[x265] [PATCH] asm:intra pred dc32 sse2

[chen]

code is right, but we have enough free register, so we may buffer r1*3 to reduce count of LEA

At 2015-03-03 12:02:46,dtyx265 at gmail.com wrote:
># HG changeset patch
># User David T Yuen <dtyx265 at gmail.com>
># Date 1425354731 28800
># Node ID b70e945ad4ce15461b1c0187dc0023b5d4a640f4
># Parent  a89ced7112fdff2c1a586f29f6bfbf584ec2f697
>asm:intra pred dc32 sse2
>
>This replaces c code for systems using ssse3 to sse2 processors
>The code is backported from intrapred dc32 sse4 high bit
>
>64-bit
>
>./test/TestBench --testbench intrapred | grep 32x32
>intra_dc_32x32[f=0]	7.41x 	 1862.41  	 13808.68
>
>32-bit
>
>./test/TestBench --testbench intrapred | grep 32x32
>intra_dc_32x32[f=0]	4.24x 	 1765.02  	 7474.88
>
>diff -r a89ced7112fd -r b70e945ad4ce source/common/x86/asm-primitives.cpp
>--- a/source/common/x86/asm-primitives.cpp	Mon Mar 02 14:32:14 2015 -0800
>+++ b/source/common/x86/asm-primitives.cpp	Mon Mar 02 19:52:11 2015 -0800
>@@ -1207,6 +1207,7 @@
>         p.cu[BLOCK_4x4].intra_pred[DC_IDX] = x265_intra_pred_dc4_sse2;
>         p.cu[BLOCK_8x8].intra_pred[DC_IDX] = x265_intra_pred_dc8_sse2;
>         p.cu[BLOCK_16x16].intra_pred[DC_IDX] = x265_intra_pred_dc16_sse2;
>+        p.cu[BLOCK_32x32].intra_pred[DC_IDX] = x265_intra_pred_dc32_sse2;
> 
>         p.cu[BLOCK_4x4].calcresidual = x265_getResidual4_sse2;
>         p.cu[BLOCK_8x8].calcresidual = x265_getResidual8_sse2;
>diff -r a89ced7112fd -r b70e945ad4ce source/common/x86/intrapred.h
>--- a/source/common/x86/intrapred.h	Mon Mar 02 14:32:14 2015 -0800
>+++ b/source/common/x86/intrapred.h	Mon Mar 02 19:52:11 2015 -0800
>@@ -29,6 +29,7 @@
> void x265_intra_pred_dc4_sse2(pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
> void x265_intra_pred_dc8_sse2(pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
> void x265_intra_pred_dc16_sse2(pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
>+void x265_intra_pred_dc32_sse2(pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
> void x265_intra_pred_dc4_sse4(pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
> void x265_intra_pred_dc8_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
> void x265_intra_pred_dc16_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
>diff -r a89ced7112fd -r b70e945ad4ce source/common/x86/intrapred8.asm
>--- a/source/common/x86/intrapred8.asm	Mon Mar 02 14:32:14 2015 -0800
>+++ b/source/common/x86/intrapred8.asm	Mon Mar 02 19:52:11 2015 -0800
>@@ -494,6 +494,48 @@
> ;---------------------------------------------------------------------------------------------
> ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
> ;---------------------------------------------------------------------------------------------
>+INIT_XMM sse2
>+cglobal intra_pred_dc32, 3, 3, 5
>+    pxor            m0,            m0
>+    movu            m1,            [r2 + 1]
>+    movu            m2,            [r2 + 17]
>+    movu            m3,            [r2 + 65]
>+    movu            m4,            [r2 + 81]
>+    psadbw          m1,            m0
>+    psadbw          m2,            m0
>+    psadbw          m3,            m0
>+    psadbw          m4,            m0
>+    paddw           m1,            m2
>+    paddw           m3,            m4
>+    paddw           m1,            m3
>+    pshufd          m2,            m1, 2
>+    paddw           m1,            m2
>+
>+    movd            r2d,           m1
>+    add             r2d,           32
>+    shr             r2d,           6     ; sum = sum / 64
>+    movd            m1,            r2d
>+    pmullw          m1,            [pw_257]
>+    pshuflw         m1,            m1, 0x00       ; m1 = byte [dc_val ...]
>+    pshufd          m1,            m1, 0x00
>+
>+%assign x 0
>+%rep 16
>+    ; store DC 16x16
>+    movu            [r0],          m1
>+    movu            [r0 + r1],     m1
>+    movu            [r0 + 16],     m1
>+    movu            [r0 + r1 + 16],m1
>+%if x < 16
>+    lea             r0,            [r0 + 2 * r1]
>+%endif
>+%assign x x+1
>+%endrep
>+    RET
>+
>+;---------------------------------------------------------------------------------------------
>+; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
>+;---------------------------------------------------------------------------------------------
> INIT_XMM sse4
> cglobal intra_pred_dc4, 5,5,3
>     inc         r2
>_______________________________________________
>x265-devel mailing list
>x265-devel at videolan.org
>https://mailman.videolan.org/listinfo/x265-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20150303/4f45df16/attachment.html>