This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
RE: [PATCH v2] Remove isinf uses that rely on signed return value
- From: "Wilco Dijkstra" <wdijkstr at arm dot com>
- To: "'Joseph Myers'" <joseph at codesourcery dot com>
- Cc: "'GNU C Library'" <libc-alpha at sourceware dot org>
- Date: Thu, 20 Aug 2015 17:39:18 +0100
- Subject: RE: [PATCH v2] Remove isinf uses that rely on signed return value
- Authentication-results: sourceware.org; auth=none
- References: <000201d0db54$373461e0$a59d25a0$ at com> <alpine dot DEB dot 2 dot 10 dot 1508201435220 dot 30940 at digraph dot polyomino dot org dot uk>
> Joseph Myers wrote:
> On Thu, 20 Aug 2015, Wilco Dijkstra wrote:
>
> > I found a few more cases where isinf is used to get the signbit. Clean
> > these up to improve the readability and maintainability and allow
> > inlining. Generated code is virtually identical once isinf is inlined
> > using the GCC builtin.
> >
> > OK for commit?
>
> What exactly are the dependencies of this patch and their review state?
>
> My understanding is that first we need to get in benchmarks for the
> classification macros. Then, given such benchmarks checked in, the change
> to use the built-in functions can be considered on the basis of evidence
> from those benchmarks. Then, with the built-in functions in use, we can
> consider patches like this based on such comparisons of code generated.
That's correct. The inlining patch was already approved, so everything
is waiting for an OK for the latest version of the math-inlines benchmark
(https://sourceware.org/ml/libc-alpha/2015-08/msg00824.html).
However there is no strict dependency here - this patch will work correctly
without inlining of isinf and cause no regressions for typical inputs.
Wilco