but this was not what we want :-)
x86_64-ucb-akaros-gcc -Ofast -mfxsr -mxsave -mxsaveopt -o fputest
fputest.c
On Tue, Mar 29, 2016 at 2:42 PM Dan Cross <[email protected]> wrote:
> FYI, please see attached patch.
>
> On Mon, Mar 28, 2016 at 6:32 PM, ron minnich <[email protected]> wrote:
>
>> The timing is per-operation, not per total run. Even if we get one timer
>> interrupt in there, it's going to be one bad data point. I'm fine with
>> that.
>>
>> But, me, i'm only taking 128 samples. I think that's more the enough and
>> it runs in maybe several dozen microseconds.
>>
>> ron
>>
>> On Mon, Mar 28, 2016 at 3:29 PM ron minnich <[email protected]> wrote:
>>
>>> I'm still testing on linux. Akaros comes later :-)
>>>
>>> On Mon, Mar 28, 2016 at 2:59 PM Barret Rhoden <[email protected]>
>>> wrote:
>>>
>>>> On 2016-03-28 at 21:46 ron minnich <[email protected]> wrote:
>>>> > https://github.com/akaros/fputest
>>>> >
>>>> >
>>>> > We're getting there. It's now using gcc intrinsics, which hopefully
>>>> > are more correct. Still trying to get sane timing for the "no op"
>>>> > case.
>>>>
>>>> i'd be concerned if the intrinsics are right and our own versions are
>>>> wrong. hopefully that won't be the case after this gets sorted out.
>>>>
>>>>
>>>> given that n = 1000000 and rdtsc alone takes O(30) ticks, you're
>>>> probably getting interference from the timer IRQ, even on Akaros Core 0.
>>>> here's two quick things to try:
>>>>
>>>> call uthread_mcp_init() somewhere in main. that'll make you an MCP.
>>>> it's a bit hacky. verify that it's working with a ps or pip.
>>>>
>>>> check your tsc overhead against __procinfo.timing_overhead or against
>>>> the console output during boot, e.g.
>>>>
>>>> TSC overhead (Min: 32, Max: 36)
>>>>
>>>> i didn't look at any of the actual FPU stuff for issues yet.
>>>>
>>>> barret
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "Akaros" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to [email protected].
>>>> To post to this group, send email to [email protected].
>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>> --
>> You received this message because you are subscribed to the Google Groups
>> "Akaros" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To post to this group, send email to [email protected].
>> For more options, visit https://groups.google.com/d/optout.
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "Akaros" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To post to this group, send email to [email protected].
> For more options, visit https://groups.google.com/d/optout.
>
--
You received this message because you are subscribed to the Google Groups
"Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.
