powerpc64le: Avoid conflicting types for f64xfmaf128 when IFUNC is not used
Checks
Context |
Check |
Description |
dj/TryBot-apply_patch |
success
|
Patch applied to master at the time it was sent
|
dj/TryBot-32bit |
success
|
Build for i686
|
Commit Message
Avoid defining f64xfmaf128 twice when building s_fmaf128.c.
This can be reproduced on powerpc64le whenever f128 functions do not
have IFUNC enabled, e.g. using "--with-cpu=power8 --disable-multi-arch", or
when using "-with-cpu=power9".
Fixes: b3f27d8150d4f ("Add narrowing fma functions")
---
sysdeps/ieee754/float128/s_fmaf128.c | 2 ++
1 file changed, 2 insertions(+)
Comments
On Thu, 23 Sep 2021, Tulio Magno Quites Machado Filho via Libc-alpha wrote:
> Avoid defining f64xfmaf128 twice when building s_fmaf128.c.
> This can be reproduced on powerpc64le whenever f128 functions do not
> have IFUNC enabled, e.g. using "--with-cpu=power8 --disable-multi-arch", or
> when using "-with-cpu=power9".
OK. This shows up that it would be desirable for such a powerpc64le
configuration to be added to build-many-glibcs.py (in a separate patch);
we already have --disable-multi-arch configurations there for some other
architectures where that makes a significant difference to what gets
built.
Joseph Myers <joseph@codesourcery.com> writes:
> OK. This shows up that it would be desirable for such a powerpc64le
> configuration to be added to build-many-glibcs.py (in a separate patch);
> we already have --disable-multi-arch configurations there for some other
> architectures where that makes a significant difference to what gets
> built.
Pushed as 54ff4f1e39067bfd04fb2141710637a11ef88862.
Paul Murphy proposed a patch to update build-many-glibcs.py here:
http://patchwork.sourceware.org/project/glibc/patch/20210923205752.358551-1-murphyp@linux.ibm.com/
@@ -1,3 +1,5 @@
#define NO_MATH_REDIRECT
+#define f64xfmaf128 __hide_f64xfmaf128
#include <float128_private.h>
+#undef f64xfmaf128
#include "../ldbl-128/s_fmal.c"