Closed MagaTailor closed 8 years ago
Compiled with default codegen (pentium4) settings:
test int::test::bench_div_1000_1000 ... bench: 7,178 ns/iter (+/- 9,400) test int::test::bench_pow_10_10 ... bench: 2,717 ns/iter (+/- 9,571)
test int::test::bench_div_1000_1000 ... bench: 7,178 ns/iter (+/- 9,400)
test int::test::bench_pow_10_10 ... bench: 2,717 ns/iter (+/- 9,571)
whereas with pentium2:
test int::test::bench_div_1000_1000 ... bench: 1,431 ns/iter (+/- 9,266) test int::test::bench_pow_10_10 ... bench: 1,576 ns/iter (+/- 9,464)
test int::test::bench_div_1000_1000 ... bench: 1,431 ns/iter (+/- 9,266)
test int::test::bench_pow_10_10 ... bench: 1,576 ns/iter (+/- 9,464)
Used rustc 1.8 (0d1cd9b) after llvm rebase. The rest of the benchmarks didn't exhibit any anomalous results.
1.8
I didn't repeat the benchmarks enough times to notice those high error bars were really causing lots of variability. My bad!
Compiled with default codegen (pentium4) settings:
test int::test::bench_div_1000_1000 ... bench: 7,178 ns/iter (+/- 9,400)
test int::test::bench_pow_10_10 ... bench: 2,717 ns/iter (+/- 9,571)
whereas with pentium2:
test int::test::bench_div_1000_1000 ... bench: 1,431 ns/iter (+/- 9,266)
test int::test::bench_pow_10_10 ... bench: 1,576 ns/iter (+/- 9,464)
Used rustc
1.8
(0d1cd9b) after llvm rebase. The rest of the benchmarks didn't exhibit any anomalous results.