topic H.264 - at least one optimized deblock function is incorrect in Intel® Integrated Performance Primitives

H.264 - at least one optimized deblock function is incorrect

registered_user — Fri, 24 Jul 2009 12:21:20 GMT

I'm on Penryn (P9500), hence _p8_ownFilterDeblockingChroma_HorEdge_H264_8u_C1IR_V8 is used, which gives different result compared to both JM and unoptimized IPP version. (both 6.0 and latest 6.1 upd1)

Bug is committed to premier support - posting here just to let others know, that output of H.264 encoder/decoder in 4:2:2 mode is corrupted for chroma on Core 2/Xeon5100 series and higher.

Re: H.264 - at least one optimized deblock function is incorrec

Vladimir_Dudnik — Tue, 28 Jul 2009 13:52:01 GMT

Thanks for informing here. When you link with IPP static libraries you may choose to dispatchCore2 optimized code (V8 libraries) instead of Core i7 optimized code where you discover the issue (P8). You may dispatch processor specific code with ippStaticInitCpu call.

Regards,
Vladimir

Re: H.264 - at least one optimized deblock function is incorrec

j_miles — Fri, 25 Sep 2009 05:36:54 GMT

Has this issue been fixed? And if so, when will it be available? I assume that the issue exists in the latest 6.1-1 as well...

- Jay

Re: H.264 - at least one optimized deblock function is incorrec

j_miles — Wed, 04 Nov 2009 09:58:17 GMT

Quoting - j_miles

Has this issue been fixed? And if so, when will it be available? I assume that the issue exists in the latest 6.1-1 as well...

- Jay

Bump!
Any news - has it been fixed? I noticed an update 2 (6.1-2) available now...

- Jay

Re: H.264 - at least one optimized deblock function is incorrec

Vladimir_Dudnik — Fri, 06 Nov 2009 18:59:33 GMT

Hi Jay,

I think it should be fixed in the latest update. Sorry, can't provide exact answer right now (do not have the issue number to check with data base).

Regards,
Vladimir

Re: H.264 - at least one optimized deblock function is incorrec

j_miles — Mon, 09 Nov 2009 20:29:19 GMT

Quoting - Vladimir Dudnik (Intel)

Hi Jay,

I think it should be fixed in the latest update. Sorry, can't provide exact answer right now (do not have the issue number to check with data base).

Regards,
Vladimir

Hi Vladimir,

Thanks. I would really appreciate that you double-checked the database.

- Jay