[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 cesar at gcc dot gnu.org changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #9 from cesar at gcc dot gnu.org --- I've updated the test in trunk and gomp-4_0-branch and that regression is no longer showing up in the power8 testlogs in gcc-testsuites.
[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 --- Comment #8 from cesar at gcc dot gnu.org --- Author: cesar Date: Thu Mar 10 22:50:40 2016 New Revision: 234124 URL: https://gcc.gnu.org/viewcvs?rev=234124=gcc=rev Log: libgomp/ PR testsuite/70009 * testsuite/libgomp.oacc-c-c++-common/vprop.c: Make test data signed. Modified: trunk/libgomp/ChangeLog trunk/libgomp/testsuite/libgomp.oacc-c-c++-common/vprop.c
[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 --- Comment #7 from Jakub Jelinek --- (In reply to cesar from comment #6) > Created attachment 37898 [details] > test fix > > I've tested this patch on an arm target and it passes now. All this patch > does is make the type macro signed. > > So the intent behind this test was to make sure that vector propagation > works on integral types less than 32-bits on nvptx targets. I was using a > signed values explicitly because the nvptx backend uses a shuffle > instruction to propagate variables, and that instruction only takes unsigned > registers. So the signed values added a little more test coverage. I didn't > realize that chars aren't signed on targets. > > This patch should fix this issue. Preapproved for trunk.
[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 --- Comment #6 from cesar at gcc dot gnu.org --- Created attachment 37898 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=37898=edit test fix I've tested this patch on an arm target and it passes now. All this patch does is make the type macro signed. So the intent behind this test was to make sure that vector propagation works on integral types less than 32-bits on nvptx targets. I was using a signed values explicitly because the nvptx backend uses a shuffle instruction to propagate variables, and that instruction only takes unsigned registers. So the signed values added a little more test coverage. I didn't realize that chars aren't signed on targets. This patch should fix this issue.
[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 James Greenhalgh changed: What|Removed |Added Target|powerpc*-*-*, aarch64-*-* |powerpc*-*-*, aarch64-*-*, ||arm*-*-* Last reconfirmed|2016-02-29 00:00:00 |2016-3-7 CC||jgreenhalgh at gcc dot gnu.org --- Comment #5 from James Greenhalgh --- Also failing on arm/aarch64 (so good further evidence of signed vs. unsigned char). Forcing the macro to use signed types clears the error for me on arm-none-linux-gnueabihf (though I don't know if this is correct).
[Bug testsuite/70009] test case libgomp.oacc-c-c++-common/vprop.c fails starting with its introduction in r233607
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70009 Thomas Schwinge changed: What|Removed |Added Keywords||openacc Component|libgomp |testsuite Assignee|unassigned at gcc dot gnu.org |cesar at gcc dot gnu.org --- Comment #4 from Thomas Schwinge --- I guess default signed vs. unsigned char.