[x265] [PATCH] asm: avx2 code for filter_vpp[16x4], filter_vps[16x4]: 303c->293c, 311c->253c
chen
chenm003 at 163.com
Wed Mar 11 23:15:05 CET 2015
right
At 2015-03-11 14:30:38,"Divya Manivannan" <divya at multicorewareinc.com> wrote:
># HG changeset patch
># User Divya Manivannan <divya at multicorewareinc.com>
># Date 1426055417 -19800
># Wed Mar 11 12:00:17 2015 +0530
># Node ID 97ff5bb38e91610bcbbef706850c777171563e8b
># Parent ad5178906527535e8f0479b7c114fd9a91cdbdb7
>asm: avx2 code for filter_vpp[16x4], filter_vps[16x4]: 303c->293c, 311c->253c
>
>diff -r ad5178906527 -r 97ff5bb38e91 source/common/x86/asm-primitives.cpp
>--- a/source/common/x86/asm-primitives.cpp Wed Mar 11 11:47:42 2015 +0530
>+++ b/source/common/x86/asm-primitives.cpp Wed Mar 11 12:00:17 2015 +0530
>@@ -1591,6 +1591,7 @@
> p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vpp = x265_interp_4tap_vert_pp_8x16_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].filter_vpp = x265_interp_4tap_vert_pp_8x32_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_12x16].filter_vpp = x265_interp_4tap_vert_pp_12x16_avx2;
>+ p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].filter_vpp = x265_interp_4tap_vert_pp_16x4_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].filter_vpp = x265_interp_4tap_vert_pp_16x8_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].filter_vpp = x265_interp_4tap_vert_pp_16x16_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_32x8].filter_vpp = x265_interp_4tap_vert_pp_32x8_avx2;
>@@ -1611,6 +1612,7 @@
> p.chroma[X265_CSP_I420].pu[CHROMA_420_8x16].filter_vps = x265_interp_4tap_vert_ps_8x16_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_8x32].filter_vps = x265_interp_4tap_vert_ps_8x32_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_12x16].filter_vps = x265_interp_4tap_vert_ps_12x16_avx2;
>+ p.chroma[X265_CSP_I420].pu[CHROMA_420_16x4].filter_vps = x265_interp_4tap_vert_ps_16x4_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_16x8].filter_vps = x265_interp_4tap_vert_ps_16x8_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_4x16].filter_vps = x265_interp_4tap_vert_ps_4x16_avx2;
> p.chroma[X265_CSP_I420].pu[CHROMA_420_16x16].filter_vps = x265_interp_4tap_vert_ps_16x16_avx2;
>diff -r ad5178906527 -r 97ff5bb38e91 source/common/x86/ipfilter8.asm
>--- a/source/common/x86/ipfilter8.asm Wed Mar 11 11:47:42 2015 +0530
>+++ b/source/common/x86/ipfilter8.asm Wed Mar 11 12:00:17 2015 +0530
>@@ -5807,6 +5807,104 @@
> FILTER_VER_CHROMA_AVX2_16x8 pp
> FILTER_VER_CHROMA_AVX2_16x8 ps
>
>+%macro FILTER_VER_CHROMA_AVX2_16x4 1
>+INIT_YMM avx2
>+cglobal interp_4tap_vert_%1_16x4, 4, 6, 8
>+ mov r4d, r4m
>+ shl r4d, 6
>+
>+%ifdef PIC
>+ lea r5, [tab_ChromaCoeffVer_32]
>+ add r5, r4
>+%else
>+ lea r5, [tab_ChromaCoeffVer_32 + r4]
>+%endif
>+
>+ lea r4, [r1 * 3]
>+ sub r0, r1
>+%ifidn %1,pp
>+ mova m7, [pw_512]
>+%else
>+ add r3d, r3d
>+ mova m7, [pw_2000]
>+%endif
>+
>+ movu xm0, [r0]
>+ vinserti128 m0, m0, [r0 + r1 * 2], 1
>+ movu xm1, [r0 + r1]
>+ vinserti128 m1, m1, [r0 + r4], 1
>+
>+ punpcklbw m2, m0, m1
>+ punpckhbw m3, m0, m1
>+ vperm2i128 m4, m2, m3, 0x20
>+ vperm2i128 m2, m2, m3, 0x31
>+ pmaddubsw m4, [r5]
>+ pmaddubsw m3, m2, [r5 + mmsize]
>+ paddw m4, m3
>+ pmaddubsw m2, [r5]
>+
>+ vextracti128 xm0, m0, 1
>+ lea r0, [r0 + r1 * 4]
>+ vinserti128 m0, m0, [r0], 1
>+
>+ punpcklbw m5, m1, m0
>+ punpckhbw m3, m1, m0
>+ vperm2i128 m6, m5, m3, 0x20
>+ vperm2i128 m5, m5, m3, 0x31
>+ pmaddubsw m6, [r5]
>+ pmaddubsw m3, m5, [r5 + mmsize]
>+ paddw m6, m3
>+ pmaddubsw m5, [r5]
>+%ifidn %1,pp
>+ pmulhrsw m4, m7 ; m4 = word: row 0
>+ pmulhrsw m6, m7 ; m6 = word: row 1
>+ packuswb m4, m6
>+ vpermq m4, m4, 11011000b
>+ vextracti128 xm6, m4, 1
>+ movu [r2], xm4
>+ movu [r2 + r3], xm6
>+%else
>+ psubw m4, m7 ; m4 = word: row 0
>+ psubw m6, m7 ; m6 = word: row 1
>+ movu [r2], m4
>+ movu [r2 + r3], m6
>+%endif
>+ lea r2, [r2 + r3 * 2]
>+
>+ movu xm4, [r0 + r1 * 2]
>+ vinserti128 m4, m4, [r0 + r1], 1
>+ vextracti128 xm1, m4, 1
>+ vinserti128 m0, m0, xm1, 0
>+
>+ punpcklbw m6, m0, m4
>+ punpckhbw m1, m0, m4
>+ vperm2i128 m0, m6, m1, 0x20
>+ vperm2i128 m6, m6, m1, 0x31
>+ pmaddubsw m0, [r5 + mmsize]
>+ paddw m5, m0
>+ pmaddubsw m6, [r5 + mmsize]
>+ paddw m2, m6
>+
>+%ifidn %1,pp
>+ pmulhrsw m2, m7 ; m2 = word: row 2
>+ pmulhrsw m5, m7 ; m5 = word: row 3
>+ packuswb m2, m5
>+ vpermq m2, m2, 11011000b
>+ vextracti128 xm5, m2, 1
>+ movu [r2], xm2
>+ movu [r2 + r3], xm5
>+%else
>+ psubw m2, m7 ; m2 = word: row 2
>+ psubw m5, m7 ; m5 = word: row 3
>+ movu [r2], m2
>+ movu [r2 + r3], m5
>+%endif
>+ RET
>+%endmacro
>+
>+FILTER_VER_CHROMA_AVX2_16x4 pp
>+FILTER_VER_CHROMA_AVX2_16x4 ps
>+
> ;-----------------------------------------------------------------------------
> ;void interp_4tap_vert_pp_24x32(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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20150312/cc192c42/attachment.html>
More information about the x265-devel
mailing list