diff options
author | 2018-09-18 15:42:44 -0700 | |
---|---|---|
committer | 2018-09-18 15:46:41 -0700 | |
commit | e1a32c98210f8ebba42a0397259d948e1433c09e (patch) | |
tree | 0a62289fe29cf2c0f481bdca90c03811a47caadf /README.md | |
parent | 6c8f6920e8bad10429ac0b88abbe0ace5a5e9a72 (diff) |
"Isolate" must-be-constant side effecting operations
I first tried to fix this issue in cr/209996730 but didn't quite fix the problem
for for XLA_* devices. A node assigned to an XLA_* device must be compiled so
the cr/209996730 fix of simply not compiling the nodes doesn't generalize to
XLA_* devices. Instead we now "isolate" these nodes, only putting them in a
trivial one-node cluster. For non-XLA devices even this trivial cluster is
ignored because of flags->tf_xla_min_cluster_size.
I was initially considering a more principled data-flow-analysis based solution
but then decided the upfront work isn't worth it until I see a clear motivating
example.
PiperOrigin-RevId: 213531437
Diffstat (limited to 'README.md')
0 files changed, 0 insertions, 0 deletions