Question about the Performance on IPP 5.2

[hjchoi]

Hello, All.

A week ago, I tested umc_video_enc_conand umc_h264_dec_con sample binaries.

In brief, I wanted h264file to be converted to yuv then converted to h263.(h264->yuv->h263).

My linux server systemwasequipedwithtwo Intel Xeon CPU@ 2.00GHz and2GByte Memory.

Sample h264 file was 6MB, CIF size,and had 12 seconds playing time.

I set one-cycle as (h264->yuv->h263) and then tested n times sequentially with shell script.

The output is as follows:

times converting time

1 4.3s

2 20.4s

330.8s

4 53.6s

5 73.6

->> I wonder why the time did not increase linearly. And the variation of converting timewas too high whe never I tested. Does the shell script matter?? Is it better to modify sample codes?

I also set one-cycle as (h264->yuv->h263) and then tested n timesconcurrently with shell script.

The output is as follows:

times converting time

1 4.2s

2 5.7s

321.8s

4 30.6s

5 35.6

->> I wonder why the time did not increase linearly. And the variation of converting timewas too high whenever I tested. Does the shell script matter?? Is it better to modify sample codes?

Please give me some answers or any comment.

Regard,

Charles Choi

[Ying_S_Intel]

Thanks for reporting this issue, we have generated this in issue 441597 via Intel Premier and our support engineer will contact you shortly.

Regards,
Ying

[Chao_Y_Intel]

Hello Choi,

Have you noticed how much IO time in your test? Also could you let us know your script file? That will help us to do some test here. You can submit the script file in your premier support issue 441597.

Thanks,

Chao