This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Fixing sparc ceil/floor/trunc implementations
- From: David Miller <davem at davemloft dot net>
- To: aurelien at aurel32 dot net
- Cc: libc-alpha at sourceware dot org
- Date: Wed, 27 Jul 2016 11:03:36 -0700 (PDT)
- Subject: Re: Fixing sparc ceil/floor/trunc implementations
- Authentication-results: sourceware.org; auth=none
- References: <20160727112051.GA21409@aurel32.net>
From: Aurelien Jarno <aurelien@aurel32.net>
Date: Wed, 27 Jul 2016 13:20:51 +0200
> Coming back about the fix saving and restoring the FSR, it appears
> it has a big impact as expected. In that case the C implementation is
> always faster than the fixed implementations.
>
> I would therefore thing that a way to fix the issue without affecting
> too much the performances is to remove the sparc specific
> implementations.
No objections. We can resurrect the sparc specific assembler from the
repo is we ever want to try and make them work properly again.
> For the sparc64 case where the floating points arguments are passed
> through floating point registers, it might make sense to compile the
> C implementation using VIS3 to benefit from the movdtox/movxtod and
> movwtos/movstow instructions.
That shouldn't be too hard to do with multiarch.