qw/void vp9_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d207_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d45_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d45_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d63_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d63_predictor_32x32/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_h_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_h_predictor_32x32 neon/, "$ssse3_x86inc";
+
+add_proto qw/void vp9_d117_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d117_predictor_32x32/;
+
+add_proto qw/void vp9_d135_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d135_predictor_32x32/;
+
+add_proto qw/void vp9_d153_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_d153_predictor_32x32/;
+
+add_proto qw/void vp9_v_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_v_predictor_32x32 neon/, "$sse2_x86inc";
+
+add_proto qw/void vp9_tm_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_tm_predictor_32x32 neon/, "$sse2_x86_64";
+
+add_proto qw/void vp9_dc_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_predictor_32x32/, "$sse2_x86inc";
+
+add_proto qw/void vp9_dc_top_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_top_predictor_32x32/;
+
+add_proto qw/void vp9_dc_left_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_left_predictor_32x32/;
+
+add_proto qw/void vp9_dc_128_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vp9_dc_128_predictor_32x32/;
+
+#
+# Loopfilter
+#
+add_proto qw/void vp9_lpf_vertical_16/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
+specialize qw/vp9_lpf_vertical_16 sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_vertical_16_dual/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh";
+specialize qw/vp9_lpf_vertical_16_dual sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_vertical_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+specialize qw/vp9_lpf_vertical_8 sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_vertical_8_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
+specialize qw/vp9_lpf_vertical_8_dual sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_vertical_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+specialize qw/vp9_lpf_vertical_4 mmx neon dspr2/;
+
+add_proto qw/void vp9_lpf_vertical_4_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
+specialize qw/vp9_lpf_vertical_4_dual sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_horizontal_16/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+specialize qw/vp9_lpf_horizontal_16 sse2 avx2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_horizontal_8/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+specialize qw/vp9_lpf_horizontal_8 sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_horizontal_8_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
+specialize qw/vp9_lpf_horizontal_8_dual sse2 neon dspr2/;
+
+add_proto qw/void vp9_lpf_horizontal_4/, "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count";
+specialize qw/vp9_lpf_horizontal_4 mmx neon dspr2/;
+
+add_proto qw/void vp9_lpf_horizontal_4_dual/, "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1";
+specialize qw/vp9_lpf_horizontal_4_dual sse2 neon dspr2/;
+
+#
+# post proc
+#
+if (vpx_config("CONFIG_VP9_POSTPROC") eq "yes") {
+add_proto qw/void vp9_mbpost_proc_down/, "uint8_t *dst, int pitch, int rows, int cols, int flimit";
+specialize qw/vp9_mbpost_proc_down mmx sse2/;
+$vp9_mbpost_proc_down_sse2=vp9_mbpost_proc_down_xmm;
+
+add_proto qw/void vp9_mbpost_proc_across_ip/, "uint8_t *src, int pitch, int rows, int cols, int flimit";
+specialize qw/vp9_mbpost_proc_across_ip sse2/;
+$vp9_mbpost_proc_across_ip_sse2=vp9_mbpost_proc_across_ip_xmm;
+
+add_proto qw/void vp9_post_proc_down_and_across/, "const uint8_t *src_ptr, uint8_t *dst_ptr, int src_pixels_per_line, int dst_pixels_per_line, int rows, int cols, int flimit";
+specialize qw/vp9_post_proc_down_and_across mmx sse2/;
+$vp9_post_proc_down_and_across_sse2=vp9_post_proc_down_and_across_xmm;
+
+add_proto qw/void vp9_plane_add_noise/, "uint8_t *Start, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int Width, unsigned int Height, int Pitch";
+specialize qw/vp9_plane_add_noise mmx sse2/;
+$vp9_plane_add_noise_sse2=vp9_plane_add_noise_wmt;
+}
+
+add_proto qw/void vp9_blend_mb_inner/, "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride";
+specialize qw/vp9_blend_mb_inner/;
+
+add_proto qw/void vp9_blend_mb_outer/, "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride";
+specialize qw/vp9_blend_mb_outer/;
+
+add_proto qw/void vp9_blend_b/, "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride";
+specialize qw/vp9_blend_b/;
+
+#
+# Sub Pixel Filters
+#
+add_proto qw/void vp9_convolve_copy/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve_copy neon dspr2/, "$sse2_x86inc";
+
+add_proto qw/void vp9_convolve_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve_avg neon dspr2/, "$sse2_x86inc";
+
+add_proto qw/void vp9_convolve8/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8 sse2 ssse3 avx2 neon dspr2/;
+
+add_proto qw/void vp9_convolve8_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8_horiz sse2 ssse3 avx2 neon dspr2/;
+
+add_proto qw/void vp9_convolve8_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8_vert sse2 ssse3 avx2 neon dspr2/;
+
+add_proto qw/void vp9_convolve8_avg/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8_avg sse2 ssse3 neon dspr2/;
+
+add_proto qw/void vp9_convolve8_avg_horiz/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8_avg_horiz sse2 ssse3 neon dspr2/;
+
+add_proto qw/void vp9_convolve8_avg_vert/, "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h";
+specialize qw/vp9_convolve8_avg_vert sse2 ssse3 neon dspr2/;
+
+#
+# dct
+#
+add_proto qw/void vp9_idct4x4_1_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct4x4_1_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct4x4_16_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct4x4_16_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct8x8_1_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct8x8_1_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct8x8_64_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct8x8_64_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct8x8_10_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct8x8_10_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct16x16_1_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct16x16_1_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct16x16_256_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct16x16_256_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct16x16_10_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct16x16_10_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct32x32_1024_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct32x32_1024_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_idct32x32_34_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct32x32_34_add sse2 neon dspr2/;
+$vp9_idct32x32_34_add_neon=vp9_idct32x32_1024_add_neon;
+
+add_proto qw/void vp9_idct32x32_1_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_idct32x32_1_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_iht4x4_16_add/, "const int16_t *input, uint8_t *dest, int dest_stride, int tx_type";
+specialize qw/vp9_iht4x4_16_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_iht8x8_64_add/, "const int16_t *input, uint8_t *dest, int dest_stride, int tx_type";
+specialize qw/vp9_iht8x8_64_add sse2 neon dspr2/;
+
+add_proto qw/void vp9_iht16x16_256_add/, "const int16_t *input, uint8_t *output, int pitch, int tx_type";
+specialize qw/vp9_iht16x16_256_add sse2 dspr2/;
+
+# dct and add
+
+add_proto qw/void vp9_iwht4x4_1_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_iwht4x4_1_add/;
+
+add_proto qw/void vp9_iwht4x4_16_add/, "const int16_t *input, uint8_t *dest, int dest_stride";
+specialize qw/vp9_iwht4x4_16_add/;
+
+#
+# Encoder functions below this point.
+#
+if (vpx_config("CONFIG_VP9_ENCODER") eq "yes") {
+
+
+# variance
+add_proto qw/unsigned int vp9_variance32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance32x16/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_variance16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance16x32/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance64x32/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_variance32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance32x64/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance32x32/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_variance64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance64x64/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_variance16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance16x16 mmx/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_variance16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance16x8 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance8x16 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance8x8 mmx/, "$sse2_x86inc";
+
+add_proto qw/void vp9_get_sse_sum_8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum";
+specialize qw/vp9_get_sse_sum_8x8 sse2/;
+$vp9_get_sse_sum_8x8_sse2=vp9_get8x8var_sse2;
+
+add_proto qw/unsigned int vp9_variance8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance8x4/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance4x8/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance4x4 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance64x64 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance64x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance64x64 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance32x64/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x64/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance32x64/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance64x32/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance64x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance64x32/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance32x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance32x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance16x32/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance16x32/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance32x32 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance32x32/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance32x32 avx2/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance16x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance16x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance8x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x16/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance8x16/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance16x8/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance16x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance16x8/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance8x8/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance8x8/, "$sse2_x86inc", "$ssse3_x86inc";
+
+# TODO(jingning): need to convert 8x4/4x8 functions into mmx/sse form
+add_proto qw/unsigned int vp9_sub_pixel_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance8x4/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance8x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance8x4/, "$sse2_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance4x8/, "$sse_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x8/, "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance4x8/, "$sse_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sub_pixel_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_variance4x4/, "$sse_x86inc", "$ssse3_x86inc";
+#vp9_sub_pixel_variance4x4_sse2=vp9_sub_pixel_variance4x4_wmt
+
+add_proto qw/unsigned int vp9_sub_pixel_avg_variance4x4/, "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred";
+specialize qw/vp9_sub_pixel_avg_variance4x4/, "$sse_x86inc", "$ssse3_x86inc";
+
+add_proto qw/unsigned int vp9_sad64x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad64x64/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x64/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad32x64/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad64x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad64x32/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad32x16/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad16x32/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x32/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad32x32/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad16x16 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad16x8 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x16/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad8x16 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad8x8 mmx/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad8x4/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad4x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad4x8/, "$sse_x86inc";
+
+add_proto qw/unsigned int vp9_sad4x4/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad";
+specialize qw/vp9_sad4x4 mmx/, "$sse_x86inc";
+
+add_proto qw/unsigned int vp9_sad64x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad64x64_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x64_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad32x64_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad64x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad64x32_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad32x16_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad16x32_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad32x32_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad32x32_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad16x16_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad16x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad16x8_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x16_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad8x16_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad8x8_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad8x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad8x4_avg/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_sad4x8_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad4x8_avg/, "$sse_x86inc";
+
+add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad";
+specialize qw/vp9_sad4x4_avg/, "$sse_x86inc";
+
+add_proto qw/unsigned int vp9_variance_halfpixvar16x16_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar16x16_h/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance_halfpixvar16x16_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar16x16_v/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance_halfpixvar16x16_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar16x16_hv/, "$sse2_x86inc";
+
+add_proto qw/unsigned int vp9_variance_halfpixvar64x64_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar64x64_h/;
+
+add_proto qw/unsigned int vp9_variance_halfpixvar64x64_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar64x64_v/;
+
+add_proto qw/unsigned int vp9_variance_halfpixvar64x64_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar64x64_hv/;
+
+add_proto qw/unsigned int vp9_variance_halfpixvar32x32_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar32x32_h/;
+
+add_proto qw/unsigned int vp9_variance_halfpixvar32x32_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar32x32_v/;
+
+add_proto qw/unsigned int vp9_variance_halfpixvar32x32_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_variance_halfpixvar32x32_hv/;
+
+add_proto qw/void vp9_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad64x64x3/;
+
+add_proto qw/void vp9_sad32x32x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad32x32x3/;
+
+add_proto qw/void vp9_sad16x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad16x16x3 sse3 ssse3/;
+
+add_proto qw/void vp9_sad16x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad16x8x3 sse3 ssse3/;
+
+add_proto qw/void vp9_sad8x16x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad8x16x3 sse3/;
+
+add_proto qw/void vp9_sad8x8x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad8x8x3 sse3/;
+
+add_proto qw/void vp9_sad4x4x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad4x4x3 sse3/;
+
+add_proto qw/void vp9_sad64x64x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad64x64x8/;
+
+add_proto qw/void vp9_sad32x32x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad32x32x8/;
+
+add_proto qw/void vp9_sad16x16x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad16x16x8 sse4/;
+
+add_proto qw/void vp9_sad16x8x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad16x8x8 sse4/;
+
+add_proto qw/void vp9_sad8x16x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad8x16x8 sse4/;
+
+add_proto qw/void vp9_sad8x8x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad8x8x8 sse4/;
+
+add_proto qw/void vp9_sad8x4x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad8x4x8/;
+
+add_proto qw/void vp9_sad4x8x8/, "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad4x8x8/;
+
+add_proto qw/void vp9_sad4x4x8/, "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array";
+specialize qw/vp9_sad4x4x8 sse4/;
+
+add_proto qw/void vp9_sad64x64x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad64x64x4d sse2/;
+
+add_proto qw/void vp9_sad32x64x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad32x64x4d sse2/;
+
+add_proto qw/void vp9_sad64x32x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad64x32x4d sse2/;
+
+add_proto qw/void vp9_sad32x16x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad32x16x4d sse2/;
+
+add_proto qw/void vp9_sad16x32x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad16x32x4d sse2/;
+
+add_proto qw/void vp9_sad32x32x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad32x32x4d sse2/;
+
+add_proto qw/void vp9_sad16x16x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad16x16x4d sse2/;
+
+add_proto qw/void vp9_sad16x8x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad16x8x4d sse2/;
+
+add_proto qw/void vp9_sad8x16x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad8x16x4d sse2/;
+
+add_proto qw/void vp9_sad8x8x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad8x8x4d sse2/;
+
+# TODO(jingning): need to convert these 4x8/8x4 functions into sse2 form
+add_proto qw/void vp9_sad8x4x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad8x4x4d sse2/;
+
+add_proto qw/void vp9_sad4x8x4d/, "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad4x8x4d sse/;
+
+add_proto qw/void vp9_sad4x4x4d/, "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array";
+specialize qw/vp9_sad4x4x4d sse/;
+
+#add_proto qw/unsigned int vp9_sub_pixel_mse16x16/, "const uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, const uint8_t *dst_ptr, int dst_pixels_per_line, unsigned int *sse";
+#specialize qw/vp9_sub_pixel_mse16x16 sse2 mmx/;
+
+add_proto qw/unsigned int vp9_mse16x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
+specialize qw/vp9_mse16x16 mmx/, "$sse2_x86inc", "$avx2_x86inc";
+
+add_proto qw/unsigned int vp9_mse8x16/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
+specialize qw/vp9_mse8x16/;
+
+add_proto qw/unsigned int vp9_mse16x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
+specialize qw/vp9_mse16x8/;
+
+add_proto qw/unsigned int vp9_mse8x8/, "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse";
+specialize qw/vp9_mse8x8/;
+
+add_proto qw/unsigned int vp9_sub_pixel_mse64x64/, "const uint8_t *src_ptr, int  source_stride, int  xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_mse64x64/;
+
+add_proto qw/unsigned int vp9_sub_pixel_mse32x32/, "const uint8_t *src_ptr, int  source_stride, int  xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
+specialize qw/vp9_sub_pixel_mse32x32/;
+
+add_proto qw/unsigned int vp9_get_mb_ss/, "const int16_t *";
+specialize qw/vp9_get_mb_ss mmx sse2/;
+# ENCODEMB INVOKE
+
+add_proto qw/int64_t vp9_block_error/, "const int16_t *coeff, const int16_t *dqcoeff, intptr_t block_size, int64_t *ssz";
+specialize qw/vp9_block_error/, "$sse2_x86inc";
+
+add_proto qw/void vp9_subtract_block/, "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride";
+specialize qw/vp9_subtract_block/, "$sse2_x86inc";
+
+add_proto qw/void vp9_quantize_b/, "const int16_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+specialize qw/vp9_quantize_b/, "$ssse3_x86_64";
+
+add_proto qw/void vp9_quantize_b_32x32/, "const int16_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan";
+specialize qw/vp9_quantize_b_32x32/, "$ssse3_x86_64";
+
+#
+# Structured Similarity (SSIM)
+#
+if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") {
+    add_proto qw/void vp9_ssim_parms_8x8/, "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
+    specialize qw/vp9_ssim_parms_8x8/, "$sse2_x86_64";
+
+    add_proto qw/void vp9_ssim_parms_16x16/, "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr";
+    specialize qw/vp9_ssim_parms_16x16/, "$sse2_x86_64";
+}
+
+# fdct functions
+add_proto qw/void vp9_fht4x4/, "const int16_t *input, int16_t *output, int stride, int tx_type";
+specialize qw/vp9_fht4x4 sse2 avx2/;
+
+add_proto qw/void vp9_fht8x8/, "const int16_t *input, int16_t *output, int stride, int tx_type";
+specialize qw/vp9_fht8x8 sse2 avx2/;
+
+add_proto qw/void vp9_fht16x16/, "const int16_t *input, int16_t *output, int stride, int tx_type";
+specialize qw/vp9_fht16x16 sse2 avx2/;
+
+add_proto qw/void vp9_fwht4x4/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fwht4x4/;
+
+add_proto qw/void vp9_fdct4x4/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fdct4x4 sse2 avx2/;
+
+add_proto qw/void vp9_fdct8x8/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fdct8x8 sse2 avx2/;
+
+add_proto qw/void vp9_fdct16x16/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fdct16x16 sse2 avx2/;
+
+add_proto qw/void vp9_fdct32x32/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fdct32x32 sse2 avx2/;
+
+add_proto qw/void vp9_fdct32x32_rd/, "const int16_t *input, int16_t *output, int stride";
+specialize qw/vp9_fdct32x32_rd sse2 avx2/;
+
+#
+# Motion search
+#
+add_proto qw/int vp9_full_search_sad/, "const struct macroblock *x, const struct mv *ref_mv, int sad_per_bit, int distance, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv, struct mv *best_mv";
+specialize qw/vp9_full_search_sad sse3 sse4_1/;
+$vp9_full_search_sad_sse3=vp9_full_search_sadx3;
+$vp9_full_search_sad_sse4_1=vp9_full_search_sadx8;
+
+add_proto qw/int vp9_refining_search_sad/, "const struct macroblock *x, struct mv *ref_mv, int sad_per_bit, int distance, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv";
+specialize qw/vp9_refining_search_sad sse3/;
+$vp9_refining_search_sad_sse3=vp9_refining_search_sadx4;
+
+add_proto qw/int vp9_diamond_search_sad/, "const struct macroblock *x, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv";
+specialize qw/vp9_diamond_search_sad sse3/;
+$vp9_diamond_search_sad_sse3=vp9_diamond_search_sadx4;
+
+add_proto qw/int vp9_full_range_search/, "const struct macroblock *x, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv";
+specialize qw/vp9_full_range_search/;
+
+add_proto qw/void vp9_temporal_filter_apply/, "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count";
+specialize qw/vp9_temporal_filter_apply sse2/;
+
+}
+# end encoder functions
+1;
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
deleted file mode 100644
index 5b44970..0000000
--- a/vp9/common/vp9_rtcd_defs.sh
+++ /dev/null
@@ -1,760 +0,0 @@
-vp9_common_forward_decls() {
-cat <<EOF
-/*
- * VP9
- */
-
-#include "vpx/vpx_integer.h"
-#include "vp9/common/vp9_enums.h"
-
-struct macroblockd;
-
-/* Encoder forward decls */
-struct macroblock;
-struct vp9_variance_vtable;
-
-#define DEC_MVCOSTS int *mvjcost, int *mvcost[2]
-struct mv;
-union int_mv;
-struct yv12_buffer_config;
-EOF
-}
-forward_decls vp9_common_forward_decls
-
-# x86inc.asm doesn't work if pic is enabled on 32 bit platforms so no assembly.
-[ "$CONFIG_USE_X86INC" = "yes" ] && mmx_x86inc=mmx && sse_x86inc=sse &&
-  sse2_x86inc=sse2 && ssse3_x86inc=ssse3 && avx_x86inc=avx && avx2_x86inc=avx2
-
-# this variable is for functions that are 64 bit only.
-[ $arch = "x86_64" ] && mmx_x86_64=mmx && sse2_x86_64=sse2 && 
-  ssse3_x86_64=ssse3 && avx_x86_64=avx && avx2_x86_64=avx2
-
-#
-# RECON
-#
-prototype void vp9_d207_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d207_predictor_4x4 $ssse3_x86inc
-
-prototype void vp9_d45_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d45_predictor_4x4 $ssse3_x86inc
-
-prototype void vp9_d63_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d63_predictor_4x4 $ssse3_x86inc
-
-prototype void vp9_h_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_4x4 $ssse3_x86inc neon dspr2
-
-prototype void vp9_d117_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d117_predictor_4x4
-
-prototype void vp9_d135_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d135_predictor_4x4
-
-prototype void vp9_d153_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d153_predictor_4x4 $ssse3_x86inc
-
-prototype void vp9_v_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_v_predictor_4x4 $sse_x86inc neon
-
-prototype void vp9_tm_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_4x4 $sse_x86inc neon dspr2
-
-prototype void vp9_dc_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_4x4 $sse_x86inc dspr2
-
-prototype void vp9_dc_top_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_top_predictor_4x4
-
-prototype void vp9_dc_left_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_left_predictor_4x4
-
-prototype void vp9_dc_128_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_128_predictor_4x4
-
-prototype void vp9_d207_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d207_predictor_8x8 $ssse3_x86inc
-
-prototype void vp9_d45_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d45_predictor_8x8 $ssse3_x86inc
-
-prototype void vp9_d63_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d63_predictor_8x8 $ssse3_x86inc
-
-prototype void vp9_h_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_8x8 $ssse3_x86inc neon dspr2
-
-prototype void vp9_d117_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d117_predictor_8x8
-
-prototype void vp9_d135_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d135_predictor_8x8
-
-prototype void vp9_d153_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d153_predictor_8x8 $ssse3_x86inc
-
-prototype void vp9_v_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_v_predictor_8x8 $sse_x86inc neon
-
-prototype void vp9_tm_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_8x8 $sse2_x86inc neon dspr2
-
-prototype void vp9_dc_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_8x8 $sse_x86inc dspr2
-
-prototype void vp9_dc_top_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_top_predictor_8x8
-
-prototype void vp9_dc_left_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_left_predictor_8x8
-
-prototype void vp9_dc_128_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_128_predictor_8x8
-
-prototype void vp9_d207_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d207_predictor_16x16 $ssse3_x86inc
-
-prototype void vp9_d45_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d45_predictor_16x16 $ssse3_x86inc
-
-prototype void vp9_d63_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d63_predictor_16x16 $ssse3_x86inc
-
-prototype void vp9_h_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_16x16 $ssse3_x86inc neon dspr2
-
-prototype void vp9_d117_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d117_predictor_16x16
-
-prototype void vp9_d135_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d135_predictor_16x16
-
-prototype void vp9_d153_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d153_predictor_16x16 $ssse3_x86inc
-
-prototype void vp9_v_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_v_predictor_16x16 $sse2_x86inc neon
-
-prototype void vp9_tm_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_16x16 $sse2_x86inc neon
-
-prototype void vp9_dc_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_16x16 $sse2_x86inc dspr2
-
-prototype void vp9_dc_top_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_top_predictor_16x16
-
-prototype void vp9_dc_left_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_left_predictor_16x16
-
-prototype void vp9_dc_128_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_128_predictor_16x16
-
-prototype void vp9_d207_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d207_predictor_32x32 $ssse3_x86inc
-
-prototype void vp9_d45_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d45_predictor_32x32 $ssse3_x86inc
-
-prototype void vp9_d63_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d63_predictor_32x32 $ssse3_x86inc
-
-prototype void vp9_h_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_32x32 $ssse3_x86inc neon
-
-prototype void vp9_d117_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d117_predictor_32x32
-
-prototype void vp9_d135_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d135_predictor_32x32
-
-prototype void vp9_d153_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_d153_predictor_32x32
-
-prototype void vp9_v_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_v_predictor_32x32 $sse2_x86inc neon
-
-prototype void vp9_tm_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_32x32 $sse2_x86_64 neon
-
-prototype void vp9_dc_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_32x32 $sse2_x86inc
-
-prototype void vp9_dc_top_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_top_predictor_32x32
-
-prototype void vp9_dc_left_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_left_predictor_32x32
-
-prototype void vp9_dc_128_predictor_32x32 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_128_predictor_32x32
-
-#
-# Loopfilter
-#
-prototype void vp9_lpf_vertical_16 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
-specialize vp9_lpf_vertical_16 sse2 neon dspr2
-
-prototype void vp9_lpf_vertical_16_dual "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh"
-specialize vp9_lpf_vertical_16_dual sse2 neon dspr2
-
-prototype void vp9_lpf_vertical_8 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
-specialize vp9_lpf_vertical_8 sse2 neon dspr2
-
-prototype void vp9_lpf_vertical_8_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
-specialize vp9_lpf_vertical_8_dual sse2 neon dspr2
-
-prototype void vp9_lpf_vertical_4 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
-specialize vp9_lpf_vertical_4 mmx neon dspr2
-
-prototype void vp9_lpf_vertical_4_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
-specialize vp9_lpf_vertical_4_dual sse2 neon dspr2
-
-prototype void vp9_lpf_horizontal_16 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
-specialize vp9_lpf_horizontal_16 sse2 avx2 neon dspr2
-
-prototype void vp9_lpf_horizontal_8 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
-specialize vp9_lpf_horizontal_8 sse2 neon dspr2
-
-prototype void vp9_lpf_horizontal_8_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
-specialize vp9_lpf_horizontal_8_dual sse2 neon dspr2
-
-prototype void vp9_lpf_horizontal_4 "uint8_t *s, int pitch, const uint8_t *blimit, const uint8_t *limit, const uint8_t *thresh, int count"
-specialize vp9_lpf_horizontal_4 mmx neon dspr2
-
-prototype void vp9_lpf_horizontal_4_dual "uint8_t *s, int pitch, const uint8_t *blimit0, const uint8_t *limit0, const uint8_t *thresh0, const uint8_t *blimit1, const uint8_t *limit1, const uint8_t *thresh1"
-specialize vp9_lpf_horizontal_4_dual sse2 neon dspr2
-
-#
-# post proc
-#
-if [ "$CONFIG_VP9_POSTPROC" = "yes" ]; then
-prototype void vp9_mbpost_proc_down "uint8_t *dst, int pitch, int rows, int cols, int flimit"
-specialize vp9_mbpost_proc_down mmx sse2
-vp9_mbpost_proc_down_sse2=vp9_mbpost_proc_down_xmm
-
-prototype void vp9_mbpost_proc_across_ip "uint8_t *src, int pitch, int rows, int cols, int flimit"
-specialize vp9_mbpost_proc_across_ip sse2
-vp9_mbpost_proc_across_ip_sse2=vp9_mbpost_proc_across_ip_xmm
-
-prototype void vp9_post_proc_down_and_across "const uint8_t *src_ptr, uint8_t *dst_ptr, int src_pixels_per_line, int dst_pixels_per_line, int rows, int cols, int flimit"
-specialize vp9_post_proc_down_and_across mmx sse2
-vp9_post_proc_down_and_across_sse2=vp9_post_proc_down_and_across_xmm
-
-prototype void vp9_plane_add_noise "uint8_t *Start, char *noise, char blackclamp[16], char whiteclamp[16], char bothclamp[16], unsigned int Width, unsigned int Height, int Pitch"
-specialize vp9_plane_add_noise mmx sse2
-vp9_plane_add_noise_sse2=vp9_plane_add_noise_wmt
-fi
-
-prototype void vp9_blend_mb_inner "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
-specialize vp9_blend_mb_inner
-
-prototype void vp9_blend_mb_outer "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
-specialize vp9_blend_mb_outer
-
-prototype void vp9_blend_b "uint8_t *y, uint8_t *u, uint8_t *v, int y1, int u1, int v1, int alpha, int stride"
-specialize vp9_blend_b
-
-#
-# Sub Pixel Filters
-#
-prototype void vp9_convolve_copy "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve_copy $sse2_x86inc neon dspr2
-
-prototype void vp9_convolve_avg "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve_avg $sse2_x86inc neon dspr2
-
-prototype void vp9_convolve8 "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8 sse2 ssse3 avx2 neon dspr2
-
-prototype void vp9_convolve8_horiz "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8_horiz sse2 ssse3 avx2 neon dspr2
-
-prototype void vp9_convolve8_vert "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8_vert sse2 ssse3 avx2 neon dspr2
-
-prototype void vp9_convolve8_avg "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8_avg sse2 ssse3 neon dspr2
-
-prototype void vp9_convolve8_avg_horiz "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8_avg_horiz sse2 ssse3 neon dspr2
-
-prototype void vp9_convolve8_avg_vert "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
-specialize vp9_convolve8_avg_vert sse2 ssse3 neon dspr2
-
-#
-# dct
-#
-prototype void vp9_idct4x4_1_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct4x4_1_add sse2 neon dspr2
-
-prototype void vp9_idct4x4_16_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct4x4_16_add sse2 neon dspr2
-
-prototype void vp9_idct8x8_1_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct8x8_1_add sse2 neon dspr2
-
-prototype void vp9_idct8x8_64_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct8x8_64_add sse2 neon dspr2
-
-prototype void vp9_idct8x8_10_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct8x8_10_add sse2 neon dspr2
-
-prototype void vp9_idct16x16_1_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct16x16_1_add sse2 neon dspr2
-
-prototype void vp9_idct16x16_256_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct16x16_256_add sse2 neon dspr2
-
-prototype void vp9_idct16x16_10_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct16x16_10_add sse2 neon dspr2
-
-prototype void vp9_idct32x32_1024_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct32x32_1024_add sse2 neon dspr2
-
-prototype void vp9_idct32x32_34_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct32x32_34_add sse2 neon dspr2
-vp9_idct32x32_34_add_neon=vp9_idct32x32_1024_add_neon
-
-prototype void vp9_idct32x32_1_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_idct32x32_1_add sse2 neon dspr2
-
-prototype void vp9_iht4x4_16_add "const int16_t *input, uint8_t *dest, int dest_stride, int tx_type"
-specialize vp9_iht4x4_16_add sse2 neon dspr2
-
-prototype void vp9_iht8x8_64_add "const int16_t *input, uint8_t *dest, int dest_stride, int tx_type"
-specialize vp9_iht8x8_64_add sse2 neon dspr2
-
-prototype void vp9_iht16x16_256_add "const int16_t *input, uint8_t *output, int pitch, int tx_type"
-specialize vp9_iht16x16_256_add sse2 dspr2
-
-# dct and add
-
-prototype void vp9_iwht4x4_1_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_iwht4x4_1_add
-
-prototype void vp9_iwht4x4_16_add "const int16_t *input, uint8_t *dest, int dest_stride"
-specialize vp9_iwht4x4_16_add
-
-#
-# Encoder functions below this point.
-#
-if [ "$CONFIG_VP9_ENCODER" = "yes" ]; then
-
-
-# variance
-prototype unsigned int vp9_variance32x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance32x16 $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_variance16x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance16x32 $sse2_x86inc
-
-prototype unsigned int vp9_variance64x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance64x32 $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_variance32x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance32x64 $sse2_x86inc
-
-prototype unsigned int vp9_variance32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance32x32 $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_variance64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance64x64 $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_variance16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance16x16 mmx $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_variance16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance16x8 mmx $sse2_x86inc
-
-prototype unsigned int vp9_variance8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance8x16 mmx $sse2_x86inc
-
-prototype unsigned int vp9_variance8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance8x8 mmx $sse2_x86inc
-
-prototype void vp9_get_sse_sum_8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, int *sum"
-specialize vp9_get_sse_sum_8x8 sse2
-vp9_get_sse_sum_8x8_sse2=vp9_get8x8var_sse2
-
-prototype unsigned int vp9_variance8x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance8x4 $sse2_x86inc
-
-prototype unsigned int vp9_variance4x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance4x8 $sse2_x86inc
-
-prototype unsigned int vp9_variance4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance4x4 mmx $sse2_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance64x64 $sse2_x86inc $ssse3_x86inc avx2
-
-prototype unsigned int vp9_sub_pixel_avg_variance64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance64x64 $sse2_x86inc $ssse3_x86inc avx2
-
-prototype unsigned int vp9_sub_pixel_variance32x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance32x64 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance32x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance32x64 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance64x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance64x32 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance64x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance64x32 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance32x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance32x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance32x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance32x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance16x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance16x32 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance16x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance16x32 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance32x32 $sse2_x86inc $ssse3_x86inc avx2
-
-prototype unsigned int vp9_sub_pixel_avg_variance32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance32x32 $sse2_x86inc $ssse3_x86inc avx2
-
-prototype unsigned int vp9_sub_pixel_variance16x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance16x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance16x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance16x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance8x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance8x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance8x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance8x16 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance16x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance16x8 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance16x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance16x8 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance8x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance8x8 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance8x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance8x8 $sse2_x86inc $ssse3_x86inc
-
-# TODO(jingning): need to convert 8x4/4x8 functions into mmx/sse form
-prototype unsigned int vp9_sub_pixel_variance8x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance8x4 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance8x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance8x4 $sse2_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance4x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance4x8 $sse_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_avg_variance4x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance4x8 $sse_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sub_pixel_variance4x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_variance4x4 $sse_x86inc $ssse3_x86inc
-#vp9_sub_pixel_variance4x4_sse2=vp9_sub_pixel_variance4x4_wmt
-
-prototype unsigned int vp9_sub_pixel_avg_variance4x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse, const uint8_t *second_pred"
-specialize vp9_sub_pixel_avg_variance4x4 $sse_x86inc $ssse3_x86inc
-
-prototype unsigned int vp9_sad64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad64x64 $sse2_x86inc
-
-prototype unsigned int vp9_sad32x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad32x64 $sse2_x86inc
-
-prototype unsigned int vp9_sad64x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad64x32 $sse2_x86inc
-
-prototype unsigned int vp9_sad32x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad32x16 $sse2_x86inc
-
-prototype unsigned int vp9_sad16x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad16x32 $sse2_x86inc
-
-prototype unsigned int vp9_sad32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad32x32 $sse2_x86inc
-
-prototype unsigned int vp9_sad16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad16x16 mmx $sse2_x86inc
-
-prototype unsigned int vp9_sad16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad16x8 mmx $sse2_x86inc
-
-prototype unsigned int vp9_sad8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad8x16 mmx $sse2_x86inc
-
-prototype unsigned int vp9_sad8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad8x8 mmx $sse2_x86inc
-
-prototype unsigned int vp9_sad8x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad8x4 $sse2_x86inc
-
-prototype unsigned int vp9_sad4x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int max_sad"
-specialize vp9_sad4x8 $sse_x86inc
-
-prototype unsigned int vp9_sad4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int max_sad"
-specialize vp9_sad4x4 mmx $sse_x86inc
-
-prototype unsigned int vp9_sad64x64_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad64x64_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad32x64_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad32x64_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad64x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad64x32_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad32x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad32x16_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad16x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad16x32_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad32x32_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad32x32_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad16x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad16x16_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad16x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad16x8_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad8x16_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad8x16_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad8x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad8x8_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad8x4_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad8x4_avg $sse2_x86inc
-
-prototype unsigned int vp9_sad4x8_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad4x8_avg $sse_x86inc
-
-prototype unsigned int vp9_sad4x4_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
-specialize vp9_sad4x4_avg $sse_x86inc
-
-prototype unsigned int vp9_variance_halfpixvar16x16_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar16x16_h $sse2_x86inc
-
-prototype unsigned int vp9_variance_halfpixvar16x16_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar16x16_v $sse2_x86inc
-
-prototype unsigned int vp9_variance_halfpixvar16x16_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar16x16_hv $sse2_x86inc
-
-prototype unsigned int vp9_variance_halfpixvar64x64_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar64x64_h
-
-prototype unsigned int vp9_variance_halfpixvar64x64_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar64x64_v
-
-prototype unsigned int vp9_variance_halfpixvar64x64_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar64x64_hv
-
-prototype unsigned int vp9_variance_halfpixvar32x32_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar32x32_h
-
-prototype unsigned int vp9_variance_halfpixvar32x32_v "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar32x32_v
-
-prototype unsigned int vp9_variance_halfpixvar32x32_hv "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_variance_halfpixvar32x32_hv
-
-prototype void vp9_sad64x64x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad64x64x3
-
-prototype void vp9_sad32x32x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad32x32x3
-
-prototype void vp9_sad16x16x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad16x16x3 sse3 ssse3
-
-prototype void vp9_sad16x8x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad16x8x3 sse3 ssse3
-
-prototype void vp9_sad8x16x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad8x16x3 sse3
-
-prototype void vp9_sad8x8x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad8x8x3 sse3
-
-prototype void vp9_sad4x4x3 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad4x4x3 sse3
-
-prototype void vp9_sad64x64x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad64x64x8
-
-prototype void vp9_sad32x32x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad32x32x8
-
-prototype void vp9_sad16x16x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad16x16x8 sse4
-
-prototype void vp9_sad16x8x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad16x8x8 sse4
-
-prototype void vp9_sad8x16x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad8x16x8 sse4
-
-prototype void vp9_sad8x8x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad8x8x8 sse4
-
-prototype void vp9_sad8x4x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
-specialize vp9_sad8x4x8
-
-prototype void vp9_sad4x8x8 "const uint8_t *src_ptr, int src_stride, const uint8_t *ref_ptr, int ref_stride, uint32_t *sad_array"
-specialize vp9_sad4x8x8
-
-prototype void vp9_sad4x4x8 "const uint8_t *src_ptr, int  src_stride, const uint8_t *ref_ptr, int  ref_stride, uint32_t *sad_array"
-specialize vp9_sad4x4x8 sse4
-
-prototype void vp9_sad64x64x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad64x64x4d sse2
-
-prototype void vp9_sad32x64x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad32x64x4d sse2
-
-prototype void vp9_sad64x32x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad64x32x4d sse2
-
-prototype void vp9_sad32x16x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad32x16x4d sse2
-
-prototype void vp9_sad16x32x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad16x32x4d sse2
-
-prototype void vp9_sad32x32x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad32x32x4d sse2
-
-prototype void vp9_sad16x16x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad16x16x4d sse2
-
-prototype void vp9_sad16x8x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad16x8x4d sse2
-
-prototype void vp9_sad8x16x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad8x16x4d sse2
-
-prototype void vp9_sad8x8x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad8x8x4d sse2
-
-# TODO(jingning): need to convert these 4x8/8x4 functions into sse2 form
-prototype void vp9_sad8x4x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
-specialize vp9_sad8x4x4d sse2
-
-prototype void vp9_sad4x8x4d "const uint8_t *src_ptr, int src_stride, const uint8_t* const ref_ptr[], int ref_stride, unsigned int *sad_array"
-specialize vp9_sad4x8x4d sse
-
-prototype void vp9_sad4x4x4d "const uint8_t *src_ptr, int  src_stride, const uint8_t* const ref_ptr[], int  ref_stride, unsigned int *sad_array"
-specialize vp9_sad4x4x4d sse
-
-#prototype unsigned int vp9_sub_pixel_mse16x16 "const uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, const uint8_t *dst_ptr, int dst_pixels_per_line, unsigned int *sse"
-#specialize vp9_sub_pixel_mse16x16 sse2 mmx
-
-prototype unsigned int vp9_mse16x16 "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse"
-specialize vp9_mse16x16 mmx $sse2_x86inc $avx2_x86inc
-
-prototype unsigned int vp9_mse8x16 "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse"
-specialize vp9_mse8x16
-
-prototype unsigned int vp9_mse16x8 "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse"
-specialize vp9_mse16x8
-
-prototype unsigned int vp9_mse8x8 "const uint8_t *src_ptr, int  source_stride, const uint8_t *ref_ptr, int  recon_stride, unsigned int *sse"
-specialize vp9_mse8x8
-
-prototype unsigned int vp9_sub_pixel_mse64x64 "const uint8_t *src_ptr, int  source_stride, int  xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_mse64x64
-
-prototype unsigned int vp9_sub_pixel_mse32x32 "const uint8_t *src_ptr, int  source_stride, int  xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
-specialize vp9_sub_pixel_mse32x32
-
-prototype unsigned int vp9_get_mb_ss "const int16_t *"
-specialize vp9_get_mb_ss mmx sse2
-# ENCODEMB INVOKE
-
-prototype int64_t vp9_block_error "const int16_t *coeff, const int16_t *dqcoeff, intptr_t block_size, int64_t *ssz"
-specialize vp9_block_error $sse2_x86inc
-
-prototype void vp9_subtract_block "int rows, int cols, int16_t *diff_ptr, ptrdiff_t diff_stride, const uint8_t *src_ptr, ptrdiff_t src_stride, const uint8_t *pred_ptr, ptrdiff_t pred_stride"
-specialize vp9_subtract_block $sse2_x86inc
-
-prototype void vp9_quantize_b "const int16_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"
-specialize vp9_quantize_b $ssse3_x86_64
-
-prototype void vp9_quantize_b_32x32 "const int16_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, int16_t *qcoeff_ptr, int16_t *dqcoeff_ptr, const int16_t *dequant_ptr, int zbin_oq_value, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"
-specialize vp9_quantize_b_32x32 $ssse3_x86_64
-
-#
-# Structured Similarity (SSIM)
-#
-if [ "$CONFIG_INTERNAL_STATS" = "yes" ]; then
-    prototype void vp9_ssim_parms_8x8 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp9_ssim_parms_8x8 $sse2_x86_64
-
-    prototype void vp9_ssim_parms_16x16 "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"
-    specialize vp9_ssim_parms_16x16 $sse2_x86_64
-fi
-
-# fdct functions
-prototype void vp9_fht4x4 "const int16_t *input, int16_t *output, int stride, int tx_type"
-specialize vp9_fht4x4 sse2 avx2
-
-prototype void vp9_fht8x8 "const int16_t *input, int16_t *output, int stride, int tx_type"
-specialize vp9_fht8x8 sse2 avx2
-
-prototype void vp9_fht16x16 "const int16_t *input, int16_t *output, int stride, int tx_type"
-specialize vp9_fht16x16 sse2 avx2
-
-prototype void vp9_fwht4x4 "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fwht4x4
-
-prototype void vp9_fdct4x4 "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fdct4x4 sse2 avx2
-
-prototype void vp9_fdct8x8 "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fdct8x8 sse2 avx2
-
-prototype void vp9_fdct16x16 "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fdct16x16 sse2 avx2
-
-prototype void vp9_fdct32x32 "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fdct32x32 sse2 avx2
-
-prototype void vp9_fdct32x32_rd "const int16_t *input, int16_t *output, int stride"
-specialize vp9_fdct32x32_rd sse2 avx2
-
-#
-# Motion search
-#
-prototype int vp9_full_search_sad "const struct macroblock *x, const struct mv *ref_mv, int sad_per_bit, int distance, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv, struct mv *best_mv"
-specialize vp9_full_search_sad sse3 sse4_1
-vp9_full_search_sad_sse3=vp9_full_search_sadx3
-vp9_full_search_sad_sse4_1=vp9_full_search_sadx8
-
-prototype int vp9_refining_search_sad "const struct macroblock *x, struct mv *ref_mv, int sad_per_bit, int distance, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv"
-specialize vp9_refining_search_sad sse3
-vp9_refining_search_sad_sse3=vp9_refining_search_sadx4
-
-prototype int vp9_diamond_search_sad "const struct macroblock *x, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv"
-specialize vp9_diamond_search_sad sse3
-vp9_diamond_search_sad_sse3=vp9_diamond_search_sadx4
-
-prototype int vp9_full_range_search "const struct macroblock *x, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, const struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv"
-specialize vp9_full_range_search
-
-prototype void vp9_temporal_filter_apply "uint8_t *frame1, unsigned int stride, uint8_t *frame2, unsigned int block_size, int strength, int filter_weight, unsigned int *accumulator, uint16_t *count"
-specialize vp9_temporal_filter_apply sse2
-
-fi
-# end encoder functions
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index 8e454e6..b507c6e 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -415,6 +415,8 @@
                                     x->sadperbit16, &num00, &v_fn_ptr,
                                     x->nmvjointcost,
                                     x->mvcost, ref_mv);
+  if (tmp_err < INT_MAX)
+    tmp_err = vp9_get_mvpred_var(x, &tmp_mv, ref_mv, &v_fn_ptr, 1);
   if (tmp_err < INT_MAX - new_mv_mode_penalty)
     tmp_err += new_mv_mode_penalty;
 
@@ -439,6 +441,8 @@
                                         &num00, &v_fn_ptr,
                                         x->nmvjointcost,
                                         x->mvcost, ref_mv);
+      if (tmp_err < INT_MAX)
+        tmp_err = vp9_get_mvpred_var(x, &tmp_mv, ref_mv, &v_fn_ptr, 1);
       if (tmp_err < INT_MAX - new_mv_mode_penalty)
         tmp_err += new_mv_mode_penalty;
 
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index 7d6fd3b..d3a9977 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -98,42 +98,23 @@
 }
 
 void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride) {
-  int len;
-  int search_site_count = 0;
+  int len, ss_count = 1;
 
-  // Generate offsets for 4 search sites per step.
-  x->ss[search_site_count].mv.col = 0;
-  x->ss[search_site_count].mv.row = 0;
-  x->ss[search_site_count].offset = 0;
-  search_site_count++;
+  x->ss[0].mv.col = x->ss[0].mv.row = 0;
+  x->ss[0].offset = 0;
 
   for (len = MAX_FIRST_STEP; len > 0; len /= 2) {
-    // Compute offsets for search sites.
-    x->ss[search_site_count].mv.col = 0;
-    x->ss[search_site_count].mv.row = -len;
-    x->ss[search_site_count].offset = -len * stride;
-    search_site_count++;
-
-    // Compute offsets for search sites.
-    x->ss[search_site_count].mv.col = 0;
-    x->ss[search_site_count].mv.row = len;
-    x->ss[search_site_count].offset = len * stride;
-    search_site_count++;
-
-    // Compute offsets for search sites.
-    x->ss[search_site_count].mv.col = -len;
-    x->ss[search_site_count].mv.row = 0;
-    x->ss[search_site_count].offset = -len;
-    search_site_count++;
-
-    // Compute offsets for search sites.
-    x->ss[search_site_count].mv.col = len;
-    x->ss[search_site_count].mv.row = 0;
-    x->ss[search_site_count].offset = len;
-    search_site_count++;
+    // Generate offsets for 4 search sites per step.
+    const MV ss_mvs[] = {{-len, 0}, {len, 0}, {0, -len}, {0, len}};
+    int i;
+    for (i = 0; i < 4; ++i) {
+      search_site *const ss = &x->ss[ss_count++];
+      ss->mv = ss_mvs[i];
+      ss->offset = ss->mv.row * stride + ss->mv.col;
+    }
   }
 
-  x->ss_count = search_site_count;
+  x->ss_count = ss_count;
   x->searches_per_step = 4;
 }
 
@@ -728,16 +709,52 @@
                                best_mv->col;
   this_mv.row = best_mv->row * 8;
   this_mv.col = best_mv->col * 8;
-  if (bestsad == INT_MAX)
-    return INT_MAX;
-
-  return vfp->vf(what, what_stride, this_offset, in_what_stride,
-                 (unsigned int *)&bestsad) +
-         use_mvcost ? mv_err_cost(&this_mv, center_mv,
-                                  x->nmvjointcost, x->mvcost, x->errorperbit)
-                    : 0;
+  return bestsad;
 }
 
+int vp9_get_mvpred_var(const MACROBLOCK *x,
+                       MV *best_mv,
+                       const MV *center_mv,
+                       const vp9_variance_fn_ptr_t *vfp,
+                       int use_mvcost) {
+  unsigned int bestsad;
+  MV this_mv;
+  const MACROBLOCKD *const xd = &x->e_mbd;
+  const uint8_t *what = x->plane[0].src.buf;
+  const int what_stride = x->plane[0].src.stride;
+  const int in_what_stride = xd->plane[0].pre[0].stride;
+  const uint8_t *base_offset = xd->plane[0].pre[0].buf;
+  const uint8_t *this_offset = base_offset + (best_mv->row * in_what_stride) +
+      best_mv->col;
+  this_mv.row = best_mv->row * 8;
+  this_mv.col = best_mv->col * 8;
+  return vfp->vf(what, what_stride, this_offset, in_what_stride, &bestsad) +
+      (use_mvcost ?  mv_err_cost(&this_mv, center_mv, x->nmvjointcost,
+                                 x->mvcost, x->errorperbit) : 0);
+}
+
+int vp9_get_mvpred_av_var(const MACROBLOCK *x,
+                          MV *best_mv,
+                          const MV *center_mv,
+                          const uint8_t *second_pred,
+                          const vp9_variance_fn_ptr_t *vfp,
+                          int use_mvcost) {
+  unsigned int bestsad;
+  MV this_mv;
+  const MACROBLOCKD *const xd = &x->e_mbd;
+  const uint8_t *what = x->plane[0].src.buf;
+  const int what_stride = x->plane[0].src.stride;
+  const int in_what_stride = xd->plane[0].pre[0].stride;
+  const uint8_t *base_offset = xd->plane[0].pre[0].buf;
+  const uint8_t *this_offset = base_offset + (best_mv->row * in_what_stride) +
+      best_mv->col;
+  this_mv.row = best_mv->row * 8;
+  this_mv.col = best_mv->col * 8;
+  return vfp->svaf(this_offset, in_what_stride, 0, 0, what, what_stride,
+                   &bestsad, second_pred) +
+      (use_mvcost ?  mv_err_cost(&this_mv, center_mv, x->nmvjointcost,
+                                 x->mvcost, x->errorperbit) : 0);
+}
 
 int vp9_hex_search(const MACROBLOCK *x,
                    MV *ref_mv,
@@ -855,182 +872,18 @@
                             square_num_candidates, square_candidates);
 };
 
-// Number of candidates in first hex search
-#define FIRST_HEX_CANDIDATES 6
-// Index of previous hex search's best match
-#define PRE_BEST_CANDIDATE 6
-// Number of candidates in following hex search
-#define NEXT_HEX_CANDIDATES 3
-// Number of candidates in refining search
-#define REFINE_CANDIDATES 4
-
 int vp9_fast_hex_search(const MACROBLOCK *x,
                         MV *ref_mv,
                         int search_param,
                         int sad_per_bit,
+                        int do_init_search,  // must be zero for fast_hex
                         const vp9_variance_fn_ptr_t *vfp,
                         int use_mvcost,
                         const MV *center_mv,
                         MV *best_mv) {
-  const MACROBLOCKD* const xd = &x->e_mbd;
-  static const MV hex[FIRST_HEX_CANDIDATES] = {
-    { -1, -2}, {1, -2}, {2, 0}, {1, 2}, { -1, 2}, { -2, 0}
-  };
-  static const MV next_chkpts[PRE_BEST_CANDIDATE][NEXT_HEX_CANDIDATES] = {
-    {{ -2, 0}, { -1, -2}, {1, -2}},
-    {{ -1, -2}, {1, -2}, {2, 0}},
-    {{1, -2}, {2, 0}, {1, 2}},
-    {{2, 0}, {1, 2}, { -1, 2}},
-    {{1, 2}, { -1, 2}, { -2, 0}},
-    {{ -1, 2}, { -2, 0}, { -1, -2}}
-  };
-  static const MV neighbors[REFINE_CANDIDATES] = {
-      {0, -1}, { -1, 0}, {1, 0}, {0, 1}
-  };
-  int i, j;
-
-  const uint8_t *what = x->plane[0].src.buf;
-  const int what_stride = x->plane[0].src.stride;
-  const int in_what_stride = xd->plane[0].pre[0].stride;
-  int br, bc;
-  MV this_mv;
-  unsigned int bestsad = 0x7fffffff;
-  unsigned int thissad;
-  const uint8_t *base_offset;
-  const uint8_t *this_offset;
-  int k = -1;
-  int best_site = -1;
-  const int max_hex_search = 512;
-  const int max_dia_search = 32;
-
-  const int *mvjsadcost = x->nmvjointsadcost;
-  int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]};
-
-  const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
-
-  // Adjust ref_mv to make sure it is within MV range
-  clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
-  br = ref_mv->row;
-  bc = ref_mv->col;
-
-  // Check the start point
-  base_offset = xd->plane[0].pre[0].buf;
-  this_offset = base_offset + (br * in_what_stride) + bc;
-  this_mv.row = br;
-  this_mv.col = bc;
-  bestsad = vfp->sdf(what, what_stride, this_offset, in_what_stride, 0x7fffffff)
-            + mvsad_err_cost(&this_mv, &fcenter_mv, mvjsadcost, mvsadcost,
-                             sad_per_bit);
-
-  // Initial 6-point hex search
-  if (check_bounds(x, br, bc, 2)) {
-    for (i = 0; i < FIRST_HEX_CANDIDATES; i++) {
-      this_mv.row = br + hex[i].row;
-      this_mv.col = bc + hex[i].col;
-      this_offset = base_offset + (this_mv.row * in_what_stride) + this_mv.col;
-      thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                         bestsad);
-      CHECK_BETTER
-    }
-  } else {
-    for (i = 0; i < FIRST_HEX_CANDIDATES; i++) {
-      this_mv.row = br + hex[i].row;
-      this_mv.col = bc + hex[i].col;
-      if (!is_mv_in(x, &this_mv))
-        continue;
-      this_offset = base_offset + (this_mv.row * in_what_stride) + this_mv.col;
-      thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                         bestsad);
-      CHECK_BETTER
-    }
-  }
-
-  // Continue hex search if we find a better match in first round
-  if (best_site != -1) {
-    br += hex[best_site].row;
-    bc += hex[best_site].col;
-    k = best_site;
-
-    // Allow search covering maximum MV range
-    for (j = 1; j < max_hex_search; j++) {
-      best_site = -1;
-
-      if (check_bounds(x, br, bc, 2)) {
-        for (i = 0; i < 3; i++) {
-          this_mv.row = br + next_chkpts[k][i].row;
-          this_mv.col = bc + next_chkpts[k][i].col;
-          this_offset = base_offset + (this_mv.row * in_what_stride) +
-              this_mv.col;
-          thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                             bestsad);
-          CHECK_BETTER
-        }
-      } else {
-        for (i = 0; i < 3; i++) {
-          this_mv.row = br + next_chkpts[k][i].row;
-          this_mv.col = bc + next_chkpts[k][i].col;
-          if (!is_mv_in(x, &this_mv))
-            continue;
-          this_offset = base_offset + (this_mv.row * in_what_stride) +
-              this_mv.col;
-          thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                             bestsad);
-          CHECK_BETTER
-        }
-      }
-
-      if (best_site == -1) {
-        break;
-      } else {
-        br += next_chkpts[k][best_site].row;
-        bc += next_chkpts[k][best_site].col;
-        k += 5 + best_site;
-        if (k >= 12) k -= 12;
-        else if (k >= 6) k -= 6;
-      }
-    }
-  }
-
-  // Check 4 1-away neighbors
-  for (j = 0; j < max_dia_search; j++) {
-    best_site = -1;
-
-    if (check_bounds(x, br, bc, 1)) {
-      for (i = 0; i < REFINE_CANDIDATES; i++) {
-        this_mv.row = br + neighbors[i].row;
-        this_mv.col = bc + neighbors[i].col;
-        this_offset = base_offset + (this_mv.row * in_what_stride) +
-            this_mv.col;
-        thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                           bestsad);
-        CHECK_BETTER
-      }
-    } else {
-      for (i = 0; i < REFINE_CANDIDATES; i++) {
-        this_mv.row = br + neighbors[i].row;
-        this_mv.col = bc + neighbors[i].col;
-        if (!is_mv_in(x, &this_mv))
-          continue;
-        this_offset = base_offset + (this_mv.row * in_what_stride) +
-            this_mv.col;
-        thissad = vfp->sdf(what, what_stride, this_offset, in_what_stride,
-                           bestsad);
-        CHECK_BETTER
-      }
-    }
-
-    if (best_site == -1) {
-      break;
-    } else {
-      br += neighbors[best_site].row;
-      bc += neighbors[best_site].col;
-    }
-  }
-
-  best_mv->row = br;
-  best_mv->col = bc;
-
-  return bestsad;
+  return vp9_hex_search(x, ref_mv, MAX(MAX_MVSEARCH_STEPS - 2, search_param),
+                        sad_per_bit, do_init_search, vfp, use_mvcost,
+                        center_mv, best_mv);
 }
 
 #undef CHECK_BETTER
@@ -1045,8 +898,6 @@
   const int what_stride = x->plane[0].src.stride;
   const uint8_t *in_what;
   const int in_what_stride = xd->plane[0].pre[0].stride;
-  const uint8_t *best_address;
-
   MV this_mv;
 
   unsigned int bestsad = INT_MAX;
@@ -1076,7 +927,6 @@
 
   // Work out the start point for the search
   in_what = xd->plane[0].pre[0].buf + ref_row * in_what_stride + ref_col;
-  best_address = in_what;
 
   // Check the starting position
   bestsad = fn_ptr->sdf(what, what_stride, in_what, in_what_stride, 0x7fffffff)
@@ -1134,20 +984,9 @@
       }
     }
   }
-
   best_mv->row += best_tr;
   best_mv->col += best_tc;
-
-  this_mv.row = best_mv->row * 8;
-  this_mv.col = best_mv->col * 8;
-
-  if (bestsad == INT_MAX)
-    return INT_MAX;
-
-  return fn_ptr->vf(what, what_stride, best_address, in_what_stride,
-                    (unsigned int *)(&thissad)) +
-                       mv_err_cost(&this_mv, center_mv,
-                                   mvjcost, mvcost, x->errorperbit);
+  return bestsad;
 }
 
 int vp9_diamond_search_sad_c(const MACROBLOCK *x,
@@ -1272,17 +1111,7 @@
       (*num00)++;
     }
   }
-
-  this_mv.row = best_mv->row * 8;
-  this_mv.col = best_mv->col * 8;
-
-  if (bestsad == INT_MAX)
-    return INT_MAX;
-
-  return fn_ptr->vf(what, what_stride, best_address, in_what_stride,
-                    (unsigned int *)(&thissad)) +
-                       mv_err_cost(&this_mv, center_mv,
-                                   mvjcost, mvcost, x->errorperbit);
+  return bestsad;
 }
 
 int vp9_diamond_search_sadx4(const MACROBLOCK *x,
@@ -1448,24 +1277,14 @@
       (*num00)++;
     }
   }
-
-  this_mv.row = best_mv->row * 8;
-  this_mv.col = best_mv->col * 8;
-
-  if (bestsad == INT_MAX)
-    return INT_MAX;
-
-  return fn_ptr->vf(what, what_stride, best_address, in_what_stride,
-                    (unsigned int *)(&thissad)) +
-                    mv_err_cost(&this_mv, center_mv,
-                                mvjcost, mvcost, x->errorperbit);
+  return bestsad;
 }
 
 /* do_refine: If last step (1-away) of n-step search doesn't pick the center
               point as the best match, we will do a final 1-away diamond
               refining search  */
 
-int vp9_full_pixel_diamond(VP9_COMP *cpi, MACROBLOCK *x,
+int vp9_full_pixel_diamond(const VP9_COMP *cpi, MACROBLOCK *x,
                            MV *mvp_full, int step_param,
                            int sadpb, int further_steps, int do_refine,
                            const vp9_variance_fn_ptr_t *fn_ptr,
@@ -1476,6 +1295,8 @@
                                         step_param, sadpb, &n,
                                         fn_ptr, x->nmvjointcost,
                                         x->mvcost, ref_mv);
+  if (bestsme < INT_MAX)
+    bestsme = vp9_get_mvpred_var(x, &temp_mv, ref_mv, fn_ptr, 1);
   *dst_mv = temp_mv;
 
   // If there won't be more n-step search, check to see if refining search is
@@ -1493,6 +1314,8 @@
                                         step_param + n, sadpb, &num00,
                                         fn_ptr, x->nmvjointcost, x->mvcost,
                                         ref_mv);
+      if (thissme < INT_MAX)
+        thissme = vp9_get_mvpred_var(x, &temp_mv, ref_mv, fn_ptr, 1);
 
       // check to see if refining search is needed.
       if (num00 > further_steps - n)
@@ -1512,12 +1335,13 @@
     thissme = cpi->refining_search_sad(x, &best_mv, sadpb, search_range,
                                        fn_ptr, x->nmvjointcost, x->mvcost,
                                        ref_mv);
+    if (thissme < INT_MAX)
+      thissme = vp9_get_mvpred_var(x, &best_mv, ref_mv, fn_ptr, 1);
     if (thissme < bestsme) {
       bestsme = thissme;
       *dst_mv = best_mv;
     }
   }
-
   return bestsme;
 }
 
@@ -1562,15 +1386,7 @@
       }
     }
   }
-
-  if (best_sad < INT_MAX) {
-    unsigned int unused;
-    const MV mv = {best_mv->row * 8, best_mv->col * 8};
-    return fn_ptr->vf(what, what_stride, best_address, in_what_stride, &unused)
-                + mv_err_cost(&mv, center_mv, mvjcost, mvcost, x->errorperbit);
-  } else {
-    return INT_MAX;
-  }
+  return best_sad;
 }
 
 int vp9_full_search_sadx3(const MACROBLOCK *x, const MV *ref_mv,
@@ -1665,17 +1481,7 @@
       c++;
     }
   }
-
-  this_mv.row = best_mv->row * 8;
-  this_mv.col = best_mv->col * 8;
-
-  if (bestsad < INT_MAX)
-    return fn_ptr->vf(what, what_stride, bestaddress, in_what_stride,
-                      (unsigned int *)(&thissad)) +
-                      mv_err_cost(&this_mv, center_mv,
-                                  mvjcost, mvcost, x->errorperbit);
-  else
-    return INT_MAX;
+  return bestsad;
 }
 
 int vp9_full_search_sadx8(const MACROBLOCK *x, const MV *ref_mv,
@@ -1798,17 +1604,7 @@
       c++;
     }
   }
-
-  this_mv.row = best_mv->row * 8;
-  this_mv.col = best_mv->col * 8;
-
-  if (bestsad < INT_MAX)
-    return fn_ptr->vf(what, what_stride, bestaddress, in_what_stride,
-                      (unsigned int *)(&thissad)) +
-                      mv_err_cost(&this_mv, center_mv,
-                                  mvjcost, mvcost, x->errorperbit);
-  else
-    return INT_MAX;
+  return bestsad;
 }
 
 int vp9_refining_search_sad_c(const MACROBLOCK *x,
@@ -1866,16 +1662,7 @@
       best_address = &in_what[ref_mv->row * in_what_stride + ref_mv->col];
     }
   }
-
-  if (bestsad < INT_MAX) {
-    unsigned int unused;
-    const MV mv = {ref_mv->row * 8, ref_mv->col * 8};
-    return fn_ptr->vf(what, what_stride, best_address, in_what_stride,
-                      &unused) +
-        mv_err_cost(&mv, center_mv, mvjcost, mvcost, x->errorperbit);
-  } else {
-    return INT_MAX;
-  }
+  return bestsad;
 }
 
 int vp9_refining_search_sadx4(const MACROBLOCK *x,
@@ -1977,17 +1764,7 @@
                       neighbors[best_site].col;
     }
   }
-
-  this_mv.row = ref_mv->row * 8;
-  this_mv.col = ref_mv->col * 8;
-
-  if (bestsad < INT_MAX)
-    return fn_ptr->vf(what, what_stride, best_address, in_what_stride,
-                      (unsigned int *)(&thissad)) +
-                      mv_err_cost(&this_mv, center_mv,
-                                  mvjcost, mvcost, x->errorperbit);
-  else
-    return INT_MAX;
+  return bestsad;
 }
 
 // This function is called when we do joint motion search in comp_inter_inter
@@ -2055,18 +1832,5 @@
       best_address = &in_what[ref_mv->row * in_what_stride + ref_mv->col];
     }
   }
-
-  this_mv.row = ref_mv->row * 8;
-  this_mv.col = ref_mv->col * 8;
-
-  if (bestsad < INT_MAX) {
-    // FIXME(rbultje, yunqing): add full-pixel averaging variance functions
-    // so we don't have to use the subpixel with xoff=0,yoff=0 here.
-    return fn_ptr->svaf(best_address, in_what_stride, 0, 0, what, what_stride,
-                        (unsigned int *)(&thissad), second_pred) +
-                        mv_err_cost(&this_mv, center_mv,
-                                    mvjcost, mvcost, x->errorperbit);
-  } else {
-    return INT_MAX;
-  }
+  return bestsad;
 }
diff --git a/vp9/encoder/vp9_mcomp.h b/vp9/encoder/vp9_mcomp.h
index 586a74c..39360f1 100644
--- a/vp9/encoder/vp9_mcomp.h
+++ b/vp9/encoder/vp9_mcomp.h
@@ -35,6 +35,19 @@
 void vp9_set_mv_search_range(MACROBLOCK *x, const MV *mv);
 int vp9_mv_bit_cost(const MV *mv, const MV *ref,
                     const int *mvjcost, int *mvcost[2], int weight);
+
+// Utility to compute variance + MV rate cost for a given MV
+int vp9_get_mvpred_var(const MACROBLOCK *x,
+                       MV *best_mv,
+                       const MV *center_mv,
+                       const vp9_variance_fn_ptr_t *vfp,
+                       int use_mvcost);
+int vp9_get_mvpred_av_var(const MACROBLOCK *x,
+                          MV *best_mv,
+                          const MV *center_mv,
+                          const uint8_t *second_pred,
+                          const vp9_variance_fn_ptr_t *vfp,
+                          int use_mvcost);
 void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride);
 void vp9_init3smotion_compensation(MACROBLOCK *x,  int stride);
 
@@ -42,47 +55,27 @@
 int vp9_init_search_range(struct VP9_COMP *cpi, int size);
 
 // Runs sequence of diamond searches in smaller steps for RD
-int vp9_full_pixel_diamond(struct VP9_COMP *cpi, MACROBLOCK *x,
+int vp9_full_pixel_diamond(const struct VP9_COMP *cpi, MACROBLOCK *x,
                            MV *mvp_full, int step_param,
                            int sadpb, int further_steps, int do_refine,
                            const vp9_variance_fn_ptr_t *fn_ptr,
                            const MV *ref_mv, MV *dst_mv);
 
-int vp9_hex_search(const MACROBLOCK *x,
-                   MV *ref_mv,
-                   int search_param,
-                   int error_per_bit,
-                   int do_init_search,
-                   const vp9_variance_fn_ptr_t *vf,
-                   int use_mvcost,
-                   const MV *center_mv,
-                   MV *best_mv);
-int vp9_bigdia_search(const MACROBLOCK *x,
-                      MV *ref_mv,
-                      int search_param,
-                      int error_per_bit,
-                      int do_init_search,
-                      const vp9_variance_fn_ptr_t *vf,
-                      int use_mvcost,
-                      const MV *center_mv,
-                      MV *best_mv);
-int vp9_square_search(const MACROBLOCK *x,
-                      MV *ref_mv,
-                      int search_param,
-                      int error_per_bit,
-                      int do_init_search,
-                      const vp9_variance_fn_ptr_t *vf,
-                      int use_mvcost,
-                      const MV *center_mv,
-                      MV *best_mv);
-int vp9_fast_hex_search(const MACROBLOCK *x,
-                        MV *ref_mv,
-                        int search_param,
-                        int sad_per_bit,
-                        const vp9_variance_fn_ptr_t *vfp,
-                        int use_mvcost,
-                        const MV *center_mv,
-                        MV *best_mv);
+typedef int (integer_mv_pattern_search_fn) (
+    const MACROBLOCK *x,
+    MV *ref_mv,
+    int search_param,
+    int error_per_bit,
+    int do_init_search,
+    const vp9_variance_fn_ptr_t *vf,
+    int use_mvcost,
+    const MV *center_mv,
+    MV *best_mv);
+
+integer_mv_pattern_search_fn vp9_hex_search;
+integer_mv_pattern_search_fn vp9_bigdia_search;
+integer_mv_pattern_search_fn vp9_square_search;
+integer_mv_pattern_search_fn vp9_fast_hex_search;
 
 typedef int (fractional_mv_step_fp) (
     const MACROBLOCK *x,
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 9ba48a1..75122bc 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -34,7 +34,7 @@
   MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
   struct buf_2d backup_yv12[MAX_MB_PLANE] = {{0}};
   int bestsme = INT_MAX;
-  int further_steps, step_param;
+  int step_param;
   int sadpb = x->sadperbit16;
   MV mvp_full;
   int ref = mbmi->ref_frame[0];
@@ -67,7 +67,6 @@
   // TODO(jingning) exploiting adaptive motion search control in non-RD
   // mode decision too.
   step_param = 6;
-  further_steps = (cpi->sf.max_step_search_steps - 1) - step_param;
 
   for (i = LAST_FRAME; i <= LAST_FRAME && cpi->common.show_frame; ++i) {
     if ((x->pred_mv_sad[ref] >> 3) > x->pred_mv_sad[i]) {
@@ -88,22 +87,28 @@
   mvp_full.row >>= 3;
 
   if (cpi->sf.search_method == FAST_HEX) {
-    bestsme = vp9_fast_hex_search(x, &mvp_full, step_param, sadpb,
+    // NOTE: this returns SAD
+    bestsme = vp9_fast_hex_search(x, &mvp_full, step_param, sadpb, 0,
                                   &cpi->fn_ptr[bsize], 1,
                                   &ref_mv.as_mv, &tmp_mv->as_mv);
   } else if (cpi->sf.search_method == HEX) {
+    // NOTE: this returns SAD
     bestsme = vp9_hex_search(x, &mvp_full, step_param, sadpb, 1,
                              &cpi->fn_ptr[bsize], 1,
                              &ref_mv.as_mv, &tmp_mv->as_mv);
   } else if (cpi->sf.search_method == SQUARE) {
+    // NOTE: this returns SAD
     bestsme = vp9_square_search(x, &mvp_full, step_param, sadpb, 1,
                                 &cpi->fn_ptr[bsize], 1,
                                 &ref_mv.as_mv, &tmp_mv->as_mv);
   } else if (cpi->sf.search_method == BIGDIA) {
+    // NOTE: this returns SAD
     bestsme = vp9_bigdia_search(x, &mvp_full, step_param, sadpb, 1,
                                 &cpi->fn_ptr[bsize], 1,
                                 &ref_mv.as_mv, &tmp_mv->as_mv);
   } else {
+    int further_steps = (cpi->sf.max_step_search_steps - 1) - step_param;
+    // NOTE: this returns variance
     bestsme = vp9_full_pixel_diamond(cpi, x, &mvp_full, step_param,
                                      sadpb, further_steps, 1,
                                      &cpi->fn_ptr[bsize],
diff --git a/vp9/encoder/vp9_ratectrl.c b/vp9/encoder/vp9_ratectrl.c
index 89aa821..8430e4b 100644
--- a/vp9/encoder/vp9_ratectrl.c
+++ b/vp9/encoder/vp9_ratectrl.c
@@ -587,6 +587,7 @@
                                                    q_adj_factor);
     }
   } else if (!rc->is_src_frame_alt_ref &&
+             !cpi->use_svc &&
              (cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame)) {
     // Use the lower of active_worst_quality and recent
     // average Q as basis for GF/ARF best Q limit unless last frame was
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 810f8ba..a6f54fd 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1848,18 +1848,30 @@
                                      sadpb, 1, v_fn_ptr, 1,
                                      &bsi->ref_mv[0]->as_mv,
                                      &new_mv->as_mv);
+            if (bestsme < INT_MAX)
+              bestsme = vp9_get_mvpred_var(x, &new_mv->as_mv,
+                                           &bsi->ref_mv[0]->as_mv,
+                                           v_fn_ptr, 1);
           } else if (cpi->sf.search_method == SQUARE) {
             bestsme = vp9_square_search(x, &mvp_full,
                                         step_param,
                                         sadpb, 1, v_fn_ptr, 1,
                                         &bsi->ref_mv[0]->as_mv,
                                         &new_mv->as_mv);
+            if (bestsme < INT_MAX)
+              bestsme = vp9_get_mvpred_var(x, &new_mv->as_mv,
+                                           &bsi->ref_mv[0]->as_mv,
+                                           v_fn_ptr, 1);
           } else if (cpi->sf.search_method == BIGDIA) {
             bestsme = vp9_bigdia_search(x, &mvp_full,
                                         step_param,
                                         sadpb, 1, v_fn_ptr, 1,
                                         &bsi->ref_mv[0]->as_mv,
                                         &new_mv->as_mv);
+            if (bestsme < INT_MAX)
+              bestsme = vp9_get_mvpred_var(x, &new_mv->as_mv,
+                                           &bsi->ref_mv[0]->as_mv,
+                                           v_fn_ptr, 1);
           } else {
             bestsme = vp9_full_pixel_diamond(cpi, x, &mvp_full, step_param,
                                              sadpb, further_steps, 0, v_fn_ptr,
@@ -2470,21 +2482,33 @@
   further_steps = (cpi->sf.max_step_search_steps - 1) - step_param;
 
   if (cpi->sf.search_method == FAST_HEX) {
-    bestsme = vp9_fast_hex_search(x, &mvp_full, step_param, sadpb,
+    bestsme = vp9_fast_hex_search(x, &mvp_full, step_param, sadpb, 0,
                                   &cpi->fn_ptr[bsize], 1,
                                   &ref_mv, &tmp_mv->as_mv);
+    if (bestsme < INT_MAX)
+      bestsme = vp9_get_mvpred_var(x, &tmp_mv->as_mv, &ref_mv,
+                                   &cpi->fn_ptr[bsize], 1);
   } else if (cpi->sf.search_method == HEX) {
     bestsme = vp9_hex_search(x, &mvp_full, step_param, sadpb, 1,
                              &cpi->fn_ptr[bsize], 1,
                              &ref_mv, &tmp_mv->as_mv);
+    if (bestsme < INT_MAX)
+      bestsme = vp9_get_mvpred_var(x, &tmp_mv->as_mv, &ref_mv,
+                                   &cpi->fn_ptr[bsize], 1);
   } else if (cpi->sf.search_method == SQUARE) {
     bestsme = vp9_square_search(x, &mvp_full, step_param, sadpb, 1,
                                 &cpi->fn_ptr[bsize], 1,
                                 &ref_mv, &tmp_mv->as_mv);
+    if (bestsme < INT_MAX)
+      bestsme = vp9_get_mvpred_var(x, &tmp_mv->as_mv, &ref_mv,
+                                   &cpi->fn_ptr[bsize], 1);
   } else if (cpi->sf.search_method == BIGDIA) {
     bestsme = vp9_bigdia_search(x, &mvp_full, step_param, sadpb, 1,
                                 &cpi->fn_ptr[bsize], 1,
                                 &ref_mv, &tmp_mv->as_mv);
+    if (bestsme < INT_MAX)
+      bestsme = vp9_get_mvpred_var(x, &tmp_mv->as_mv, &ref_mv,
+                                   &cpi->fn_ptr[bsize], 1);
   } else {
     bestsme = vp9_full_pixel_diamond(cpi, x, &mvp_full, step_param,
                                      sadpb, further_steps, 1,
@@ -2610,6 +2634,9 @@
                                        x->nmvjointcost, x->mvcost,
                                        &ref_mv[id].as_mv, second_pred,
                                        pw, ph);
+    if (bestsme < INT_MAX)
+      bestsme = vp9_get_mvpred_av_var(x, &tmp_mv.as_mv, &ref_mv[id].as_mv,
+                                      second_pred, &cpi->fn_ptr[bsize], 1);
 
     x->mv_col_min = tmp_col_min;
     x->mv_col_max = tmp_col_max;
diff --git a/vp9/vp9_common.mk b/vp9/vp9_common.mk
index 9fb6115..b4ab99b 100644
--- a/vp9/vp9_common.mk
+++ b/vp9/vp9_common.mk
@@ -45,7 +45,7 @@
 VP9_COMMON_SRCS-yes += common/vp9_reconinter.h
 VP9_COMMON_SRCS-yes += common/vp9_reconintra.h
 VP9_COMMON_SRCS-yes += common/vp9_rtcd.c
-VP9_COMMON_SRCS-yes += common/vp9_rtcd_defs.sh
+VP9_COMMON_SRCS-yes += common/vp9_rtcd_defs.pl
 VP9_COMMON_SRCS-yes += common/vp9_scale.h
 VP9_COMMON_SRCS-yes += common/vp9_scale.c
 VP9_COMMON_SRCS-yes += common/vp9_seg_common.h
@@ -145,4 +145,4 @@
 VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_save_reg_neon$(ASM)
 VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_reconintra_neon$(ASM)
 
-$(eval $(call rtcd_h_template,vp9_rtcd,vp9/common/vp9_rtcd_defs.sh))
+$(eval $(call rtcd_h_template,vp9_rtcd,vp9/common/vp9_rtcd_defs.pl))
diff --git a/vpx_scale/vpx_scale.mk b/vpx_scale/vpx_scale.mk
index 50d3e9d..51a0ec9 100644
--- a/vpx_scale/vpx_scale.mk
+++ b/vpx_scale/vpx_scale.mk
@@ -7,7 +7,7 @@
 SCALE_SRCS-$(CONFIG_SPATIAL_RESAMPLING) += generic/gen_scalers.c
 SCALE_SRCS-yes += vpx_scale_asm_offsets.c
 SCALE_SRCS-yes += vpx_scale_rtcd.c
-SCALE_SRCS-yes += vpx_scale_rtcd.sh
+SCALE_SRCS-yes += vpx_scale_rtcd.pl
 
 #neon
 SCALE_SRCS-$(HAVE_NEON)  += arm/neon/vp8_vpxyv12_copyframe_func_neon$(ASM)
@@ -24,4 +24,4 @@
 $(eval $(call asm_offsets_template,\
 	         vpx_scale_asm_offsets.asm, vpx_scale/vpx_scale_asm_offsets.c))
 
-$(eval $(call rtcd_h_template,vpx_scale_rtcd,vpx_scale/vpx_scale_rtcd.sh))
+$(eval $(call rtcd_h_template,vpx_scale_rtcd,vpx_scale/vpx_scale_rtcd.pl))
diff --git a/vpx_scale/vpx_scale_rtcd.pl b/vpx_scale/vpx_scale_rtcd.pl
new file mode 100644
index 0000000..28e168e
--- /dev/null
+++ b/vpx_scale/vpx_scale_rtcd.pl
@@ -0,0 +1,35 @@
+sub vpx_scale_forward_decls() {
+print <<EOF
+struct yv12_buffer_config;
+EOF
+}
+forward_decls qw/vpx_scale_forward_decls/;
+
+# Scaler functions
+if (vpx_config("CONFIG_SPATIAL_RESAMPLING") eq "yes") {
+    add_proto qw/void vp8_horizontal_line_5_4_scale/, "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width";
+    add_proto qw/void vp8_vertical_band_5_4_scale/, "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width";
+    add_proto qw/void vp8_horizontal_line_5_3_scale/, "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width";
+    add_proto qw/void vp8_vertical_band_5_3_scale/, "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width";
+    add_proto qw/void vp8_horizontal_line_2_1_scale/, "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width";
+    add_proto qw/void vp8_vertical_band_2_1_scale/, "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width";
+    add_proto qw/void vp8_vertical_band_2_1_scale_i/, "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width";
+}
+
+add_proto qw/void vp8_yv12_extend_frame_borders/, "struct yv12_buffer_config *ybf";
+specialize qw/vp8_yv12_extend_frame_borders neon/;
+
+add_proto qw/void vp8_yv12_copy_frame/, "const struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc";
+specialize qw/vp8_yv12_copy_frame neon/;
+
+add_proto qw/void vpx_yv12_copy_y/, "const struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc";
+specialize qw/vpx_yv12_copy_y neon/;
+
+if (vpx_config("CONFIG_VP9") eq "yes") {
+    add_proto qw/void vp9_extend_frame_borders/, "struct yv12_buffer_config *ybf";
+    specialize qw/vp9_extend_frame_borders dspr2/;
+
+    add_proto qw/void vp9_extend_frame_inner_borders/, "struct yv12_buffer_config *ybf";
+    specialize qw/vp9_extend_frame_inner_borders dspr2/;
+}
+1;
diff --git a/vpx_scale/vpx_scale_rtcd.sh b/vpx_scale/vpx_scale_rtcd.sh
deleted file mode 100644
index c26208c..0000000
--- a/vpx_scale/vpx_scale_rtcd.sh
+++ /dev/null
@@ -1,34 +0,0 @@
-vpx_scale_forward_decls() {
-cat <<EOF
-struct yv12_buffer_config;
-EOF
-}
-forward_decls vpx_scale_forward_decls
-
-# Scaler functions
-if [ "$CONFIG_SPATIAL_RESAMPLING" = "yes" ]; then
-    prototype void vp8_horizontal_line_5_4_scale "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width"
-    prototype void vp8_vertical_band_5_4_scale "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width"
-    prototype void vp8_horizontal_line_5_3_scale "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width"
-    prototype void vp8_vertical_band_5_3_scale "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width"
-    prototype void vp8_horizontal_line_2_1_scale "const unsigned char *source, unsigned int source_width, unsigned char *dest, unsigned int dest_width"
-    prototype void vp8_vertical_band_2_1_scale "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width"
-    prototype void vp8_vertical_band_2_1_scale_i "unsigned char *source, unsigned int src_pitch, unsigned char *dest, unsigned int dest_pitch, unsigned int dest_width"
-fi
-
-prototype void vp8_yv12_extend_frame_borders "struct yv12_buffer_config *ybf"
-specialize vp8_yv12_extend_frame_borders neon
-
-prototype void vp8_yv12_copy_frame "const struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc"
-specialize vp8_yv12_copy_frame neon
-
-prototype void vpx_yv12_copy_y "const struct yv12_buffer_config *src_ybc, struct yv12_buffer_config *dst_ybc"
-specialize vpx_yv12_copy_y neon
-
-if [ "$CONFIG_VP9" = "yes" ]; then
-    prototype void vp9_extend_frame_borders "struct yv12_buffer_config *ybf"
-    specialize vp9_extend_frame_borders dspr2
-
-    prototype void vp9_extend_frame_inner_borders "struct yv12_buffer_config *ybf"
-    specialize vp9_extend_frame_inner_borders dspr2
-fi