Re: compilation of static binaries for Linux & Windows

Bill Allombert <[email protected]> Mon, 20 Jan 2025 22:58:37 +0100
Newsgroups gmane.comp.mathematics.pari.devel
Message-ID <Z47HDTEf1OAxzR6A@seventeen>
On Mon, Jan 20, 2025 at 10:23:54PM +0100, Beckert wrote:
> Thank you!
> I have now a full static gp for Linux.
> 
> Is there some processor optimizations in the source for special cpu instructions like e.g. AVX512? So it would make
> sense for me to compile on such a host and distribute it to such hosts.

Yes, we have support for SSE2, AVX and AVX2 in some functions.
But even for generic code, PARI is faster when build with -march=native.
However you do not to compile on a particular host, you can force
-march=sse2, -march=avx, -march=avx2 in CFLAGS.

Cheers,
Bill.