26 Jan
2007
26 Jan
'07
4:42 p.m.
At Fri, 26 Jan 2007 07:23:26 -0800, David Roundy wrote:
I would think that what we'd want to benchmark would be clean, optimized actually-used code.
Maybe there should be two versions of each benchmark: 1) an clean, simple, idiomatic version, aka the code we would like to write if performance was not an issue. 2) a super-hand optimized version This would hopefully show, which benchmarks have the biggest room for improvement. And, it would provide information to the people working on the compilers as to what optimizations are actually needed for that test case. j.