digitalmars.D - Dhrystone Benchmark
Tower Ty <towerty msn.com.au> writes:
I ran the above programs in D on my machine and got wildly varying results as soDhrystone Benchmark, Version 2.1 (Language: D) Register option not selected. Microseconds for one loop: 1.8 Dhrystones per second: 552669.4 VAX MIPS rating: 314.553 Dhrystone Benchmark, Version 2.1 (Language: D) Register option not selected. Microseconds for one loop: 0.3 Dhrystones per second: 3333333.3 VAX MIPS rating: 1897.173 It settled on Register option selected? NO Microseconds for one run through Dhrystone: 0.3 Dhrystones per Second: 3333333.3 VAX MIPS rating = 1897.173 My question is why the massive variation and why does the benchmark take the highest ?
Mar 19 2008
renoX <renosky free.fr> writes:
Tower Ty a écrit : [cut]My question is why the massive variation and why does the benchmark take the highest ?
Well I don't know this particular benchmark so that's just a guess but maybe the variation is caused by the scheduler or by the GC. Maybe it's possible to disable the GC to see if the variation is reduced? As for taking the highest figure: if the variation is caused by process scheduling then it make sense to show the best performance as the result of the benchmark if you're trying to measure the runtime performance of your binary not the scheduler's performance.. Regards, renoX
Mar 19 2008