Hi Matt, Thanks a lot for the info. I also tried building with -O2 and did some testing. Seems that linux/gcc is safe to use -O2.
Would it benefits if runConfigure regard linux as a recognized platform and set debugflag also to "-w -O2"? (Manually setting the bunch of 10+ environment variables, esp the -D's in CFLAGS/CXXFLAGS, and then using ./configure is troublesome and more difficult to maintain than using ./runConfigure.) cheers, Chris On Thu, 19 May 2005 [EMAIL PROTECTED] wrote: > I compile with GCC 3.4.3 and max optimizations on Solaris and I do not have > any problems. > > Try it and see if it works with -O3 -fexpensive-optimizations. > > FYI, in my case, it only got about 2 secs faster on a 1.5 minute test run. > > HTH! > Matt > > -----Original Message----- > From: Chris Cheung [mailto:[EMAIL PROTECTED] > Sent: Thursday, May 19, 2005 2:04 AM > To: [email protected] > Subject: -O2 on Linux > > > Hi, > > In Xerces-C++'s runConfigure script, > > if test $platform = "os400"; then > echo "NATIVE OS400 BUILD" > debugflag=""; > elif test $platform = "irix"; then > debugflag="-w -O2"; > elif test $platform = "aix"; then > debugflag="-w -O2"; > elif test $platform = "os390"; then > debugflag=""; > else > debugflag="-w -O"; > fi > > I notice that only some platforms use -O2, and the default is -O. For > linux (i686/Fedora Core 3/gcc 3.4.2 and x86_64/SuSe 9/gcc 3.3.3), is there > any danger in using -O2? Does anyone has experience in using -O2? > successful or problematic? > > My compilation configuration is: > > ./runConfigure -plinux -minmem -nsocket -tnative -rpthread > > I add -b64 in x86_64. > > Thanks a lot in advance. > > -- Best Regards, Chris Cheung
