Does Loop Fission Work in Single Cores?

Question

When does it make sense to use Loop fission/distribution if I am compiling for a single core processor?

Answer 1

Performance wise? No, it will almost certainly introduce overhead.

Answer 2

See comp.compilers for the standard answers.

In addition to that, an odd-ball case that I could think up might be if there is a potential that one of the iterations might block. (e.g. does dynamic memory allocation)

Note that that case is more "abuse" of an easy language "paralel for" than a real numerical reason. (easier to use "for" than to manually do it using threads, even if the FOR waits for all threads to complete)

Answer 3

Got wonderful answers at comp.compiler