Add V210 SIMD

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

1 files changed, 72 insertions, 17 deletions

diff --git a/libavcodec/v210dec.c b/libavcodec/v210dec.c
index ecd88be22b..4f40e0872f 100644
--- a/libavcodec/v210dec.c
+++ b/libavcodec/v210dec.c
@@ -22,10 +22,35 @@
  */
 
 #include "avcodec.h"
+#include "v210dec.h"
 #include "libavutil/bswap.h"
+#include "libavutil/x86/timer.h"
+
+#define READ_PIXELS(a, b, c)         \
+    do {                             \
+        val  = av_le2ne32(*src++);   \
+        *a++ =  val & 0x3FF;         \
+        *b++ = (val >> 10) & 0x3FF;  \
+        *c++ = (val >> 20) & 0x3FF;  \
+    } while (0)
+
+static void v210_planar_unpack_c(const uint32_t *src, uint16_t *y, uint16_t *u, uint16_t *v, int width)
+{
+    uint32_t val;
+    int i;
+
+    for( i = 0; i < width-5; i += 6 ){
+        READ_PIXELS(u, y, v);
+        READ_PIXELS(y, u, y);
+        READ_PIXELS(v, y, u);
+        READ_PIXELS(y, v, y);
+    }
+}
 
 static av_cold int decode_init(AVCodecContext *avctx)
 {
+    V210DecContext *s = avctx->priv_data;
+
     if (avctx->width & 1) {
         av_log(avctx, AV_LOG_ERROR, "v210 needs even width\n");
         return -1;
@@ -35,18 +60,37 @@ static av_cold int decode_init(AVCodecContext *avctx)
 
     avctx->coded_frame         = avcodec_alloc_frame();
 
+    s->unpack_frame            = v210_planar_unpack_c;
+
+    if (HAVE_MMX)
+        v210_x86_init(s);
+
     return 0;
 }
 
 static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
                         AVPacket *avpkt)
 {
-    int h, w;
+    V210DecContext *s = avctx->priv_data;
+
+    int h, w, stride, aligned_input;
     AVFrame *pic = avctx->coded_frame;
     const uint8_t *psrc = avpkt->data;
     uint16_t *y, *u, *v;
-    int aligned_width = ((avctx->width + 47) / 48) * 48;
-    int stride = aligned_width * 8 / 3;
+
+    if (s->custom_stride )
+        stride = s->custom_stride;
+    else {
+        int aligned_width = ((avctx->width + 47) / 48) * 48;
+        stride = aligned_width * 8 / 3;
+    }
+
+    aligned_input = !((uintptr_t)psrc & 0xf) && !(stride & 0xf);
+    if (aligned_input != s->aligned_input) {
+        s->aligned_input = aligned_input;
+        if (HAVE_MMX)
+            v210_x86_init(s);
+    }
 
     if (pic->data[0])
         avctx->release_buffer(avctx, pic);
@@ -66,23 +110,18 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
     pic->pict_type = AV_PICTURE_TYPE_I;
     pic->key_frame = 1;
 
-#define READ_PIXELS(a, b, c)         \
-    do {                             \
-        val  = av_le2ne32(*src++);   \
-        *a++ =  val & 0x3FF;         \
-        *b++ = (val >> 10) & 0x3FF;  \
-        *c++ = (val >> 20) & 0x3FF;  \
-    } while (0)
-
     for (h = 0; h < avctx->height; h++) {
         const uint32_t *src = (const uint32_t*)psrc;
         uint32_t val;
-        for (w = 0; w < avctx->width - 5; w += 6) {
-            READ_PIXELS(u, y, v);
-            READ_PIXELS(y, u, y);
-            READ_PIXELS(v, y, u);
-            READ_PIXELS(y, v, y);
-        }
+
+        w = (avctx->width / 6) * 6;
+        s->unpack_frame(src, y, u, v, w);
+
+        y += w;
+        u += w >> 1;
+        v += w >> 1;
+        src += (w << 1) / 3;
+
         if (w < avctx->width - 1) {
             READ_PIXELS(u, y, v);
 
@@ -120,13 +159,29 @@ static av_cold int decode_close(AVCodecContext *avctx)
     return 0;
 }
 
+#define V210DEC_FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
+static const AVOption v210dec_options[] = {
+    {"custom_stride", "Custom V210 stride", offsetof(V210DecContext, custom_stride), FF_OPT_TYPE_INT,
+     {.dbl = 0}, INT_MIN, INT_MAX, V210DEC_FLAGS},
+    {NULL}
+};
+
+static const AVClass v210dec_class = {
+    "V210 Decoder",
+    av_default_item_name,
+    v210dec_options,
+    LIBAVUTIL_VERSION_INT,
+};
+
 AVCodec ff_v210_decoder = {
     .name           = "v210",
     .type           = AVMEDIA_TYPE_VIDEO,
     .id             = CODEC_ID_V210,
+    .priv_data_size = sizeof(V210DecContext),
     .init           = decode_init,
     .close          = decode_close,
     .decode         = decode_frame,
     .capabilities   = CODEC_CAP_DR1,
     .long_name = NULL_IF_CONFIG_SMALL("Uncompressed 4:2:2 10-bit"),
+    .priv_class     = &v210dec_class,
 };