Merge remote-tracking branch 'qatar/master'

* qatar/master:
  proresdec: support mixed interlaced/non-interlaced content
  vp3/5: move put_no_rnd_pixels_l2 from dsputil to VP3DSPContext.

Merged-by: Michael Niedermayer <michaelni@gmx.at>

10 files changed, 102 insertions, 46 deletions

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index ffb508535f..c52c8f1703 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -2946,7 +2946,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->clear_blocks                  = FUNCC(clear_blocks ## dct   , depth);\
     c->add_pixels8                   = FUNCC(add_pixels8  ## dct   , depth);\
     c->add_pixels4                   = FUNCC(add_pixels4  ## dct   , depth);\
-    c->put_no_rnd_pixels_l2          = FUNCC(put_no_rnd_pixels8_l2 , depth);\
 \
     c->put_h264_chroma_pixels_tab[0] = FUNCC(put_h264_chroma_mc8   , depth);\
     c->put_h264_chroma_pixels_tab[1] = FUNCC(put_h264_chroma_mc4   , depth);\
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index 5bac9dfd2c..8c52e5118a 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -308,8 +308,6 @@ typedef struct DSPContext {
      */
     op_pixels_func avg_no_rnd_pixels_tab[4][4];
 
-    void (*put_no_rnd_pixels_l2)(uint8_t *block/*align 8*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
-
     /**
      * Thirdpel motion compensation with rounding (a+b+1)>>1.
      * this is an array[12] of motion compensation functions for the 9 thirdpe
diff --git a/libavcodec/dsputil_template.c b/libavcodec/dsputil_template.c
index 26167d808f..5162304d05 100644
--- a/libavcodec/dsputil_template.c
+++ b/libavcodec/dsputil_template.c
@@ -582,10 +582,6 @@ PIXOP2(put, op_put)
 #define put_no_rnd_pixels8_c  put_pixels8_c
 #define put_no_rnd_pixels16_c put_pixels16_c
 
-static void FUNCC(put_no_rnd_pixels8_l2)(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){
-    FUNC(put_no_rnd_pixels8_l2)(dst, a, b, stride, stride, stride, h);
-}
-
 #define H264_CHROMA_MC(OPNAME, OP)\
 static void FUNCC(OPNAME ## h264_chroma_mc2)(uint8_t *p_dst/*align 8*/, uint8_t *p_src/*align 1*/, int stride, int h, int x, int y){\
     pixel *dst = (pixel*)p_dst;\
diff --git a/libavcodec/proresdec_lgpl.c b/libavcodec/proresdec_lgpl.c
index 4fa2817e86..a25ecca3c9 100644
--- a/libavcodec/proresdec_lgpl.c
+++ b/libavcodec/proresdec_lgpl.c
@@ -164,6 +164,8 @@ static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
     if (ctx->frame_type) {      /* if interlaced */
         ctx->picture.interlaced_frame = 1;
         ctx->picture.top_field_first  = ctx->frame_type & 1;
+    } else {
+        ctx->picture.interlaced_frame = 0;
     }
 
     avctx->color_primaries = buf[14];
diff --git a/libavcodec/vp3.c b/libavcodec/vp3.c
index 7b470dcccf..11436e5772 100644
--- a/libavcodec/vp3.c
+++ b/libavcodec/vp3.c
@@ -1570,7 +1570,7 @@ static void render_slice(Vp3DecodeContext *s, int slice)
                                 motion_source, stride, 8);
                         }else{
                             int d= (motion_x ^ motion_y)>>31; // d is 0 if motion_x and _y have the same sign, else -1
-                            s->dsp.put_no_rnd_pixels_l2(
+                            s->vp3dsp.put_no_rnd_pixels_l2(
                                 output_plane + first_pixel,
                                 motion_source - d,
                                 motion_source + stride + 1 + d,
diff --git a/libavcodec/vp3dsp.c b/libavcodec/vp3dsp.c
index 6502a158e6..6db80c8d00 100644
--- a/libavcodec/vp3dsp.c
+++ b/libavcodec/vp3dsp.c
@@ -274,8 +274,27 @@ static void vp3_h_loop_filter_c(uint8_t *first_pixel, int stride,
     }
 }
 
+static void put_no_rnd_pixels_l2(uint8_t *dst, const uint8_t *src1,
+                                 const uint8_t *src2, ptrdiff_t stride, int h)
+{
+    int i;
+
+    for (i = 0; i < h; i++) {
+        uint32_t a, b;
+
+        a = AV_RN32A(&src1[i * stride]);
+        b = AV_RN32A(&src2[i * stride]);
+        AV_WN32A(&dst[i * stride], no_rnd_avg32(a, b));
+        a = AV_RN32A(&src1[i * stride + 4]);
+        b = AV_RN32A(&src2[i * stride + 4]);
+        AV_WN32A(&dst[i * stride + 4], no_rnd_avg32(a, b));
+    }
+}
+
 av_cold void ff_vp3dsp_init(VP3DSPContext *c, int flags)
 {
+    c->put_no_rnd_pixels_l2 = put_no_rnd_pixels_l2;
+
     c->idct_put      = vp3_idct_put_c;
     c->idct_add      = vp3_idct_add_c;
     c->idct_dc_add   = vp3_idct_dc_add_c;
diff --git a/libavcodec/vp3dsp.h b/libavcodec/vp3dsp.h
index bc651997d2..35bd95c684 100644
--- a/libavcodec/vp3dsp.h
+++ b/libavcodec/vp3dsp.h
@@ -19,10 +19,26 @@
 #ifndef AVCODEC_VP3DSP_H
 #define AVCODEC_VP3DSP_H
 
+#include <stddef.h>
 #include <stdint.h>
 #include "dsputil.h"
 
 typedef struct VP3DSPContext {
+    /**
+     * Copy 8xH pixels from source to destination buffer using a bilinear
+     * filter with no rounding (i.e. *dst = (*a + *b) >> 1).
+     *
+     * @param dst destination buffer, aligned by 8
+     * @param a first source buffer, no alignment
+     * @param b second source buffer, no alignment
+     * @param stride distance between two lines in source/dest buffers
+     * @param h height
+     */
+    void (*put_no_rnd_pixels_l2)(uint8_t *dst,
+                                 const uint8_t *a,
+                                 const uint8_t *b,
+                                 ptrdiff_t stride, int h);
+
     void (*idct_put)(uint8_t *dest, int line_size, DCTELEM *block);
     void (*idct_add)(uint8_t *dest, int line_size, DCTELEM *block);
     void (*idct_dc_add)(uint8_t *dest, int line_size, DCTELEM *block);
diff --git a/libavcodec/vp56.c b/libavcodec/vp56.c
index 13bbb02481..7df5adade3 100644
--- a/libavcodec/vp56.c
+++ b/libavcodec/vp56.c
@@ -373,9 +373,9 @@ static void vp56_mc(VP56Context *s, int b, int plane, uint8_t *src,
             s->filter(s, dst, src_block, src_offset, src_offset+overlap_offset,
                       stride, s->mv[b], mask, s->filter_selection, b<4);
         else
-            s->dsp.put_no_rnd_pixels_l2(dst, src_block+src_offset,
-                                        src_block+src_offset+overlap_offset,
-                                        stride, 8);
+            s->vp3dsp.put_no_rnd_pixels_l2(dst, src_block+src_offset,
+                                           src_block+src_offset+overlap_offset,
+                                           stride, 8);
     } else {
         s->dsp.put_pixels_tab[1][0](dst, src_block+src_offset, stride, 8);
     }
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 1e0f8a168c..3bc3d61c89 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -1839,39 +1839,6 @@ void ff_avg_vc1_mspel_mc00_mmxext(uint8_t *dst, const uint8_t *src,
     avg_pixels8_mmxext(dst, src, stride, 8);
 }
 
-/* only used in VP3/5/6 */
-static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h)
-{
-//    START_TIMER
-    MOVQ_BFE(mm6);
-    __asm__ volatile(
-        "1:                             \n\t"
-        "movq   (%1), %%mm0             \n\t"
-        "movq   (%2), %%mm1             \n\t"
-        "movq   (%1,%4), %%mm2          \n\t"
-        "movq   (%2,%4), %%mm3          \n\t"
-        PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)
-        "movq   %%mm4, (%3)             \n\t"
-        "movq   %%mm5, (%3,%4)          \n\t"
-
-        "movq   (%1,%4,2), %%mm0        \n\t"
-        "movq   (%2,%4,2), %%mm1        \n\t"
-        "movq   (%1,%5), %%mm2          \n\t"
-        "movq   (%2,%5), %%mm3          \n\t"
-        "lea    (%1,%4,4), %1           \n\t"
-        "lea    (%2,%4,4), %2           \n\t"
-        PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)
-        "movq   %%mm4, (%3,%4,2)        \n\t"
-        "movq   %%mm5, (%3,%5)          \n\t"
-        "lea    (%3,%4,4), %3           \n\t"
-        "subl   $4, %0                  \n\t"
-        "jnz    1b                      \n\t"
-        :"+r"(h), "+r"(a), "+r"(b), "+r"(dst)
-        :"r"((x86_reg)stride), "r"((x86_reg)3L*stride)
-        :"memory");
-//    STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx")
-}
-
 #if CONFIG_DIRAC_DECODER
 #define DIRAC_PIXOP(OPNAME, EXT)\
 void ff_ ## OPNAME ## _dirac_pixels8_ ## EXT(uint8_t *dst, const uint8_t *src[5], int stride, int h)\
@@ -2115,8 +2082,6 @@ static void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
 
     c->add_bytes = add_bytes_mmx;
 
-    c->put_no_rnd_pixels_l2= put_vp_no_rnd_pixels8_l2_mmx;
-
     if (CONFIG_H263_DECODER || CONFIG_H263_ENCODER) {
         c->h263_v_loop_filter = h263_v_loop_filter_mmx;
         c->h263_h_loop_filter = h263_h_loop_filter_mmx;
diff --git a/libavcodec/x86/vp3dsp_init.c b/libavcodec/x86/vp3dsp_init.c
index 245b60f6fa..e425f503ac 100644
--- a/libavcodec/x86/vp3dsp_init.c
+++ b/libavcodec/x86/vp3dsp_init.c
@@ -1,4 +1,6 @@
 /*
+ * Copyright (c) 2009 David Conrad <lessen42@gmail.com>
+ *
  * This file is part of FFmpeg.
  *
  * FFmpeg is free software; you can redistribute it and/or
@@ -21,6 +23,7 @@
 #include "libavutil/attributes.h"
 #include "libavutil/cpu.h"
 #include "libavutil/x86/cpu.h"
+#include "libavutil/x86/asm.h"
 #include "libavcodec/avcodec.h"
 #include "libavcodec/vp3dsp.h"
 #include "config.h"
@@ -39,10 +42,68 @@ void ff_vp3_v_loop_filter_mmxext(uint8_t *src, int stride,
 void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride,
                                  int *bounding_values);
 
+#if HAVE_INLINE_ASM
+
+#define MOVQ_BFE(regd)                                  \
+    __asm__ volatile (                                  \
+        "pcmpeqd %%"#regd", %%"#regd"   \n\t"           \
+        "paddb   %%"#regd", %%"#regd"   \n\t" ::)
+
+#define PAVGBP_MMX_NO_RND(rega, regb, regr,  regc, regd, regp)   \
+    "movq  "#rega", "#regr"             \n\t"                    \
+    "movq  "#regc", "#regp"             \n\t"                    \
+    "pand  "#regb", "#regr"             \n\t"                    \
+    "pand  "#regd", "#regp"             \n\t"                    \
+    "pxor  "#rega", "#regb"             \n\t"                    \
+    "pxor  "#regc", "#regd"             \n\t"                    \
+    "pand    %%mm6, "#regb"             \n\t"                    \
+    "pand    %%mm6, "#regd"             \n\t"                    \
+    "psrlq      $1, "#regb"             \n\t"                    \
+    "psrlq      $1, "#regd"             \n\t"                    \
+    "paddb "#regb", "#regr"             \n\t"                    \
+    "paddb "#regd", "#regp"             \n\t"
+
+static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h)
+{
+//    START_TIMER
+    MOVQ_BFE(mm6);
+    __asm__ volatile(
+        "1:                             \n\t"
+        "movq   (%1), %%mm0             \n\t"
+        "movq   (%2), %%mm1             \n\t"
+        "movq   (%1,%4), %%mm2          \n\t"
+        "movq   (%2,%4), %%mm3          \n\t"
+        PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)
+        "movq   %%mm4, (%3)             \n\t"
+        "movq   %%mm5, (%3,%4)          \n\t"
+
+        "movq   (%1,%4,2), %%mm0        \n\t"
+        "movq   (%2,%4,2), %%mm1        \n\t"
+        "movq   (%1,%5), %%mm2          \n\t"
+        "movq   (%2,%5), %%mm3          \n\t"
+        "lea    (%1,%4,4), %1           \n\t"
+        "lea    (%2,%4,4), %2           \n\t"
+        PAVGBP_MMX_NO_RND(%%mm0, %%mm1, %%mm4,   %%mm2, %%mm3, %%mm5)
+        "movq   %%mm4, (%3,%4,2)        \n\t"
+        "movq   %%mm5, (%3,%5)          \n\t"
+        "lea    (%3,%4,4), %3           \n\t"
+        "subl   $4, %0                  \n\t"
+        "jnz    1b                      \n\t"
+        :"+r"(h), "+r"(a), "+r"(b), "+r"(dst)
+        :"r"((x86_reg)stride), "r"((x86_reg)3L*stride)
+        :"memory");
+//    STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx")
+}
+#endif /* HAVE_INLINE_ASM */
+
 av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
 {
     int cpuflags = av_get_cpu_flags();
 
+#if HAVE_INLINE_ASM
+    c->put_no_rnd_pixels_l2 = put_vp_no_rnd_pixels8_l2_mmx;
+#endif /* HAVE_INLINE_ASM */
+
 #if ARCH_X86_32
     if (EXTERNAL_MMX(cpuflags)) {
         c->idct_put  = ff_vp3_idct_put_mmx;