swscale/la: Add output_lasx.c file.

ffmpeg -i 1_h264_1080p_30fps_3Mbps.mp4 -f rawvideo -s 640x480 -pix_fmt
rgb24 -y /dev/null -an
before: 150fps
after:  183fps

Signed-off-by: Hao Chen <chenhao@loongson.cn>
Reviewed-by: yinshiyou-hf@loongson.cn
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libswscale/loongarch/Makefile

@@ -2,4 +2,5 @@ OBJS-$(CONFIG_SWSCALE)      += loongarch/swscale_init_loongarch.o
 LASX-OBJS-$(CONFIG_SWSCALE) += loongarch/swscale_lasx.o \
                                loongarch/input_lasx.o   \
                                loongarch/yuv2rgb_lasx.o \
-                               loongarch/rgb2rgb_lasx.o
+                               loongarch/rgb2rgb_lasx.o \
+                               loongarch/output_lasx.o

libswscale/loongarch/swscale_init_loongarch.c

@@ -28,6 +28,7 @@ av_cold void ff_sws_init_swscale_loongarch(SwsContext *c)
 {
     int cpu_flags = av_get_cpu_flags();
     if (have_lasx(cpu_flags)) {
+        ff_sws_init_output_loongarch(c);
         if (c->srcBpc == 8) {
             if (c->dstBpc <= 14) {
                 c->hyScale = c->hcScale = ff_hscale_8_to_15_lasx;
@@ -47,6 +48,8 @@ av_cold void ff_sws_init_swscale_loongarch(SwsContext *c)
             }
             break;
         }
+        if (c->dstBpc == 8)
+            c->yuv2planeX = ff_yuv2planeX_8_lasx;
     }
 }
 

libswscale/loongarch/swscale_loongarch.h

@@ -69,4 +69,10 @@ void ff_interleave_bytes_lasx(const uint8_t *src1, const uint8_t *src2,
                               uint8_t *dest, int width, int height,
                               int src1Stride, int src2Stride, int dstStride);
 
+av_cold void ff_sws_init_output_loongarch(SwsContext *c);
+
+void ff_yuv2planeX_8_lasx(const int16_t *filter, int filterSize,
+                          const int16_t **src, uint8_t *dest, int dstW,
+                          const uint8_t *dither, int offset);
+
 #endif /* SWSCALE_LOONGARCH_SWSCALE_LOONGARCH_H */