public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* add alignment to enable store merging in strict-alignment targets
@ 2021-01-05  7:46 Alexandre Oliva
  2021-01-05 17:16 ` Jeff Law
  0 siblings, 1 reply; 2+ messages in thread
From: Alexandre Oliva @ 2021-01-05  7:46 UTC (permalink / raw)
  To: gcc-patches


In g++.dg/opt/store-merging-2.C, the natural alignment of types T and
S is a single byte, so we shouldn't expect store merging on
strict-alignment platforms.  Indeed, without something like the
adjust-alignment pass to bump up the alignment of the automatic
variable, as in GCC 10, the optimization does not occur.

This patch adjusts the test so that the required alignment is
expressly stated, and so we don't rely on its accidentally being there
to get the desired optimization.

Regstrapped on x86_64-linux-gnu, also tested on x-arm-wrs-vxworks7r2.
Ok to install?


for  gcc/testsuite/ChangeLog

	* g++.dg/opt/store-merging-2.C: Add the required alignment.
---
 gcc/testsuite/g++.dg/opt/store-merging-2.C |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/gcc/testsuite/g++.dg/opt/store-merging-2.C b/gcc/testsuite/g++.dg/opt/store-merging-2.C
index 3c17033764b75..b1ad7018a1789 100644
--- a/gcc/testsuite/g++.dg/opt/store-merging-2.C
+++ b/gcc/testsuite/g++.dg/opt/store-merging-2.C
@@ -4,7 +4,9 @@
 // { dg-options "-O2 -flifetime-dse=2 -fdump-tree-store-merging-details" }
 // { dg-final { scan-tree-dump "New sequence of 2 stores to replace old one of 3 stores" "store-merging" } }
 
-struct T { char a[128]; };
+/* The alignment is necessary for store-merging to take place on
+   strict-alignment targets.  */
+struct __attribute__ ((__aligned__ (4))) T { char a[128]; };
 struct S { S () : a () { a.a[12] = 0; a.a[13] = 1; a.a[14] = 0; a.a[15] = 6; } T a; };
 void foo (S &);
 void bar (void) { S s; foo (s); }

-- 
Alexandre Oliva, happy hacker  https://FSFLA.org/blogs/lxo/
   Free Software Activist         GNU Toolchain Engineer
        Vim, Vi, Voltei pro Emacs -- GNUlius Caesar

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: add alignment to enable store merging in strict-alignment targets
  2021-01-05  7:46 add alignment to enable store merging in strict-alignment targets Alexandre Oliva
@ 2021-01-05 17:16 ` Jeff Law
  0 siblings, 0 replies; 2+ messages in thread
From: Jeff Law @ 2021-01-05 17:16 UTC (permalink / raw)
  To: Alexandre Oliva, gcc-patches



On 1/5/21 12:46 AM, Alexandre Oliva wrote:
> In g++.dg/opt/store-merging-2.C, the natural alignment of types T and
> S is a single byte, so we shouldn't expect store merging on
> strict-alignment platforms.  Indeed, without something like the
> adjust-alignment pass to bump up the alignment of the automatic
> variable, as in GCC 10, the optimization does not occur.
>
> This patch adjusts the test so that the required alignment is
> expressly stated, and so we don't rely on its accidentally being there
> to get the desired optimization.
>
> Regstrapped on x86_64-linux-gnu, also tested on x-arm-wrs-vxworks7r2.
> Ok to install?
>
>
> for  gcc/testsuite/ChangeLog
>
> 	* g++.dg/opt/store-merging-2.C: Add the required alignment.
OK
jeff


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-01-05 17:16 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-05  7:46 add alignment to enable store merging in strict-alignment targets Alexandre Oliva
2021-01-05 17:16 ` Jeff Law

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).