[TF:XLA] Generalize existing support for keeping variables on an XLA device in reshap...
authorPeter Hawkins <phawkins@google.com>
Tue, 15 May 2018 16:47:50 +0000 (09:47 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Tue, 15 May 2018 16:50:47 +0000 (09:50 -0700)
commitc515b31314f0c4b551bded4961e096db98a998fc
tree6c258f70f2a306b4dff9aaa26ee340d551aa22fd
parent6f971fb184560ba7e5aada79e539588f73450d16
[TF:XLA] Generalize existing support for keeping variables on an XLA device in reshaped form, instead allowing XLA devices to keep all tensors in a reshaped form outside an XLA computation.

PiperOrigin-RevId: 196683444
20 files changed:
tensorflow/compiler/aot/tests/tfcompile_test.cc
tensorflow/compiler/jit/kernels/xla_launch_op.cc
tensorflow/compiler/jit/xla_cpu_device.cc
tensorflow/compiler/jit/xla_device.cc
tensorflow/compiler/jit/xla_device.h
tensorflow/compiler/jit/xla_device_context.cc
tensorflow/compiler/jit/xla_device_context.h
tensorflow/compiler/jit/xla_gpu_device.cc
tensorflow/compiler/jit/xla_launch_util.cc
tensorflow/compiler/tests/BUILD
tensorflow/compiler/tests/xla_device_gpu_test.py [new file with mode: 0644]
tensorflow/compiler/tests/xla_device_test.py
tensorflow/compiler/tf2xla/BUILD
tensorflow/compiler/tf2xla/kernels/retval_op.cc
tensorflow/compiler/tf2xla/xla_compiler.cc
tensorflow/compiler/tf2xla/xla_compiler.h
tensorflow/compiler/tf2xla/xla_compiler_test.cc
tensorflow/compiler/tf2xla/xla_context.cc
tensorflow/compiler/tf2xla/xla_context.h
tensorflow/compiler/tf2xla/xla_op_kernel.cc