diff mbox series

[testsuite] Fix PR92464 by adjust test case loop bound

Message ID eee4ccb6-983e-f947-9c84-966cf8195299@linux.ibm.com
State New
Headers show
Series [testsuite] Fix PR92464 by adjust test case loop bound | expand

Commit Message

Kewen.Lin Nov. 13, 2019, 7:31 a.m. UTC 
Hi,

As PR92464 shows, the recent vectorization cost adjustment on load
insns is responsible for this regression.  It leads the profitable
min iteration count to change from 19 to 12.  The case happens to
hit the threshold.  By actual runtime performance evaluation, the
vectorized version perform on par with non vectorized version
(before).  So the vectorization on 12 is actually fine.  To keep
the case sensitive on high peeling cost, this patch is to adjust
the loop bound from 16 to 14.

Verified on ppc64-redhat-linux (BE P7) and powerpc64le-linux-gnu
(LE P8). 


BR,
Kewen

-----

gcc/testsuite/ChangeLog

2019-11-13  Kewen Lin  <linkw@gcc.gnu.org>

	PR target/92464
	* gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c: Adjust
	loop bound due to load cost adjustment.

Comments

Segher Boessenkool Nov. 13, 2019, 10:42 a.m. UTC | #1 
Hi!

On Wed, Nov 13, 2019 at 03:31:11PM +0800, Kewen.Lin wrote:
> As PR92464 shows, the recent vectorization cost adjustment on load
> insns is responsible for this regression.  It leads the profitable
> min iteration count to change from 19 to 12.  The case happens to
> hit the threshold.  By actual runtime performance evaluation, the
> vectorized version perform on par with non vectorized version
> (before).  So the vectorization on 12 is actually fine.  To keep
> the case sensitive on high peeling cost, this patch is to adjust
> the loop bound from 16 to 14.

This is fine, okay for trunk, thanks!

Could you add a comment to the testcase saying what it is trying to test
exactly?  It wasn't obvious to me.  Just one or two lines is fine.


Segher
Kewen.Lin Nov. 14, 2019, 6 a.m. UTC | #2 
Hi Segher,

on 2019/11/13 下午6:42, Segher Boessenkool wrote:
> Hi!
> 
> On Wed, Nov 13, 2019 at 03:31:11PM +0800, Kewen.Lin wrote:
>> As PR92464 shows, the recent vectorization cost adjustment on load
>> insns is responsible for this regression.  It leads the profitable
>> min iteration count to change from 19 to 12.  The case happens to
>> hit the threshold.  By actual runtime performance evaluation, the
>> vectorized version perform on par with non vectorized version
>> (before).  So the vectorization on 12 is actually fine.  To keep
>> the case sensitive on high peeling cost, this patch is to adjust
>> the loop bound from 16 to 14.
> 
> This is fine, okay for trunk, thanks!
> 
> Could you add a comment to the testcase saying what it is trying to test
> exactly?  It wasn't obvious to me.  Just one or two lines is fine.
> 

Thanks!  Done with additional comments, committed in r278195.


BR,
Kewen
diff mbox series

Patch

diff --git a/gcc/testsuite/gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c b/gcc/testsuite/gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c
index 4a7da2e..1bb064e 100644
--- a/gcc/testsuite/gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c
+++ b/gcc/testsuite/gcc.dg/vect/costmodel/ppc/costmodel-vect-76b.c
@@ -4,7 +4,7 @@ 
 #include <stdarg.h>
 #include "../../tree-vect.h"

-#define N 16
+#define N 14
 #define OFF 4

 /* Check handling of accesses for which the "initial condition" -