Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CudaUVMSpace should track # allocations, due to CUDA limit on # UVM allocations #300

Closed
hcedwar opened this issue May 25, 2016 · 2 comments
Assignees
Labels
Feature Request Create new capability; will potentially require voting
Milestone

Comments

@hcedwar
Copy link
Contributor

hcedwar commented May 25, 2016

There is a limit of 64k Cuda UVM allocations after which an "out of memory" error is generated from the cuda malloc function. The Kokkos::CudaUVMSpace should track the number of allocations and report when this threshold is exceeded instead of simply "out of memory."

@hcedwar hcedwar added the Feature Request Create new capability; will potentially require voting label May 25, 2016
@hcedwar hcedwar added this to the Summer 2016 milestone May 25, 2016
@etphipp
Copy link
Contributor

etphipp commented May 25, 2016

Actually it is probably 2^16 = 65536

@etphipp
Copy link
Contributor

etphipp commented May 25, 2016

Oh, I guess that is 64k when k=1024. Sorry.

@gmackey gmackey self-assigned this Jun 15, 2016
@mhoemmen mhoemmen changed the title Cuda UVM allocation limit CudaUVMSpace should track # allocations, due to CUDA limit on # UVM allocations Aug 17, 2016
@hcedwar hcedwar modified the milestones: Fall 2016, Summer 2016 Aug 31, 2016
ndellingwood added a commit to ndellingwood/kokkos that referenced this issue Oct 26, 2016
This refers to Kokkos Github issue kokkos#300.
There is a limit of 64k i.e. 65536 Cuda UVM allocations after which an
"out of memory" error is issue.
This commit adds tracking of the number of allocations to CudaUVM space
and throws a runtime error when the max number of UVM allocations is
reached, along with a message for users.
	modified:   Cuda/Kokkos_CudaSpace.cpp
	modified:   Kokkos_CudaSpace.hpp
ndellingwood added a commit to ndellingwood/kokkos that referenced this issue Oct 26, 2016
This refers to Kokkos Github issue kokkos#300.
There is a limit of 64k i.e. 65536 Cuda UVM allocations after which an
"out of memory" error is issue.
This commit adds tracking of the number of allocations to CudaUVM space
and throws a runtime error when the max number of UVM allocations is
reached, along with a message for users.
	modified:   Cuda/Kokkos_CudaSpace.cpp
	modified:   Kokkos_CudaSpace.hpp
ndellingwood added a commit to ndellingwood/kokkos that referenced this issue Oct 26, 2016
This refers to Kokkos Github issue kokkos#300.
There is a limit of 64k i.e. 65536 Cuda UVM allocations after which an
"out of memory" error is issue.
This commit adds tracking of the number of allocations to CudaUVM space
and throws a runtime error when the max number of UVM allocations is
reached, along with a message for users.
	modified:   Cuda/Kokkos_CudaSpace.cpp
	modified:   Kokkos_CudaSpace.hpp
ndellingwood added a commit to ndellingwood/kokkos that referenced this issue Oct 26, 2016
This refers to Kokkos Github issue kokkos#300.
There is a limit of 64k i.e. 65536 Cuda UVM allocations after which an
"out of memory" error is issue.
This commit adds tracking of the number of allocations to CudaUVM space
and throws a runtime error when the max number of UVM allocations is
reached, along with a message for users.

	modified:   core/src/Cuda/Kokkos_CudaSpace.cpp
	modified:   core/src/Kokkos_CudaSpace.hpp
	modified:   core/unit_test/cuda/TestCuda_Spaces.cpp
@crtrott crtrott closed this as completed Oct 30, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Feature Request Create new capability; will potentially require voting
Projects
None yet
Development

No branches or pull requests

5 participants