| dfb59da4ede1daf163a167da590ac70c447eb41a |
|
26-Jan-2018 |
A. Unique TensorFlower <gardener@tensorflow.org> |
[XLA:GPU] Implement conditional as a sequence of thunks in the GPU backend.
This also includes the following fixes:
(1) Update buffer assignment for conditionals so that the buffers corresponding to the true operand and the true computation parameter are colocated, and similarly, the buffers corresponding to the false operand and the false computation parameter are colocated.
(2) Update GPU copy insertion pass to insert copies when constants appear as operands of conditional instructions.
PiperOrigin-RevId: 183297282
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 5bf26acd87d3d44183fc28cb9576cda10c0255ca |
|
02-Jan-2018 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Automated g4 rollback of changelist 180000981
PiperOrigin-RevId: 180581912
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| c0c2775ce3de682f7913d1aeaf50bbc4d1521934 |
|
23-Dec-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Automated g4 rollback of changelist 179983419
PiperOrigin-RevId: 180000981
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 7d1072dd3374a0aa22637a0fd4a17a4ddd064110 |
|
23-Dec-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Adds FFT for XLA: CPU via Eigen, GPU via cuFFT.
GPU support includes plan reuse with new scratch allocator per execution in fft_thunk.
PiperOrigin-RevId: 179983419
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 6a9a9ed0e1f5eded19d793b2be125d2d845cf079 |
|
22-Dec-2017 |
Justin Lebar <jlebar@google.com> |
[XLA:GPU] Implement BatchNormThunk as a call into cudnn.
Using cudnn for these calls is disabled by default, because it's not a
performance win on our benchmarks.
PiperOrigin-RevId: 179882911
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 1988732f81bc5f61cd97c20952d5359fc0bf627f |
|
20-Dec-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
[XLA:GPU] Make the use of scratch allocator in convolution_thunk safe.
Add member function Thunk::ShouldFutureScheduledThunksDependOn for
convolution_thunk to tell thunk executor that all future scheduled thunks
should wait for convolution_thunk. This can ensure that the use of scratch
allocator in convolution_thunk is safe.
PiperOrigin-RevId: 179628764
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 1a786ab335aabe9020cff4f0ab69a5844de70fbc |
|
06-Dec-2017 |
Justin Lebar <jlebar@google.com> |
[XLA:GPU] Don't autotune while other kernels are running.
XLA:GPU autotunes gemm and conv thunks, trying multiple algorithms in
sequence and picking the fastest one.
If other work is running concurrently with our autotuning, this can mess
up the results. In particular, even if the GPU is totally
deterministic, the concurrent work may finish before we finish
autotuning, giving an unfair advantage to the later algorithms.
To address this, we modify GpuExecutable to wait until the GPU is
quiescent before executing a thunk which performs autotuning. We then
cross our fingers and hope that whatever is fastest while the GPU is
quiescent will also be fastest in the "real world", with (potentially)
concurrent work going on.
PiperOrigin-RevId: 178041481
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 8cb5e9867482a8e05f756fad35634e1674fe7f16 |
|
25-May-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Preliminary Infeed support for GPU backend.
** GPU transfer manager and GPU specific infeed manager/infeed buffer implementation
** Infeed thunk
PiperOrigin-RevId: 157054373
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|
| 1e67c90e2caceeff82d09793d1ef5fa0300d219b |
|
09-Jan-2017 |
Peter Hawkins <phawkins@google.com> |
Initial open-source release of XLA: Accelerated Linear Algebra.
XLA is a compiler-based linear algebra execution engine that targets CPUs, GPUs and custom accelerators.
XLA is still experimental; we are releasing it early to get the community involved.
Change: 143990941
/external/tensorflow/tensorflow/compiler/xla/service/gpu/thunk.h
|