--- Comment #3 from ubizjak at gmail dot com 2008-04-16 06:21 ---
(In reply to comment #2)
The tree dump for the kind=8 and kind=10 are exactly identical except for the
call to __builtin_fmodl for kind=10 and __builtin_fmod for kind=8.
It looks like fmodl is broken.
Can somebody
--- Comment #4 from ubizjak at gmail dot com 2008-04-16 11:28 ---
(In reply to comment #3)
Can somebody provide equivalent c testcase? I will look into this problem.
I have a patch in testing.
--
ubizjak at gmail dot com changed:
What|Removed
--- Comment #5 from uros at gcc dot gnu dot org 2008-04-16 15:54 ---
Subject: Bug 35944
Author: uros
Date: Wed Apr 16 15:53:59 2008
New Revision: 134348
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=134348
Log:
PR target/35944
* config/i386/i386.md (fmodxf3):
--- Comment #1 from tkoenig at gcc dot gnu dot org 2008-04-15 21:02 ---
Confirmed, this used to work for 4.2.
$ gfortran-4.2 foo.f90
$ ./a.out
kind = 10
1 -1.40 -1.40 0.00
2 -0.40 -0.40 0.00
3 -0.40 -0.40 0.00
4 -1.40
--- Comment #2 from jvdelisle at gcc dot gnu dot org 2008-04-16 03:51
---
The tree dump for the kind=8 and kind=10 are exactly identical except for the
call to __builtin_fmodl for kind=10 and __builtin_fmod for kind=8.
It looks like fmodl is broken.
--