diff options
| author |  Justin Lebar <jlebar@google.com> | 2018-08-01 16:00:38 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2018-08-01 16:06:11 -0700 |
| commit | bed00207976a70370cb6e3615c7ad56a5547cf45 (patch) | |
| tree | b154beccd7cbabc839cc72ac6291ad082055879b /third_party | |
| parent | 626317cb35524e4deb0851a65ca5dd5ca61d431f (diff) | |
[SE] Allow context reuse in CreatedContexts::Add.
It's possible for an already-existing context to be returned by
cuDevicePrimaryCtxRetain. Previously, this would be handled incorrectly
by CreatedContexts::Add, which was assuming that inserts into the map
always succeeded.
This makes XLA work with
TF_CUDA_PLATFORM_GPU_DEVICE_SCHEDULE=blocking_sync, although exactly how
that flag is related to this bug is unclear to me. It seems like some
sort of race condition, maybe?
PiperOrigin-RevId: 207010059
Diffstat (limited to 'third_party')
0 files changed, 0 insertions, 0 deletions