Surprising benchmark result

[tapi...@gmail.com]

The code: https://play.golang.org/p/DxUj6kBqf8k

The Filter4 function has one more assignment statement than Filter3.

But the benchmark result shows Filter4 is faster than Filter3.

The result:

cpu: Intel(R) Core(TM) i5-4210U CPU @ 1.70GHz
BenchmarkFilter3-4        3575656           980.2 ns/op           0 B/op           0 allocs/op
BenchmarkFilter4-4        3956533           916.8 ns/op           0 B/op           0 allocs/op

[jake...@gmail.com]

FWIW I do not get the same result:

goos: windows
goarch: amd64
cpu: AMD Phenom(tm) II X4 830 Processor
BenchmarkFilter3-4        599912              1902 ns/op               0 B/op          0 allocs/op
BenchmarkFilter4-4        569143              2118 ns/op               0 B/op          0 allocs/op

PASS

[peterGo]

name       time/op
Filter3-4  1.38µs ± 0%
Filter4-4  1.39µs ± 0%

On Monday, June 7, 2021 at 10:57:19 AM UTC-4 tapi...@gmail.com wrote:

[tapi...@gmail.com]

On Monday, June 7, 2021 at 1:01:49 PM UTC-4 peterGo wrote:

name       time/op
Filter3-4  1.38µs ± 0%
Filter4-4  1.39µs ± 0%

What is your CPU model?

[peterGo]

i7-7500U

name       time/op
Filter3-4  1.38µs ± 0%

Filter4-4  1.39µs ± 0%

i5-8250U
name       time/op
Filter3-8  1.37µs ± 0%
Filter4-8  1.39µs ± 0%

[tapi...@gmail.com]

It is found that the test file is not written correctly.

The filtered data should be reset in each loop.

The corrected one: https://play.golang.org/p/yJMweZLIXAz

;D

On Monday, June 7, 2021 at 10:57:19 AM UTC-4 tapi...@gmail.com wrote:

[peterGo]

Revised results using the same input data for each iteration without a copy just a Filter function.

i7-7500U
name       time/op
Filter3-4  2.29µs ± 0%
Filter4-4  1.51µs ± 0%

i5-8250U
name       time/op
Filter3-8  2.30µs ± 0%
Filter4-8  1.53µs ± 0%

Peter