[PPL-devel] [GIT] ppl/ppl(floating_point): Use ROUND_DOWN instead of ROUND_UP to compute the negation of the upper
Fabio Bossi
bossi at cs.unipr.it
Thu Sep 17 10:50:09 CEST 2009
Module: ppl/ppl
Branch: floating_point
Commit: b9f5aee8d9e95cda8126480538154b2b929bb163
URL: http://www.cs.unipr.it/git/gitweb.cgi?p=ppl/ppl.git;a=commit;h=b9f5aee8d9e95cda8126480538154b2b929bb163
Author: Fabio Bossi <bossi at cs.unipr.it>
Date: Thu Sep 17 10:52:37 2009 +0200
Use ROUND_DOWN instead of ROUND_UP to compute the negation of the upper
bound of -var.
---
src/Octagonal_Shape.templates.hh | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/src/Octagonal_Shape.templates.hh b/src/Octagonal_Shape.templates.hh
index 1d93abf..0c6efd2 100644
--- a/src/Octagonal_Shape.templates.hh
+++ b/src/Octagonal_Shape.templates.hh
@@ -5339,7 +5339,7 @@ linear_form_upper_bound(const Linear_Form< Interval<T, Interval_Info> >& lf,
assign_r(curr_var_ub, matrix[n_var+1][n_var], ROUND_NOT_NEEDED);
div_2exp_assign_r(curr_var_ub, curr_var_ub, 1, ROUND_UP);
neg_assign_r(curr_minus_var_ub, matrix[n_var][n_var+1], ROUND_NOT_NEEDED);
- div_2exp_assign_r(curr_minus_var_ub, curr_minus_var_ub, 1, ROUND_UP);
+ div_2exp_assign_r(curr_minus_var_ub, curr_minus_var_ub, 1, ROUND_DOWN);
// Optimize the most common case: curr = +/-[1;1]
if (curr_lb == 1 && curr_ub == 1) {
add_assign_r(result, result, std::max(curr_var_ub, curr_minus_var_ub),
More information about the PPL-devel
mailing list