On 5 Dec 2002, Chris Kloiber wrote: >> I downloaded VMware-workstation-3.2.0-2230.i386.rpm from VMWare site, >> installed it and when running "vmware-config.pl" it obviously told me >> that there are no precompiled modules for my running system and that it >> was going to do some. >> >> What really gets my attention, is that when it is compiling the modules, >> I get these warnings from gcc: >> >> make[2]: Entering directory >> `/tmp/vmware-config0/vmmon-only/driver-2.4.18-18.8.0' >> `-m486' is deprecated. Use `-march=i486' or `-mcpu=i486' instead. >(snip) >> Anyway, it compiles and runs "fine", but I don't know why does it insist >> in using -m486, even if I do have a i686 kernel: > >It's stating that the compiler flag '-m486' should not be used anymore, >that in the very short future that flag will go away completely, and the >warning messages you are seeing now will become error messages that halt >the compiler. VMware needs to fix their code. > >> And in a side note, it really runs SLOWLY in a PIV 1.4 GHz, with win98. >> I remember that I installed VMWare (don't know which version) in a PIII >> 800 MHz a year ago or so, with Windows XP on it. It used to run very >> fine! I am thinking that probably those "optimizations" for 486 are >> doing something tricky to my system. > >No, they aren't optimizations really. IIRC, it means that only >instructions common to i486 cpu's and higher will be used. This permits >VMware to run on processors from i486-i686 theoretically, although in >practice it's so that you can run it on AMD K5's and Cyrix 586 cpu's I >would guess. I don't believe for a minute a real i486 would run VMware. > >> Oh, and yes, I do load the libnice.so before running VMWare. > >libnice.so? Perhaps I'm spoiled hardware wise, but I never nice my >VMware. 486 optimizations are probably the worst possible optimization level to use for any processor except a real i486. Later processors are very different. If for some reason the actual code requires 486 specific instructions that aren't available on the original i386 instruction set, such as BSWAP, then using something like "-march=i486 -mcpu=i686" would provide access to the i486 class instructions, but optimize scheduling for i686 class CPUs. In the case of VMware however, I seriously doubt that it is even remotely useable on a real i486 class machine, and it would arguably be not very useful on i586 class hardware either. It might make more sense to just require i686 class hardware and have it optimized directly for i686. Of course, there may be some other technical reason I'm overlooking... TTYL -- Mike A. Harris ftp://people.redhat.com/mharris OS Systems Engineer - XFree86 maintainer - Red Hat -- Psyche-list mailing list Psyche-list@redhat.com https://listman.redhat.com/mailman/listinfo/psyche-list
