similar to: [ARM][FFT][NEON] Integrate Ne10 into Opus?

Hi Ralph, I have pushed patches to enable radix 3 and radix 5. Github: https://github.com/projectNe10/Ne10/releases/tag/v1.2.0 Best Regards, Phil Wang > Date: Thu, 11 Dec 2014 10:46:50 -0800 > From: Ralph Giles <giles at thaumas.net> > Subject: Re: [opus] [ARM][FFT][NEON] Integrate Ne10 into Opus? > To: opus at xiph.org > Message-ID: <5489E69A.5000305 at thaumas.net>

Hi everyone, I am working on Ne10 project. Ne10 provides NEON optimized FFT routines that are much faster (compared to those without NEON), on most ARMv7-A and all ARMv8-A devices. How about integrate it into Opus? I am not familiar with configure script, but I find "Optinal Packages" in it. If we provides --with-ne10-fft option, the one extra thing that users need to do is to

Hi, I am working on DSP module of Ne10. I see there are fixed-point and floating-point FFT inside Opus. Is fixed-point FFT only a fall back for CPU without VFP? On ARMv7-A and ARMv8-A, benchmark result shows that fixed-point (int32) and floating-point (float32) FFT have similar performance. I guess fixed-point version is not often used on these platforms. Is it worth the effort to NEON-optimize

On 29 April 2015 at 17:22, Timothy B. Terriberry <tterribe at xiph.org> wrote: > > Viswanath Puttagunta wrote: >> >> This patch series is follow up on work I posted on [1]. >> In addition to what was posted on [1], this patch series mainly >> integrates Fixed point FFT implementations in NE10 library into opus. >> You can view my opus wip code at [2]. >

I'm trying to get these cleaned up and landed, but I'm running into some trouble with this patch. Using commit a08b29d88e3c (July 21) of Ne10, I'm seeing test failures for 60-point FFTs: nfft=60 inverse=0,snr = -3.312408 ** poor snr: -3.312408 ** nfft=60 inverse=1,snr = -16.079597 ** poor snr: -16.079597 ** All other sizes tested appear to work fine (84 to 140 dB of SNR). This

Hi Timothy, Sorry for late reply. I have upstreamed the patch to fix the regression here: https://github.com/projectNe10/Ne10/commit/ee5d856cd9cb8c4a15ace567df4239f4e788d043 I have tested it with Vish's branch: http://git.linaro.org/people/viswanath.puttagunta/opus.git/shortlog/refs/heads/rfcv3_fft_fixed) Both unit test dft and unit test mdct passed on ARM v7/v8, floating point/fixed

Hello Jean-Marc, Yep, that was it.. with your patch, test_unit_mdct passes for all nfft. So, what you do you suggest the next step here is? Regards, Vish On 8 May 2015 at 12:30, Jean-Marc Valin <jmvalin at jmvalin.ca> wrote: > Hi, > > Can you apply this change to the MDCT test and run it again. See if more > (all) sizes pass. Given the results, I strongly suspect an

Hello Jean-Marc, Below are the results that show test_unit_dft passes, but test_unit_mdct fails (only for nfft=480, 960, 1920) Note: Tested on BeagleboneBlack(Cortex-A8) fixed point on branch [1] ./test_unit_dft nfft=32 inverse=0,snr = 88.394372 nfft=32 inverse=1,snr = 93.896470 nfft=128 inverse=0,snr = 89.185895 nfft=128 inverse=1,snr = 93.537021 nfft=256 inverse=0,snr = 88.353151 nfft=256

Hi Jean-Marc, I have implemented fixed-point FFT with 32-bit twiddles. Now I want to evaluate the accuracy, what method does Opus use? I use function implemented inside Ne10 to calculate SNR. Any comment? | size | SNR (dB) | | 16 | 82.558587 | | 32 | 83.530298 | | 60 | 80.292433 | | 64 | 82.752950 | | 120 | 79.625077 | | 128 | 83.091260 | | 240 | 79.555263 | | 256 |

Uses NEON optimized fixed point fft routines in NE10 library Signed-off-by: Viswanath Puttagunta <viswanath.puttagunta at linaro.org> Signed-off-by: Jonathan Lennox <jonathan at vidyo.com> --- Makefile.am | 12 +- celt/arm/arm_celt_map.c | 46 ++-- celt/arm/celt_ne10_fft.c | 98 +++++---- celt/arm/fft_arm.h |

Viswanath Puttagunta wrote: > if OPUS_ARM_NEON_INTR > CELT_ARM_NEON_INTR_OBJ = $(CELT_SOURCES_ARM_NEON_INTR:.c=.lo) \ > - %test_unit_rotation.o %test_unit_mathops.o > -$(CELT_ARM_NEON_INTR_OBJ): CFLAGS += $(OPUS_ARM_NEON_INTR_CPPFLAGS) > + $(CELT_SOURCES_ARM_NE10:.c=.lo) \ > + %test_unit_rotation.o %test_unit_mathops.o \ > +

Using GCC 4.9.2, decoding the opus test vector set 10 times. All tests pass. I will do longer tests later with a larger test set, but it looks good so far. I am having a lot of trouble with the Ne10 detection. Using the precompiled Ne10 binaries at http://people.linaro.org/~viswanath.puttagunta/opus/NE10_root/ Both NE10 and Ne10 capitalizations seem to be in use. Also, where should the NE10

Hello Timothy, Great to hear from you! Fired up my hardware today and this issue looks like a regression in Ne10 library. The commit in Ne10 [1] that I tested to be working successfully back in May 5b63074db45000f9688460990ee3f5e147d93782 which is the Patch Phil at ARM added to fix the overflow issue in nfft=60 case. After git-bisect, looks like the culprit patch in Ne10 [1] is

Hello Timothy, Just FYI, Phil at ARM is still looking into why mdct is failing.. will keep you posted. In the mean time, do you want me to disable NE10 for mdct_forward and re-submit the patchset so we may make progress? Regards, Vish On 30 April 2015 at 09:33, Viswanath Puttagunta <viswanath.puttagunta at linaro.org> wrote: > > On 29 April 2015 at 17:22, Timothy B. Terriberry

Hello Jean-Marc, **Resending.. not sure why subject got removed earlier** Below are the results that show test_unit_dft passes, but test_unit_mdct fails (only for nfft=480, 960, 1920) Note: Tested on BeagleboneBlack(Cortex-A8) fixed point on branch [1] ./test_unit_dft nfft=32 inverse=0,snr = 88.394372 nfft=32 inverse=1,snr = 93.896470 nfft=128 inverse=0,snr = 89.185895 nfft=128 inverse=1,snr =

There is definitely some use for a Neon fixed-point FFT. How much exactly I'm not sure. Fixed-point is a bit more than just a fall-back for CPUs with no FPU. There are CPUs for which fixed-point is still faster. It depends on the exact model but also on what you run. For example, even on x86 I believe that SILK encoding is slightly faster in fixed-point, even though CELT is faster in float.

> I am working on DSP module of Ne10. I see there are fixed-point and > floating-point FFT inside Opus. Is fixed-point FFT only a fall back for CPU > without VFP? On ARMv7-A and ARMv8-A, benchmark result shows that fixed-point > (int32) and floating-point (float32) FFT have similar performance. I guess > fixed-point version is not often used on these platforms. Is it worth the >

Viswanath Puttagunta wrote: > Can we please have review on RFCv2? We have quite a few optimizations > (Eg: ifft/mdct_backwards, fixed point fft/ifft mdct_forward/backward > etc) that are in my pipeline that depend on this patch series being > accepted. So, trying to make progress on this... On an armv7l board running Ubuntu, you've broken the build with just --enable-intrinsics

Changes from RFC PATCH v2 - fixed compile issue when just compiling for --enable-intrinsics for ARMv7 without NE10 - Notes for NE10: - All compile/link warnings are now in upstream NE10 - Only patch pending upstream in NE10 is the one that needs to add -funsafe-math-optimizations for ARMv7 targets. - Phil Wang @ ARM is working on getting this fixed. - Note that even without

Hello Timothy / Jean-Marc / opus-dev, This patch series is follow up on work I posted on [1]. In addition to what was posted on [1], this patch series mainly integrates Fixed point FFT implementations in NE10 library into opus. You can view my opus wip code at [2]. Note that while I found some issues both with the NE10 library(fixed fft) and with Linaro toolchain (armv8 intrinsics), the work