nvenc: drop the hard dependency on CUDA

[ffmpeg] / libavcodec / lagarith.c

diff --git a/libavcodec/lagarith.c b/libavcodec/lagarith.c
index 292d952a7d2e38b8139be6c85fde69a758acba5f..55e10bce55f8389bf5732b6fed9877247bfd7d9b 100644 (file)
--- a/libavcodec/lagarith.c
+++ b/libavcodec/lagarith.c
@@ -25,10 +25,12 @@
  * @author Nathan Caldwell
  */
 
+#include <inttypes.h>
+
 #include "avcodec.h"
 #include "get_bits.h"
 #include "mathops.h"
-#include "dsputil.h"
+#include "huffyuvdsp.h"
 #include "lagarithrac.h"
 #include "thread.h"
 
@@ -48,19 +50,20 @@ enum LagarithFrameType {
 
 typedef struct LagarithContext {
     AVCodecContext *avctx;
-    DSPContext dsp;
+    HuffYUVDSPContext hdsp;
     int zeros;                  /**< number of consecutive zero bytes encountered */
     int zeros_rem;              /**< number of zero bytes remaining to output */
     uint8_t *rgb_planes;
+    int      rgb_planes_allocated;
     int rgb_stride;
 } LagarithContext;
 
 /**
- * Compute the 52bit mantissa of 1/(double)denom.
+ * Compute the 52-bit mantissa of 1/(double)denom.
  * This crazy format uses floats in an entropy coder and we have to match x86
  * rounding exactly, thus ordinary floats aren't portable enough.
  * @param denom denominator
- * @return 52bit mantissa
+ * @return 52-bit mantissa
  * @see softfloat_mul
  */
 static uint64_t softfloat_reciprocal(uint32_t denom)
@@ -77,9 +80,9 @@ static uint64_t softfloat_reciprocal(uint32_t denom)
 /**
  * (uint32_t)(x*f), where f has the given mantissa, and exponent 0
  * Used in combination with softfloat_reciprocal computes x/(double)denom.
- * @param x 32bit integer factor
+ * @param x 32-bit integer factor
  * @param mantissa mantissa of f with exponent 0
- * @return 32bit integer value (x*f)
+ * @return 32-bit integer value (x*f)
  * @see softfloat_reciprocal
  */
 static uint32_t softfloat_mul(uint32_t x, uint64_t mantissa)
@@ -222,8 +225,8 @@ static void add_lag_median_prediction(uint8_t *dst, uint8_t *src1,
                                       uint8_t *diff, int w, int *left,
                                       int *left_top)
 {
-    /* This is almost identical to add_hfyu_median_prediction in dsputil.h.
-     * However the &0xFF on the gradient predictor yealds incorrect output
+    /* This is almost identical to add_hfyu_median_pred in huffyuvdsp.h.
+     * However the &0xFF on the gradient predictor yields incorrect output
      * for lagarith.
      */
     int i;
@@ -250,8 +253,7 @@ static void lag_pred_line(LagarithContext *l, uint8_t *buf,
     if (!line) {
         int i, align_width = (width - 1) & ~31;
         /* Left prediction only for first line */
-        L = l->dsp.add_hfyu_left_prediction(buf + 1, buf + 1,
-                                            align_width, buf[0]);
+        L = l->hdsp.add_hfyu_left_pred(buf + 1, buf + 1, align_width, buf[0]);
         for (i = align_width + 1; i < width; i++)
             buf[i] += buf[i - 1];
     } else {
@@ -286,7 +288,7 @@ static void lag_pred_line_yuy2(LagarithContext *l, uint8_t *buf,
         }
 
         align_width = (width - 1) & ~31;
-        l->dsp.add_hfyu_left_prediction(buf + 1, buf + 1, align_width, buf[0]);
+        l->hdsp.add_hfyu_left_pred(buf + 1, buf + 1, align_width, buf[0]);
 
         for (i = align_width + 1; i < width; i++)
             buf[i] += buf[i - 1];
@@ -311,8 +313,7 @@ static void lag_pred_line_yuy2(LagarithContext *l, uint8_t *buf,
     } else {
         TL = buf[width - (2 * stride) - 1];
         L  = buf[width - stride - 1];
-        l->dsp.add_hfyu_median_prediction(buf, buf - stride, buf, width,
-                                          &L, &TL);
+        l->hdsp.add_hfyu_median_pred(buf, buf - stride, buf, width, &L, &TL);
     }
 }
 
@@ -448,7 +449,7 @@ static int lag_decode_arith_plane(LagarithContext *l, uint8_t *dst,
 
         if (read > length)
             av_log(l->avctx, AV_LOG_WARNING,
-                   "Output more bytes than length (%d of %d)\n", read,
+                   "Output more bytes than length (%d of %"PRIu32")\n", read,
                    length);
     } else if (esc_count < 8) {
         esc_count -= 4;
@@ -564,13 +565,12 @@ static int lag_decode_frame(AVCodecContext *avctx,
         offs[1] = offset_gu;
         offs[2] = offset_ry;
 
+        l->rgb_stride = FFALIGN(avctx->width, 16);
+        av_fast_malloc(&l->rgb_planes, &l->rgb_planes_allocated,
+                       l->rgb_stride * avctx->height * planes + 1);
         if (!l->rgb_planes) {
-            l->rgb_stride = FFALIGN(avctx->width, 16);
-            l->rgb_planes = av_malloc(l->rgb_stride * avctx->height * planes + 1);
-            if (!l->rgb_planes) {
-                av_log(avctx, AV_LOG_ERROR, "cannot allocate temporary buffer\n");
-                return AVERROR(ENOMEM);
-            }
+            av_log(avctx, AV_LOG_ERROR, "cannot allocate temporary buffer\n");
+            return AVERROR(ENOMEM);
         }
         for (i = 0; i < planes; i++)
             srcs[i] = l->rgb_planes + (i + 1) * l->rgb_stride * avctx->height - l->rgb_stride;
@@ -666,7 +666,7 @@ static int lag_decode_frame(AVCodecContext *avctx,
         break;
     default:
         av_log(avctx, AV_LOG_ERROR,
-               "Unsupported Lagarith frame type: %#x\n", frametype);
+               "Unsupported Lagarith frame type: %#"PRIx8"\n", frametype);
         return -1;
     }
 
@@ -680,7 +680,7 @@ static av_cold int lag_decode_init(AVCodecContext *avctx)
     LagarithContext *l = avctx->priv_data;
     l->avctx = avctx;
 
-    ff_dsputil_init(&l->dsp, avctx);
+    ff_huffyuvdsp_init(&l->hdsp);
 
     return 0;
 }
@@ -703,5 +703,5 @@ AVCodec ff_lagarith_decoder = {
     .init           = lag_decode_init,
     .close          = lag_decode_end,
     .decode         = lag_decode_frame,
-    .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
+    .capabilities   = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
 };