View Single Post
Old 2019-01-26, 20:46   #13
danaj
 
"Dana Jacobsen"
Feb 2011
Bangkok, TH

38D16 Posts
Default

Paul,

Could you send or point me to your test programs? I'd like to run on my machine to compare identical programs.

I was debating the merits of doing a comparison graph at different sizes, e.g. (1000+250n for n 0 to 12). A simple way would be creating a file with 100 or 1000 random primes of the given size, then time running a program that processes them all. E.g. PFGW with the file. That seems like a fairly realistic test of performance that removes most of the startup overhead.

If you made a program that did some various tests (e.g. PRP, SPRP, BPSW and/or your tests) given a file with a number per line, would that seem like a good test? For very small inputs it would be worth pulling out even that overhead (e.g. read them all into an array of inputs, then start a timer around the testing, so we don't count overhead of I/O and string->bignum conversion).
danaj is offline   Reply With Quote