- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi experts, we were using INTEL IPP from 6.x for our video conferencing application, we encountered the H.264 decoder is very slow issue, this issue also reported by other users:
http://software.intel.com/en-us/forums/showpost.php?p=36228
http://software.intel.com/en-us/forums/showthread.php?t=67755
In above threads that INTEL experts provided a hack way to avoid the latency that works fine with 6.x.
But last month we updated IPP from 6.x to "7 update 5", same hack way is not effect, we got about 2-3 seconds latency in LAN, the CPU is i7-2620M 2.7G, 8GB RAM, this is big problem to us, could you provide us a new way to solve it ? We don't want the buffering for decode, just want the realtime, the short latency is really important for video conferencing application.
This is urgent, please help me.
http://software.intel.com/en-us/forums/showpost.php?p=36228
http://software.intel.com/en-us/forums/showthread.php?t=67755
In above threads that INTEL experts provided a hack way to avoid the latency that works fine with 6.x.
But last month we updated IPP from 6.x to "7 update 5", same hack way is not effect, we got about 2-3 seconds latency in LAN, the CPU is i7-2620M 2.7G, 8GB RAM, this is big problem to us, could you provide us a new way to solve it ? We don't want the buffering for decode, just want the realtime, the short latency is really important for video conferencing application.
This is urgent, please help me.
Link Copied
4 Replies
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
Is this problem is related to this one:
http://software.intel.com/en-us/forums/showthread.php?t=102065
Jeff is looking some more information, and hopefully we have some update.
Btw, I noticed a few users tried to improve the latency in the sample by the following code change:
At function Status TaskSupplier::RunDecoding(bool force, H264DecoderFrame ** )?
Instead of:
umcRes = m_pSegmentDecoder[0]->ProcessSegment();
use:
do
{umcRes = m_pSegmentDecoder[0]->ProcessSegment();
} while (umcRes == UMC_OK);
Possibly you can have some check how it works for you there.
Thanks,
Chao
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Thanks Chao, we have been solved this issue as this way:
http://software.intel.com/en-us/forums/showthread.php?t=73188
Could INTEL IPP will has an official way(method) in the next release or future release to disable the DPB for the realtime application ? This is really useful.
http://software.intel.com/en-us/forums/showthread.php?t=73188
Could INTEL IPP will has an official way(method) in the next release or future release to disable the DPB for the realtime application ? This is really useful.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Good day.
Chao, this loop were written in 7.0.0 as i remember, it should be in 7.0.5.free1006, I will try to add DPB control in future.
PS:
Meanwile I had made some measurements and found that old hack works fine for me (It is too late but I think this can be interesting for someone):
Video: 720x576
no hack, no threads:
452/440 - (0.800000 + 0.005428)=0.805428
hack(m_maxDecFrameBuffering=0), no threads:
311/310 - (0.066667 + 0.005466)=0.072133
hack(m_maxDecFrameBuffering=1), 2 threads:
343/340 - (0.199804 + 0.002888)=0.202692
Frame delay + decoding delay in seconds. About 70-80ms lag in best case.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Pavel, the old way really not effect for us in 7 update 5.
Geat news to hearing that you consider add the DPB control in future.
Geat news to hearing that you consider add the DPB control in future.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page