MCPcopy Create free account
hub / github.com/benoitsteiner/tensorflow-opencl / PersistentTensor

Class PersistentTensor

tensorflow/core/framework/op_kernel.h:202–222  ·  view source on GitHub ↗

Wraps a tensor that is held by an Op across calls to Compute(). For memory safety when using asynchronous devices like GPUs, the system must be notified when a Tensor is used inside an Op execution. The wrapper ensures that all uses of the Tensor are tracked, because in order to retrieve the Tensor the caller must use AccessTensor which notifies the context.

Source from the content-addressed store, hash-verified

source not stored for this graph (policy: none)

Callers 8

allocate_persistentMethod · 0.85
LockedReadMethod · 0.85
TryEnqueueMethod · 0.85
TryEnqueueMethod · 0.85
MutableDenseHashTableMethod · 0.85
LOCKS_EXCLUDEDMethod · 0.85
TryEnqueueMethod · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected