- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
there is a table of color conversions that are supported by VPP, e.g. YUY2 to NV12 or Y210 to NV12, etc.
I would like to know which of these conversion are handled by HW (the SFC block of Gen Media), which are done on EUs of Gen Gfx and which are maybe even done on the CPU with IPP.
where can I find the information about this or can somebody tell me ?
- Tags:
- Development Tools
- Graphics
- Intel® Media SDK
- Intel® Media Server Studio
- Media Processing
- Optimization
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Michael,
If you check the table 8 in page 27 of the following dev guide, does it solve your problem?
https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.pdf
Mark
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Mark,
to be honest , no. This is a very interesting document I was not aware yet, but it does not tell me which conversion is hw acclerated and which not. I would like to know even more, functions can are avaialbel as hw preprocessing and which are postrpocessing.
regards
Michael
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Michael,
Sorry for the late response. I have checked and as far as I can tell, the Media SDK VPP color conversion is done by the hardware, we called VEBox in the GPU.
Mark
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Mark,
thanks, very helpful.
This means for all the color conversions functions we could expect to get the same performance as for VDBox, as it could be pipelined with VEBox, correct ? which means 4K60 in the case of Apollolake.
When I let sample_vpp run the performance I measure is ~20 MPix/sec, which is far off the 497 MPix/s that correspond to 4K60. Any idea what is limiting here ?
I would assume that the measurements of the example exclude file i/o timeand other potential overhead , true ?
Michael
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thanks Michael,
Do you have the command line to reproduce the issues? I want to try it to check.
I think the measurement should not include the file i/o but I want to check it before I answer you.
Mark
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page