x86: Don't use explicitly aligned versions of SAD on AVX CPUs

On modern CPUs movdqu isn't slower than movdqa when used on aligned data and using the same code in both cases saves cache.

This was already done for the high bit-depth AVX2 implementation but the aligned version still exists as dead code so remove that.

diff --git a/common/pixel.c b/common/pixel.c
index 2b5e4317bbcc294112d5f83614f6d877abf527ad..f5db938bbae72cd624ed080645fad7f926c76ce2 100644 (file)
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -966,6 +966,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     }
     if( cpu&X264_CPU_AVX )
     {
+        INIT5_NAME( sad_aligned, sad, _ssse3 ); /* AVX-capable CPUs doesn't benefit from an aligned version */
         INIT_ADS( _avx );
         INIT6( satd, _avx );
         pixf->satd[PIXEL_4x16] = x264_pixel_satd_4x16_avx;
@@ -1234,6 +1235,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
 
     if( cpu&X264_CPU_AVX )
     {
+        INIT2_NAME( sad_aligned, sad, _sse2 ); /* AVX-capable CPUs doesn't benefit from an aligned version */
         INIT8( satd, _avx );
         INIT7( satd_x3, _avx );
         INIT7( satd_x4, _avx );

diff --git a/common/x86/pixel.h b/common/x86/pixel.h
index 0571d74d8af38725dd6ad457c24d4ad7b1d0d623..aca996ad0334f186d20ab29e4f1a37140559d3c7 100644 (file)
--- a/common/x86/pixel.h
+++ b/common/x86/pixel.h
@@ -53,7 +53,6 @@ DECL_X1( sad, sse2_aligned )
 DECL_X1( sad, ssse3 )
 DECL_X1( sad, ssse3_aligned )
 DECL_X1( sad, avx2 )
-DECL_X1( sad, avx2_aligned )
 DECL_X4( sad, mmx2 )
 DECL_X4( sad, sse2 )
 DECL_X4( sad, sse3 )

diff --git a/common/x86/sad16-a.asm b/common/x86/sad16-a.asm
index 8c48723a6a486c369a51a62c1ec36da747e62495..7606da381596d1c103acda5257ba2f0597d694ea 100644 (file)
--- a/common/x86/sad16-a.asm
+++ b/common/x86/sad16-a.asm
@@ -205,9 +205,6 @@ SAD  8,  8
 INIT_YMM avx2
 SAD 16, 16
 SAD 16,  8
-INIT_YMM avx2, aligned
-SAD 16, 16
-SAD 16,  8
 
 ;=============================================================================
 ; SAD x3/x4