search for: armcpu

...ge many function signatures in the call stack to make this happen. 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 helps avoid needing to make too many changes to function signatures. --- celt/arm/armcpu.c | 19 +++++++++++++++---- celt/x86/x86cpu.c | 22 ++++++++++++++-------- 2 files changed, 29 insertions(+), 12 deletions(-) diff --git a/celt/arm/armcpu.c b/celt/arm/armcpu.c index 1768525..26aae09 100644 --- a/celt/arm/armcpu.c +++ b/celt/arm/armcpu.c @@ -151,24 +151,35 @@ opus_uint32 opus_...

...X86_MAY_HAVE_SSE2) || defined(OPUS_X86_MAY_HAVE_SSE4_1) # include "x86/x86cpu.c" -#elif defined(OPUS_HAVE_RTCD) && \ - (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#elif defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # include "arm/armcpu.c" # include "celt_lpc.c" # include "pitch.c" diff --git a/celt/tests/test_unit_mathops.c b/celt/tests/test_unit_mathops.c index 5b446b7..fd3319d 100644 --- a/celt/tests/test_unit_mathops.c +++ b/celt/tests/test_unit_mathops.c @@ -63,8 +63,7 @@ # include "x86/celt_...

...X86_MAY_HAVE_SSE2) || defined(OPUS_X86_MAY_HAVE_SSE4_1) # include "x86/x86cpu.c" -#elif defined(OPUS_HAVE_RTCD) && \ - (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#elif defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # include "arm/armcpu.c" # include "celt_lpc.c" # include "pitch.c" diff --git a/celt/tests/test_unit_mathops.c b/celt/tests/test_unit_mathops.c index 5b446b7..fd3319d 100644 --- a/celt/tests/test_unit_mathops.c +++ b/celt/tests/test_unit_mathops.c @@ -63,8 +63,7 @@ # include "x86/celt_...

Fixes build failure on platforms with MAY_HAVE_SSE4_1 (but not PRESUME_SSE4_1) with --enable-intrinsics. --- silk/fixed/main_FIX.h | 11 +++++++++++ silk/x86/x86_silk_map.c | 2 ++ 2 files changed, 13 insertions(+) diff --git a/silk/fixed/main_FIX.h b/silk/fixed/main_FIX.h index ffeb4f3..375b5eb 100644 --- a/silk/fixed/main_FIX.h +++ b/silk/fixed/main_FIX.h @@ -97,6 +97,17 @@ void

Fixes build failure on platforms with MAY_HAVE_SSE4_1 (but not PRESUME_SSE4_1) with --enable-intrinsics. --- silk/fixed/main_FIX.h | 11 +++++++++++ silk/x86/x86_silk_map.c | 2 ++ 2 files changed, 13 insertions(+) diff --git a/silk/fixed/main_FIX.h b/silk/fixed/main_FIX.h index ffeb4f3..375b5eb 100644 --- a/silk/fixed/main_FIX.h +++ b/silk/fixed/main_FIX.h @@ -97,6 +97,17 @@ void

...diff --git a/opus_sources.mk b/opus_sources.mk index e4eeb91..1e9791b 100644 --- a/opus_sources.mk +++ b/opus_sources.mk @@ -4,7 +4,8 @@ src/opus_encoder.c \ src/opus_multistream.c \ src/opus_multistream_encoder.c \ src/opus_multistream_decoder.c \ -src/repacketizer.c +src/repacketizer.c \ +src/armcpu.c OPUS_SOURCES_FLOAT = \ src/analysis.c \ diff --git a/src/armcpu.c b/src/armcpu.c new file mode 100644 index 0000000..10a2905 --- /dev/null +++ b/src/armcpu.c @@ -0,0 +1,160 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source...

Jonathan Lennox wrote: > +opus_int32 silk_noise_shape_quantizer_short_prediction_neon(const opus_int32 *buf32, const opus_int32 *coef32) > +{ > + int32x4_t coef0 = vld1q_s32(coef32); > + int32x4_t coef1 = vld1q_s32(coef32 + 4); > + int32x4_t coef2 = vld1q_s32(coef32 + 8); > + int32x4_t coef3 = vld1q_s32(coef32 + 12); > + > + int32x4_t a0 = vld1q_s32(buf32 -

...on. Apply Neon short prediction optimization to silk_noise_shape_quantizer_del_dec. Add Neon intrinsics for Silk noise shape feedback loop. Add Neon fixed-point implementation of xcorr_kernel. Makefile.am | 5 +- celt/arm/arm_celt_map.c | 17 ++++++ celt/arm/armcpu.c | 35 +++++++---- celt/arm/armcpu.h | 6 ++ celt/arm/celt_neon_intr.c | 61 ++++++++++++++++++- celt/arm/pitch_arm.h | 31 +++++++++- silk/NSQ.c | 57 ++++++----------- silk/NSQ.h | 97 ++++++++++++++++++++++++...

...work. Opus codec works great under many platforms. I have found a small performance issue under iOS platform. If the macro OPUS_HAVE_RTCD is not set, then encoder doesn't use some _neon functions at low bitrates (up to 64k). If the macro is set, then the compiler hits the error at opus/celt/arm/armcpu.c:153 (a function for detection missed). Being compared to Android version performance degradation is ~30%. Detection NEON under iOS is a bit tricky, because there is no API for it. I added compiler-time detection. I made a commit into github repository. Attached please find patch for it. Regards,...

Hi, Attached is a fix related to ARM optimization jump table compiling error. Thanks, Linfeng Zhang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.xiph.org/pipermail/opus/attachments/20170720/661d96b5/attachment.html> -------------- next part -------------- A non-text attachment was scrubbed... Name:

...hard to track down and update. I see your point ? what do you suggest instead? As the comment mentions, this can?t use the usual IMPL table implementation, because the parameters are different (due to the transformed coefficients.) Would something like an OPUS_ARCH_ARM_NEON #define in celt/arm/armcpu.h be okay? It?d be a bit confusing with OPUS_CPU_ARM_NEON in armcpu.c, but I could probably organize it to be sensible. > (also, I realize libopus doesn't have a line-length restriction, but a > few newlines in here might be a mercy to those of us who work in > 80-column terminals)...

...great under many platforms. > I have found a small performance issue under iOS platform. > If the macro OPUS_HAVE_RTCD is not set, then encoder doesn't use some > _neon functions at low bitrates (up to 64k). If the macro is set, then the > compiler hits the error at opus/celt/arm/armcpu.c:153 (a function for > detection missed). Being compared to Android version performance > degradation is ~30%. > Detection NEON under iOS is a bit tricky, because there is no API for it. > I added compiler-time detection. > I made a commit into github repository. Attached please...

...0ba5995 100644 --- a/celt/dump_modes/Makefile +++ b/celt/dump_modes/Makefile @@ -19,7 +19,8 @@ INCLUDES += -I$(NE10_INCDIR) -DHAVE_ARM_NE10 -DOPUS_ARM_PRESUME_NEON_INTR LIBDIR = -l:$(NE10_LIBDIR)/libNE10.so SOURCES += ../arm/celt_ne10_fft.c \ dump_modes_arm_ne10.c \ - ../arm/armcpu.c + ../arm/armcpu.c \ + ../entcode.c endif all: dump_modes diff --git a/celt/dump_modes/dump_modes_arch.h b/celt/dump_modes/dump_modes_arch.h index 1436926..59073ee 100644 --- a/celt/dump_modes/dump_modes_arch.h +++ b/celt/dump_modes/dump_modes_arch.h @@ -28,11 +28,17 @@ #i...

...THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(CELT_LPC_ARM_H) +# define CELT_LPC_ARM_H + +# include "armcpu.h" + +# if defined(FIXED_POINT) + +# if defined(OPUS_ARM_MAY_HAVE_NEON) +void celt_fir_neon( + const opus_val16 *_x, + const opus_val16 *num, + opus_val16 *_y, + int N, + int ord, + int arch); +# endif + +# if !defined(OPUS_HAVE_RTCD) +# def...

...t_unit_dft.c @@ -45,8 +45,7 @@ #include "mathops.c" #include "entcode.c" -#if defined(OPUS_HAVE_RTCD) && \ - (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) || defined(OPUS_ARM_ASM) #include "arm/armcpu.c" #if !defined(FIXED_POINT) #if defined(HAVE_ARM_NE10) diff --git a/celt/tests/test_unit_mathops.c b/celt/tests/test_unit_mathops.c index a1cf2f7..2e43e07 100644 --- a/celt/tests/test_unit_mathops.c +++ b/celt/tests/test_unit_mathops.c @@ -65,17 +65,18 @@ #include "x86/celt_lpc_sse.c&...

Hi all, This is Linfeng Zhang from Google. I'll work on ARM NEON optimization in the next few months. I'm submitting 2 patches in the following couple of emails, which have the new created celt_fir_neon(). I revised celt_fir_c() to not pass in argument "mem" in Patch 1. If there are concerns to this change, please let me know. Many thanks to your comments. Linfeng Zhang

...th compilers with non-GCC-flavor flags for enabling architecture options. * Hopefully makes the configuration and ifdef?s easier to follow and understand. Reviewed-by: Viswanath Puttagunta <viswanath.puttagunta at linaro.org> --- Makefile.am | 38 ++-- celt/arm/armcpu.c | 6 +- celt/arm/pitch_arm.h | 4 +- celt/bands.c | 6 +- celt/celt.c | 16 +- celt/celt.h | 12 +- celt/celt_decoder.c | 6 +- celt/celt...

...th compilers with non-GCC-flavor flags for enabling architecture options. * Hopefully makes the configuration and ifdef?s easier to follow and understand. Reviewed-by: Viswanath Puttagunta <viswanath.puttagunta at linaro.org> --- Makefile.am | 38 ++-- celt/arm/armcpu.c | 6 +- celt/arm/pitch_arm.h | 4 +- celt/bands.c | 6 +- celt/celt.c | 16 +- celt/celt.h | 12 +- celt/celt_decoder.c | 6 +- celt/celt...

...D_ARM64_H + +#include <arm_neon.h> + +#undef SIG2WORD16 +#define SIG2WORD16(x) (vqmovns_s32((x))) + +#endif diff --git a/celt_headers.mk b/celt_headers.mk index 0eca6e6..c9df94b 100644 --- a/celt_headers.mk +++ b/celt_headers.mk @@ -36,6 +36,7 @@ celt/static_modes_fixed_arm_ne10.h \ celt/arm/armcpu.h \ celt/arm/fixed_armv4.h \ celt/arm/fixed_armv5e.h \ +celt/arm/fixed_arm64.h \ celt/arm/kiss_fft_armv4.h \ celt/arm/kiss_fft_armv5e.h \ celt/arm/pitch_arm.h \ -- 2.4.9 (Apple Git-60)

...codec works great under many platforms. > I have found a small performance issue under iOS platform. > If the macro OPUS_HAVE_RTCD is not set, then encoder doesn't use some _neon functions at low bitrates (up to 64k). If the macro is set, then the compiler hits the error at opus/celt/arm/armcpu.c:153 (a function for detection missed). Being compared to Android version performance degradation is ~30%. > Detection NEON under iOS is a bit tricky, because there is no API for it. I added compiler-time detection. > I made a commit into github repository. Attached please find patch for it....