[x265] [PATCH] asm: interp_4tap_horiz_ps sse3
dave
dtyx265 at gmail.com
Fri May 22 04:11:38 CEST 2015
On 05/21/2015 07:07 PM, chen wrote:
> lea srcq, [srcq + srcstrideq]
> why not ADD?
That's carried over from the sse4 code. I can change it to add if that
is what you want.
> At 2015-05-22 09:27:57,dtyx265 at gmail.com wrote:
> ># HG changeset patch
> ># User David T Yuen <dtyx265 at gmail.com>
> ># Date 1432258035 25200
> ># Node ID 4e79322b066536eed735a7aafd296b07e9195296
> ># Parent 234bc93bd51698801fad77cc861177ed019f5113
> >asm: interp_4tap_horiz_ps sse3
> >
> >This replaces c code for all of interp_4tap_horiz_ps for sse3
> >
> >diff -r 234bc93bd516 -r 4e79322b0665 source/common/x86/asm-primitives.cpp
> >--- a/source/common/x86/asm-primitives.cpp Thu May 21 16:34:48 2015 +0530
> >+++ b/source/common/x86/asm-primitives.cpp Thu May 21 18:27:15 2015 -0700
> >@@ -1614,6 +1614,9 @@
> > ALL_CHROMA_420_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
> > ALL_CHROMA_422_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
> > ALL_CHROMA_444_PU(filter_hpp, interp_4tap_horiz_pp, sse3);
> >+ ALL_CHROMA_420_PU(filter_hps, interp_4tap_horiz_ps, sse3);
> >+ ALL_CHROMA_422_PU(filter_hps, interp_4tap_horiz_ps, sse3);
> >+ ALL_CHROMA_444_PU(filter_hps, interp_4tap_horiz_ps, sse3);
> > }
> > if (cpuMask & X265_CPU_SSSE3)
> > {
> >diff -r 234bc93bd516 -r 4e79322b0665 source/common/x86/ipfilter8.asm
> >--- a/source/common/x86/ipfilter8.asm Thu May 21 16:34:48 2015 +0530
> >+++ b/source/common/x86/ipfilter8.asm Thu May 21 18:27:15 2015 -0700
> >@@ -641,6 +641,123 @@
> > IPFILTER_CHROMA_sse3 48, 64
> > IPFILTER_CHROMA_sse3 64, 16
> >
> >+%macro FILTER_2 2
> >+ movd m3, [srcq + %1]
> >+ movd m4, [srcq + 1 + %1]
> >+ punpckldq m3, m4
> >+ punpcklbw m3, m0
> >+ pmaddwd m3, m1
> >+ packssdw m3, m3
> >+ pshuflw m4, m3, q2301
> >+ paddw m3, m4
> >+ psrldq m3, 2
> >+ psubw m3, m2
> >+ movd [dstq + %2], m3
> >+%endmacro
> >+
> >+%macro FILTER_4 2
> >+ movd m3, [srcq + %1]
> >+ movd m4, [srcq + 1 + %1]
> >+ punpckldq m3, m4
> >+ punpcklbw m3, m0
> >+ pmaddwd m3, m1
> >+ movd m4, [srcq + 2 + %1]
> >+ movd m5, [srcq + 3 + %1]
> >+ punpckldq m4, m5
> >+ punpcklbw m4, m0
> >+ pmaddwd m4, m1
> >+ packssdw m3, m4
> >+ pshuflw m4, m3, q2301
> >+ pshufhw m4, m4, q2301
> >+ paddw m3, m4
> >+ psrldq m3, 2
> >+ pshufd m3, m3, q3120
> >+ psubw m3, m2
> >+ movh [dstq + %2], m3
> >+%endmacro
> >+
> >+%macro FILTER_4TAP_HPS_sse3 2
> >+INIT_XMM sse3
> >+cglobal interp_4tap_horiz_ps_%1x%2, 4, 7, 6, src, srcstride, dst, dststride
> >+ mov r4d, r4m
> >+ add dststrided, dststrided
> >+ mova m2, [pw_2000]
> >+ pxor m0, m0
> >+
> >+%ifdef PIC
> >+ lea r6, [tabw_ChromaCoeff]
> >+ movddup m1, [r6 + r4 * 8]
> >+%else
> >+ movddup m1, [tabw_ChromaCoeff + r4 * 8]
> >+%endif
> >+
> >+ mov r4d, %2
> >+ cmp r5m, byte 0
> >+ je .loopH
> >+ sub srcq, srcstrideq
> >+ add r4d, 3
> >+
> >+.loopH:
> >+%assign x -1
> >+%assign y 0
> >+%rep %1/4
> >+ FILTER_4 x,y
> >+%assign x x+4
> >+%assign y y+8
> >+%endrep
> >+%rep (%1 % 4)/2
> >+ FILTER_2 x,y
> >+%endrep
> >+ lea srcq, [srcq + srcstrideq]
> >+ lea dstq, [dstq + dststrideq]
> >+
> >+ dec r4d
> >+ jnz .loopH
> >+ RET
> >+
> >+%endmacro
> >+
> >+ FILTER_4TAP_HPS_sse3 2, 4
> >+ FILTER_4TAP_HPS_sse3 2, 8
> >+ FILTER_4TAP_HPS_sse3 2, 16
> >+ FILTER_4TAP_HPS_sse3 4, 2
> >+ FILTER_4TAP_HPS_sse3 4, 4
> >+ FILTER_4TAP_HPS_sse3 4, 8
> >+ FILTER_4TAP_HPS_sse3 4, 16
> >+ FILTER_4TAP_HPS_sse3 4, 32
> >+ FILTER_4TAP_HPS_sse3 6, 8
> >+ FILTER_4TAP_HPS_sse3 6, 16
> >+ FILTER_4TAP_HPS_sse3 8, 2
> >+ FILTER_4TAP_HPS_sse3 8, 4
> >+ FILTER_4TAP_HPS_sse3 8, 6
> >+ FILTER_4TAP_HPS_sse3 8, 8
> >+ FILTER_4TAP_HPS_sse3 8, 12
> >+ FILTER_4TAP_HPS_sse3 8, 16
> >+ FILTER_4TAP_HPS_sse3 8, 32
> >+ FILTER_4TAP_HPS_sse3 8, 64
> >+ FILTER_4TAP_HPS_sse3 12, 16
> >+ FILTER_4TAP_HPS_sse3 12, 32
> >+ FILTER_4TAP_HPS_sse3 16, 4
> >+ FILTER_4TAP_HPS_sse3 16, 8
> >+ FILTER_4TAP_HPS_sse3 16, 12
> >+ FILTER_4TAP_HPS_sse3 16, 16
> >+ FILTER_4TAP_HPS_sse3 16, 24
> >+ FILTER_4TAP_HPS_sse3 16, 32
> >+ FILTER_4TAP_HPS_sse3 16, 64
> >+ FILTER_4TAP_HPS_sse3 24, 32
> >+ FILTER_4TAP_HPS_sse3 24, 64
> >+ FILTER_4TAP_HPS_sse3 32, 8
> >+ FILTER_4TAP_HPS_sse3 32, 16
> >+ FILTER_4TAP_HPS_sse3 32, 24
> >+ FILTER_4TAP_HPS_sse3 32, 32
> >+ FILTER_4TAP_HPS_sse3 32, 48
> >+ FILTER_4TAP_HPS_sse3 32, 64
> >+ FILTER_4TAP_HPS_sse3 48, 64
> >+ FILTER_4TAP_HPS_sse3 64, 16
> >+ FILTER_4TAP_HPS_sse3 64, 32
> >+ FILTER_4TAP_HPS_sse3 64, 48
> >+ FILTER_4TAP_HPS_sse3 64, 64
> >+
> > %macro FILTER_H8_W8_sse2 0
> > movh m1, [r0 + x - 3]
> > movh m4, [r0 + x - 2]
> >diff -r 234bc93bd516 -r 4e79322b0665 source/common/x86/ipfilter8.h
> >--- a/source/common/x86/ipfilter8.h Thu May 21 16:34:48 2015 +0530
> >+++ b/source/common/x86/ipfilter8.h Thu May 21 18:27:15 2015 -0700
> >@@ -854,6 +854,47 @@
> > void x265_interp_4tap_horiz_pp_64x32_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> > void x265_interp_4tap_horiz_pp_64x48_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> > void x265_interp_4tap_horiz_pp_64x64_sse3(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> >+void x265_interp_4tap_horiz_ps_2x4_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_2x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_2x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_4x2_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_4x4_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_4x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_4x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_4x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_6x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_6x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x2_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x4_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x6_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x12_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_8x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_12x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_12x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_12x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x4_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x12_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x24_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_16x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_24x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_24x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x8_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x24_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x48_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_32x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_64x16_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_64x32_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_64x48_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_64x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> >+void x265_interp_4tap_horiz_ps_48x64_sse3(const pixel* src, intptr_t srcStride, int16_t* dst, intptr_t dstStride, int coeffIdx, int isRowExt);
> > void x265_interp_8tap_horiz_pp_4x4_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> > void x265_interp_8tap_horiz_pp_4x8_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> > void x265_interp_8tap_horiz_pp_4x16_sse2(const pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
> >_______________________________________________
> >x265-devel mailing list
> >x265-devel at videolan.org
> >https://mailman.videolan.org/listinfo/x265-devel
>
>
> _______________________________________________
> 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/20150521/05df2de3/attachment.html>
More information about the x265-devel
mailing list