dan@ece.arizona.edu (Dan Filiberti) (04/01/91)
I just glanced through the new edition of NextWorld mag... It has benchmark tests for Next, Sparcstation 2, and Mac IIfx in a comparison article...take a look :) Daniel Filiberti dan@ece.arizona.edu [:)}
cmaz514@ut-emx.uucp (Raghu Ramachandran) (04/02/91)
In article <1991Apr1.001408@ece.arizona.edu> dan@ece.arizona.edu (Dan Filiberti) writes: > >I just glanced through the new edition of NextWorld mag... > > >It has benchmark tests for Next, Sparcstation 2, and Mac IIfx in a comparison >article...take a look :) > > > Daniel Filiberti > dan@ece.arizona.edu [:)} Did anyone else notice that in Table 1 the IIfx has a '030 processor, but in Table 2 the benchmarks were compiled using the '020 option! The co-processor was also steped down. I wouldn't trade my NeXT for a Mac, but this kind of "reporting" doesn't help Nextworld nor NeXT. Just my $0.02. Raghu
edwardj@microsoft.UUCP (Edward JUNG) (04/10/91)
In article <46541@ut-emx.uucp> cmaz514@ut-emx.uucp (Raghu Ramachandran) writes: >Did anyone else notice that in Table 1 the IIfx has a '030 processor, but in >Table 2 the benchmarks were compiled using the '020 option! The co-processor >was also steped down. > >I wouldn't trade my NeXT for a Mac, but this kind of "reporting" doesn't help >Nextworld nor NeXT. The compilers used in the comparison for the Mac IIfx were the MPW 3.1 C compiler and the MPW 3.2 beta C compiler. Both MPW C compilers have an option called "68020" and "68881", which utilize 68020-specific instructions and full 32-bit address arithmetic, and in-line floating point math respectively. These flags enable 68030-optimized code generation in recent MPW compilers (there are very few 68020 machines, after all), but to my knowledge there still is no Apple-supplied compiler that optimizes for the 68882 (different pipelining from the 68881, therefore slightly different instruction sequences are best). Perhaps this should be explained in an addendum? I could insert various caveats about benchmarking and using the compiler you are given, etc., but you all know that! The point is that I did not purposefully hamstring the MPW compiler. If you have further comments on the article, I welcome them via email. Thanks! -- Edward Jung Microsoft Corp. My opinions do not reflect any policy of my employer.