topic Re: MPEG-2 Decoding Sample Memory Usage in Intel® Integrated Performance Primitives

MPEG-2 Decoding Sample Memory Usage

jere_jones1 — Tue, 08 Feb 2005 00:59:45 GMT

Hi all,

I compiled the MPEG-2 Decoding sample using MSVS.NET 2003 and IPP 4.1 on a Pentium-4M machine.It is really fast at the beginning but it uses a lot of memory. On my 512M system, the decoder's memory usage steadily climbs until it reaches 420-430M.

I tracked the memory allocation to this macro:

Code:

#define BSWAP(x)    
    (Ipp32u)(((x) << 24) + (((x)&0xff00) << 8) + 
     (((x) >> 8)&0xff00) + ((x) >> 24))

Every 6th time that macro is called, the memory usage increases by 4k. Themacro is called via:

Code:

return BSWAP(*code);

I have isolated the macro from the return and the allocation follows the macro.

I don't understand why this allocates memory or how to fix it.

I appreciate any help.

Jere

Re: MPEG-2 Decoding Sample Memory Usage

jere_jones1 — Mon, 14 Feb 2005 09:51:23 GMT

For those that are curios, the problem was in the bitstream routines. Apparently they try to map the ENTIRE mpeg into memory at once.

Memory was being allocated as soon as the byte was needed (i.e. read).

A rewrite of those routines fixed it.

Jere

Re: MPEG-2 Decoding Sample Memory Usage

Vladimir_Dudnik — Thu, 17 Feb 2005 06:28:43 GMT

Hi Jere,

we are trying to keep sample code as simple as possible, it was the reason for using such approach. If you feel it is not appropriate you can submit this issue to premier.intel.com

Regards,

Vladimir

Re: MPEG-2 Decoding Sample Memory Usage

jere_jones1 — Thu, 17 Feb 2005 07:11:23 GMT

It is an appropriate example of mpeg-2 decoding routines. It is not an appropriate example of a bitstream. But that's ok. It's not advertised as a bitstream example.

I learned quite a bit fromthe sample(much more than I expected) and appreciate it.

BTW, thanks for all your effort responding to all the questions!

Jere

Re: MPEG-2 Decoding Sample Memory Usage

Vladimir_Dudnik — Thu, 17 Feb 2005 07:17:39 GMT

Hi Jere,

yes, you are right, this particular sample can be better. Actually we are trying to keep continuously improve our samples, so I hope in the next versions it will became more clean.

Thanks you for your interest to IPP.

Regards,

Vladimir