| Message ID | 4C4A9AB8.3070302@net-b.de |
|---|---|
| State | New |
| Headers | show |
On Sat, Jul 24, 2010 at 9:48 AM, Tobias Burnus <burnus@net-b.de> wrote: > Tobias Burnus wrote: >> I will change the tests to use "abs(a-b) > epsilon(a)" instead of "(a >> /= b)". >> > > For integer_exponentiation_3.F90 I could not use "eps" as the check > contains nearest.Thus, I use -ffloat-store. > > Committed as Rev. 162498. Maybe use float-store only for x86 in 32 bits mode?
Steven Bosscher wrote: > On Sat, Jul 24, 2010 at 9:48 AM, Tobias Burnus <burnus@net-b.de> wrote: >> >> For integer_exponentiation_3.F90 I could not use "eps" as the check >> contains nearest.Thus, I use -ffloat-store. > > Maybe use float-store only for x86 in 32 bits mode? You mean something like the following? Tobias --- gcc/testsuite/gfortran.dg/integer_exponentiation_3.F90 (revision 162498) +++ gcc/testsuite/gfortran.dg/integer_exponentiation_3.F90 (working copy) @@ -1,8 +1,7 @@ ! { dg-do run { xfail spu-*-* } } ! FAILs on SPU because of wrong compile-time rounding mode -! { dg-options "-ffloat-store" } +! { dg-options "-ffloat-store" { target i?86-*-* } } ! -! -ffloat-store needed for x87 ! module mod_check implicit none
On Sat, Jul 24, 2010 at 1:59 PM, Tobias Burnus <burnus@net-b.de> wrote: > Steven Bosscher wrote: >> On Sat, Jul 24, 2010 at 9:48 AM, Tobias Burnus <burnus@net-b.de> wrote: >>> >>> For integer_exponentiation_3.F90 I could not use "eps" as the check >>> contains nearest.Thus, I use -ffloat-store. >> >> Maybe use float-store only for x86 in 32 bits mode? > > You mean something like the following? Yes. Ciao! Steven
Index: gcc/testsuite/ChangeLog =================================================================== --- gcc/testsuite/ChangeLog (Revision 162497) +++ gcc/testsuite/ChangeLog (Arbeitskopie) @@ -1,5 +1,10 @@ 2010-07-24 Tobias Burnus <burnus@net-b.de> + * gfortran.dg/entry_13.f90: Use abs() > eps instead of "/=". + * gfortran.dg/integer_exponentiation_3.F90: Use -ffloat-store. + +2010-07-24 Tobias Burnus <burnus@net-b.de> + * gfortran.dg/func_decl_4.f90: Split test into two ... * gfortran.dg/func_decl_5.f90: ... parts. * gfortran.dg/common_resize_1.f: xfail two warnings (cf. PR 45045). Index: gcc/testsuite/gfortran.dg/entry_13.f90 =================================================================== --- gcc/testsuite/gfortran.dg/entry_13.f90 (Revision 162497) +++ gcc/testsuite/gfortran.dg/entry_13.f90 (Arbeitskopie) @@ -67,14 +67,14 @@ type(z) z1 z1 = x1//y1 - if (z1%x .ne. 19.0_4 + 7.0_4) call abort () + if (abs(z1%x - (19.0_4 + 7.0_4)) > epsilon(x1%x)) call abort () z1 = y1//x1 - if (z1%x .ne. 19.0_4 - 7.0_4) call abort () + if (abs(z1%x - (19.0_4 - 7.0_4)) > epsilon(x1%x)) call abort () z1 = x1==y1 - if (z1%x .ne. 19.0_4/7.0_4) call abort () + if (abs(z1%x - 19.0_4/7.0_4) > epsilon(x1%x)) call abort () z1 = y1==x1 - if (z1%x .ne. 19.0_4/7.0_4) call abort () + if (abs(z1%x - 19.0_4/7.0_4) > epsilon(x1%x)) call abort () end program test ! { dg-final { cleanup-modules "type_mod" } } Index: gcc/testsuite/gfortran.dg/integer_exponentiation_3.F90 =================================================================== --- gcc/testsuite/gfortran.dg/integer_exponentiation_3.F90 (Revision 162497) +++ gcc/testsuite/gfortran.dg/integer_exponentiation_3.F90 (Arbeitskopie) @@ -1,6 +1,9 @@ ! { dg-do run { xfail spu-*-* } } ! FAILs on SPU because of wrong compile-time rounding mode -! { dg-options "" } +! { dg-options "-ffloat-store" } +! +! -ffloat-store needed for x87 +! module mod_check implicit none