🚀 go-pugleaf

jbuck@forney.berkeley.edu (Joe Buck) writes:

>In article <1993Apr23.200335.4875@beaver.cs.washington.edu> mef@cs.washington.edu (Marc Fiuczynski) writes:
>>In an article of PC MAGAZINE (April 27, 93) "Apps Ready for Pentium"
>>they mention that applications that have been recompiled for Pentium
>>chips will also run 10 - 15 percent faster on a 486, too. 
>>
>>1) Does this mean that the various 486 compilers for DOS machines don't
>>   support instruction scheduling?
>>2) Does GCC have the inst. scheduling stuff in there to make it work
>>   faster.

>GCC does instruction scheduling.  Also, it has a flag -m486 to tell it to
>generate code assuming a 486, which will affect some optimization

        I have compiled (on 48633dx running os2 2.0) a variety of things
        unger OS2 with the -m486 option and without.  In practice, the
        MAJORITY of apps (including povray) ran faster without it.
        
        From (foggy) memory only a recursive factorial program actually
        ran faster.
        BTW: don't take my work for it, try yourself. It'd be great if I
                was wrong.

                                                        Kingsley Turner
                                                        NSW Australia