A bug in ippiFilterDeblockingChroma_HorEdge_H264_8u_C1IR

[nagataki]

Hello.

I've found a bug of ippiFilterDeblockingChroma_HorEdge_H264_8u_C1IR() in IPP5.1.
The following code can prove it.
If you replace ippCpuSSE2 with ippCpuPMX, the result of the function
has no trouble.

ippiFilterDeblockingChroma_VerEdge_(MBAFF)_H264_8u_C1IR() may also have such a
problem like this.

Thanks.

--
#include

int main(void)
{
Ipp8u pSrcDst[8*8] =
{
0x78, 0x77, 0x76, 0x74, 0x73, 0x73, 0x74, 0x75,
0x78, 0x77, 0x76, 0x74, 0x73, 0x73, 0x75, 0x79,
0x76, 0x76, 0x76, 0x75, 0x75, 0x75, 0x7c, 0x7e,
0x75, 0x75, 0x75, 0x76, 0x76, 0x76, 0x7c, 0x7e,
0x78, 0x79, 0x7a, 0x7c, 0x7d, 0x7e, 0x81, 0x82,
0x78, 0x79, 0x7a, 0x7c, 0x7d, 0x7e, 0x81, 0x82,
0x7a, 0x7b, 0x7c, 0x7e, 0x7e, 0x7f, 0x81, 0x82,
0x7d, 0x7e, 0x7f, 0x81, 0x81, 0x82, 0x82, 0x82
};
Ipp8u pAlpha[2] = {25,25};
Ipp8u pBeta[2] = {8,8};
Ipp8u pThreshold[16] = {255,255,255,255,255,255,1,1,204,204,204,204,204,204,204,204};
Ipp8u pBS[16] = {0,0,0,0,0,0,0,2,0,0,1,2,0,0,0,0};

ippStaticInitCpu(ippCpuSSE2);

ippiFilterDeblockingChroma_HorEdge_H264_8u_C1IR(pSrcDst,
8,
pAlpha,
pBeta,
pThreshold,
pBS);
}

[Vladimir_Dudnik]

thank you a lot for informing us about that. I will pass this info to our engineers and will inform you about status of that issue. Please note, you also can submit your bug report to Intel Premier Support to get notification about status of that issue automatically.

Regards,
Vladimir

[Vladimir_Dudnik]

Thanks for your report, our develoopers have verified and confirmed that there is a bug in optimized version of that function. We will fix this in our nearest release.

Regards,
Vladimir

[Vladimir_Dudnik]

Hello,

after looking at this issue more carefully, our expert commented that:

according H.264 specification (see table 8-17) threshold values should be in 0..25 range. So, it look like you use wrong parameters in the function call. Could you please clarify where from did you take these values?

Regards,
Vladimir