From a1b6bd47bcbad3f188a78616ca8c1b6134a113b5 Mon Sep 17 00:00:00 2001 From: Kieran Kunhya Date: Fri, 21 Oct 2022 04:18:11 +0100 Subject: [PATCH] RFC: v210enc optimisations and initial AVX-512 --- libavcodec/x86/v210enc.asm | 59 ++++++++++++++++++++++------------- libavcodec/x86/v210enc_init.c | 7 +++++ 2 files changed, 45 insertions(+), 21 deletions(-) diff --git a/libavcodec/x86/v210enc.asm b/libavcodec/x86/v210enc.asm index 965f2bea3c..2d0827bbd0 100644 --- a/libavcodec/x86/v210enc.asm +++ b/libavcodec/x86/v210enc.asm @@ -38,13 +38,16 @@ cextern pb_1 cextern pb_FE %define v210_enc_max_8 pb_FE -v210_enc_luma_shuf_8: times 2 db 6,-1,7,-1,8,-1,9,-1,10,-1,11,-1,-1,-1,-1,-1 -v210_enc_luma_mult_8: times 2 dw 16,4,64,16,4,64,0,0 +v210_enc_mult_8: times 8 dw 4,64 +v210_enc_shift_8: times 8 dw 2,6 +v210_enc_mask_8: times 8 db 0x00,0xff,0x00,0x00 +v210_enc_mask2_8: times 8 db 0xff,0x03,0xff,0xff -v210_enc_chroma_shuf1_8: times 2 db 0,-1,1,-1,2,-1,3,-1,8,-1,9,-1,10,-1,11,-1 -v210_enc_chroma_shuf2_8: times 2 db 3,-1,4,-1,5,-1,7,-1,11,-1,12,-1,13,-1,15,-1 +v210_enc_luma_shuf1_8: times 2 db -1,0,-1,-1,1,-1,2,-1,-1,3,-1,-1,4,-1,5,-1 +v210_enc_luma_shuf2_8: times 2 db -1,6,-1,-1,7,-1,8,-1,-1,9,-1,-1,10,-1,11,-1 -v210_enc_chroma_mult_8: times 2 dw 4,16,64,0,64,4,16,0 +v210_enc_chroma_shuf1_8: times 2 db 0,-1,8,-1,-1,1,-1,-1,9,-1,2,-1,-1,10,-1,-1 +v210_enc_chroma_shuf2_8: times 2 db 3,-1,11,-1,-1,4,-1,-1,12,-1,5,-1,-1,13,-1,-1 SECTION .text @@ -115,7 +118,7 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width mova m4, [v210_enc_min_8] mova m5, [v210_enc_max_8] - pxor m6, m6 + mova m6, [v210_enc_mask_8] .loop: movu xm1, [yq+widthq*2] @@ -124,16 +127,6 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width %endif CLIPUB m1, m4, m5 - punpcklbw m0, m1, m6 - ; can't unpack high bytes in the same way because we process - ; only six bytes at a time - pshufb m1, [v210_enc_luma_shuf_8] - - pmullw m0, [v210_enc_luma_mult_8] - pmullw m1, [v210_enc_luma_mult_8] - pshufb m0, [v210_enc_luma_shuf_10] - pshufb m1, [v210_enc_luma_shuf_10] - movq xm3, [uq+widthq] movhps xm3, [vq+widthq] %if cpuflag(avx2) @@ -143,14 +136,33 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width %endif CLIPUB m3, m4, m5 - ; shuffle and multiply to get the same packing as in 10-bit + ; vpermi2b is obvious choice but too slow + pshufb m0, m1, [v210_enc_luma_shuf1_8] + pshufb m1, [v210_enc_luma_shuf2_8] + pshufb m2, m3, [v210_enc_chroma_shuf1_8] pshufb m3, [v210_enc_chroma_shuf2_8] - pmullw m2, [v210_enc_chroma_mult_8] - pmullw m3, [v210_enc_chroma_mult_8] - pshufb m2, [v210_enc_chroma_shuf_10] - pshufb m3, [v210_enc_chroma_shuf_10] + por m0, m2 + por m1, m3 + + ; TODO: avx-512 masked mov? + pand m2, m6, m0 + pand m3, m6, m1 + + pslld m2, 4 + pslld m3, 4 + +%if cpuflag(avx512) + vpsllvw m0, [v210_enc_shift_8] + vpsllvw m1, [v210_enc_shift_8] +%else + pmullw m0, [v210_enc_mult_8] + pmullw m1, [v210_enc_mult_8] +%endif + + pand m0, [v210_enc_mask2_8] + pand m1, [v210_enc_mask2_8] por m0, m2 por m1, m3 @@ -182,3 +194,8 @@ v210_planar_pack_8 INIT_YMM avx2 v210_planar_pack_8 %endif + +%if HAVE_AVX512_EXTERNAL +INIT_YMM avx512 +v210_planar_pack_8 +%endif diff --git a/libavcodec/x86/v210enc_init.c b/libavcodec/x86/v210enc_init.c index 13a351dd1d..095ed5e913 100644 --- a/libavcodec/x86/v210enc_init.c +++ b/libavcodec/x86/v210enc_init.c @@ -27,6 +27,8 @@ void ff_v210_planar_pack_8_avx(const uint8_t *y, const uint8_t *u, const uint8_t *v, uint8_t *dst, ptrdiff_t width); void ff_v210_planar_pack_8_avx2(const uint8_t *y, const uint8_t *u, const uint8_t *v, uint8_t *dst, ptrdiff_t width); +void ff_v210_planar_pack_8_avx512(const uint8_t *y, const uint8_t *u, + const uint8_t *v, uint8_t *dst, ptrdiff_t width); void ff_v210_planar_pack_10_ssse3(const uint16_t *y, const uint16_t *u, const uint16_t *v, uint8_t *dst, ptrdiff_t width); @@ -52,4 +54,9 @@ av_cold void ff_v210enc_init_x86(V210EncContext *s) s->sample_factor_10 = 2; s->pack_line_10 = ff_v210_planar_pack_10_avx2; } + + if (EXTERNAL_AVX512(cpu_flags)) { + s->sample_factor_8 = 2; + s->pack_line_8 = ff_v210_planar_pack_8_avx512; + } } -- 2.24.1.windows.2