On Thu, Aug 08, 2024 at 10:01:14PM +0200, Alejandro Colomar wrote:
> Here are the suspects:
>
> ./gcc/testsuite/gcc.target/powerpc/sse3-addsubps.c:80:
> for (i = 0; i < sizeof (vals) / sizeof (vals); i += 8)
This is an obvious typo as can be seen in what the test does, what similar
tests committed in the same commit do (all the others use sizeof (vals) /
sizeof (vals[0])) and what the test originates from (i386/sse3-addsubps.c
uses there constant 80, which is that sizeof (vals) / sizeof (vals[0])).
Tested on powerpc64-linux (where the test is UNSUPPORTED) and
powerpc64le-linux, where the test passes before (in that case it tests just
one vector rather than all 10) and after the change.
Committed to trunk as obvious.
2024-08-10 Jakub Jelinek <ja...@redhat.com>
* gcc.target/powerpc/sse3-addsubps.c (TEST): Divide by
sizeof (vals[0]) rather than sizeof (vals).
--- gcc/testsuite/gcc.target/powerpc/sse3-addsubps.c.jj3 2024-03-18
11:02:16.152884555 +0000
+++ gcc/testsuite/gcc.target/powerpc/sse3-addsubps.c 2024-08-10
08:46:57.259430503 +0000
@@ -77,7 +77,7 @@ TEST (void)
int i;
int fail = 0;
- for (i = 0; i < sizeof (vals) / sizeof (vals); i += 8)
+ for (i = 0; i < sizeof (vals) / sizeof (vals[0]); i += 8)
{
p1[0] = vals[i+0];
p1[1] = vals[i+1];
Jakub
