Commit 3836f404 authored by James Darnley's avatar James Darnley

avcodec/v210: add avx2 version of the 8-bit line encoder

Around 35% faster than the avx version.
Signed-off-by: 's avatarHenrik Gramner <henrik@gramner.com>
parent 61625dcc
......@@ -86,6 +86,7 @@ av_cold void ff_v210enc_init(V210EncContext *s)
{
s->pack_line_8 = v210_planar_pack_8_c;
s->pack_line_10 = v210_planar_pack_10_c;
s->sample_factor = 1;
if (ARCH_X86)
ff_v210enc_init_x86(s);
......@@ -172,13 +173,13 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
const uint8_t *v = pic->data[2];
for (h = 0; h < avctx->height; h++) {
uint32_t val;
w = (avctx->width / 12) * 12;
w = (avctx->width / (12 * s->sample_factor)) * 12 * s->sample_factor;
s->pack_line_8(y, u, v, dst, w);
y += w;
u += w >> 1;
v += w >> 1;
dst += (w / 12) * 32;
dst += (w / (12 * s->sample_factor)) * 32 * s->sample_factor;
for (; w < avctx->width - 5; w += 6) {
WRITE_PIXELS8(u, y, v);
......
......@@ -28,6 +28,7 @@ typedef struct V210EncContext {
const uint8_t *v, uint8_t *dst, ptrdiff_t width);
void (*pack_line_10)(const uint16_t *y, const uint16_t *u,
const uint16_t *v, uint8_t *dst, ptrdiff_t width);
int sample_factor;
} V210EncContext;
void ff_v210enc_init(V210EncContext *s);
......
......@@ -74,7 +74,8 @@ DECLARE_ALIGNED(32, const ymm_reg, ff_pb_3) = { 0x0303030303030303ULL, 0x030
0x0303030303030303ULL, 0x0303030303030303ULL };
DECLARE_ALIGNED(32, const xmm_reg, ff_pb_15) = { 0x0F0F0F0F0F0F0F0FULL, 0x0F0F0F0F0F0F0F0FULL };
DECLARE_ALIGNED(16, const xmm_reg, ff_pb_80) = { 0x8080808080808080ULL, 0x8080808080808080ULL };
DECLARE_ALIGNED(16, const xmm_reg, ff_pb_FE) = { 0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL };
DECLARE_ALIGNED(32, const ymm_reg, ff_pb_FE) = { 0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL,
0xFEFEFEFEFEFEFEFEULL, 0xFEFEFEFEFEFEFEFEULL };
DECLARE_ALIGNED(8, const uint64_t, ff_pb_FC) = 0xFCFCFCFCFCFCFCFCULL;
DECLARE_ALIGNED(16, const xmm_reg, ff_ps_neg) = { 0x8000000080000000ULL, 0x8000000080000000ULL };
......
......@@ -57,7 +57,7 @@ extern const ymm_reg ff_pb_1;
extern const ymm_reg ff_pb_2;
extern const ymm_reg ff_pb_3;
extern const xmm_reg ff_pb_80;
extern const xmm_reg ff_pb_FE;
extern const ymm_reg ff_pb_FE;
extern const uint64_t ff_pb_FC;
extern const xmm_reg ff_ps_neg;
......
......@@ -21,30 +21,30 @@
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
SECTION_RODATA 32
cextern pw_4
%define v210_enc_min_10 pw_4
v210_enc_max_10: times 8 dw 0x3fb
v210_enc_max_10: times 16 dw 0x3fb
v210_enc_luma_mult_10: dw 4,1,16,4,1,16,0,0
v210_enc_luma_shuf_10: db -1,0,1,-1,2,3,4,5,-1,6,7,-1,8,9,10,11
v210_enc_luma_mult_10: times 2 dw 4,1,16,4,1,16,0,0
v210_enc_luma_shuf_10: times 2 db -1,0,1,-1,2,3,4,5,-1,6,7,-1,8,9,10,11
v210_enc_chroma_mult_10: dw 1,4,16,0,16,1,4,0
v210_enc_chroma_shuf_10: db 0,1,8,9,-1,2,3,-1,10,11,4,5,-1,12,13,-1
v210_enc_chroma_mult_10: times 2 dw 1,4,16,0,16,1,4,0
v210_enc_chroma_shuf_10: times 2 db 0,1,8,9,-1,2,3,-1,10,11,4,5,-1,12,13,-1
cextern pb_1
%define v210_enc_min_8 pb_1
cextern pb_FE
%define v210_enc_max_8 pb_FE
v210_enc_luma_shuf_8: db 6,-1,7,-1,8,-1,9,-1,10,-1,11,-1,-1,-1,-1,-1
v210_enc_luma_mult_8: dw 16,4,64,16,4,64,0,0
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_chroma_shuf1_8: db 0,-1,1,-1,2,-1,3,-1,8,-1,9,-1,10,-1,11,-1
v210_enc_chroma_shuf2_8: db 3,-1,4,-1,5,-1,7,-1,11,-1,12,-1,13,-1,15,-1
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_chroma_mult_8: dw 4,16,64,0,64,4,16,0
v210_enc_chroma_mult_8: times 2 dw 4,16,64,0,64,4,16,0
SECTION .text
......@@ -103,7 +103,10 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width
pxor m6, m6
.loop:
movu m1, [yq+2*widthq]
movu xm1, [yq+widthq*2]
%if cpuflag(avx2)
vinserti128 m1, m1, [yq+widthq*2+12], 1
%endif
CLIPUB m1, m4, m5
punpcklbw m0, m1, m6
......@@ -116,8 +119,13 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width
pshufb m0, [v210_enc_luma_shuf_10]
pshufb m1, [v210_enc_luma_shuf_10]
movq m3, [uq+widthq]
movhps m3, [vq+widthq]
movq xm3, [uq+widthq]
movhps xm3, [vq+widthq]
%if cpuflag(avx2)
movq xm2, [uq+widthq+6]
movhps xm2, [vq+widthq+6]
vinserti128 m3, m3, xm2, 1
%endif
CLIPUB m3, m4, m5
; shuffle and multiply to get the same packing as in 10-bit
......@@ -132,11 +140,15 @@ cglobal v210_planar_pack_8, 5, 5, 7, y, u, v, dst, width
por m0, m2
por m1, m3
movu [dstq], m0
movu [dstq+mmsize], m1
movu [dstq], xm0
movu [dstq+16], xm1
%if cpuflag(avx2)
vextracti128 [dstq+32], m0, 1
vextracti128 [dstq+48], m1, 1
%endif
add dstq, 2*mmsize
add widthq, 6
add widthq, (mmsize*3)/8
jl .loop
RET
......@@ -146,3 +158,5 @@ INIT_XMM ssse3
v210_planar_pack_8
INIT_XMM avx
v210_planar_pack_8
INIT_YMM avx2
v210_planar_pack_8
......@@ -24,6 +24,8 @@ void ff_v210_planar_pack_8_ssse3(const uint8_t *y, const uint8_t *u,
ptrdiff_t width);
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_10_ssse3(const uint16_t *y, const uint16_t *u,
const uint16_t *v, uint8_t *dst,
ptrdiff_t width);
......@@ -39,4 +41,9 @@ av_cold void ff_v210enc_init_x86(V210EncContext *s)
if (EXTERNAL_AVX(cpu_flags))
s->pack_line_8 = ff_v210_planar_pack_8_avx;
if (EXTERNAL_AVX2(cpu_flags)) {
s->pack_line_8 = ff_v210_planar_pack_8_avx2;
s->sample_factor = 2;
}
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment