Open 4e6 opened 6 years ago
Benchmark in the collinearity exercise should include some kind of warmup to show the correct execution time.
Here I time two identical collinearity_float calls and the first one is always about 5 times slower.
collinearity_float
It is not the case with the warmup phase:
Benchmark with warmup phase shows correct ratio for collinearity_float vs collinearity_int -- float arithmetic is only 2 times slower instead of 8
collinearity_int
Benchmark in the collinearity exercise should include some kind of warmup to show the correct execution time.
Issue
Here I time two identical
collinearity_float
calls and the first one is always about 5 times slower.It is not the case with the warmup phase:
Correct benchmark
Benchmark with warmup phase shows correct ratio for
collinearity_float
vscollinearity_int
-- float arithmetic is only 2 times slower instead of 8