| Message ID | 54E72A3F.5020105@mentor.com |
|---|---|
| State | New |
| Headers | show |
On Fri, Feb 20, 2015 at 01:36:15PM +0100, Tom de Vries wrote: > On 20-02-15 10:42, Jakub Jelinek wrote: > >On Fri, Feb 20, 2015 at 10:25:54AM +0100, Tom de Vries wrote: > >>this patch reverses the abort logic in pr30957-1.c, such that it aborts on > >>failure rather than on success. > > > >That sounds really weird. From the description it looks like it is a known bug > >that we don't return -0.0. > >If 0.0 is the right return value instead, I'd the test should be written as > >if (__builtin_copysignf (1.0, foo (0.0 / -5.0, 10)) != 1.0) > > abort (); > >to make it clear you are expecting positive 0. > > > > Updated patch accordingly. OK for stage1? If it is a known bug, it should better stay as xfail, rather than pretending the wrong behavior is correct. Jakub
On Feb 20, 2015, at 4:36 AM, Tom de Vries <Tom_deVries@mentor.com> wrote: > On 20-02-15 10:42, Jakub Jelinek wrote: >> On Fri, Feb 20, 2015 at 10:25:54AM +0100, Tom de Vries wrote: >>> this patch reverses the abort logic in pr30957-1.c, such that it aborts on >>> failure rather than on success. >> >> That sounds really weird. From the description it looks like it is a known bug >> that we don't return -0.0. >> If 0.0 is the right return value instead, I'd the test should be written as >> if (__builtin_copysignf (1.0, foo (0.0 / -5.0, 10)) != 1.0) >> abort (); >> to make it clear you are expecting positive 0. >> > > Updated patch accordingly. OK for stage1? I’ve tried to read through the bug report and all the patches, who did them and why… The entire thread is messier than I’d like, which makes dealing with this whole thing messy. The bug report I marked as fixed, as I think it now works as the bug reporter expects. Seems like a mistake it wasn’t closed a while ago. I now see why you went with this patch. Why wait for stage 1… Lets just put it in now and put an end to the misery. Ok for trunk now.
2015-02-20 Tom de Vries <tom@codesourcery.com> * gcc.dg/pr30957-1.c: Make pr30957-1.c pass rather xfail. --- gcc/testsuite/gcc.dg/pr30957-1.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/gcc/testsuite/gcc.dg/pr30957-1.c b/gcc/testsuite/gcc.dg/pr30957-1.c index 65b98fa..0ed150d 100644 --- a/gcc/testsuite/gcc.dg/pr30957-1.c +++ b/gcc/testsuite/gcc.dg/pr30957-1.c @@ -1,10 +1,6 @@ -/* { dg-do run { xfail *-*-* } } */ -/* We don't (and don't want to) perform this optimisation on soft-float - targets, where each addition is a library call. This test requires - -fassociative-math for enabling the variable-expansion as well as - -fsigned-zeros for honoring the sign of zero; but - they can not co-exist; also under -funsafe-math-optimizations, so we - expect it to fail. */ +/* { dg-do run } */ +/* We don't (and don't want to) perform this optimisation on soft-float targets, + where each addition is a library call. / /* { dg-require-effective-target hard_float } */ /* { dg-options "-O2 -funroll-loops -funsafe-math-optimizations -fvariable-expansion-in-unroller -fdump-rtl-loop2_unroll" } */ @@ -26,12 +22,14 @@ foo (float d, int n) int main () { - if (__builtin_copysignf (1.0, foo (0.0 / -5.0, 10)) != -1.0) + /* When compiling standard compliant we expect foo to return -0.0. But the + variable expansion during unrolling optimization (for this testcase enabled + by -funsafe-math-optimization) instantiates copy(s) of the accumulator + which it initializes with +0.0. Hence we expect that foo returns +0.0. */ + if (__builtin_copysignf (1.0, foo (0.0 / -5.0, 10)) != 1.0) abort (); exit (0); } /* { dg-final { scan-rtl-dump "Expanding Accumulator" "loop2_unroll" } } */ /* { dg-final { cleanup-rtl-dump "loop*" } } */ - - -- 1.9.1