Guys, Just wanted to wrap up loose ends on this.
It turned out the arctan2 call was not the problem – most of the time was spent averaging the nodal values onto the cells, but a separate thread has enabled me to call vtkPointDataToCellData for selected variables within my Programmable Filter. The filter is now ~ 15 times faster than before and my users have stopped grumbling (won’t last though!!) Thanks for all the support! Dennis From: Dennis Conklin Sent: Monday, April 30, 2018 2:08 PM To: 'Cory Quammen' <[email protected]>; kenichiro yoshimi <[email protected]> Cc: Paraview ([email protected]) <[email protected]> Subject: RE: [Paraview] [EXT] RE: [EXTERNAL] Calculating cylindrical coordinates Cory, No, I’m using numpy.arctan2 – sorry, always think of it as atan2. What I’m doing now is: # Undeformed Sector Angles (degrees about Y-axis, 0=FP) numPts=block.GetNumberOfPoints() node_angles=zeros( (numPts,1) ) x_undef,y_undef,z_undef=calc_node_undeformed_coords(block) node_angles=arctan2(x_undef,z_undef)*180.0/pi return node_angles So, I’m not looping, as you suggest. I am later doing a custom point-data-to-cell-data by looping over millions of cells and finding each cells nodes and averaging them, but I’ve failed to get vtk.vtkPointDataToCellData to work inside my Programmable filter, so I don’t have a good alternative. I keep meaning to take out the cell angles and see how much that speeds it up, but haven’t done that yet. Thanks for looking at this. Dennis From: Cory Quammen [mailto:[email protected]] Sent: Monday, April 30, 2018 1:44 PM To: kenichiro yoshimi <[email protected]<mailto:[email protected]>> Cc: Dennis Conklin <[email protected]<mailto:[email protected]>>; Paraview ([email protected]<mailto:[email protected]>) <[email protected]<mailto:[email protected]>> Subject: Re: [Paraview] [EXT] RE: [EXTERNAL] Calculating cylindrical coordinates Dennis, First, do you mean numpy.arctan2? I don't see that numpy.atan2 exists,but math.atan2 exists. Second, I suspect you are looping over millions of points in Python and are calling numpy.arctan2 on single X, Y values. Looping in Python like that will be very slow. Instead, you can get all your X values into one numpy array and do the same with your Y values, then make a single call to numpy.atan2. Here's an example: >>> from paraview.simple import * >>> import numpy >>> x = [1, 2, 3, 4] >>> y = [5, 6, 7, 8] >>> numpy.arctan2(y, x) array([ 1.37340077, 1.24904577, 1.16590454, 1.10714872]) To get the X and Y arrays, you can use the numpy adapter: >>> from vtk.numpy_interface import dataset_adapter as dsa >>> wrappedData = dsa.WrapDataObject(polydata) >>> X = wrappedData.Points[:,0] >>> Y = wrappedData.Points[:,1] >>> result = numpy.arctan2(Y, X) Hope that helps, Cory On Sat, Apr 28, 2018 at 1:25 AM, kenichiro yoshimi <[email protected]<mailto:[email protected]>> wrote: Hi Dennis, You can calculate cylindrical coordinates by utilizing vtkCylindricalTransform within the python programmable filter. It is something like below and faster than numpy. --- import vtk input = self.GetInput() output = self.GetOutput() transform = vtk.vtkCylindricalTransform() transformFilter = vtk.vtkTransformFilter() transformFilter.SetInputData(input) transformFilter.SetTransform(transform.GetInverse()) transformFilter.Update() output.ShallowCopy(transformFilter.GetOutput()) --- This converts (x,y,z) coordinates to (r,theta,z) coordinates by GetInverse method, where the angles are calculated by the following equation to change range to [0, 2*pi]: theta = pi + atan2(-y, -x). Regards 2018-04-28 5:25 GMT+09:00 Dennis Conklin <[email protected]<mailto:[email protected]>>: Alan, Yes, I am calling it many times at the same angular location, but it’s tough to know you’re at the same angular location without calculating the angle. Also, C++ is definitely orders of magnitude quicker, but once I have to compile routines into Paravew life gets very complicated. We barely survived compiling for cluster use and I haven’t written C++ for at least 20 years. So, if possible, I would like a clever external add-on that wouldn’t require us to re-compile PV everytime we update our version. I have visions of library errors in my nightmares! Dennis From: Scott, W Alan [mailto:[email protected]<mailto:[email protected]>] Sent: Friday, April 27, 2018 4:00 PM To: Dennis Conklin <[email protected]<mailto:[email protected]>>; Paraview ([email protected]<mailto:[email protected]>) <[email protected]<mailto:[email protected]>> Subject: [EXT] RE: [EXTERNAL] [Paraview] Calculating cylindrical coordinates CAUTION: EXTERNAL email. Please think before clicking on any links or attachments. Are there places on the cylinder you are calling atan2 with the same inputs, returning the same data, lots of times? Alternatively, could you calculate this in the simulation, and just add it to the simulation output? I am speculating that a C++ call to atan2 may be faster than numpy? Alan From: ParaView [mailto:[email protected]] On Behalf Of Dennis Conklin Sent: Friday, April 27, 2018 1:39 PM To: Paraview ([email protected]<mailto:[email protected]>) <[email protected]<mailto:[email protected]>> Subject: [EXTERNAL] [Paraview] Calculating cylindrical coordinates All, If I wanted to run atan2(X,Y) on millions and millions of nodes to calculate cylindrical coordinates, but found that numpy.atan2 in a Programmable Filter was taking 45 minutes to run, what should I do to make it much faster? Thanks for any hints Dennis _______________________________________________ Powered by www.kitware.com<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.kitware.com&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=FVhh30nhp0wVhcyd%2F4hFpRUuM%2BAOFMt7QoDOE54BMoQ%3D&reserved=0> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.kitware.com%2Fopensource%2Fopensource.html&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=3ldRUwa4mjFAej8qHdU%2Bm3hPNt1qi%2Fi%2F2UB8Edjp0uA%3D&reserved=0> Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fparaview.org%2FWiki%2FParaView&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=X2tWI8BBiUD2n9JXLfMqrB8ZVU2zDCE44%2F86YpebIrc%3D&reserved=0> Search the list archives at: http://markmail.org/search/?q=ParaView<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmarkmail.org%2Fsearch%2F%3Fq%3DParaView&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=LrjHRj3n56BoYZVKKbZ4BQIasalVoDbwqK4Esrl7NIU%3D&reserved=0> Follow this link to subscribe/unsubscribe: https://public.kitware.com/mailman/listinfo/paraview<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpublic.kitware.com%2Fmailman%2Flistinfo%2Fparaview&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=d27eFwhKD9HViwciR9Xa42vLICS34tLZLmL81ZPL%2Bf8%3D&reserved=0> _______________________________________________ Powered by www.kitware.com<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.kitware.com&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=FVhh30nhp0wVhcyd%2F4hFpRUuM%2BAOFMt7QoDOE54BMoQ%3D&reserved=0> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.kitware.com%2Fopensource%2Fopensource.html&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=3ldRUwa4mjFAej8qHdU%2Bm3hPNt1qi%2Fi%2F2UB8Edjp0uA%3D&reserved=0> Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fparaview.org%2FWiki%2FParaView&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=X2tWI8BBiUD2n9JXLfMqrB8ZVU2zDCE44%2F86YpebIrc%3D&reserved=0> Search the list archives at: http://markmail.org/search/?q=ParaView<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmarkmail.org%2Fsearch%2F%3Fq%3DParaView&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=LrjHRj3n56BoYZVKKbZ4BQIasalVoDbwqK4Esrl7NIU%3D&reserved=0> Follow this link to subscribe/unsubscribe: https://public.kitware.com/mailman/listinfo/paraview<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpublic.kitware.com%2Fmailman%2Flistinfo%2Fparaview&data=01%7C01%7Cdennis_conklin%40goodyear.com%7C4547c889b286410afb1c08d5aec1fb03%7C939e896692854a9a9f040887efe8aae0%7C0&sdata=d27eFwhKD9HViwciR9Xa42vLICS34tLZLmL81ZPL%2Bf8%3D&reserved=0> -- Cory Quammen Staff R&D Engineer Kitware, Inc.
_______________________________________________ Powered by www.kitware.com Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html Please keep messages on-topic and check the ParaView Wiki at: http://paraview.org/Wiki/ParaView Search the list archives at: http://markmail.org/search/?q=ParaView Follow this link to subscribe/unsubscribe: https://public.kitware.com/mailman/listinfo/paraview
