Simple opaque transcoding demo GPU performance not captured by System Analyzer

Simple opaque transcoding demo GPU performance not captured by System Analyzer

Folks,

We've been trying to use System Analyzer to examine the performance of simple_transcode_opaque_async_vppresize demo. 

For some reason, however, the MFX Engine Usage percent shows almost no activity, and similar situation is with GPU overall usage.

Is there a reason (such as mismatch between the System Analyzer tool and API 1.3) for the inability to get the actual usage ? 

As a comparison point, we also run simple_2_decode demo, which showed the expected activity on both items (MFX Engine Usage/ GPU overall usage).

We use the Intel GPA Monitor 2013 R2 (13.2.200837)

Thanks,

Mamil

2 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Hi Mamil,

I checked the same setup locally and I do not see any issues with what Intel GPA System Analyzer reports. MFX and GPU usage data looks correct.

Make sure that your system supports HW acceleration and that the "simple_transcode_opaque_async_vppresize" sample executes as expected. Keep in mind that this specific sample requires command line input.

When running the "simple_transcode_opaque_async_vppresize" sample on my machine I also noticed a behavior I had not seen before in this sample. EncodeFrameAsync sometime returns MFX_ERR_MORE_DATA repeatedly at start of pipeline execution (this is expected), but since the following action for this async use case is to read more data into bit stream buffer, the ReadBitStreamData occasionally misinterprets the case when the buffer is already full as if it has reached the end of the input file. If this is the case for you, just add the statement "MSDK_IGNORE_MFX_STS(sts, MFX_ERR_MORE_DATA)", after the EncodeFrameAsync call.

Regards,
Petter 

Leave a Comment

Please sign in to add a comment. Not a member? Join today