[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] autoconf, --enable-sse, SIMD, AVX

From: Øystein Schønning-Johansen
Subject: Re: [Bug-gnubg] autoconf, --enable-sse, SIMD, AVX
Date: Thu, 20 Jun 2013 22:16:53 +0200

Totally agree. SSE as a term is even architecture dependent. SIMD is not. When the whole world has started using ARM processors, we have the configure option --enable-simd=neon ready!


On Thu, Jun 20, 2013 at 9:47 PM, Michael Petch <address@hidden> wrote:
Howdy all,

Some may not be aware that Philippe committed support for the AVX SIMD
instruction set. I added and cleaned up support code for cputest builds
(that now support 32bit and 64bit with in and without -fpic option when
built for an Intel processor with a GNUC compiler or equivalent).

I have also begun cleaning up some of the autoconf files (configure.ac
etc). I have added better build time checking for the CPU extensions of
the machine being built on. It should detect SSE,SSE2,AVX, and No SIMD
support on Intel machines using a GNUC (gcc) or equivalent compiler.

During cleanup I realized there are some things that may bother people,
or that I'd like to change. First the autoconf magic has been adjusted
to an OPT-OUT mechanism. SIMD (SSE/AVX etc) builds are on by default (if
available on the build machines CPU), and configure options can
change/disable it via --enable-sse=type/--disable-sse . Previously I
believe this was OPT-IN.

When no --enable-sse is present as an option (or someone uses
--enable-sse=yes ) the build process looks at the CPU capabilities of
the machine being built on. It will detect SSE2,SSE,AVX,NO SSE in that
order if they are present. After AVX support comes out of the
experimental stage I will change it to AVX,SSE2,SSE,No SSE .

--enable-sse is now a misnomer with AVX etc. I'd like to actually change
it to be --enable-simd=type/--disable-simd=type .

Please feel free to comment.

Michael Petch
GNU Backgammon Maintainer / Developer
OpenPGP FingerPrint=D81C 6A0D 987E 7DA5 3219 6715 466A 2ACE 5CAE 3304

Bug-gnubg mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]