This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR70457 (ICE on incompatible call to built-in pow)
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Bill Schmidt <wschmidt at linux dot vnet dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 4 Apr 2016 17:16:43 +0200
- Subject: Re: [PATCH] Fix PR70457 (ICE on incompatible call to built-in pow)
- Authentication-results: sourceware.org; auth=none
- References: <1459727027 dot 6430 dot 52 dot camel at oc8801110288 dot ibm dot com> <20160404065739 dot GA19207 at tucnak dot redhat dot com> <1459782164 dot 6430 dot 64 dot camel at oc8801110288 dot ibm dot com>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Mon, Apr 04, 2016 at 10:02:44AM -0500, Bill Schmidt wrote:
> Sorry again for the misunderstanding. Here are revised patches and
> change logs for trunk, GCC 5, and GCC 4.9. Note that GCC 5 and GCC 4.9
> have additional exposures in tree-ssa-math-opts.c that I've repaired
> similarly. The only differences between 5 and 4.9 are the names of the
> affected functions.
>
> All have passed regstrap on powerpc64le-unknown-linux-gnu. Are these
> ok?
Yes, with small nits:
> +extern double pow (double __x, double __y) __attribute__ ((__nothrow__ , __leaf__));
> +extern double __pow (double __x, double __y) __attribute__ ((__nothrow__ , __leaf__));
No need for __pow prototype, you can also remove the __x and __y argument
names.
> +typedef int int64_t __attribute__ ((__mode__ (__DI__)));
Please remove the above typedef and just use:
> +
> +typedef struct {
> + int64_t data;
long long data;
here. Not sure if we support DImode on all targets, e.g. those with
32-bit BITS_PER_UNIT etc. (where DImode would be 256-bit integer).
Or just remove the data field altogether if it reproduces without it too.
Jakub