v210: Add avx2 version of the 8-bit line encoder

Around 35% faster than the avx version.
Signed-off-by: Henrik Gramner <henrik@gramner.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

libavcodec/v210enc.c

@@ -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);

libavcodec/v210enc.h

@@ -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);

libavcodec/x86/v210enc.asm

@@ -21,27 +21,26 @@
 
 %include "libavutil/x86/x86util.asm"
 
-SECTION_RODATA
+SECTION_RODATA 32
 
-v210_enc_min_10: times 8 dw 0x4
-v210_enc_max_10: times 8 dw 0x3fb
+v210_enc_min_10: times 16 dw 0x4
+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
 
-v210_enc_min_8: times 16 db 0x1
-v210_enc_max_8: times 16 db 0xfe
+v210_enc_min_8: times 32 db 0x1
+v210_enc_max_8: times 32 db 0xfe
 
-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_mult_8: times 2 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_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_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
+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
 
 SECTION .text
 
@@ -102,7 +101,10 @@ 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+2*widthq]
+%if cpuflag(avx2)
+    vinserti128 m1, m1, [yq+2*widthq+12], 1
+%endif
     CLIPUB  m1, m4, m5
 
     punpcklbw m0, m1, m6
@@ -115,8 +117,13 @@ 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
@@ -131,11 +138,15 @@ 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
@@ -149,3 +160,7 @@ v210_planar_pack_8
 INIT_XMM avx
 v210_planar_pack_8
 %endif
+%if HAVE_AVX2_EXTERNAL
+INIT_YMM avx2
+v210_planar_pack_8
+%endif

libavcodec/x86/v210enc_init.c

@@ -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->sample_factor = 2;
+        s->pack_line_8   = ff_v210_planar_pack_8_avx2;
+    }
 }