thanks,

i was guessing something like that, these are very short pieces of linear 
code.

so i did try; using returned values, repeating the test 10x in the loop, 
and adding stop/start.

the code's on github, (https://github.com/splace/signals) but this is a 
selective test, so its really buried, if i cant get it to work after 
reading up, i'll snip it out and post.



On Sunday, 19 June 2016 00:32:32 UTC+1, Dave Cheney wrote:
>
> Without seeing the code (hint, hint) I'm guessing that your benchmark got 
> optimised away.
>
> Here are some links with suggestions on how to make your benchmark 
> reliable.
>
>
> http://dave.cheney.net/2013/06/30/how-to-write-benchmarks-in-go#compiler-optimisation
>
> http://talks.godoc.org/github.com/davecheney/presentations/writing-high-performance-go.slide#19
>
>
>
> On Sunday, 19 June 2016 09:17:31 UTC+10, simon place wrote:
>>
>> any ideas? i've run out.
>>
>> this only happens with one program, and only in go1.7beta1(and 2).
>>
>> (i notice PASS is now at the end rather than the start. FWIW)
>>
>> see below 
>> /*  Hal3 Sat Jun 18 21:52:37 BST 2016  go version go1.7beta1 linux/amd64
>>
>> BenchmarkPCM8bitEncode-2        2000000000             0.39 ns/op
>> BenchmarkPCM8bitDecode-2        2000000000             0.39 ns/op
>> BenchmarkPCM16bitEncode-2       2000000000             0.39 ns/op
>> BenchmarkPCM16bitDecode-2       2000000000             0.39 ns/op
>> BenchmarkPCM24bitEncode-2       2000000000             0.39 ns/op
>> BenchmarkPCM24bitDecode-2       2000000000             0.39 ns/op
>> BenchmarkPCM32bitEncode-2       2000000000             0.39 ns/op
>> BenchmarkPCM32bitDecode-2       2000000000             0.39 ns/op
>> PASS
>> ok      _/home/simon/Dropbox/github/working/signals    6.570s
>> Sat Jun 18 21:52:49 BST 2016 */
>>
>> /*  Hal3 Sat Jun 18 22:07:30 BST 2016 go version go1.5.1 linux/amd64
>> PASS
>> BenchmarkPCM8bitEncode-2     300000000             4.25 ns/op
>> BenchmarkPCM8bitDecode-2     300000000             5.58 ns/op
>> BenchmarkPCM16bitEncode-2    300000000             5.57 ns/op
>> BenchmarkPCM16bitDecode-2    200000000             8.33 ns/op
>> BenchmarkPCM24bitEncode-2    200000000             8.10 ns/op
>> BenchmarkPCM24bitDecode-2    100000000            12.3 ns/op
>> BenchmarkPCM32bitEncode-2    100000000            10.7 ns/op
>> BenchmarkPCM32bitDecode-2    100000000            16.2 ns/op
>> ok      _/home/simon/Dropbox/github/working/signals    15.140s
>> Sat Jun 18 22:07:47 BST 2016 */
>> /*  Hal3 Sat Jun 18 23:35:10 BST 2016  go version go1.6.2 linux/amd64
>>
>> PASS
>> BenchmarkPCM8bitEncode-2     300000000             4.25 ns/op
>> BenchmarkPCM8bitDecode-2     300000000             5.59 ns/op
>> BenchmarkPCM16bitEncode-2    300000000             5.56 ns/op
>> BenchmarkPCM16bitDecode-2    200000000             8.37 ns/op
>> BenchmarkPCM24bitEncode-2    200000000             8.12 ns/op
>> BenchmarkPCM24bitDecode-2    100000000            12.4 ns/op
>> BenchmarkPCM32bitEncode-2    100000000            10.7 ns/op
>> BenchmarkPCM32bitDecode-2    100000000            16.2 ns/op
>> ok      _/home/simon/Dropbox/github/working/signals    15.145s
>> Sat Jun 18 23:35:27 BST 2016 */
>>
>>

-- 
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.

Reply via email to