This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 0/6] Generate Complex functions from a common template.
- From: "Paul E. Murphy" <murphyp at linux dot vnet dot ibm dot com>
- To: Joseph Myers <joseph at codesourcery dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Thu, 30 Jun 2016 20:17:01 -0500
- Subject: Re: [PATCH 0/6] Generate Complex functions from a common template.
- Authentication-results: sourceware.org; auth=none
- References: <cover dot 1467323999 dot git dot murphyp at linux dot vnet dot ibm dot com> <alpine dot DEB dot 2 dot 20 dot 1606302231240 dot 21713 at digraph dot polyomino dot org dot uk>
On 06/30/2016 05:32 PM, Joseph Myers wrote:
> On Thu, 30 Jun 2016, Paul E. Murphy wrote:
>
>> math-type-macros.h hides much of the complexity of the
>> aliasing for the common case, with cpow10 beings the
>> exception. Likewise, some heavily used functions
>
> There's no such function as cpow10.
Whoops, that should read clog10.
>
>> Finally, a few functions have some very minor drift
>> between the float, double, and ldouble variants. I
>> have preserved these in order to keep my sanity
>> when validating the generated code has not changed
>> meaningifully.
>
> In such cases, the right thing to do is to have preliminary patches that
> only address the drift, so that the comparison is clean in the subsequent
> stages.
I believe there are only 4 such cases, and are very limited.
Spurious usage of glibc_likely in two cases, and two cases
of testing for FE_INVALID macro before trying to raise it.
The behavior is exclusive to s_ccosh*.c and s_csinh*.c.
The diff isn't much obfuscated by either. If you can
tell me which is the desired behavior, I can propose
and submit those patches independently.