From mboxrd@z Thu Jan 1 00:00:00 1970 To: 9fans@cse.psu.edu From: "ozan s. yigit" Message-ID: <4da3d9af.0203042219.50978166@posting.google.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Subject: [9fans] gcc bug qed [plan or side effect] Date: Tue, 5 Mar 2002 09:41:50 +0000 Topicbox-Message-UUID: 5fa61896-eaca-11e9-9e20-41e7f4b1d025 in case thomas cares, bug in the gcc optimizer for 3.0.4 under i386. sparc version (egcs 2.91) does not have the bug. took about half an hour (a few tries) to show that it exists in several versions of the compiler. it will take longer to find out exactly where, and why. i could not care less, so i'll leave this as an exercise to those who need it. here is the clue: paranoia in netlib, by kahan. optimizer fails the floting point tests. done. yes, the bug is seen in g77 3.0.4 as well, given -O. oz --- fabricati diem, pvnc.