search for: silk_biquad_alt

Hi, Attached are 2 patches related to silk_biquad_alt() optimization. Please review. Thanks, Linfeng Zhang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.xiph.org/pipermail/opus/attachments/20170419/f08f5030/attachment-0001.html> -------------- next part -------------- A non-text attachment was scr...

...at rz.uni-regensburg.de> wrote: > >>> Linfeng Zhang <linfengz at google.com> schrieb am 19.04.2017 um 18:29 in > Nachricht > <CAKoqLCDX3eCUGbnZFvRzhiCV1Mbo2ksbj8K+pcVu60Dvit7WCQ at mail.gmail.com>: > > Hi, > > > > Attached are 2 patches related to silk_biquad_alt() optimization. Please > > review. > > Out of curiosity: How do you test "The optimization is bit exact with C > function."? Use one example, or use a few random examples, or use a defined > test suite? Or is is done per code inspection? > > Regards, > Ulrich &g...

>>> Linfeng Zhang <linfengz at google.com> schrieb am 19.04.2017 um 18:29 in Nachricht <CAKoqLCDX3eCUGbnZFvRzhiCV1Mbo2ksbj8K+pcVu60Dvit7WCQ at mail.gmail.com>: > Hi, > > Attached are 2 patches related to silk_biquad_alt() optimization. Please > review. Out of curiosity: How do you test "The optimization is bit exact with C function."? Use one example, or use a few random examples, or use a defined test suite? Or is is done per code inspection? Regards, Ulrich > > Thanks, > Linfeng Zhang

Hi Linfeng, Thanks for the patches. I'll have a look and get back to you. What kind of speedup are you getting for these functions? On what command line? Cheers, Jean-Marc On 19/04/17 12:29 PM, Linfeng Zhang wrote: > Hi, > > Attached are 2 patches related to silk_biquad_alt() optimization. Please > review. > > Thanks, > Linfeng Zhang > > > > _______________________________________________ > opus mailing list > opus at xiph.org > http://lists.xiph.org/mailman/listinfo/opus >

...rote: > >> >>> Linfeng Zhang <linfengz at google.com> schrieb am 19.04.2017 um 18:29 in >> Nachricht >> <CAKoqLCDX3eCUGbnZFvRzhiCV1Mbo2ksbj8K+pcVu60Dvit7WCQ at mail.gmail.com>: >> > Hi, >> > >> > Attached are 2 patches related to silk_biquad_alt() optimization. Please >> > review. >> >> Out of curiosity: How do you test "The optimization is bit exact with C >> function."? Use one example, or use a few random examples, or use a defined >> test suite? Or is is done per code inspection? >> >...

...get back to you. What kind > of speedup are you getting for these functions? On what command line? > > Cheers, > > Jean-Marc > > On 19/04/17 12:29 PM, Linfeng Zhang wrote: > > Hi, > > > > Attached are 2 patches related to silk_biquad_alt() optimization. > Please > > review. > > > > Thanks, > > Linfeng Zhang > > > > > > > > _______________________________________________ > > opus mailing list > > opus at xiph.org <mailto:opu...

...ches. I'll have a look and get back to you. What kind > of speedup are you getting for these functions? On what command line? > > Cheers, > > Jean-Marc > > On 19/04/17 12:29 PM, Linfeng Zhang wrote: > > Hi, > > > > Attached are 2 patches related to silk_biquad_alt() optimization. Please > > review. > > > > Thanks, > > Linfeng Zhang > > > > > > > > _______________________________________________ > > opus mailing list > > opus at xiph.org > > http://lists.xiph.org/mailman/listinfo/opus > >...

On 25/04/17 01:37 PM, Linfeng Zhang wrote: > Is that gain due to Neon or simply due to computing two channels in > parallel? For example, if you make a special case in the C code to > handle both channels in the same loop, what kind of performance do > you get? > > > Tested Complexity 8, it's half half, i.e., 0.8% faster if handling both > channels in

...ting difference accumulates in S[] and its > > difference cannot be easily bounded. The difference in out[] may > > somehow be bounded to 5 in my quick testing, but is not guaranteed > > to other inputs. So maybe comparing bit exactness with the > > following silk_biquad_alt_c_MulSingleAQ28() is better. > > > > Please let me know the decision (whether keeping the original NEON > > (stride 2 only) or choosing the new NEON (both stride 1 and 2) which > > optimizes following silk_biquad_alt_c_MulSingleAQ28()), and I'll > >...

...ode). > > Unfortunately, the truncating difference accumulates in S[] and its > difference cannot be easily bounded. The difference in out[] may somehow be > bounded to 5 in my quick testing, but is not guaranteed to other inputs. So > maybe comparing bit exactness with the following silk_biquad_alt_c_ > MulSingleAQ28() is better. > > Please let me know the decision (whether keeping the original NEON (stride > 2 only) or choosing the new NEON (both stride 1 and 2) which optimizes > following silk_biquad_alt_c_MulSingleAQ28()), and I'll wrap up the patch. > > Here attac...

On Mon, Apr 24, 2017 at 5:52 PM, Jean-Marc Valin <jmvalin at jmvalin.ca> wrote: > On 24/04/17 08:03 PM, Linfeng Zhang wrote: > > Tested on my chromebook, when stride (channel) == 1, the optimization > > has no gain compared with C function. > > You mean that the Neon code is the same speed as the C code for > stride==1? This is not terribly surprising for an IIRC

...tunately, the truncating difference accumulates in S[] and its > difference cannot be easily bounded. The difference in out[] may > somehow be bounded to 5 in my quick testing, but is not guaranteed > to other inputs. So maybe comparing bit exactness with the > following silk_biquad_alt_c_MulSingleAQ28() is better. > > Please let me know the decision (whether keeping the original NEON > (stride 2 only) or choosing the new NEON (both stride 1 and 2) which > optimizes following silk_biquad_alt_c_MulSingleAQ28()), and I'll > wrap up the patch. >...

...ared to the original C code). Unfortunately, the truncating difference accumulates in S[] and its difference cannot be easily bounded. The difference in out[] may somehow be bounded to 5 in my quick testing, but is not guaranteed to other inputs. So maybe comparing bit exactness with the following silk_biquad_alt_c_MulSingleAQ28() is better. Please let me know the decision (whether keeping the original NEON (stride 2 only) or choosing the new NEON (both stride 1 and 2) which optimizes following silk_biquad_alt_c_MulSingleAQ28()), and I'll wrap up the patch. Here attached the corresponding C code silk_...

.../* for memset(), memcpy(), memmove() */ @@ -48,6 +49,7 @@ extern "C" #endif #if (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#include "arm/LPC_analysis_filter_arm.h" #include "arm/LPC_inv_pred_gain_arm.h" #endif @@ -111,7 +113,7 @@ void silk_biquad_alt( ); /* Variable order MA prediction error filter. */ -void silk_LPC_analysis_filter( +void silk_LPC_analysis_filter_c( opus_int16 *out, /* O Output signal */ const opus_int16 *in,...