[PATCH] PR fortran/66864 -- Use the right precision in FLOOR()

[PATCH] PR fortran/66864 -- Use the right precision in FLOOR()

[Steve Kargl]

[-- Attachment #1: Type: text/plain, Size: 566 bytes --]

The attached patch fixes an issue in the constant 
folding of FLOOR().  In the old code, the precision
of an temporary MPFR variable was set according to
the requested kind of the INTEGER result, ie., the
wrong precision.  The patch sets the precision to
the precision of the arg of FLOOR().  Regression
tested on trunk.  OK to commit.

2015-07-14  Steven G. Kargl  <kargl@gcc.gnu.org>

	* simplify.c (gfc_simplify_floor): Set precision of temporary to
	that of arg.

2015-07-14  Steven G. Kargl  <kargl@gcc.gnu.org>

	gfortran.dg/pr66864.f90: New test.


-- 
Steve

[-- Attachment #2: pr66864.diff --]
[-- Type: text/x-diff, Size: 1214 bytes --]

Index: gcc/fortran/simplify.c
===================================================================
--- gcc/fortran/simplify.c	(revision 225784)
+++ gcc/fortran/simplify.c	(working copy)
@@ -2351,9 +2351,7 @@ gfc_simplify_floor (gfc_expr *e, gfc_exp
   if (e->expr_type != EXPR_CONSTANT)
     return NULL;
 
-  gfc_set_model_kind (kind);
-
-  mpfr_init (floor);
+  mpfr_init2 (floor, mpfr_get_prec (e->value.real));
   mpfr_floor (floor, e->value.real);
 
   result = gfc_get_constant_expr (BT_INTEGER, kind, &e->where);
Index: gcc/testsuite/gfortran.dg/pr66864.f90
===================================================================
--- gcc/testsuite/gfortran.dg/pr66864.f90	(revision 0)
+++ gcc/testsuite/gfortran.dg/pr66864.f90	(working copy)
@@ -0,0 +1,16 @@
+! { dg-do run }
+! PR fortran/66864
+!
+program t
+   implicit none
+   real(8) x
+   x = 2.0d0**26.5d0
+   if (floor(x) /= 94906265) call abort
+   if (floor(2.0d0**26.5d0)/= 94906265) call abort
+   x = 777666555.6d0
+   if (floor(x) /= 777666555) call abort
+   if (floor(777666555.6d0) /= 777666555) call abort
+   x = 2000111222.6d0
+   if (floor(x) /= 2000111222) call abort
+   if (floor(2000111222.6d0) /= 2000111222) call abort
+end program t

Re: [PATCH] PR fortran/66864 -- Use the right precision in FLOOR()

[Janne Blomqvist]

On Tue, Jul 14, 2015 at 10:20 PM, Steve Kargl
<sgk@troutmask.apl.washington.edu> wrote:
> The attached patch fixes an issue in the constant
> folding of FLOOR().  In the old code, the precision
> of an temporary MPFR variable was set according to
> the requested kind of the INTEGER result, ie., the
> wrong precision.  The patch sets the precision to
> the precision of the arg of FLOOR().  Regression
> tested on trunk.  OK to commit.
>
> 2015-07-14  Steven G. Kargl  <kargl@gcc.gnu.org>
>
>         * simplify.c (gfc_simplify_floor): Set precision of temporary to
>         that of arg.
>
> 2015-07-14  Steven G. Kargl  <kargl@gcc.gnu.org>
>
>         gfortran.dg/pr66864.f90: New test.
>
>
> --
> Steve

Ok, thanks.

-- 
Janne Blomqvist