On Wednesday 26 January 2005 2:12, Michael Obster wrote: > > > As for reading GCC Documentation, what about it? Did you ever think that > > maybe... just maybe... the build environment might be optimizing for the > > specific platform you are compiling on? Jeez. > > Thats the point. There are many build kits out which do optimization for > specific architectures. For example look at Gentoo or ROCK Linux. The > build system optimizes the complete build environment for a selected > cpu. The only exception on x86 is pentium 4, because this is still > experimental *g*. > If you look at the assembler output ,gcc creates, there is much > difference between the different platforms. For example parallel working > on cpu registers, using SSE, using 3dnow or mmx and so on. Also a speed > up is between this different optimization. > This features are in the gcc documentation for example ;-). > A hardcoded -march=pentiumpro does not recognizes such things, even I > know qmake is a little bit tricky here. Personally I prefer autotools > for building software. > So to get an optimal optimization, it is better to let the build > determine which option is needed ;-).
Actually, if you read through the mailing list archives you'd find that -march=pentiumpro actually generates better performing object code than some of the more processor-specific (i.e., -march=athlon-xp) optimizations. It all has to do with the type of operations being performed. Also, as far as non-x86 architectures are concerned... MythTV is not supported "out of the box" for anything other than Linux on x86. There is an ongoing effort to support Mac OS X out of the box, as you'll see if you examine the settings.pro file. As for other architectures like Linux PPC or *BSD, I believe most of the code has been patched (mostly endianness issues), but the build configuration is not supported by default (yet). -JAC _______________________________________________ mythtv-users mailing list [email protected] http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
