time SIMT?

[daydreamer]

Hi
of course you could use the usual clock cycle timings on each proc first
but split execution in separate threads on modern cpus I hope to utilize that I can use several caches that is per core = 2 or more innerloops aligned into 64byte caches
simt also represent splitting between threads ,for example 2 threads one is overutilized while the second one underutilized
ofcourse clock cycle timings on all thread/mutex/semaphore related apis are helpful
a long running test timed in milliseconds I think