On Tue, Oct 23, 2001 at 09:54:49AM -0700, Christine Ames wrote: > Snipped from "man gcc": I hope you realise that the gcc manual page is NOT the authorative reference. The gcc info files are authorative. > -O2 > Optimize even more. Nearly all supported optimizations that do not involve a > space-speed tradeoff are performed. Loop unrolling and function inlining are > not done, for example. As compared to -O, this option increases both > compilation time and the performance of the generated code. The most important is the compiler uses -finline in this particular case, so it does indeed pay attention to the 'inline' keyword. Without that, it won't and that's exactly the reason why modules compiled without -O2 fail to link or load. > -O3 > Optimize yet more. This turns on everything -O2 does, along with also > turning on -finline-func*- tions. And it also enables -funroll-all-loops which isn't too smart. In most of the cases -O3 actually generates *slower* code than -O2 because unrolling all loops makes the code larger so it won't fit in the I-cache anymore. To give you an idea how expensive a cache miss is: on StrongARM CPU a single cache miss costs you about 60 cycles. -Os is also interesting, it optimises for size. The ARM gcc compiler will still compile correct kernels, I don't know what will happen with other architectures. An intersting sidenote is that -Os sometimes even generates faster code than -O2, because the code fits better in the I-cache. Erik -- J.A.K. (Erik) Mouw, Information and Communication Theory Group, Department of Electrical Engineering, Faculty of Information Technology and Systems, Delft University of Technology, PO BOX 5031, 2600 GA Delft, The Netherlands Phone: +31-15-2783635 Fax: +31-15-2781843 Email: J.A.K.Mouw@its.tudelft.nl WWW: http://www-ict.its.tudelft.nl/~erik/ - Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ IRC Channel: irc.openprojects.net / #kernelnewbies Web Page: http://www.kernelnewbies.org/
