This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: k_sinl bug fix for ldbl-128 and ldbl-128ibm
From: "Joseph S. Myers" <joseph@codesourcery.com>
Date: Wed, 21 Mar 2012 21:31:36 +0000 (UTC)
> On Fri, 16 Mar 2012, Joseph S. Myers wrote:
>
>> On Fri, 16 Mar 2012, Joseph S. Myers wrote:
>>
>> > On Thu, 15 Mar 2012, David Miller wrote:
>> >
>> > > This gives new ULPs on sparc, this is test-ldouble.out:
>> >
>> > Hopefully it also reduces some ULPs (which is less visible) so being on
>> > average better (as was the case when making this fix for ldbl-96).
>> >
>> > I had the idea last night that this issue only affected k_sinl, but on
>> > looking at it again I think k_cosl is affected as well (and k_sincosl,
>> > which I didn't pay much attention to because I'm keeping the existing
>> > ldbl-96 s_sincosl for now which doesn't use k_sincosl). So I'll post
>> > revised versions of both patches that fix those other functions as well.
>>
>> Like so. The corresponding cosl fix for ldbl-96 reduces the number of
>> ulps changes for test-ldouble on x86 from 20 to 16 (fixing sinl having
>> reduced the number from 30 to 20). (These are figures in the presence of
>> the testsuite changes from the ldbl-96 patch - so include both ULPs for
>> tests that patch newly enables as well as ULPs for tests that were already
>> being run for long double.)
>
> I have now tested this patch
> <http://sourceware.org/ml/libc-alpha/2012-03/msg00513.html> for powerpc,
> with the existing ulps removed so it's tested fairly rather than on a
> basis where whatever random errors the existing code has are expected and
> whatever random errors the new code has aren't. With the patch applied,
> there were 272 errors in test-ldouble.out compared to 276 without (and two
> values for max errors of particular functions went down).
>
> On that basis I think we have further evidence from testing that this is a
> sound change - is the patch OK?
I'm personally fine with it, and I'll scrutinize the resulting sparc ULPs
similarly to how you did for the powerpc ones once you commit this.