|
View:
New views
7 Messages
—
Rating Filter:
Alert me
|
|
|
64-bit CUDAHi
I have made an attempt at performing 64 to 32 bit conversion for the unsigned int arrays, but I have not been able to test it as I do not have a 64bit installation at my disposal. Is there a way to force the size of k3d::uint_t to 64bits without too much trouble for testing? Then on a side note - the failing test CUDASubdivideEdges.py is due to the difference in numeric precision between the CUDA and CPU versions - I have had a look at the MeshDiff plugin (a very quick one), but have not been able to adjust the threshold to allow the test to pass - will have another look at it as soon as possible. In terms of 64-bit support, I could install a 64bit version of ubuntu on my laptop, although then I will only be able to test in emulation mode - which may be sufficient. Evan ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAOn Monday 21 July 2008 16:25:48 Evan Lezar wrote:
> Hi > > I have made an attempt at performing 64 to 32 bit conversion for the > unsigned int arrays, but I have not been able to test it as I do not have a > 64bit installation at my disposal. After fixing a small compilation error (in SVN), I am still getting an incorrect mesh. There are probably some non-converted arrays left. I'll see if I can hun them down :) > Is there a way to force the size of k3d::uint_t to 64bits without too much > trouble for testing? I wouldn't bother with this, that's what the dashboard is there for. > Then on a side note - the failing test CUDASubdivideEdges.py is due to the > difference in numeric precision between the CUDA and CPU versions - I have > had a look at the MeshDiff plugin (a very quick one), but have not been > able to adjust the threshold to allow the test to pass - will have another > look at it as soon as possible. I think the proper way to solve that now, assuming there are no topological differences, is to add a new testmesh in the tests/meshes directory, named mesh.modifier.SubdivideEdges.reference.2.k3d This second mesh should be generated with the CUDA plugin. The testing system will then try both the .1 and .2 mesh, and if one of them matches, the test passes. Cheers, Bart ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAOn Monday 21 July 2008 16:25:48 Evan Lezar wrote:
> I have made an attempt at performing 64 to 32 bit conversion for the > unsigned int arrays, but I have not been able to test it as I do not have a > 64bit installation at my disposal. I fixed the remaining problems. It does indeed help if one can experiment a bit ;) While the idea of doing the conversion on the GPU is interesting, maybe we should also compare performance to conversion on the CPU, as I did before in cuda_mesh_subdivide_edges on line 379. Cheers, Bart ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAHi Bart
I have committed an updated test and reference mesh for CUDASubdivideEdges. I will have a look at setting up a comparative benchmark between the two conversion methods - but obviously it will have to be tested on your side :) Cheers Evan On Mon, Jul 21, 2008 at 11:42 PM, Bart Janssens <bart.janssens@...> wrote:
------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAOn 7/22/08, Evan Lezar <evanlezar@...> wrote:
> I have committed an updated test and reference mesh for CUDASubdivideEdges. > I will have a look at setting up a comparative benchmark between the two > conversion methods - but obviously it will have to be tested on your side :) OK! Seems I still need to do some more tests with SubdivideEdges on 64bit: the mesh conversion test passes, but SubdivideEdgesSimple has wrong edge_points and clockwise_edges arrays. Cheers, -- Bart ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAOn 7/22/08, Bart Janssens <bart.janssens@...> wrote:
> OK! Seems I still need to do some more tests with SubdivideEdges on > 64bit: the mesh conversion test passes, but SubdivideEdgesSimple has > wrong edge_points and clockwise_edges arrays. Good news! Looks like we had only overlooked the index_map array, which is fixed now. All CUDASubdivideEdges tests pass on my machine now. Cheers, -- Bart ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
|
|
Re: 64-bit CUDAThats good to hear!
Sorry that I've been so quite the last couple of days - my GF is leaving for Japan on Friday morning and I'm helping her pack etc. Will do some significant work this weekend! Evan On Wed, Jul 23, 2008 at 9:38 PM, Bart Janssens <bart.janssens@...> wrote:
------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ K3d-development mailing list K3d-development@... https://lists.sourceforge.net/lists/listinfo/k3d-development |
| Free Forum Powered by Nabble | Forum Help |