https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
--- Comment #5 from Michael Meissner ---
Author: meissner
Date: Wed Mar 15 00:25:10 2017
New Revision: 246150
URL: https://gcc.gnu.org/viewcvs?rev=246150&root=gcc&view=rev
Log:
[gcc]
2017-03-14 Michael Meissner
PR target/79947
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
--- Comment #4 from Michael Meissner ---
Created attachment 40976
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40976&action=edit
Proposed patch to fix the problem
The tARGET_RSQRTES macro needed a guard to require -mpowerpc-gfxopt.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
--- Comment #3 from Michael Meissner ---
The problem is the -mno-powerpc-gfxopt option disables floating point
conditional moves, which is needed to use the floating point reciprocal
estimate instructions.
The macro TARGET_FRSQRTES did not have
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
Michael Meissner changed:
What|Removed |Added
Status|UNCONFIRMED |ASSIGNED
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
--- Comment #2 from Martin Liška ---
There's reduced test-case:
/tmp/ppc-ice.c
float a, b;
void
c ()
{
a = __builtin_sqrtf (b);
}
$ ppc64-linux-gnu-gcc -Ofast -mno-powerpc-gfxopt -mcmpb -c ppc-ice.c --verbose
Using built-in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org
--- Comment #1
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79947
Richard Biener changed:
What|Removed |Added
Priority|P3 |P2
Target Milestone|---
