SSSE3 idct8x8 functions for highbitdpeth build

This commit changes SSSE3 optimized idct8x8 functions to work with
highbitdepth build.

With this commit and the previous one that enabled SSSE3 idct32x32
functions, tests showed virtually no difference on decoding speed for
file fdJc1_IBKJA.248.webm for the build with -enable-vp9-highbitdpeth
option and the build without the option.

Change-Id: Ibe0634149ec70e8b921e6b30171664b8690a9c45

diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl
index d3c4971..9abbe91 100644 (file)
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -762,7 +762,7 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
 
     add_proto qw/void vpx_idct32x32_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
     specialize qw/vpx_idct32x32_1_add/;
-    
+
     add_proto qw/void vpx_highbd_idct4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int bd";
     specialize qw/vpx_highbd_idct4x4_16_add/;
 
@@ -785,10 +785,10 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
     specialize qw/vpx_idct4x4_1_add sse2/;
 
     add_proto qw/void vpx_idct8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
-    specialize qw/vpx_idct8x8_64_add sse2/;
+    specialize qw/vpx_idct8x8_64_add sse2/, "$ssse3_x86_64_x86inc";
 
     add_proto qw/void vpx_idct8x8_12_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
-    specialize qw/vpx_idct8x8_12_add sse2/;
+    specialize qw/vpx_idct8x8_12_add sse2/, "$ssse3_x86_64_x86inc";
 
     add_proto qw/void vpx_idct8x8_1_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride";
     specialize qw/vpx_idct8x8_1_add sse2/;

diff --git a/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm b/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
index 7cb9a41..20baf82 100644 (file)
--- a/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
+++ b/vpx_dsp/x86/inv_txfm_ssse3_x86_64.asm
@@ -220,7 +220,24 @@ cglobal idct8x8_64_add, 3, 5, 13, input, output, stride
   mova    m12, [pw_11585x2]
 
   lea      r3, [2 * strideq]
-
+%if CONFIG_VP9_HIGHBITDEPTH
+  mova     m0, [inputq +   0]
+  packssdw m0, [inputq +  16]
+  mova     m1, [inputq +  32]
+  packssdw m1, [inputq +  48]
+  mova     m2, [inputq +  64]
+  packssdw m2, [inputq +  80]
+  mova     m3, [inputq +  96]
+  packssdw m3, [inputq + 112]
+  mova     m4, [inputq + 128]
+  packssdw m4, [inputq + 144]
+  mova     m5, [inputq + 160]
+  packssdw m5, [inputq + 176]
+  mova     m6, [inputq + 192]
+  packssdw m6, [inputq + 208]
+  mova     m7, [inputq + 224]
+  packssdw m7, [inputq + 240]
+%else
   mova     m0, [inputq +   0]
   mova     m1, [inputq +  16]
   mova     m2, [inputq +  32]
@@ -229,7 +246,7 @@ cglobal idct8x8_64_add, 3, 5, 13, input, output, stride
   mova     m5, [inputq +  80]
   mova     m6, [inputq +  96]
   mova     m7, [inputq + 112]
-
+%endif
   TRANSPOSE8X8  0, 1, 2, 3, 4, 5, 6, 7, 9
   IDCT8_1D
   TRANSPOSE8X8  0, 1, 2, 3, 4, 5, 6, 7, 9
@@ -254,10 +271,21 @@ cglobal idct8x8_12_add, 3, 5, 13, input, output, stride
 
   lea        r3, [2 * strideq]
 
+%if CONFIG_VP9_HIGHBITDEPTH
+  mova       m0, [inputq +   0]
+  packssdw   m0, [inputq +  16]
+  mova       m1, [inputq +  32]
+  packssdw   m1, [inputq +  48]
+  mova       m2, [inputq +  64]
+  packssdw   m2, [inputq +  80]
+  mova       m3, [inputq +  96]
+  packssdw   m3, [inputq + 112]
+%else
   mova       m0, [inputq +  0]
   mova       m1, [inputq + 16]
   mova       m2, [inputq + 32]
   mova       m3, [inputq + 48]
+%endif
 
   punpcklwd  m0, m1
   punpcklwd  m2, m3