I'm actually struggling to come up with a toy example that identically
reproduces the results. I wonder if there are some compiler optimizations
happening here.
The example isn't a lot of source, but has some IP and code that's not
exactly in a state I wish to post publicly (rough draft). Is there
somewhere I can post privately? If not, I can continue to try to reproduce
with a non-IPish example.
On Thursday, 10 November 2016 17:08:15 UTC-8, Dave Cheney wrote:
>
> Can you post a runable sample so that other can try to reproduce your
> issue?
>
> On Friday, 11 November 2016 12:06:21 UTC+11, Evan Digby wrote:
>>
>> Is it expected that if multiple sub-benchmarks are run in the same
>> benchmark, the cost of the setup will impact the results from the first
>> benchmark but not the second?
>>
>> func BenchmarkIntersection(b *testing.B) {
>> // Long setup -- ~5.5 seconds
>>
>> b.Run("basic 1", func(b *testing.B) {
>> for i := 0; i < b.N; i++ {
>> // Do benchmark stuff
>> }
>> })
>> b.Run("basic 2", func(b *testing.B) {
>> for i := 0; i < b.N; i++ {
>> // Do benchmark stuff (EXACTLY THE SAME AS TEST 1)
>> }
>> })
>> }
>>
>> Results in:
>>
>> BenchmarkIntersection/basic_1-4 1 5521938867 ns/op
>> BenchmarkIntersection/basic_2-4 200000000 8.89 ns/op
>>
>> Am I doing this wrong?
>>
>> I don't see an issue raised against this--should I be raising one?
>>
>>
>>
--
You received this message because you are subscribed to the Google Groups
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
