Message ID | 56741838.5080706@mentor.com |
---|---|
State | New |
Headers | show |
Hi Nathan,
> c-c++-common/attr-simd-3.c's use of -fcilkplus means it requires pthreads.
why would this matter for a compile test? What happens without?
Rainer
On 12/18/15 09:36, Rainer Orth wrote: > Hi Nathan, > >> c-c++-common/attr-simd-3.c's use of -fcilkplus means it requires pthreads. > > why would this matter for a compile test? What happens without? Executing on host: nvptx-none-gcc /scratch/nsidwell/openacc/trunk-ptx-shared/src/gcc-mainline/gcc/testsuite/c-c++-common/attr-simd-3.c -fno-diagnostics-show-caret -fdiagnostics-color=never -Wc++-compat -fcilkplus -DSTACK_SIZE=8192 -DNO_TRAMPOLINES -DNO_LABEL_VALUES -DSIGNAL_SUPPRESS -S -o attr-simd-3.s (timeout = 600) nvptx-none-gcc: error: unrecognized command line option '-pthread' compiler exited with status 1 output is: nvptx-none-gcc: error: unrecognized command line option '-pthread' FAIL: c-c++-common/attr-simd-3.c -Wc++-compat PR68158 (test for errors, line 5) FAIL: c-c++-common/attr-simd-3.c -Wc++-compat (test for excess errors) Excess errors: nvptx-none-gcc: error: unrecognized command line option '-pthread' nathan
Hi Nathan, > On 12/18/15 09:36, Rainer Orth wrote: >> Hi Nathan, >> >>> c-c++-common/attr-simd-3.c's use of -fcilkplus means it requires pthreads. >> >> why would this matter for a compile test? What happens without? > > Executing on host: nvptx-none-gcc > /scratch/nsidwell/openacc/trunk-ptx-shared/src/gcc-mainline/gcc/testsuite/c-c++-common/attr-simd-3.c > -fno-diagnostics-show-caret -fdiagnostics-color=never -Wc++-compat > -fcilkplus -DSTACK_SIZE=8192 -DNO_TRAMPOLINES -DNO_LABEL_VALUES > -DSIGNAL_SUPPRESS -S -o attr-simd-3.s (timeout = 600) > nvptx-none-gcc: error: unrecognized command line option '-pthread' > > compiler exited with status 1 > output is: > nvptx-none-gcc: error: unrecognized command line option '-pthread' > > > FAIL: c-c++-common/attr-simd-3.c -Wc++-compat PR68158 (test for errors, > line 5) > FAIL: c-c++-common/attr-simd-3.c -Wc++-compat (test for excess errors) > Excess errors: > nvptx-none-gcc: error: unrecognized command line option '-pthread' ok, I see. I'm uncomfortable declaring the requirement this indirectly/unobviously. I'd rather add a requirement on cilkplus (we have that effective-target keyword already) and update check_effective_target_cilkplus in target-supports.exp for nvptx. Rainer
On Fri, Dec 18, 2015 at 03:49:59PM +0100, Rainer Orth wrote: > Hi Nathan, > > > On 12/18/15 09:36, Rainer Orth wrote: > >> Hi Nathan, > >> > >>> c-c++-common/attr-simd-3.c's use of -fcilkplus means it requires pthreads. > >> > >> why would this matter for a compile test? What happens without? > > > > Executing on host: nvptx-none-gcc > > /scratch/nsidwell/openacc/trunk-ptx-shared/src/gcc-mainline/gcc/testsuite/c-c++-common/attr-simd-3.c > > -fno-diagnostics-show-caret -fdiagnostics-color=never -Wc++-compat > > -fcilkplus -DSTACK_SIZE=8192 -DNO_TRAMPOLINES -DNO_LABEL_VALUES > > -DSIGNAL_SUPPRESS -S -o attr-simd-3.s (timeout = 600) > > nvptx-none-gcc: error: unrecognized command line option '-pthread' > > > > compiler exited with status 1 > > output is: > > nvptx-none-gcc: error: unrecognized command line option '-pthread' > > > > > > FAIL: c-c++-common/attr-simd-3.c -Wc++-compat PR68158 (test for errors, > > line 5) > > FAIL: c-c++-common/attr-simd-3.c -Wc++-compat (test for excess errors) > > Excess errors: > > nvptx-none-gcc: error: unrecognized command line option '-pthread' > > ok, I see. I'm uncomfortable declaring the requirement this > indirectly/unobviously. I'd rather add a requirement on cilkplus (we > have that effective-target keyword already) and update > check_effective_target_cilkplus in target-supports.exp for nvptx. That is what we have in the tree until yesterday and it has been very buggy (disabled all Cilk+ testing on all architectures). See PR68629 or the gcc-patches discussions about this for details. Jakub
Hi Jakub, >> ok, I see. I'm uncomfortable declaring the requirement this >> indirectly/unobviously. I'd rather add a requirement on cilkplus (we >> have that effective-target keyword already) and update >> check_effective_target_cilkplus in target-supports.exp for nvptx. > > That is what we have in the tree until yesterday and it has been very buggy > (disabled all Cilk+ testing on all architectures). > See PR68629 or the gcc-patches discussions about this for details. I didn't mean to introduce a compile test like this, but simply return 0 for nvptx-*-* like we already do for avr-*-*. The compile test is the correct approach, actually, but would need to be done right... Besides, doing the -fcilkplus as a link test had been wrong, anyway, given that many cilkplus tests are compile only... Rainer
2015-12-18 Nathan Sidwell <nathan@acm.org> * c-c++-common/attr-simd-3.c: Requires pthreads. Index: c-c++-common/attr-simd-3.c =================================================================== --- c-c++-common/attr-simd-3.c (revision 231815) +++ c-c++-common/attr-simd-3.c (working copy) @@ -1,5 +1,6 @@ /* { dg-do compile } */ /* { dg-options "-fcilkplus" } */ /* { dg-prune-output "undeclared here \\(not in a function\\)|\[^\n\r\]* was not declared in this scope" } */ +/* { dg-require-effective-target pthread } */ void f () __attribute__((__simd__, __vector__)); /* { dg-error "in the same function marked as a Cilk Plus" "PR68158" } */