real.c (real_trunc, [...]): New functions to implement trunc, floor and ceil respecti...
authorRoger Sayle <roger@eyesopen.com>
Thu, 3 Jul 2003 21:38:55 +0000 (21:38 +0000)
committerRoger Sayle <sayle@gcc.gnu.org>
Thu, 3 Jul 2003 21:38:55 +0000 (21:38 +0000)
commit0a9530a9d7cf389493916027f9980fcbf028b602
tree6f8671ee7d010f3f877a3ed92690c0d270f47544
parent7516d73631a96bcc28178f057bee4e713022b7d5
real.c (real_trunc, [...]): New functions to implement trunc, floor and ceil respectively.

* real.c (real_trunc, real_floor, real_ceil): New functions
to implement trunc, floor and ceil respectively.
* real.h (real_trunc, real_floor, real_ceil): Prototype here.
* builtins.c (integer_valued_real_p): New function to test if
a floating point expression has an integer valued result.
(fold_trunc_transparent_mathfn): Optimize foo(foo(x)) as
foo(x) where foo is an integer rounding function.  Similarly,
optimize foo(bar(x)) as bar(x), and foo((double)(int)x) as
(double)(int)x when both foo and bar are integer rounding
functions and we don't need to honor errno.
(fold_builtin_trunc, fold_builtin_floor, fold_builtin_ceil):
New functions to fold trunc, floor and ceil.
(fold_builtin): Use fold_builtin_trunc to fold BUILT_IN_TRUNC*,
fold_builtin_floor to fold BUILT_IN_FLOOR* and fold_builtin_ceil
to fold BUILT_IN_CEIL*.
* fold-const.c (tree_expr_nonnegative_p): Handle FLOAT_EXPR and
the remaining integer rounding functions.

* gcc.dg/builtins-25.c: New testcase.
* gcc.dg/builtins-26.c: New testcase.

From-SVN: r68903
gcc/ChangeLog
gcc/builtins.c
gcc/fold-const.c
gcc/real.c
gcc/real.h
gcc/testsuite/ChangeLog
gcc/testsuite/gcc.dg/builtins-25.c [new file with mode: 0644]
gcc/testsuite/gcc.dg/builtins-26.c [new file with mode: 0644]