Date: Sun, 5 Dec 2010 22:24:12 -0800 From: Garrett Cooper <gcooper@FreeBSD.org> To: Julian Elischer <julian@freebsd.org> Cc: freebsd-current@freebsd.org, Steve Kargl <sgk@troutmask.apl.washington.edu> Subject: Re: Process accounting/timing has broken recently Message-ID: <AANLkTimk11=75bBjGzgYwH6niifBjCf%2B%2BXCNN8c9GCJt@mail.gmail.com> In-Reply-To: <4CFC812B.9060505@freebsd.org> References: <20101205231829.GA68156@troutmask.apl.washington.edu> <4CFC27A0.8000406@freebsd.org> <20101206061230.GA69477@troutmask.apl.washington.edu> <AANLkTikVDUPFM6Twdz%2B367APxLPS4fswfbqv0m_KLMan@mail.gmail.com> <4CFC812B.9060505@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Dec 5, 2010 at 10:22 PM, Julian Elischer <julian@freebsd.org> wrote=
:
> On 12/5/10 10:19 PM, Garrett Cooper wrote:
>>
>> On Sun, Dec 5, 2010 at 10:12 PM, Steve Kargl
>> <sgk@troutmask.apl.washington.edu> =A0wrote:
>>>
>>> On Sun, Dec 05, 2010 at 04:00:32PM -0800, Julian Elischer wrote:
>>>>
>>>> On 12/5/10 3:18 PM, Steve Kargl wrote:
>>>>>
>>>>> Sometime in the last 7-10 days, some one made a
>>>>> change that has broken process accounting/timing.
>>>>>
>>>>> laptop:kargl[42] foreach i ( 0 1 2 3 4 5 6 7 8 9 )
>>>>> foreach? time ./testf
>>>>> foreach? end
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A069.55 real =A0 =A0 =A0 =A038.39 user =A0 =A0 =A0 =A030=
.94 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A068.82 real =A0 =A0 =A0 =A040.95 user =A0 =A0 =A0 =A027=
.60 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A069.14 real =A0 =A0 =A0 =A038.90 user =A0 =A0 =A0 =A030=
.02 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A068.79 real =A0 =A0 =A0 =A040.59 user =A0 =A0 =A0 =A027=
.99 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A068.93 real =A0 =A0 =A0 =A039.76 user =A0 =A0 =A0 =A028=
.96 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A068.71 real =A0 =A0 =A0 =A041.21 user =A0 =A0 =A0 =A027=
.29 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A069.05 real =A0 =A0 =A0 =A039.68 user =A0 =A0 =A0 =A029=
.15 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A068.99 real =A0 =A0 =A0 =A039.98 user =A0 =A0 =A0 =A028=
.80 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A069.02 real =A0 =A0 =A0 =A039.64 user =A0 =A0 =A0 =A029=
.16 sys
>>>>> Max ULP: 0.501607 for x in [-18.000000:88.709999] with dx =3D
>>>>> 1.067100e-04
>>>>> =A0 =A0 =A0 =A069.38 real =A0 =A0 =A0 =A037.49 user =A0 =A0 =A0 =A031=
.67 sys
>>>>>
>>>>> testf is a numerically intensive program that tests the
>>>>> accuracy of expf() in a tight loop. =A0User time varies
>>>>> by ~3 seconds on my lightly loaded 2 GHz core2 duo processor.
>>>>> I'm fairly certain that the code does not suddenly grow/loose
>>>>> 6 GFLOP of operations.
>>>>>
>>>> I know it's a lot to ask but it may be something that you can help
>>>> with if you
>>>> had the time to triangulate in on the change that did it..
>>>> I presume that since you are an "old hand" you can check out sources
>>>> at different revisions..
>>>
>>> I was hoping that someone (possibly the person responsible) would
>>> recognize the symptoms and recommend a revision or two to revert.
>>> Otherwise, doing a binary search will take some time in that it
>>> takes 4+ hours for a buildworld/kernel cycle on my laptop.
>>
>> =A0 =A0 If you can provide the source for the application you're running
>> above and instructions on how to compile it, I can at least give you a
>> bit of a head start :).
>> Thanks,
>> -Garrett
>>
> plus which probably just
> `cd /sys/amd64/conf config GENERIC;cd ../compile/GENERIC; make kernel`
> =A0would be enough...
But couldn't it be libthr changes? There have been a handful of
those that have been committed recently by davidxu.
HTH,
-Garrett
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTimk11=75bBjGzgYwH6niifBjCf%2B%2BXCNN8c9GCJt>