search for: opus_select_arch

Currently, opus_select_arch() is being called during initial setup of encoder/decoder structures and then stored. However, this "arch" variable does not always get passed to every function that may need it for architecture specific optimization. As a result, when a certain function is to be optimized for a particul...

Viswanath Puttagunta wrote: > Instead, just optimize the opus_select_arch() such that only > the first call to it takes more time, but subsequent calls to > it are much faster. This has been suggested before, but it is a bad idea. The patch posted below is not thread-safe or even re-entrant. You can argue that the results should always be the same, so it won'...

Yes, Thank you. 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

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 failures, because we mask with OPUS_ARCHMASK, but it does mean that a CPU with AVX will invoke the C versions of every function instead of any potential SSE, SSE2...

...t 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 failures, because we mask with OPUS_ARCHMASK, but it does mean that a CPU with AVX will invoke the C versions of every function instead of any potential SSE, SSE2, or...

Hi, I'm announcing two new releases of opus, along with the first release of the new libopusenc library. First for opus, we just fixed a bad bug in the surround encoding code that would cause very bad quality on signals beyond a certain amplitude. Thanks to Franziska Trojahn and others from HfT Leipzig for finding this issue through their listening test[1] (the quality of the two bad files

...signalling; + int arch; /* Everything beyond this point gets cleared on a reset */ #define DECODER_RESET_START rng @@ -159,6 +161,7 @@ OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMod st->signalling = 1; st->loss_count = 0; + st->arch = opus_select_arch(); opus_custom_decoder_ctl(st, OPUS_RESET_STATE); @@ -430,7 +433,7 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, opus_val16 * OPUS_R { VARDECL( opus_val16, lp_pitch_buf ); ALLOC( lp_pitch_buf, DECODE_BUFFER_SIZE>>1, opus_val16 ); -...

...Vish [1]: https://docs.google.com/a/linaro.org/document/d/1avz20b3DOnD3IwxiKTmUfyUK89hUwL9K2PYMh7dlkNg/edit# [2]: https://bugs.linaro.org/show_bug.cgi?id=1044 [3]: http://people.linaro.org/~viswanath.puttagunta/opus/NE10_root/NE10_root.tar.gz Viswanath Puttagunta (2): Optimize repeated calls to opus_select_arch armv7(float): Optimize encode usecase using NE10 library Makefile.am | 30 +-- celt/arm/arm_celt_ne10_fft_map.c | 65 ++++++ celt/arm/arm_celt_ne10_mdct_map.c | 53 +++++ celt/arm/armcpu.c | 19 +- celt/arm/celt_ne10_fft.c |...

...ILD/opus-1.2-beta/silk/tests/test_unit_LPC_inv_pred_gain.c:114: > undefined reference to `silk_LPC_inverse_pred_gain_c' > > Second is a few of them have this error: > > /builddir/build/BUILD/opus-1.2-beta/silk/tests/test_unit_LPC_inv_pred_gain.c:91: > undefined reference to `opus_select_arch' > > all the build logs for the various arches can be found here: > https://koji.fedoraproject.org/koji/taskinfo?taskID=19718776 > > Let me know if you've got any queries. > > Peter >

..._c2c_1d_float32_neon((ne10_fft_cpx_float32_t *)fout, > + (ne10_fft_cpx_float32_t *)tempin, > + cfg, 0); More mis-alignment. > + /* N/4 complex FFT, does not downscale anymore */ > + opus_fft(st, f2, (kiss_fft_cpx *)f, opus_select_arch()); Because you removed the scaling from the above loop, this comment is inaccurate. > diff --git a/celt/arm/fft_arm.h b/celt/arm/fft_arm.h > new file mode 100644 > index 0000000..16f008b > --- /dev/null > +++ b/celt/arm/fft_arm.h > @@ -0,0 +1,65 @@ > +/* Copyright (c) 2015-...

...e = *yp++; + im = *yp++; + yr = S_MUL(re,t0) - S_MUL(im,t1); + yi = S_MUL(im,t0) + S_MUL(re,t1); + yc.r = yr; + yc.i = yi; + f2[i] = yc; + } + } + + /* N/4 complex FFT, does not downscale anymore */ + opus_fft(st, f2, (kiss_fft_cpx *)f, opus_select_arch()); + + /* Post-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_cpx * OPUS_RESTRICT fp = (kiss_fft_cpx *)f; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + kiss_fft_scalar * OPUS_RESTRICT yp2 = out+stride*(N2-1...

..._float32_t *)fout, >> + (ne10_fft_cpx_float32_t *)tempin, >> + cfg, 0); > > > More mis-alignment. > >> + /* N/4 complex FFT, does not downscale anymore */ >> + opus_fft(st, f2, (kiss_fft_cpx *)f, opus_select_arch()); > > > Because you removed the scaling from the above loop, this comment is > inaccurate. > >> diff --git a/celt/arm/fft_arm.h b/celt/arm/fft_arm.h >> new file mode 100644 >> index 0000000..16f008b >> --- /dev/null >> +++ b/celt/arm/fft_arm.h >>...

The ARM implementation of opus_select_arch wants to verify that a given CPU architecture also implies all lower-capability architectures. Unfortunately, the ARM opus_cpu_capabilities function only currently checks for those architectures that are enabled in the configuration, and thus —disable-asm — or a floating-point build — doesn’t chec...

...IR_IMPL' /home/derf/src/xiph/git/xiph/opus/build/../celt/tests/test_unit_optimization_lpc.c:77: undefined reference to `celt_fir_c' tests/test_unit_optimization.o: In function `main': /home/derf/src/xiph/git/xiph/opus/build/../tests/test_unit_optimization.c:51: undefined reference to `opus_select_arch' collect2: error: ld returned 1 exit status

...ll arches: /builddir/build/BUILD/opus-1.2-beta/silk/tests/test_unit_LPC_inv_pred_gain.c:114: undefined reference to `silk_LPC_inverse_pred_gain_c' Second is a few of them have this error: /builddir/build/BUILD/opus-1.2-beta/silk/tests/test_unit_LPC_inv_pred_gain.c:91: undefined reference to `opus_select_arch' all the build logs for the various arches can be found here: https://koji.fedoraproject.org/koji/taskinfo?taskID=19718776 Let me know if you've got any queries. Peter

Hi, I got libopus 1.2 from the download page. I compiled it using visual studio 2015 with your configuration (Release). I integrated opus.lib and the new include files in my own solution, but when I compile, I found 28 link errors (lnk 2001): - silk_Encode - ec_enc_init - celt_inner_prod_sse - opus_select_arch - silk_InitEncoder - ec_enc_shrink - silk_log2lin - ec_enc_bit_logp - celt_encoder_get_size - celt_encoder_init - silk_lin2log - opus_custom_encoder_ctl - ec_enc_uint - silk_Get_Encoder_Size - celt_encode_with_ec - ec_enc_done - opus_fft_c - silk_Decode - ec_dec_init - ec_dec_uint - celt_decode_wit...

.../derf/src/xiph/git/xiph/opus/build/../celt/tests/test_unit_optimization_lpc.c:77: > undefined reference to `celt_fir_c' > tests/test_unit_optimization.o: In function `main': > /home/derf/src/xiph/git/xiph/opus/build/../tests/test_unit_optimization.c:51: > undefined reference to `opus_select_arch' > collect2: error: ld returned 1 exit status > > _______________________________________________ > opus mailing list > opus at xiph.org > http://lists.xiph.org/mailman/listinfo/opus > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http:...

Created corresponding unit test, and the optimization is bit exact with C function. --- silk/SigProc_FIX.h | 7 ++- silk/arm/arm_silk_map.c | 12 ++++ silk/arm/inner_prod_aligned_arm.h | 58 +++++++++++++++++++ silk/arm/inner_prod_aligned_neon_intr.c | 66 ++++++++++++++++++++++ silk/enc_API.c

...static const opus_int16 eband5ms[] = { /*0 200 400 600 800 1k 1.2 1.4 1.6 2k 2.4 2.8 3.2 4k 4.8 5.6 6.8 8k 9.6 12k 15.6 */ @@ -229,6 +230,7 @@ CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) opus_val16 *window; opus_int16 *logN; int LM; + int arch = opus_select_arch(); ALLOC_STACK; #if !defined(VAR_ARRAYS) && !defined(USE_ALLOCA) if (global_stack==NULL) @@ -389,7 +391,7 @@ CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) compute_pulse_cache(mode, mode->maxLM); if (clt_mdct_init(&mode->mdct, 2*mod...

...static const opus_int16 eband5ms[] = { /*0 200 400 600 800 1k 1.2 1.4 1.6 2k 2.4 2.8 3.2 4k 4.8 5.6 6.8 8k 9.6 12k 15.6 */ @@ -229,6 +230,7 @@ CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) opus_val16 *window; opus_int16 *logN; int LM; + int arch = opus_select_arch(); ALLOC_STACK; #if !defined(VAR_ARRAYS) && !defined(USE_ALLOCA) if (global_stack==NULL) @@ -389,7 +391,7 @@ CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) compute_pulse_cache(mode, mode->maxLM); if (clt_mdct_init(&mode->mdct, 2*mod...