From: dcf@interactive.net ("David C. Feldmeier") Subject: floating point optimization bug, beta version 16 of GCC 19 Oct 1996 13:46:52 -0700 Sender: daemon@cygnus.com Approved: cygnus.gnu-win32@cygnus.com Distribution: cygnus Message-ID: <1.5.4.32.19961019195447.006b8254.cygnus.gnu-win32@interactive.net> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" X-Sender: dcf@interactive.net X-Mailer: Windows Eudora Light Version 1.5.4 (32) Original-To: gnu-win32@cygnus.com Original-Cc: dcf@onyx.interactive.net Original-Sender: owner-gnu-win32@cygnus.com I have been using your GCC for several months now, and I've generally had good luck with it. However, I seems to have found an optimizer bug. Here is some code that shows the bug: ?* START OF CODE */ # /* bug.c * * David C. Feldmeier * October 19, 1996 * */ #include #include static void func_B( unsigned long redundancy) { /* percentage of redundancy to use */ printf ("inv redund %f, test %f\n", 100.0/(double)redundancy, 100.0/(double)300L); } int main() { unsigned long redundancy = 300L; func_B(redundancy); } /* END OF CODE */ Here is the list of commands given to bash: bash$ gcc -o bug.exe bug.c bash$ bug inv redund 0.333333, test 0.333333 bash$ gcc -O1 -o bug.exe bug.c bash$ bug inv redund 0.000000, test 0.333333 bash$ As you can see, when the optimizer is running, it the code generates an incorrect floating point result. The code which generates this bug is no different than code that I ran sucessfully using a previous version (13, I believe) of GCC. Also, this bug may be related to a floating point problem reported by Peter Fan to the mailing list on April 30, 1996. I am running the lastest version of the GCC compiler, beta release 16. My computer is a 133 MHz Pentium Micron Millenia with 16M RAM. I am running GCC using the DOS emulator under Windows95. If you need any other information, please let me know. Thanks. -Dave - For help on using this list, send a message to "gnu-win32-request@cygnus.com" with one line of text: "help".