[opus] Antw: Re: [PATCH] Optimize silk_warped_autocorrelation_FIX() for ARM NEON

[Ulrich Windl]

>>> Linfeng Zhang <linfengz at google.com> schrieb am 05.04.2017 um 20:13 in Nachricht
<CAKoqLCA2ECpr9dgWZ0Ay_R5oatHS_4Na4Z=2W64p8c3kv26yyA at mail.gmail.com>:
> Thank Jean-Marc!
> 
> The speedup percentages are all relative to the entire encoder.
> 
> Comparing to master, this optimization patch speeds up fixed-point SILK
> encoder on NEON as following: Complexity 5: 6.1% Complexity 6: 5.8%
> Complexity 8: 5.5% Complexity 10: 4.0%

[...]

As a side-note: Linfeng said he's using gcc. In the past I wrote some utility that provided build information to an executable. If you are tight on memory, you may not want it, but my system provides information like this (example):
"$Build: 0.3.3 (Git:master-36611c9) #109 on rksapv04 (x86_64) with gcc 4.3.4 (4.3.4 [gcc-4_3-branch revision 152973]) (-I. -Wall -Wextra -Wshadow -pipe -O2 -DDEBUG_LEVEL=0  -lrt -lm) at Mon Sep 12 13:24:58 CEST 2016 by windl $"

In the auto-generated buildlevel.c the information is created like this (so individual components are available as well):
/* RCS ident compatible identification string */
const char      build_id[] =
        "@(#) $Build: " BUILD_VERSION  " #" BUILD_NUMBER
        " on " BUILD_HOST " (" BUILD_ARCH ")"
        " with " BUILD_COMPILER " (" BUILD_CFLAGS ")"
        " at " BUILD_DATE " by " BUILD_USER " $";

When comparing performance, such information (if accurate) seem helpful.
Personally I added it to help me find out which of my many test binaries is which... ;-)

Regards,
Ulrich