| Message ID | 20170303045420.5036-1-jamrial@gmail.com |
|---|---|
| State | Accepted |
| Commit | e2b7ae4b198c1dc001b3b28476608eaf4daf726c |
| Headers | show |
On Fri, Mar 03, 2017 at 01:54:19AM -0300, James Almer wrote: > This makes ubsan happy and also considerably increases performance on > big endian systems. please include the difference in speed in the commit message if you tested (it sounds like you did) should be fine otherwise thx [...]
On 3/3/2017 12:34 PM, Michael Niedermayer wrote: > On Fri, Mar 03, 2017 at 01:54:19AM -0300, James Almer wrote: >> This makes ubsan happy and also considerably increases performance on >> big endian systems. > > please include the difference in speed in the commit message if you > tested (it sounds like you did) I didn't, Nicolas George did some time ago back the first time i sent this patch. https://ffmpeg.org/pipermail/ffmpeg-devel/2016-February/190027.html > > should be fine otherwise > > thx Pushed. > > [...] > > > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel >
diff --git a/libavutil/md5.c b/libavutil/md5.c index 8c36aa80c4..d3698dcb1d 100644 --- a/libavutil/md5.c +++ b/libavutil/md5.c @@ -86,14 +86,14 @@ static const uint32_t T[64] = { // T[i]= fabs(sin(i+1)<<32) \ if (i < 32) { \ if (i < 16) \ - a += (d ^ (b & (c ^ d))) + X[ i & 15]; \ + a += (d ^ (b & (c ^ d))) + AV_RL32(X+( i & 15));\ else \ - a += ((d & b) | (~d & c)) + X[(1 + 5*i) & 15]; \ + a += ((d & b) | (~d & c)) + AV_RL32(X+((1 + 5*i) & 15));\ } else { \ if (i < 48) \ - a += (b ^ c ^ d) + X[(5 + 3*i) & 15]; \ + a += (b ^ c ^ d) + AV_RL32(X+((5 + 3*i) & 15));\ else \ - a += (c ^ (b | ~d)) + X[( 7*i) & 15]; \ + a += (c ^ (b | ~d)) + AV_RL32(X+(( 7*i) & 15));\ } \ a = b + (a << t | a >> (32 - t)); \ } while (0) @@ -112,11 +112,6 @@ static void body(uint32_t ABCD[4], uint32_t *src, int nblocks) X = src + n * 16; -#if HAVE_BIGENDIAN - for (i = 0; i < 16; i++) - X[i] = av_bswap32(X[i]); -#endif - #if CONFIG_SMALL for (i = 0; i < 64; i++) { CORE(i, a, b, c, d); @@ -173,7 +168,7 @@ void av_md5_update(AVMD5 *ctx, const uint8_t *src, int len) } end = src + (len & ~63); - if (HAVE_BIGENDIAN || (!HAVE_FAST_UNALIGNED && ((intptr_t)src & 3))) { + if (!HAVE_FAST_UNALIGNED && ((intptr_t)src & 3)) { while (src < end) { memcpy(ctx->block, src, 64); body(ctx->ABCD, (uint32_t *) ctx->block, 1);
This makes ubsan happy and also considerably increases performance on big endian systems. Signed-off-by: James Almer <jamrial@gmail.com> --- libavutil/md5.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-)