Sign in to follow this  

Benchmarks

This topic is 3293 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hello there. I hope this isn't repost, but I made a search and nothing satisfied me. So, want to test my c++ program(interactive RayTracer) and I'm not sure what to test and why. For now I want to test Time of execution, pixels/second, rays/second, instructions/sec, Cache miss rate, the amount of data going through memories, GFLOPS. I' not asking for software to test this. What I'm asking is if there should be more things that I should test, and the motives why I have to test so many things. I've decided to use PAPI (http://icl.cs.utk.edu/papi/) and probably valgrind (http://valgrind.org/). Actually if you know of better software feel free to enlighten me :) Thanks in advance, Threethazz

Share this post


Link to post
Share on other sites
Well, what are your requirements?

Benchmarks aren't really tests unless they're judging the suitability of some speed/throughput/compatibility requirement. Otherwise they're experiments or marketing blurbs. If you're experimenting, then devise what you want to know. Do you suspect the raytracer has problems in complex environments? With reflection/refaction? With many entities, with big entities? When memory is limited?

Devise a baseline (usually an empty room or something similar), then devise deviations from the baseline and measure the difference.

For marketing blurbs, it's useful to use known maps/setups. What other trials do benchmarks for raytracers use? Do those so you (and customers) can do competitive analysis.

Share this post


Link to post
Share on other sites
Thank you for your reply, Telastyn. But some stuff you wrote are the same I've been reading in dozens of webpages. It's all theory. My question was very practical.

I want to identify possible bottlenecks, cache misses and that sort of stuff. I was hopping someone here had experience on this so I don't start testing random stuff and get lame results out of it.

Threethazz

Share this post


Link to post
Share on other sites
Then you might want to search for profiling, not benchmarking. And really, without knowing the structure of your program, having some concept of where the major parts are; where the areas of likely problems lie, I doubt you'll get very practical answers.

Share this post


Link to post
Share on other sites

This topic is 3293 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this