--- Comment #8 from ubizjak at gmail dot com 2008-05-08 06:56 ---
Fixed.
--
ubizjak at gmail dot com changed:
What|Removed |Added
Status|ASSIGNED
--- Comment #1 from ubizjak at gmail dot com 2008-05-07 16:39 ---
Hm... strange, because my patch changed x86 target specific MMX and SSE vector
builtins only. I don't see any __builtin_X usage in gfc_simplify_set_exponent
that would trigger codepaths that were changed.
Can you do a
--- Comment #2 from dominiq at lps dot ens dot fr 2008-05-07 17:06 ---
Can you do a backtrace of the failure?
I tried, but my knowledge of gdb is too limited. I get the error, but backtrace
gives no stack.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36169
--- Comment #3 from ubizjak at gmail dot com 2008-05-07 18:08 ---
Created an attachment (id=15610)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15610action=view)
P
Can you try attached patch that fixes some patterns only at expand time?
--
--- Comment #4 from ubizjak at gmail dot com 2008-05-07 18:32 ---
I see the problem:
define_insn_and_split *fixuns_truncmode_1 is a post-reload splitter that
calls ix86_split_convert_uns_si_sse after reload. There we have:
gen_sse2_loadlpd (value, CONST0_RTX (V2DFmode), input)
and
--- Comment #5 from hjl dot tools at gmail dot com 2008-05-07 19:04 ---
This may be related to PR 36174.
--
hjl dot tools at gmail dot com changed:
What|Removed |Added
--- Comment #6 from ubizjak at gmail dot com 2008-05-07 19:11 ---
Should be fixed now.
Sorry for the breakage, I didn't notice one postreload usage of loadlpd.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36169
--- Comment #7 from dominiq at lps dot ens dot fr 2008-05-07 19:54 ---
Should be fixed now.
I am now at stage 3, so it seems fixed.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36169
