qsvdec: do not sync PIX_FMT_QSV surfaces

[ffmpeg] / libavcodec / wmv2dsp.c

diff --git a/libavcodec/wmv2dsp.c b/libavcodec/wmv2dsp.c
index 85a9db1d75958eb68a896f1199213f3a0985cf20..90073e4b4a6d892e59afba2f27c3c5253d41ad4d 100644 (file)
--- a/libavcodec/wmv2dsp.c
+++ b/libavcodec/wmv2dsp.c
@@ -19,6 +19,8 @@
 #include "libavutil/attributes.h"
 #include "libavutil/common.h"
 #include "avcodec.h"
+#include "idctdsp.h"
+#include "mathops.h"
 #include "wmv2dsp.h"
 
 #define W0 2048
@@ -91,7 +93,7 @@ static void wmv2_idct_col(short * b)
     b[8 * 7] = (a0 + a2 - a1 - a5 + (1 << 13)) >> 14;
 }
 
-static void wmv2_idct_add_c(uint8_t *dest, int line_size, DCTELEM *block)
+static void wmv2_idct_add_c(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
 {
     int i;
 
@@ -114,7 +116,7 @@ static void wmv2_idct_add_c(uint8_t *dest, int line_size, DCTELEM *block)
     }
 }
 
-static void wmv2_idct_put_c(uint8_t *dest, int line_size, DCTELEM *block)
+static void wmv2_idct_put_c(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
 {
     int i;
 
@@ -137,9 +139,127 @@ static void wmv2_idct_put_c(uint8_t *dest, int line_size, DCTELEM *block)
     }
 }
 
+static void wmv2_mspel8_h_lowpass(uint8_t *dst, const uint8_t *src,
+                                  int dstStride, int srcStride, int h)
+{
+    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
+    int i;
+
+    for (i = 0; i < h; i++) {
+        dst[0] = cm[(9 * (src[0] + src[1]) - (src[-1] + src[2]) + 8) >> 4];
+        dst[1] = cm[(9 * (src[1] + src[2]) - (src[0]  + src[3]) + 8) >> 4];
+        dst[2] = cm[(9 * (src[2] + src[3]) - (src[1]  + src[4]) + 8) >> 4];
+        dst[3] = cm[(9 * (src[3] + src[4]) - (src[2]  + src[5]) + 8) >> 4];
+        dst[4] = cm[(9 * (src[4] + src[5]) - (src[3]  + src[6]) + 8) >> 4];
+        dst[5] = cm[(9 * (src[5] + src[6]) - (src[4]  + src[7]) + 8) >> 4];
+        dst[6] = cm[(9 * (src[6] + src[7]) - (src[5]  + src[8]) + 8) >> 4];
+        dst[7] = cm[(9 * (src[7] + src[8]) - (src[6]  + src[9]) + 8) >> 4];
+        dst   += dstStride;
+        src   += srcStride;
+    }
+}
+
+static void wmv2_mspel8_v_lowpass(uint8_t *dst, const uint8_t *src,
+                                  int dstStride, int srcStride, int w)
+{
+    const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
+    int i;
+
+    for (i = 0; i < w; i++) {
+        const int src_1 = src[-srcStride];
+        const int src0  = src[0];
+        const int src1  = src[srcStride];
+        const int src2  = src[2 * srcStride];
+        const int src3  = src[3 * srcStride];
+        const int src4  = src[4 * srcStride];
+        const int src5  = src[5 * srcStride];
+        const int src6  = src[6 * srcStride];
+        const int src7  = src[7 * srcStride];
+        const int src8  = src[8 * srcStride];
+        const int src9  = src[9 * srcStride];
+        dst[0 * dstStride] = cm[(9 * (src0 + src1) - (src_1 + src2) + 8) >> 4];
+        dst[1 * dstStride] = cm[(9 * (src1 + src2) - (src0  + src3) + 8) >> 4];
+        dst[2 * dstStride] = cm[(9 * (src2 + src3) - (src1  + src4) + 8) >> 4];
+        dst[3 * dstStride] = cm[(9 * (src3 + src4) - (src2  + src5) + 8) >> 4];
+        dst[4 * dstStride] = cm[(9 * (src4 + src5) - (src3  + src6) + 8) >> 4];
+        dst[5 * dstStride] = cm[(9 * (src5 + src6) - (src4  + src7) + 8) >> 4];
+        dst[6 * dstStride] = cm[(9 * (src6 + src7) - (src5  + src8) + 8) >> 4];
+        dst[7 * dstStride] = cm[(9 * (src7 + src8) - (src6  + src9) + 8) >> 4];
+        src++;
+        dst++;
+    }
+}
+
+static void put_mspel8_mc10_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    uint8_t half[64];
+
+    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
+    ff_put_pixels8_l2_8(dst, src, half, stride, stride, 8, 8);
+}
+
+static void put_mspel8_mc20_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
+}
+
+static void put_mspel8_mc30_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    uint8_t half[64];
+
+    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
+    ff_put_pixels8_l2_8(dst, src + 1, half, stride, stride, 8, 8);
+}
+
+static void put_mspel8_mc02_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
+}
+
+static void put_mspel8_mc12_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    uint8_t halfH[88];
+    uint8_t halfV[64];
+    uint8_t halfHV[64];
+
+    wmv2_mspel8_h_lowpass(halfH, src - stride, 8, stride, 11);
+    wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
+    wmv2_mspel8_v_lowpass(halfHV, halfH + 8, 8, 8, 8);
+    ff_put_pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);
+}
+
+static void put_mspel8_mc32_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    uint8_t halfH[88];
+    uint8_t halfV[64];
+    uint8_t halfHV[64];
+
+    wmv2_mspel8_h_lowpass(halfH, src - stride, 8, stride, 11);
+    wmv2_mspel8_v_lowpass(halfV, src + 1, 8, stride, 8);
+    wmv2_mspel8_v_lowpass(halfHV, halfH + 8, 8, 8, 8);
+    ff_put_pixels8_l2_8(dst, halfV, halfHV, stride, 8, 8, 8);
+}
+
+static void put_mspel8_mc22_c(uint8_t *dst, const uint8_t *src, ptrdiff_t stride)
+{
+    uint8_t halfH[88];
+
+    wmv2_mspel8_h_lowpass(halfH, src - stride, 8, stride, 11);
+    wmv2_mspel8_v_lowpass(dst, halfH + 8, stride, 8, 8);
+}
+
 av_cold void ff_wmv2dsp_init(WMV2DSPContext *c)
 {
     c->idct_add  = wmv2_idct_add_c;
     c->idct_put  = wmv2_idct_put_c;
-    c->idct_perm = FF_NO_IDCT_PERM;
+    c->idct_perm = FF_IDCT_PERM_NONE;
+
+    c->put_mspel_pixels_tab[0] = ff_put_pixels8x8_c;
+    c->put_mspel_pixels_tab[1] = put_mspel8_mc10_c;
+    c->put_mspel_pixels_tab[2] = put_mspel8_mc20_c;
+    c->put_mspel_pixels_tab[3] = put_mspel8_mc30_c;
+    c->put_mspel_pixels_tab[4] = put_mspel8_mc02_c;
+    c->put_mspel_pixels_tab[5] = put_mspel8_mc12_c;
+    c->put_mspel_pixels_tab[6] = put_mspel8_mc22_c;
+    c->put_mspel_pixels_tab[7] = put_mspel8_mc32_c;
 }