The Trunk: Kernel-nice.491.mcz

> Nicolas Cellier uploaded a new version of Kernel to project The Trunk:
> http://source.squeak.org/trunk/Kernel-nice.491.mcz
>
> ==================== Summary ====================
>
> Name: Kernel-nice.491
> Author: nice
> Time: 10 September 2010, 2:33:29.294 pm
> UUID: 00d7b4ff-5219-aa44-b82d-be4c7f18c761
> Ancestors: Kernel-ar.490
>
> Prevent bench to display 14 digits or so.
> 3 significative digits should be enough for a human brain (at least mine).
>
> =============== Diff against Kernel-ar.490 ===============
>
> Item was changed:
>  ----- Method: BlockClosure>>bench (in category 'evaluating') -----
>  bench
>        "See how many times I can value in 5 seconds.  I'll answer a meaningful description."
>
> +       | startTime endTime count roundTo3Digits |
> +       roundTo3Digits := [:num | num roundTo: (10 raisedTo: (num numberOfDigitsInBase: 10) - 3)].
> -       | startTime endTime count |
>        count := 0.
>        endTime := Time millisecondClockValue + 5000.
>        startTime := Time millisecondClockValue.
>        [ Time millisecondClockValue > endTime ] whileFalse: [ self value.  count := count + 1 ].
>        endTime := Time millisecondClockValue.
>        ^count = 1
> +               ifTrue: [ (roundTo3Digits value: (endTime - startTime) // 1000) printString, ' seconds.' ]
> -               ifTrue: [ ((endTime - startTime) // 1000) printString, ' seconds.' ]
>                ifFalse:
> +                       [ ((roundTo3Digits value: (count * 1000 * 100) // (endTime - startTime)) / 100.0) printString, ' per second.' ]!
> -                       [ ((count * 1000) / (endTime - startTime)) asFloat printString, ' per second.' ]!
>
>
>

> Hi, I've been meaning to update this for a while too.  A comment and a question.
>
> Re: displaying only 3 digits, isn't roundTo: 0.001 supposed to accomplish this?
>
> Also, I am thinking that #bench should display the total time it took
> to run the block if the block could only be run once in the
> five-second limit.
>
> IOW, how sensible is it for it to display:   '0.877 per second.'  ??
>
> Probably, instead, it should just say, "4.123 seconds." shouldn't it?
>
> (Again, this is only in the case where it could not run more than once
> in 5 seconds).
>
>  - Chris
>
>
> On Fri, Sep 10, 2010 at 7:34 AM,  <[hidden email]> wrote:
>> Nicolas Cellier uploaded a new version of Kernel to project The Trunk:
>> http://source.squeak.org/trunk/Kernel-nice.491.mcz
>>
>> ==================== Summary ====================
>>
>> Name: Kernel-nice.491
>> Author: nice
>> Time: 10 September 2010, 2:33:29.294 pm
>> UUID: 00d7b4ff-5219-aa44-b82d-be4c7f18c761
>> Ancestors: Kernel-ar.490
>>
>> Prevent bench to display 14 digits or so.
>> 3 significative digits should be enough for a human brain (at least mine).
>>
>> =============== Diff against Kernel-ar.490 ===============
>>
>> Item was changed:
>>  ----- Method: BlockClosure>>bench (in category 'evaluating') -----
>>  bench
>>        "See how many times I can value in 5 seconds.  I'll answer a meaningful description."
>>
>> +       | startTime endTime count roundTo3Digits |
>> +       roundTo3Digits := [:num | num roundTo: (10 raisedTo: (num numberOfDigitsInBase: 10) - 3)].
>> -       | startTime endTime count |
>>        count := 0.
>>        endTime := Time millisecondClockValue + 5000.
>>        startTime := Time millisecondClockValue.
>>        [ Time millisecondClockValue > endTime ] whileFalse: [ self value.  count := count + 1 ].
>>        endTime := Time millisecondClockValue.
>>        ^count = 1
>> +               ifTrue: [ (roundTo3Digits value: (endTime - startTime) // 1000) printString, ' seconds.' ]
>> -               ifTrue: [ ((endTime - startTime) // 1000) printString, ' seconds.' ]
>>                ifFalse:
>> +                       [ ((roundTo3Digits value: (count * 1000 * 100) // (endTime - startTime)) / 100.0) printString, ' per second.' ]!
>> -                       [ ((count * 1000) / (endTime - startTime)) asFloat printString, ' per second.' ]!
>>
>>
>>
>
>

> The problem is for micro benchmarks where the display is more like
>
> 1.40983424386e6 per second
>
> so roundTo: 0.001 does not solve anything (the last digits are not
> meaningfull at all).
> Using three digits, I get
>
> 1.41e6 per second
>
> which is already better (less pollution).
>
> For the other suggestions, I don't know, just did the simplest modification...

> On 9/10/2010 10:41 AM, Nicolas Cellier wrote:
>>
>> The problem is for micro benchmarks where the display is more like
>>
>> 1.40983424386e6 per second
>>
>> so roundTo: 0.001 does not solve anything (the last digits are not
>> meaningfull at all).
>> Using three digits, I get
>>
>> 1.41e6 per second
>>
>> which is already better (less pollution).
>>
>> For the other suggestions, I don't know, just did the simplest
>> modification...
>
> 1.40983424386e6 truncated asStringWithCommas
> => 1,409,834 per second
>
> Cheers,
>  - Andreas
>
>