On 08/03/2016 03:28 PM, Richard Biener wrote:
> On Wed, Aug 3, 2016 at 11:54 AM, Martin Liška <mli...@suse.cz> wrote:
>> Hi.
>>
>> I've just grabbed patch a that was suggested in the PR (and IMHO makes
>> sense).
>>
>> Patch can bootstrap on ppc64le-redhat-linux and survives regression tests.
>>
>> Ready to be installed?
>
> Err, but what was suggested, not enable it with -Os is already done - you
> disable it unconditionally?
Sorry, I was wrong ;) Looks the bug is fixed since r222033, where following
hunk was added:
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 110ec4a..b442da9 100644
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -4168,6 +4168,7 @@ ix86_option_override_internal (bool main_args_p,
if (opts->x_flag_prefetch_loop_arrays < 0
&& HAVE_prefetch
&& (opts->x_optimize >= 3 || opts->x_flag_profile_use)
+ && !opts->x_optimize_size
&& TARGET_SOFTWARE_PREFETCHING_BENEFICIAL)
opts->x_flag_prefetch_loop_arrays = 1;
I'm going to close the PR.
Martin
>
> Richard.
>
>> Martin
