Vintage vs. SIMH Dhrystones

On Jan 3, 2013, at 10:05 PM, Jonathan Katz <jon at jonworld.com> wrote: I found the following C version of Dhrystone 1.1 at the following URL: http://classes.soe.ucsc.edu/cmpe202/benchmarks/standard/dhrystone.c I have simh-780 running BSD 4.3 found via these instructions: http://gunkies.org/wiki/Installing_4.3_BSD_on_SIMH It's hosted on an OpenBSD VM on my MacBook Pro. It has one CPU assigned: cpu0: Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz, 2278.19 MHz I used the BSD 4.3 compiler (not gcc), HZ=60, TIMES, and I did NOT use the register setting. According to the source code similar software configurations on a true 11/780 did around 1400 Dhrystones (the version 1.0 of the code had an error, so I'm going off of the 1.1 results in the file.) kremvax# ./dhry Dhrystone(1.1) time for 500000 passes = 12 This machine benchmarks at 40376 dhrystones/second I upped it to LOOPS at 50000000 to get a number that may make more sense, since we're comparing GHZ to MHZ and I think CPUs these days have more on-die cache than some of these systems had in RAM. kremvax# ./dhry Dhrystone(1.1) time for 50000000 passes = 1266 This machine benchmarks at -17034 dhrystones/second OOPS! I think I need to go find a long I left behind somewhere. OK, one more time with LOOPS set to 5000000. ./dhry Dhrystone(1.1) time for 5000000 passes = 133 This machine benchmarks at 37425 dhrystones/second OK, interesting. Finally, I did one more with the REGS setting. kremvax# ./dhry Dhrystone(1.1) time for 5000000 passes = 119 This machine benchmarks at 41811 dhrystones/second So we're looking at 29 times faster. -Jon