search for: velea

...e busy I would like to resend this on the mailing list. I've created a pull request https://github.com/xiph/opus/pull/5 to add the testing infrastructure for the changes before adding the actual code. A draft for the rest of the code for the AVX functions can be found here: https://github.com/rvelea/opus/commits/workspace. Timothy advised me to break them into smaller commits before trying to push them so as not to break anything. If anyone is interested in reviewing this please let me know and I will follow up with the rest of the code via pull requests. Thanks, Radu -------------- next p...

Hello, I'm currently looking to see if Opus could benefit from AVX/AVX2 instructions and I've made some experiments optimizing xcorr_kernel_* function in celt using AVX2 intrinsics and got about 1.5x-1.6x speedup over SSE floating point version in pitch_sse.c. Have AVX optimizations been tried in Opus? Is there someone I could talk to about optimizing compute intensive tasks in Opus using

...view. Thanks, Radu [1] https://github.com/xiph/opus/pull/3 -----Original Message----- From: opus-bounces at xiph.org [mailto:opus-bounces at xiph.org] On Behalf Of Timothy B. Terriberry Sent: Wednesday, October 21, 2015 11:29 AM To: opus at xiph.org Subject: Re: [opus] AVX Optimizations "Velea, Radu" <radu.velea at intel.com> wrote: > Have AVX optimizations been tried in Opus? Is there someone I could > talk to about optimizing compute intensive tasks in Opus using AVX? No, I'm not aware of anyone trying out AVX. If you want a more real-time exchange, the IRC ch...

Hello, I've integrated the tests and AVX pitch code into my fork: https://github.com/rvelea/opus/commits/master I ran test_opus_decode and it seems to be running OK. I'll do some more measurements with VTUNE and the new pitch test in the following days and get back to you with a pull request and the data. In the meantime if you have time to give it a quick look and maybe point out an...

...od observation. Please go ahead and correct the patch. Thanks, Radu -----Original Message----- From: opus-bounces at xiph.org [mailto:opus-bounces at xiph.org] On Behalf Of Timothy B. Terriberry Sent: Thursday, November 5, 2015 11:08 AM To: opus at xiph.org Subject: Re: [opus] AVX Optimizations Velea, Radu wrote: > Yes, > > Thank you. I'll follow up with the AVX code and tests for pitch code. Actually, I lied. Because you update opus_select_arch(), you can now return a value for arch (4) that is larger than the maximum we currently support (3). This doesn't actually cause fail...

"Velea, Radu" <radu.velea at intel.com> wrote: > Have AVX optimizations been tried in Opus? Is there someone I could > talk to about optimizing compute intensive tasks in Opus using AVX? No, I'm not aware of anyone trying out AVX. If you want a more real-time exchange, the IRC ch...

...here are different version of AVX; maybe clarify which versions you use, and how you detect the availability of the instructions needed. Also you could point out which CPUs are expected to be supported. Also consider that besides Intel there still exists AMD ;-) Regards, Ulrich >>> "Velea, Radu" <radu.velea at intel.com> schrieb am 20.01.2016 um 19:02 in Nachricht <81B494F3CB9BA140BC4E7A506215EB0C5C442CFE at irsmsx105.ger.corp.intel.com>: > Hello, > > I had talked earlier with 'Timothy B. Terriberry' <tterribe at xiph.org>, about > adding...

...I'll follow up with the AVX code and tests for pitch code. Radu -----Original Message----- From: opus-bounces at xiph.org [mailto:opus-bounces at xiph.org] On Behalf Of Timothy B. Terriberry Sent: Thursday, November 5, 2015 10:31 AM To: opus at xiph.org Subject: Re: [opus] AVX Optimizations Velea, Radu wrote: > I've created a pull request[1] to enable configuration to search for AVX support and set OPUS_X86_MAY_HAVE_AVX accordingly. > Please review. These commits both look fine. Would you like me to land them? _______________________________________________ opus mailing list opu...

Velea, Radu wrote: > I've created a pull request[1] to enable configuration to search for AVX support and set OPUS_X86_MAY_HAVE_AVX accordingly. > Please review. These commits both look fine. Would you like me to land them?

Velea, Radu wrote: > Sorry. I missed that. Good observation. > Please go ahead and correct the patch. Pushed: https://git.xiph.org/?p=opus.git;a=commitdiff;h=bb0e1e0d6f6b https://git.xiph.org/?p=opus.git;a=commitdiff;h=1632152b83b8

Velea, Radu wrote: > In the meantime if you have time to give it a quick look and maybe point out any errors or problems at this stage I'd greatly appreciate it. I took a brief look and didn't see any obvious bugs, but we will want you to organize the commits so that each one is independentl...

Velea, Radu wrote: > Yes, > > Thank you. I'll follow up with the AVX code and tests for pitch code. Actually, I lied. Because you update opus_select_arch(), you can now return a value for arch (4) that is larger than the maximum we currently support (3). This doesn't actually cause fa...