CompuServe Thread

I just uploaded Dhrystone program for the Benchmark M2 compiler to DL12. Here are the stats for both TDI and Benchmark (running on an A1000 with 68000 and 2.5 Meg). Please note that the idea was to introduce the least number of changes into the TDI version in order to get it working with Benchmark. ———————————————————————— System CPU / FPP Clock Cache Memory Time Dhrystones/sec (MHz) (bits) (sec ) ———————————————————————— TDI Modula-2 rel. 3.00a Amiga (68000/—–) 7.16 —– 16 —– 378 1) Amiga (68000/—–) 7.16 —– 16 87.4 571 2) Amiga (68000/—–) 7.16 —– 16 —– 613 3) Amiga (68000/—–) 7.16 —– 16 —– 1064 4) ———————————————————————— 1) Using the released Strings.Compare procedure 2) Using a recoded Strings.Compare, still full general procedure. This is probably the figure, that should be compared to the results above. 3) Using the Compare30 procedure below, similar to compiler generated inline code from many compilers 4) Using no string comparison at all ———————————————————————— Benchmark Modula-2 rel. 1.02 Amiga (68000/—–) 7.16 —– 16 79.3 630 1) Amiga (68000/—–) 7.16 —– 16 83.5 598 2) Amiga (68000/—–) 7.16 —– 16 86.8 576 3) Amiga (68000/—–) 7.16 —– 16 49.3 1012 4) ———————————————————————— 1) same as note 1 for TDI 2) same as note 1, except using base relative addressing (the default), as opposed to absolute addressing which was used for entries 1,3,4 3) same as note 3 for TDI 4) hopefully same as note 4 for TDI, but I'm not sure, as I don't quite know what was actually done in the TDI test. All I did was comment out the call to StringCompare and substituted it with a FALSE. Please Note: The TDI stats and notes are not mine (they're probably Joergen's)