This is the mail archive of the
mailing list for the glibc project.
Re: [PATCH v2] Add math-inline benchmark
- From: "Carlos O'Donell" <carlos at redhat dot com>
- To: Wilco Dijkstra <wdijkstr at arm dot com>, 'Ondřej Bílka' <neleai at seznam dot cz>
- Cc: "'GNU C Library'" <libc-alpha at sourceware dot org>
- Date: Fri, 17 Jul 2015 09:48:42 -0400
- Subject: Re: [PATCH v2] Add math-inline benchmark
- Authentication-results: sourceware.org; auth=none
- References: <002001d0bfb8$b36fa330$1a4ee990$ at com> <20150716225056 dot GA24479 at domone> <002501d0c094$3ea04cd0$bbe0e670$ at com>
On 07/17/2015 09:26 AM, Wilco Dijkstra wrote:
> It shouldn't be necessary to use assembly code, GCC is perfectly capable of
> generating efficient code for EXTRACT_WORDS. If GCC doesn't for x64 then you
> should report it and ensure it gets fixed. I don't understand why people keep
> using target-specific hacks and assembly rather than fixing the underlying
> issues and use generic code that works well on all targets...
Agreed. The compiler can't schedule the assembly at all, and worse.
>> So at least on x64 we should publish math_private inlines instead using
>> slow builtins.
> Well it was agreed we are going to use the GCC built-ins and then improve
> those. If you want to propose additional patches with special inlines for
> x64 then please go ahead, but my plan is to improve the builtins.