[FFmpeg-trac] #9149(swscale:new): bugs in Altivec code

FFmpeg trac at avcodec.org
Wed Mar 17 13:29:25 EET 2021

#9149: bugs in Altivec code
             Reporter:  ilyakurdyukov  |                    Owner:
                 Type:  defect         |                   Status:  new
             Priority:  normal         |                Component:  swscale
              Version:  unspecified    |               Resolution:
             Keywords:  ppc altivec    |               Blocked By:
             Blocking:                 |  Reproduced by developer:  0
Analyzed by developer:  0              |

Comment (by ilyakurdyukov):

 "If the issues you see are reproducible with current FFmpeg git head"

 As I said, I can't test this because I don't have Altivec hardware to test
 it, but I'm pretty sure this code is wrong as I ported this code to a
 different SIMD architecture and changed these parts with the correct code.

 And I have no intention of fixing the Altivec code, when I told my
 colleagues that I found bugs in the Altivec code during porting, they
 recommended reporting this to the ffmpeg bug tracker, which I did.

 Also I found strange code in "libswscale/ppc/yuv2rgb_altivec.c":

 yuv2packedX_altivec() has some code after the loop that calculates the
 extra remaining bytes that are less than 16x RGB blocks. But this code
 will never be called, because of "i < dstW" conditional, it should be "i +
 15 < dstW".

     for (i = 0; i < dstW; i += 16) {

     if (i < dstW) {
         i -= 16;
         memcpy(&((uint32_t *) dest)[i], scratch, (dstW - i) / 4);

 And altivec_uyvy_rgb32() in the same source file - seems to be never
 called in the current version of ffmpeg.

Ticket URL: <https://trac.ffmpeg.org/ticket/9149#comment:3>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker

More information about the FFmpeg-trac mailing list