| Commit message (Expand) | Author | Age |
... | |
* | | [XLA] Fix some outdated comments referring to FlatMap | Benjamin Kramer | 2018-10-02 |
* | | Add a hint parameter to TransferLiteralToDeviceAsync that the implementation ... | A. Unique TensorFlower | 2018-10-02 |
* | | Fix layout assignment for cross module all reduce | A. Unique TensorFlower | 2018-10-02 |
* | | Merge pull request #21958 from MattConley:CudaOccupancy | TensorFlower Gardener | 2018-10-01 |
|\ \ |
|
* | | | [XLA] Migrate from gtl::FlatSet to absl::flat_hash_set | Benjamin Kramer | 2018-10-01 |
* | | | [XLA] Add kAllToAll and kCollectivePermute to EffectiveOperandPrecisionIsOutp... | A. Unique TensorFlower | 2018-10-01 |
* | | | Improve error message in transpose shape inference. | Tayo Oguntebi | 2018-10-01 |
* | | | [XLA] Migrate from gtl::FlatMap to absl::flat_hash_map | Benjamin Kramer | 2018-10-01 |
* | | | Name fusion parameters simply "param_X". | Mark Heffernan | 2018-10-01 |
* | | | [HloOrdering] Make parameter always defined before other instructions. | Yunxing Dai | 2018-09-30 |
* | | | [XLA] Use a result cache to speed up InstructionFusion::CanFuseOnAllPaths() | Yuanzhong Xu | 2018-09-27 |
* | | | Add opaque field to custom call. | Mark Heffernan | 2018-09-27 |
* | | | [XLA] Don't use NumUniqueInstructionIds() as a proxy for instruction_count() | Michael Kuperstein | 2018-09-26 |
* | | | [XLA] In HloEvaluator, fix an issue where the return type and native type are... | Kay Zhu | 2018-09-25 |
* | | | [XLA:GPU] Pad convolution features of size 3 up to 4. | Justin Lebar | 2018-09-25 |
* | | | Add a new pass after convolution rewriter and pad insertion, to pattern | Tim Shen | 2018-09-25 |
* | | | [XLA] Avoid recursion in global decreasing size best-fit heap. | Yuanzhong Xu | 2018-09-25 |
* | | | [XLA] Make HloComputation::instruction_count() constant-time. | Michael Kuperstein | 2018-09-25 |
* | | | [XLA] Use \n rather than <br/> for linebreaks in graphviz tooltips. | Justin Lebar | 2018-09-24 |
* | | | Add cuDNN fused convolution forward support. | Tim Shen | 2018-09-24 |
* | | | Short-circuit AllOf as well. This fixes a crash in ConstantScalar, as it | Tim Shen | 2018-09-24 |
* | | | Fix Hlo pattern matcher's AnyOf, so that a sub-pattern doesn't capture | Tim Shen | 2018-09-24 |
* | | | Remove the public uses of CreateCudnnConv* in the favor of | Tim Shen | 2018-09-24 |
* | | | Fix typo in error message. | Sanjoy Das | 2018-09-24 |
* | | | [XLA] Update Scatter to ignore out-of-bound indices. | A. Unique TensorFlower | 2018-09-24 |
* | | | Further simplify the cuDNN wrappers. Instead of passing around | Tim Shen | 2018-09-24 |
* | | | Generalize sort implementation in the HloEvaluator. | Adrian Kuegel | 2018-09-24 |
* | | | Add a flag that lets users override the number of host "devices" as seen by XLA | Sanjoy Das | 2018-09-21 |
* | | | Move winograd algorithm workaround to stream executor. | Tim Shen | 2018-09-21 |
* | | | Ensure that no capture is done unless Match() return true. Otherwise the | Tim Shen | 2018-09-21 |
* | | | [XLA] Add a global decreasing size best-fit buffer allocation algorithm, whic... | Yuanzhong Xu | 2018-09-21 |
* | | | Simplify XLA pattern matcher to explicitly construct AllOf relations, | Tim Shen | 2018-09-21 |
* | | | It is more computationally efficient to represent resize bilinear as a | Blake Hechtman | 2018-09-20 |
* | | | [XLA] Don't create mixed precision operations accidentally | David Majnemer | 2018-09-20 |
* | | | Automated rollback of commit da3357ecbdd6772413e8bbceeab8238971be11ce | Adrian Kuegel | 2018-09-20 |
* | | | Internal change. | A. Unique TensorFlower | 2018-09-20 |
* | | | Add VerifiedHloModule class. | Mark Heffernan | 2018-09-19 |
* | | | Simplify ir_emitter_unnested so that it doesn't take a look at conv | Tim Shen | 2018-09-19 |
* | | | Remove non-determinism in model-parallel compilation | HyoukJoong Lee | 2018-09-19 |
* | | | [XLA:CPU] Add an emitter for erfinv(double) and erfinv(half). | Benjamin Kramer | 2018-09-19 |
* | | | Treat kDomain instruction as a pure pass-through in HloValue | A. Unique TensorFlower | 2018-09-19 |
* | | | Add interface for HLO passes which run on HloModuleGroup. | Mark Heffernan | 2018-09-19 |
* | | | Run CPU tests remotely. | A. Unique TensorFlower | 2018-09-19 |
* | | | Replace DLOG(FATAL) with an Unimplemented error. | Adrian Kuegel | 2018-09-19 |
* | | | Implement sort op for CPU. | Adrian Kuegel | 2018-09-19 |
* | | | [TF:XLA] Change HloPtrComparator to work across HLO modules. Declaring the me... | Dimitris Vardoulakis | 2018-09-18 |
* | | | Add layout information to logging. | Tim Shen | 2018-09-18 |
* | | | Add ConstantScalar, WithPredicate, Disjunction, and OpAnyOrder (where Op | Tim Shen | 2018-09-18 |
* | | | [TF:XLA] Document that the order of control predecessors matters. | Dimitris Vardoulakis | 2018-09-18 |
* | | | Remove unnecessary side-effect test, since HLO liveness now reports correct | A. Unique TensorFlower | 2018-09-17 |