On Wed, Mar 10, 2021 at 7:04 AM Alexandre Oliva <ol...@adacore.com> wrote:
>
>
> ppc configurations that have -mstrict-align enabled by default fail
> gcc.dg/strlenopt-80.c, because some memcpy calls don't get turned into
> MEM_REFs, which defeats the tested-for strlen optimization.
>
> This was regstrapped on x86_64-linux-gnu, tested with a cross to a
> ppc64-vxworks7r2 configured with -mstrict-align enabled by default,
> and I'm now also regstrapping on ppc64-linux-gnu just to be sure.
> Ok to install?
Btw, I asked during two stage1s whether relaxing the folding to also
apply to strict-align targets would be OK. But I didn't receive any
answers and thus dropped the ball ...
Basically for the folding turning integer-mode sized copies into
separate load + store also allow the load/store to be not aligned
according to their mode (and rely on RTL expansion to perform
bitfield load/store ops for them).
Richard.
>
> for gcc/testsuite/ChangeLog
>
> * gcc.dg/strlenopt-80.c: Add -mno-strict-align on ppc.
> ---
> gcc/testsuite/gcc.dg/strlenopt-80.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/gcc/testsuite/gcc.dg/strlenopt-80.c
> b/gcc/testsuite/gcc.dg/strlenopt-80.c
> index 9124fe4740a1b..9978f5cf3bc56 100644
> --- a/gcc/testsuite/gcc.dg/strlenopt-80.c
> +++ b/gcc/testsuite/gcc.dg/strlenopt-80.c
> @@ -5,7 +5,11 @@
> such a store.
> { dg-do compile { target aarch64*-*-* i?86-*-* powerpc*-*-* x86_64-*-* } }
>
> - { dg-options "-O2 -Wall -fdump-tree-optimized" } */
> + { dg-options "-O2 -Wall -fdump-tree-optimized" }
> +
> + On powerpc configurations that have -mstrict-align by default,
> + the memcpy calls for ncpylog >= 3 are not turned into MEM_REFs.
> + { dg-additional-options "-mno-strict-align" { target powerpc*-*-* } } */
>
> #define CHAR_BIT __CHAR_BIT__
> #define SIZE_MAX __SIZE_MAX__
>
>
> --
> Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/
> Free Software Activist GNU Toolchain Engineer
> Vim, Vi, Voltei pro Emacs -- GNUlius Caesar
