Research Study - Effects of Compiler Optimization on pattern collection and Spatter benchmark runs

[jyoung3131]

Some researchers have noted that our GS Patterns tool may generate vastly different pattern outputs depending on the enabled optimization level. This study would look at the effects of optimization level on the amount of sparse accesses within a target application along with the performance differences in these patterns when run on a standard Spatter backend (CPU or GPU).

[jyoung3131]

We may want to check on the effects of inlining and loop unrolling (which may duplicate addresses) across compilers.

One way to do this would be to check the optimization report for the compiler used.