Map a set of HIP built-ins referenced by HeCBench benchmarks to compiler built-ins. The myocyte-hip benchmark seems to benefit from this the most - most likely due to:
HIP built-in calls with compile time values which are constant-folded.
HIP built-in calls which are simplified such as pow(x, 2) --> x*x.
Map a set of HIP built-ins referenced by HeCBench benchmarks to compiler built-ins. The myocyte-hip benchmark seems to benefit from this the most - most likely due to:
HIP built-in calls with compile time values which are constant-folded.
HIP built-in calls which are simplified such as
pow(x, 2) --> x*x
.