Optimize silk_warped_autocorrelation_FIX() for ARM NEON
authorLinfeng Zhang <linfengz@google.com>
Wed, 13 Jul 2016 23:40:48 +0000 (16:40 -0700)
committerJean-Marc Valin <jmvalin@jmvalin.ca>
Thu, 13 Apr 2017 22:30:27 +0000 (18:30 -0400)
commit8e19536baa5064a89abe93570fcc99412b6c68a1
tree95eb80319a5cc30193dfe78726ca070f4365a7c8
parentc43633d1a9d3747f96fc30a32f6fdaa322195aed
Optimize silk_warped_autocorrelation_FIX() for ARM NEON

The optimization is bit exact with C function.

This optimization speeds up fixed-point SILK encoder on NEON about
5% to 8%. (Tested on Acer Chromebook, ARMv7 Processor rev 3 (v7l).)

Change-Id: I582f6f3585b7946149e16a2ad3084ebc0ae79a4f

Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>
Makefile.am
silk/arm/arm_silk_map.c
silk/fixed/arm/warped_autocorrelation_FIX_arm.h [new file with mode: 0644]
silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c [new file with mode: 0644]
silk/fixed/main_FIX.h
silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h
silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h
silk/fixed/noise_shape_analysis_FIX.c
silk/fixed/warped_autocorrelation_FIX.c
silk_headers.mk
silk_sources.mk