| Message ID | 657de55b-a1ec-c46b-f76b-5ef97483303d@codesourcery.com |
|---|---|
| State | New |
| Headers | show |
| Series | testsuite/libgomp.fortran/pr66199-2.f90 – remove dg-options | expand |
On Mon, Nov 04, 2019 at 01:18:58PM +0100, Tobias Burnus wrote: > This test case had from the beginning both > ! { dg-do run } > ! { dg-options "-O2" } > > However, 'dg-do run' implies torture, i.e. multiple optimization options, > while 'dg-options "-O2"' implies -O2, only. > > This means that the test case is compiled 4 times with > "-O0", "-O1", "-O3 -f…" and "-Os" – followed by "-O2". > Hence, it is run 4 times with nearly identical options > (With "-O3" [followed by "-O2"], the "-fomit-frame-pointer -funroll-loops -fpeel-loops > -ftracer -finline-functions" options do get added.) > > Hence, we can either remove "dg-do run" or "dg-options". I suggest to remove > the latter. > > OK? Note, pr66199-1.f90 has the same issue. And I'd probably go for the removal of dg-do run instead. Jakub
On 11/4/19 1:29 PM, Jakub Jelinek wrote: > Note, pr66199-1.f90 has the same issue. And I'd probably go for the > removal of dg-do run instead. Well, there are even more (ignoring those which have -std=legacy or -fno-inline or … in dg-options): libgomp.fortran/examples-4/declare_target-5.f90, libgomp.fortran/taskloop[234].f90 libgomp.fortran/examples-4/declare_target-5.f90:! { dg-do run { target vect_simd_clones } } I assume, you prefer no "dg-do run" for those as well: How about the attached patch? Tobias
On Mon, Nov 04, 2019 at 02:02:38PM +0100, Tobias Burnus wrote: > On 11/4/19 1:29 PM, Jakub Jelinek wrote: > > Note, pr66199-1.f90 has the same issue. And I'd probably go for the > > removal of dg-do run instead. > > Well, there are even more (ignoring those which have -std=legacy or > -fno-inline or … in dg-options): > libgomp.fortran/examples-4/declare_target-5.f90, > libgomp.fortran/taskloop[234].f90 > > libgomp.fortran/examples-4/declare_target-5.f90:! { dg-do run { target vect_simd_clones } } That vect_simd_clones can't be dropped of course. > I assume, you prefer no "dg-do run" for those as well: > How about the attached patch? But your patch doesn't do that, so ok for trunk. > 2019-11-04 Tobias Burnus <tobias@codesourcery.com> > > * testsuite/libgomp.fortran/pr66199-1.f90: Remove > 'dg-do run' (implies torture test) as 'dg-options "O2"' is used. > * testsuite/libgomp.fortran/pr66199-2.f90: Ditto. > * testsuite/libgomp.fortran/taskloop2.f90: Ditto. > * testsuite/libgomp.fortran/taskloop3.f90: Ditto. > * testsuite/libgomp.fortran/taskloop4.f90: Ditto. Jakub
2019-11-04 Tobias Burnus <tobias@codesourcery.com> * testsuite/libgomp.fortran/pr66199-2.f90: Remove dg-options from torture test. diff --git a/libgomp/testsuite/libgomp.fortran/pr66199-2.f90 b/libgomp/testsuite/libgomp.fortran/pr66199-2.f90 index 0cc0fa5097f..5ee74149089 100644 --- a/libgomp/testsuite/libgomp.fortran/pr66199-2.f90 +++ b/libgomp/testsuite/libgomp.fortran/pr66199-2.f90 @@ -1,6 +1,5 @@ ! PR middle-end/66199 ! { dg-do run } -! { dg-options "-O2" } integer :: u(1024), v(1024), w(1024), a, b, c, d, e, a1, b1, a2, b2, d1, d2 a = 1
This test case had from the beginning both ! { dg-do run } ! { dg-options "-O2" } However, 'dg-do run' implies torture, i.e. multiple optimization options, while 'dg-options "-O2"' implies -O2, only. This means that the test case is compiled 4 times with "-O0", "-O1", "-O3 -f…" and "-Os" – followed by "-O2". Hence, it is run 4 times with nearly identical options (With "-O3" [followed by "-O2"], the "-fomit-frame-pointer -funroll-loops -fpeel-loops -ftracer -finline-functions" options do get added.) Hence, we can either remove "dg-do run" or "dg-options". I suggest to remove the latter. OK? Tobias