PhasePlot give inconsistent results with different core count and fieldc
The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end. From G.S. #----------------------------------------------------------------------------------- from yt.mods import * i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data() fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones' pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file) #---------------------------------------------------------------------------------------
Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the
ProfileND objects not conducting their parallel reduction operation.
A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#----------------------------------------------------------------------------------- from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file) #---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi Geoffrey,
The operation to combine profile and phase plot data after running in
parallel was missing, so what you were seeing was the results from only one
of the processors. I have just filed a pull request to fix this. Now, I
get the same results independent of processor number. Please. have a look
at the PR herehttps://bitbucket.org/yt_analysis/yt/pull-request/782/adding-parallel-suppor...
and
test it out if you can.
Britton
On Mon, Mar 31, 2014 at 2:14 PM, Matthew Turk
Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the ProfileND objects not conducting their parallel reduction operation. A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
wrote: The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#-----------------------------------------------------------------------------------
from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file)
#---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi again,
The pull request has been accepted. You can do "yt update" and you should
be all set.
Britton
On Mon, Mar 31, 2014 at 4:03 PM, Britton Smith
Hi Geoffrey,
The operation to combine profile and phase plot data after running in parallel was missing, so what you were seeing was the results from only one of the processors. I have just filed a pull request to fix this. Now, I get the same results independent of processor number. Please. have a look at the PR herehttps://bitbucket.org/yt_analysis/yt/pull-request/782/adding-parallel-suppor... and test it out if you can.
Britton
On Mon, Mar 31, 2014 at 2:14 PM, Matthew Turk
wrote: Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the ProfileND objects not conducting their parallel reduction operation. A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
wrote: The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#-----------------------------------------------------------------------------------
from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file)
#---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi,
Thanks for the quick fix. My question is did the bug affect anything else
besides PhasePlot (ProjectionPlot, SlicePlot etc)? My old results using
PlotCollection and add_phase in parallel should be OK right?
From
G.S.
On Mon, Mar 31, 2014 at 10:13 AM, Britton Smith
Hi again,
The pull request has been accepted. You can do "yt update" and you should be all set.
Britton
On Mon, Mar 31, 2014 at 4:03 PM, Britton Smith
wrote: Hi Geoffrey,
The operation to combine profile and phase plot data after running in parallel was missing, so what you were seeing was the results from only one of the processors. I have just filed a pull request to fix this. Now, I get the same results independent of processor number. Please. have a look at the PR herehttps://bitbucket.org/yt_analysis/yt/pull-request/782/adding-parallel-suppor... and test it out if you can.
Britton
On Mon, Mar 31, 2014 at 2:14 PM, Matthew Turk
wrote: Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the ProfileND objects not conducting their parallel reduction operation. A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
wrote: The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#-----------------------------------------------------------------------------------
from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file)
#---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Hi Geoffrey,
Only PhasePlot and ProfilePlot have been affected.
Britton
On Mon, Mar 31, 2014 at 7:32 PM, Geoffrey So
Hi,
Thanks for the quick fix. My question is did the bug affect anything else besides PhasePlot (ProjectionPlot, SlicePlot etc)? My old results using PlotCollection and add_phase in parallel should be OK right?
From G.S.
On Mon, Mar 31, 2014 at 10:13 AM, Britton Smith
wrote: Hi again,
The pull request has been accepted. You can do "yt update" and you should be all set.
Britton
On Mon, Mar 31, 2014 at 4:03 PM, Britton Smith
wrote: Hi Geoffrey,
The operation to combine profile and phase plot data after running in parallel was missing, so what you were seeing was the results from only one of the processors. I have just filed a pull request to fix this. Now, I get the same results independent of processor number. Please. have a look at the PR herehttps://bitbucket.org/yt_analysis/yt/pull-request/782/adding-parallel-suppor... and test it out if you can.
Britton
On Mon, Mar 31, 2014 at 2:14 PM, Matthew Turk
wrote: Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the ProfileND objects not conducting their parallel reduction operation. A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
wrote: The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#-----------------------------------------------------------------------------------
from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file)
#---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
Thanks Britton, glad to know.
From
G.S.
On Mon, Mar 31, 2014 at 11:33 AM, Britton Smith
Hi Geoffrey,
Only PhasePlot and ProfilePlot have been affected.
Britton
On Mon, Mar 31, 2014 at 7:32 PM, Geoffrey So
wrote: Hi,
Thanks for the quick fix. My question is did the bug affect anything else besides PhasePlot (ProjectionPlot, SlicePlot etc)? My old results using PlotCollection and add_phase in parallel should be OK right?
From G.S.
On Mon, Mar 31, 2014 at 10:13 AM, Britton Smith
wrote: Hi again,
The pull request has been accepted. You can do "yt update" and you should be all set.
Britton
On Mon, Mar 31, 2014 at 4:03 PM, Britton Smith
wrote: Hi Geoffrey,
The operation to combine profile and phase plot data after running in parallel was missing, so what you were seeing was the results from only one of the processors. I have just filed a pull request to fix this. Now, I get the same results independent of processor number. Please. have a look at the PR herehttps://bitbucket.org/yt_analysis/yt/pull-request/782/adding-parallel-suppor... and test it out if you can.
Britton
On Mon, Mar 31, 2014 at 2:14 PM, Matthew Turk
wrote: Hi Geoffrey,
Britton and I verified you are correct, there is a problem with the ProfileND objects not conducting their parallel reduction operation. A fix is forthcoming.
-Matt
On Mon, Mar 31, 2014 at 5:31 AM, Geoffrey So
wrote: The following script gives me different results in the 2d phase plot when I use multiple cores and switch the fieldc parameter. The results are consistent if I only use a single core. Some of the plots produced seems like it is missing data points or under-sampled, so I'm guessing maybe not all the data from all the cores were collected at the end.
From G.S.
#-----------------------------------------------------------------------------------
from yt.mods import *
i=2 file = "RD%04i" % i fn = file+"/"+file pf=load(fn) dd=pf.h.all_data()
fieldx='Density' fieldy='Temperature' fieldc='CellVolumeCode' #fieldc='CellVolume' #fieldc='CellMassMsun' #fieldc='CellMass' #fieldc='Ones'
pplot = PhasePlot(dd, fieldx, fieldy, [fieldc],weight_field=None) pplot.save('example/'+file)
#---------------------------------------------------------------------------------------
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
_______________________________________________ yt-users mailing list yt-users@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
participants (3)
-
Britton Smith
-
Geoffrey So
-
Matthew Turk