Re: [Starpu-devel] Problem with performance model calibration

[Mirko Myllykoski <mirkom@cs.umu.se>]

Hi,

On 2020-06-03 20:41, Samuel Thibault wrote:

> Ok. When that happens, what are the values of the other fields?  We have
> some code that tries to get rid of outliers, which just flushes the
> content of the entry, in which case the mean will indeed be reset to
> 0.0, but also all other values (sum, sum2, nsample).

It appears that most of other field are zero:

$1 = {mean = 0, deviation = 0, sum = 0, sum2 = 0, nsample = 0, nerror = 
0, footprint = 3995150445, size = 31433, flops = 62866,
  duration = 0, tag = 0, parameters = 0x0}
$2 = {mean = 0, deviation = 0, sum = 0, sum2 = 0, nsample = 0, nerror = 
0, footprint = 3929775567, size = 11492, flops = 22984,
  duration = 0, tag = 0, parameters = 0x0}
$3 = {mean = 0, deviation = 0, sum = 0, sum2 = 0, nsample = 0, nerror = 
0, footprint = 1444706215, size = 14400, flops = 28800,
  duration = 0, tag = 0, parameters = 0x0}
$4 = {mean = 0, deviation = 0, sum = 0, sum2 = 0, nsample = 0, nerror = 
0, footprint = 1856779794, size = 60000, flops = 120000,
  duration = 0, tag = 0, parameters = 0x0}

I was simply looking at the size and mean fields...

I checked and I could not detect any entries with nsample != 0 and mean 
== 0.0.

Am I understanding correctly that these mean == 0.0 entries are the 
result of a cleanup performed before the models gets calibrate?

 - Mirko

--
Mirko Myllykoski, Ph.D.
Senior research engineer, Department of Computing Science
Application expert, HPC2N
Umeå University

By sending an email to Umeå University, the University processes your 
personal data. For more information, please read: https://www.umu.se/en/gdpr