Commit 446d62f0 authored by Michael Niedermayer's avatar Michael Niedermayer

Merge commit '69c25c92'

* commit '69c25c92':
  dnxhdenc: fix invalid reads in dnxhd_mb_var_thread().
  x86: h264qpel: Move stray comment to the right spot and clarify it
  atrac3: use correct loop variable in add_tonal_components()

Conflicts:
	tests/ref/vsynth/vsynth1-dnxhd-1080i
	tests/ref/vsynth/vsynth2-dnxhd-1080i
Merged-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parents d235d240 69c25c92
......@@ -629,14 +629,35 @@ static void dnxhd_setup_threads_slices(DNXHDEncContext *ctx)
static int dnxhd_mb_var_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
{
DNXHDEncContext *ctx = avctx->priv_data;
int mb_y = jobnr, mb_x;
int mb_y = jobnr, mb_x, x, y;
int partial_last_row = (mb_y == ctx->m.mb_height - 1) &&
((avctx->height >> ctx->interlaced) & 0xF);
ctx = ctx->thread[threadnr];
if (ctx->cid_table->bit_depth == 8) {
uint8_t *pix = ctx->thread[0]->src[0] + ((mb_y<<4) * ctx->m.linesize);
for (mb_x = 0; mb_x < ctx->m.mb_width; ++mb_x, pix += 16) {
unsigned mb = mb_y * ctx->m.mb_width + mb_x;
int sum = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
int varc = (ctx->m.dsp.pix_norm1(pix, ctx->m.linesize) - (((unsigned)sum*sum)>>8)+128)>>8;
int sum;
int varc;
if (!partial_last_row && mb_x * 16 <= avctx->width - 16) {
sum = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
varc = ctx->m.dsp.pix_norm1(pix, ctx->m.linesize);
} else {
int bw = FFMIN(avctx->width - 16 * mb_x, 16);
int bh = FFMIN((avctx->height >> ctx->interlaced) - 16 * mb_y, 16);
sum = varc = 0;
for (y = 0; y < bh; y++) {
for (x = 0; x < bw; x++) {
uint8_t val = pix[x + y * ctx->m.linesize];
sum += val;
varc += val * val;
}
}
}
varc = (varc - (((unsigned)sum * sum) >> 8) + 128) >> 8;
ctx->mb_cmp[mb].value = varc;
ctx->mb_cmp[mb].mb = mb;
}
......
......@@ -2309,8 +2309,6 @@ static void dsputil_init_avx(DSPContext *c, AVCodecContext *avctx, int mm_flags)
const int bit_depth = avctx->bits_per_raw_sample;
if (bit_depth == 10) {
// AVX implies !cache64.
// TODO: Port cache(32|64) detection from x264.
if (CONFIG_H264CHROMA) {
c->put_h264_chroma_pixels_tab[0] = ff_put_h264_chroma_mc8_10_avx;
c->avg_h264_chroma_pixels_tab[0] = ff_avg_h264_chroma_mc8_10_avx;
......
......@@ -611,6 +611,10 @@ void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth)
}
if (EXTERNAL_AVX(mm_flags)) {
/* AVX implies 64 byte cache lines without the need to avoid unaligned
* memory accesses that cross the boundary between two cache lines.
* TODO: Port X264_CPU_CACHELINE_32/64 detection from x264 to avoid
* having to treat SSE2 functions with such properties as AVX. */
if (bit_depth == 10) {
H264_QPEL_FUNCS_10(1, 0, sse2);
H264_QPEL_FUNCS_10(2, 0, sse2);
......
9a4781b0a052d9efaafbaf8893db9632 *tests/data/fate/vsynth1-dnxhd-1080i.mov
124c991ee3ac0caef39a58a45287a762 *tests/data/fate/vsynth1-dnxhd-1080i.mov
3031911 tests/data/fate/vsynth1-dnxhd-1080i.mov
e55bf857297ba4d911a9d17a984b125d *tests/data/fate/vsynth1-dnxhd-1080i.out.rawvideo
a09132c6db44f415e831dcaa630a351b *tests/data/fate/vsynth1-dnxhd-1080i.out.rawvideo
stddev: 6.29 PSNR: 32.15 MAXDIFF: 64 bytes: 7603200/ 760320
93b878dcf8f2ecc9798d0e0885c9eec9 *tests/data/fate/vsynth2-dnxhd-1080i.mov
5d7ab75ce6e547ed63a7a0eacf18f078 *tests/data/fate/vsynth2-dnxhd-1080i.mov
3031911 tests/data/fate/vsynth2-dnxhd-1080i.mov
27edc8dfe2ca19097c7f9119705b3a60 *tests/data/fate/vsynth2-dnxhd-1080i.out.rawvideo
744ba46da5d4c19a28562ea31061d170 *tests/data/fate/vsynth2-dnxhd-1080i.out.rawvideo
stddev: 1.31 PSNR: 45.77 MAXDIFF: 23 bytes: 7603200/ 760320
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment