Tracing memory accesses to see which loop has better temporal locality... #hpc
@niconiconi What do you mean by kernel, kernel function on heterogeneous accelerators or operating system kernel?
@fulkrum@qoto.org kernel function, as in "the inner loop of a simulation that most processor cycles are spent on running, possibly runs on an accelerator". I simply deleted all arithmetic statements and replaced all array indexing with printf() to test different N-D array layouts.
@niconiconi Thanks for the description. I hope my confusion is reasonable（
@niconiconi @fulkrum
Never underestimate the power of sprinkling printf() through your code.
@fulkrum@qoto.org It's just a mock-up kernel with a dummy inner loop and printf()...