I'm trying to use add_phase_object to get a 2D distribution and summing over the y axis to see the total quantity in each x bins. I am using the old plot collection to do the phase distribution because it is parallelized, the binned2D is serial I believe.
I'm checking the numbers by using TotalQuantity which sums all the cells in the simulation, and I should get the same number when I sum over the color axis of the phase plot (I'm assuming this is true, can someone also verify the logic?), but I am getting slightly different number. More curious, is when the y axis is the same as the color axis, I get different numbers as when the axis names are different. I first saw the discrepancy using other fields, but in this script I am demonstrating with fields that everyone should have access to in an Enzo simulation.
http://paste.yt-project.org/show/3222/
Can someone run the script very quick and verify that they get different numbers printed on screen?
Is there an easier way to do what I'm trying to do in parallel?
From G.S.
PS: This issue is in both dev-yt I have on my laptop and on Gordon
my laptop uses: changeset: 6566:6794b2798b5a branch: yt tag: tip user: Matthew Turk matthewturk@gmail.com date: Wed Oct 24 15:58:34 2012 -0400 summary: Swap out get_smallest_dx for 8192. Should fix the buildbot. Gordon I have: changeset: 7124:4151cc9ff9b2 branch: yt tag: tip user: Kacper Kowalik xarthisius.kk@gmail.com date: Sat Feb 02 11:03:07 2013 +0100 files: yt/utilities/logger.py description: Add `NullHandler` to `ytLogger` when stream logging is suppresed. Fixes #492
Hi Geoffrey
On Wed, Mar 6, 2013 at 4:38 PM, Geoffrey So gsiisg@gmail.com wrote:
I'm trying to use add_phase_object to get a 2D distribution and summing over the y axis to see the total quantity in each x bins. I am using the old plot collection to do the phase distribution because it is parallelized, the binned2D is serial I believe.
If you specify lazy_reader = True it will be run in parallel.
I'm checking the numbers by using TotalQuantity which sums all the cells in the simulation, and I should get the same number when I sum over the color axis of the phase plot (I'm assuming this is true, can someone also verify the logic?), but I am getting slightly different number. More curious, is when the y axis is the same as the color axis, I get different numbers as when the axis names are different. I first saw the discrepancy using other fields, but in this script I am demonstrating with fields that everyone should have access to in an Enzo simulation.
http://paste.yt-project.org/show/3222/
Can someone run the script very quick and verify that they get different numbers printed on screen?
I can't right now, but what if you specify a dtype in the call to sum() ? For instance:
.sum(dtype="float64")
-Matt
Is there an easier way to do what I'm trying to do in parallel?
From G.S.
PS: This issue is in both dev-yt I have on my laptop and on Gordon
my laptop uses: changeset: 6566:6794b2798b5a branch: yt tag: tip user: Matthew Turk matthewturk@gmail.com date: Wed Oct 24 15:58:34 2012 -0400 summary: Swap out get_smallest_dx for 8192. Should fix the buildbot. Gordon I have: changeset: 7124:4151cc9ff9b2 branch: yt tag: tip user: Kacper Kowalik xarthisius.kk@gmail.com date: Sat Feb 02 11:03:07 2013 +0100 files: yt/utilities/logger.py description: Add `NullHandler` to `ytLogger` when stream logging is suppresed. Fixes #492
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
The issue is resolved.
It turns out that PlotCollection uses the min/max values of the x,y axes as bin edges so the edge values may sometimes not get included correctly.
By using BinnedProfile2D and add_field with weight=None, I can get consistent answers as compared to using TotalQuantity by either:
1) expanding the min/max by a tiny fraction 2) switch end_collect=True (which places values outside the bins to the edges)
Thanks Matt, mystery solved.
From G.S.
On Wed, Mar 6, 2013 at 4:41 PM, Matthew Turk matthewturk@gmail.com wrote:
Hi Geoffrey
On Wed, Mar 6, 2013 at 4:38 PM, Geoffrey So gsiisg@gmail.com wrote:
I'm trying to use add_phase_object to get a 2D distribution and summing
over
the y axis to see the total quantity in each x bins. I am using the old plot collection to do the phase distribution because it is parallelized,
the
binned2D is serial I believe.
If you specify lazy_reader = True it will be run in parallel.
I'm checking the numbers by using TotalQuantity which sums all the cells
in
the simulation, and I should get the same number when I sum over the
color
axis of the phase plot (I'm assuming this is true, can someone also
verify
the logic?), but I am getting slightly different number. More curious,
is
when the y axis is the same as the color axis, I get different numbers as when the axis names are different. I first saw the discrepancy using
other
fields, but in this script I am demonstrating with fields that everyone should have access to in an Enzo simulation.
http://paste.yt-project.org/show/3222/
Can someone run the script very quick and verify that they get different numbers printed on screen?
I can't right now, but what if you specify a dtype in the call to sum() ? For instance:
.sum(dtype="float64")
-Matt
Is there an easier way to do what I'm trying to do in parallel?
From G.S.
PS: This issue is in both dev-yt I have on my laptop and on Gordon
my laptop uses: changeset: 6566:6794b2798b5a branch: yt tag: tip user: Matthew Turk matthewturk@gmail.com date: Wed Oct 24 15:58:34 2012 -0400 summary: Swap out get_smallest_dx for 8192. Should fix the buildbot. Gordon I have: changeset: 7124:4151cc9ff9b2 branch: yt tag: tip user: Kacper Kowalik xarthisius.kk@gmail.com date: Sat Feb 02 11:03:07 2013 +0100 files: yt/utilities/logger.py description: Add `NullHandler` to `ytLogger` when stream logging is suppresed. Fixes
#492
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
yt-dev mailing list yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
-
Geoffrey So -
Matthew Turk