Path: utzoo!attcan!uunet!tank!uxc.cso.uiuc.edu!csd4.milw.wisc.edu!bionet!agate!shelby!unix!quintus!arisia!mdixon From: mdixon@thelonius.PARC.xerox.com (Mike Dixon) Newsgroups: comp.sys.next Subject: poor Mathematica performance Message-ID: Date: 5 Jul 89 18:09:15 GMT Sender: news@arisia.Xerox.COM Distribution: comp Organization: Xerox PARC, Palo Alto, CA Lines: 10 i tried a simple root-finding benchmark and got 40s for my NeXT vs. 30s on a mac IIx. given that the mac is running the same hardware slower, this is presumably a combination of (a) higher OS overhead, and (b) better performance-tuning in the mac version. on the other hand, the OS overhead shouldn't be *that* high (none of the other processes were compute bound, and there's 16MB of memory; i ran the test several times and got quite consistent timings). and i don't understand why the code in question (kernel) isn't identical between the two machines (what's the point of separating kernel & front end if you don't share the kernel code??). so what's happening?