[x265] [PATCH 2 of 8] asm: interp_4tap_horiz_pp_2x8_sse3

[dtyx265 at gmail.com]

# HG changeset patch
# User David T Yuen <dtyx265 at gmail.com>
# Date 1429287024 25200
# Node ID 3d43e6ad9a958794a2809fe46e949539c28523ce
# Parent  34b0ec5fbc8695afd640821f8a9aba0e30a4253a
asm: interp_4tap_horiz_pp_2x8_sse3

This replaces c code.

64-bit

./test/TestBench --testbench interp | grep "chroma_hpp\[  2x8\]"
chroma_hpp[  2x8]	1.83x 	 1104.89  	 2026.66
chroma_hpp[  2x8]	1.84x 	 1102.43  	 2025.46

32-bit

./test/TestBench --testbench interp | grep "chroma_hpp\[  2x8\]"
chroma_hpp[  2x8]	1.93x 	 1252.42  	 2419.18
chroma_hpp[  2x8]	1.93x 	 1252.46  	 2417.46

diff -r 34b0ec5fbc86 -r 3d43e6ad9a95 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Fri Apr 17 09:04:59 2015 -0700
+++ b/source/common/x86/asm-primitives.cpp	Fri Apr 17 09:10:24 2015 -0700
@@ -1357,6 +1357,8 @@
     if (cpuMask & X265_CPU_SSE3)
     {
         p.chroma[X265_CSP_I420].pu[CHROMA_420_2x4].filter_hpp = x265_interp_4tap_horiz_pp_2x4_sse3;
+        p.chroma[X265_CSP_I420].pu[CHROMA_420_2x8].filter_hpp = x265_interp_4tap_horiz_pp_2x8_sse3;
+        p.chroma[X265_CSP_I422].pu[CHROMA_422_2x8].filter_hpp = x265_interp_4tap_horiz_pp_2x8_sse3;
     }
     if (cpuMask & X265_CPU_SSSE3)
     {
diff -r 34b0ec5fbc86 -r 3d43e6ad9a95 source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm	Fri Apr 17 09:04:59 2015 -0700
+++ b/source/common/x86/ipfilter8.asm	Fri Apr 17 09:10:24 2015 -0700
@@ -361,6 +361,33 @@
 
 RET
 
+;-----------------------------------------------------------------------------
+; void interp_4tap_horiz_pp_2x8(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
+;-----------------------------------------------------------------------------
+INIT_XMM sse3
+cglobal interp_4tap_horiz_pp_2x8, 4, 6, 6, src, srcstride, dst, dststride
+mov         r4d,        r4m
+mova        m5,         [pw_32]
+
+%ifdef PIC
+lea         r5,          [tabw_ChromaCoeff]
+movddup     m4,         [r5 + r4 * 8]
+%else
+movddup     m4,         [tabw_ChromaCoeff + r4 * 8]
+%endif
+
+%assign x 1
+%rep 4
+FILTER_H4_w2_2_sse2
+%if x < 4
+lea         srcq,       [srcq + srcstrideq * 2]
+lea         dstq,       [dstq + dststrideq * 2]
+%endif
+%assign x x+1
+%endrep
+
+RET
+
 %macro FILTER_H4_w2_2 3
     movh        %2, [srcq - 1]
     pshufb      %2, %2, Tm0
diff -r 34b0ec5fbc86 -r 3d43e6ad9a95 source/common/x86/ipfilter8.h
--- a/source/common/x86/ipfilter8.h	Fri Apr 17 09:04:59 2015 -0700
+++ b/source/common/x86/ipfilter8.h	Fri Apr 17 09:10:24 2015 -0700
@@ -764,6 +764,7 @@
 void x265_filterPixelToShort_48x64_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
 void x265_filterPixelToShort_24x32_avx2(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride);
 void x265_interp_4tap_horiz_pp_2x4_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
+void x265_interp_4tap_horiz_pp_2x8_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
 #undef LUMA_FILTERS
 #undef LUMA_SP_FILTERS
 #undef LUMA_SS_FILTERS