search for: opus_arm_neon_intr

...floating point. Changes from RFCv2: - Changes recommended by Timothy for celt_neon_intr.c everything except, left the unrolled loop still unrolled - configure.ac - use AC_LINK_IFELSE instead of AC_COMPILE_IFELSE - Moved compile flags into Makefile.am - OPUS_ARM_NEON_INR --> typo --> OPUS_ARM_NEON_INTR Viswanath Puttagunta (1): armv7: celt_pitch_xcorr: Introduce ARM neon intrinsics Makefile.am | 12 ++ celt/arm/arm_celt_map.c | 15 ++- celt/arm/celt_neon_intr.c | 256 +++++++++++++++++++++++++++++++++++++++ celt/arm/pitch_arm.h | 13 +- cel...

...+++++++++---------------------- 12 files changed, 67 insertions(+), 67 deletions(-) diff --git a/Makefile.am b/Makefile.am index 3a75740..8bd7447 100644 --- a/Makefile.am +++ b/Makefile.am @@ -47,7 +47,7 @@ if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) endif @@ -286,7 +286,7 @@ SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \ $(SSE4_1_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS) endif -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_ARM_NEON_INTR_OBJ = $...

....am | 4 ++-- configure.ac | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/Makefile.am b/Makefile.am index 4d3a888..d256b45 100644 --- a/Makefile.am +++ b/Makefile.am @@ -47,7 +47,7 @@ if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) endif @@ -294,7 +294,7 @@ SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \ $(SSE4_1_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS) endif -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_ARM_NEON_INTR_OBJ = $...

....am | 4 ++-- configure.ac | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/Makefile.am b/Makefile.am index 4d3a888..d256b45 100644 --- a/Makefile.am +++ b/Makefile.am @@ -47,7 +47,7 @@ if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) endif @@ -294,7 +294,7 @@ SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \ $(SSE4_1_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS) endif -if OPUS_ARM_NEON_INTR +if HAVE_ARM_NEON_INTR CELT_ARM_NEON_INTR_OBJ = $...

...| 4 +- 13 files changed, 205 insertions(+), 25 deletions(-) create mode 100644 celt/arm/celt_neon_intr.c diff --git a/Makefile.am b/Makefile.am index e20f7b4..0e9e120 100644 --- a/Makefile.am +++ b/Makefile.am @@ -38,6 +38,12 @@ CELT_SOURCES += $(CELT_SOURCES_SSE) endif endif +if OPUS_ARM_NEON_INTR +noinst_LTLIBRARIES = libarmneon.la +libarmneon_la_SOURCES = $(CELT_SOURCES_ARM_NEON_INTR) +libarmneon_la_CPPFLAGS = $(OPUS_ARM_NEON_INTR_CPPFLAGS) -I$(top_srcdir)/include +endif + if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) @@ -59,6 +65,9 @@ include opus_h...

Hi, Optimizes celt_pitch_xcorr for ARM NEON floating point. Changes from RFCv3: - celt_neon_intr.c - removed warnings due to not having constant pointers - Put simpler loop to take care of corner cases. Unrolling using intrinsics was not really mapping well to what was done in celt_pitch_xcorr_arm.s - Makefile.am Removed explicit -O3 optimization - test_unit_mathops.c,

Hello, I received feedback from engineers working on NE10 [1] that it would be better to use NE10 [1] for FFT optimizations for opus use cases. However, these FFT patches are currently in review and haven't been integrated into NE10 yet. While the FFT functions in NE10 are getting baked, I wanted to optimize the celt_pitch_xcorr (floating point only) and use it to introduce ARM NEON

...="$CFLAGS" >> + save_CFLAGS="$CFLAGS"; CFLAGS="-mfpu=neon $CFLAGS" >> + AC_COMPILE_IFELSE( >> + [AC_LANG_PROGRAM([[#include <arm_neon.h>]], [])], >> + [ >> + OPUS_ARM_NEON_INTR=1 >> + OPUS_ARM_NEON_INTR_CPPFLAGS="-mfpu=neon -O3" >> + AC_SUBST(OPUS_ARM_NEON_INTR_CPPFLAGS) >> + ], >> + [ >> + OPUS_ARM_NEON_INTR=0 >> + ]) &...

...insertions(+), 19 deletions(-) create mode 100644 celt/arm/celt_neon_intr.c diff --git a/Makefile.am b/Makefile.am index e20f7b4..95323ca 100644 --- a/Makefile.am +++ b/Makefile.am @@ -41,6 +41,12 @@ endif if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) + +if OPUS_ARM_NEON_INTR +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) +OPUS_ARM_NEON_INTR_CPPFLAGS = -mfpu=neon +endif + if OPUS_ARM_EXTERNAL_ASM nodist_libopus_la_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) BUILT_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ @@ -260,3 +266,9 @@ if HAVE_SSE2 $(SSE_OBJ): CFLAGS += -ms...

...insertions(+), 19 deletions(-) create mode 100644 celt/arm/celt_neon_intr.c diff --git a/Makefile.am b/Makefile.am index e20f7b4..be53754 100644 --- a/Makefile.am +++ b/Makefile.am @@ -41,6 +41,12 @@ endif if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) + +if OPUS_ARM_NEON_INTR +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) +OPUS_ARM_NEON_INTR_CPPFLAGS = -mfpu=neon -O3 +endif + if OPUS_ARM_EXTERNAL_ASM nodist_libopus_la_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) BUILT_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ @@ -260,3 +266,9 @@ if HAVE_SSE2 $(SSE_OBJ): CFLAGS +=...

...insertions(+), 19 deletions(-) create mode 100644 celt/arm/celt_neon_intr.c diff --git a/Makefile.am b/Makefile.am index e20f7b4..0687814 100644 --- a/Makefile.am +++ b/Makefile.am @@ -41,6 +41,11 @@ endif if CPU_ARM CELT_SOURCES += $(CELT_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM) + +if OPUS_ARM_NEON_INTR +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) +endif + if OPUS_ARM_EXTERNAL_ASM nodist_libopus_la_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) BUILT_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ @@ -260,3 +265,9 @@ if HAVE_SSE2 $(SSE_OBJ): CFLAGS += -msse2 endif endif + +if OPUS_ARM_NEON_INTR...

Hi, Optimizes celt_pitch_xcorr for floating point. Changes from RFCv1: - Rebased on top of commit aad281878: Fix celt_pitch_xcorr_c signature. which got rid of ugly code around CELT_PITCH_XCORR_IMPL passing of "arch" parameter. - Unified with --enable-intrinsics used by x86 - Modified algorithm to be more in-line with algorithm in celt_pitch_xcorr_arm.s Viswanath Puttagunta

...t/arm/celt_neon_intr.c > > diff --git a/Makefile.am b/Makefile.am > index e20f7b4..95323ca 100644 > --- a/Makefile.am > +++ b/Makefile.am > @@ -41,6 +41,12 @@ endif > if CPU_ARM > CELT_SOURCES += $(CELT_SOURCES_ARM) > SILK_SOURCES += $(SILK_SOURCES_ARM) > + > +if OPUS_ARM_NEON_INTR > +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) > +OPUS_ARM_NEON_INTR_CPPFLAGS = -mfpu=neon > +endif > + > if OPUS_ARM_EXTERNAL_ASM > nodist_libopus_la_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) > BUILT_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ > @@ -260,3 +266,9 @@...

From: Viswanath Puttagunta <viswanath.puttagunta at linaro.org> Hi, Optimizes celt_pitch_xcorr for floating point. Changes from RFCv1: - Rebased on top of commit aad281878: Fix celt_pitch_xcorr_c signature. which got rid of ugly code around CELT_PITCH_XCORR_IMPL passing of "arch" parameter. - Unified with --enable-intrinsics used by x86 - Modified algorithm to be more

Almost there... just a few nits left. Viswanath Puttagunta wrote: > +if OPUS_ARM_NEON_INTR > +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) > +OPUS_ARM_NEON_INTR_CPPFLAGS = -mfpu=neon -O3 I'll repeat: I don't think you should change the optimization level here. > + /* Just unroll the rest of the loop */ I saw you decided to keep this unrolled, but you didn't act...

Review comments inline. > +if OPUS_ARM_NEON_INTR > +noinst_LTLIBRARIES = libarmneon.la > +libarmneon_la_SOURCES = $(CELT_SOURCES_ARM_NEON_INTR) > +libarmneon_la_CPPFLAGS = $(OPUS_ARM_NEON_INTR_CPPFLAGS) -I$(top_srcdir)/include > +endif I don't think these should be in a separate library. It brings with it lots of complications (...

...sumi += xi[i] * yi[i]; > + *sum = sumi; This bounces things through the stack into vfp registers, which is a huge stall. I'd continue to use NEON here with vld1_dup_f32()/vmla_f32()/etc. > +} > +#if ((defined(OPUS_ARM_ASM) && defined(FIXED_POINT)) \ > + || defined(OPUS_ARM_NEON_INTR)) > #include "arm/arm_celt_map.c" > #endif > > + Unrelated whitespace change. > - inline_optimization="No ASM for your platform, please send patches" > + inline_optimization="No in-line ASM for your platform, please send patches" "in...

...t it as PATCHv1.. since the patch is almost there and there are no major objections any more.. Regards, Vish On 17 December 2014 at 23:59, Timothy B. Terriberry <tterribe at xiph.org> wrote: > > Almost there... just a few nits left. > > Viswanath Puttagunta wrote: > > +if OPUS_ARM_NEON_INTR > > +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) > > +OPUS_ARM_NEON_INTR_CPPFLAGS = -mfpu=neon -O3 > > I'll repeat: I don't think you should change the optimization level here. After you feedback, I moved this part from configure.ac to Makefile.am.. If it doesn't be...

Hi All, Since I continue to base my work on top of Jonathan's patch, and my previous Ne10 fft/ifft/mdct_forward/backward patches, I thought it would be better to just post all new patches as a patch series. Please let me know if anyone disagrees with this approach. You can see wip branch of all latest patches at https://git.linaro.org/people/viswanath.puttagunta/opus.git Branch:

Hello Timothy, Appreciate the thorough review. Have a few questions before I re-spin the patch in-line. On 28 November 2014 at 15:52, Timothy B. Terriberry <tterribe at xiph.org> wrote: > Review comments inline. > >> +if OPUS_ARM_NEON_INTR >> +noinst_LTLIBRARIES = libarmneon.la >> +libarmneon_la_SOURCES = $(CELT_SOURCES_ARM_NEON_INTR) >> +libarmneon_la_CPPFLAGS = $(OPUS_ARM_NEON_INTR_CPPFLAGS) -I$(top_srcdir)/include >> +endif > > I don't think these should be in a separate library. It brings with it...