aboutsummaryrefslogtreecommitdiffhomepage
path: root/tensorflow/compiler/xla/service/transpose_folding.cc
Commit message (Collapse)AuthorAge
* [XLA] Make kConvolution, kDot HLO attributes mandatoryGravatar David Majnemer2018-09-04
| | | | | | | | HLO transformations would forget to propagate the feature depth attribute. Making these attributes mandatory, while slightly less convenient for tests, makes HLO transformations more robust. PiperOrigin-RevId: 211490160
* [XLA] Expose a way to control dot/conv precisionGravatar David Majnemer2018-08-22
| | | | | | | | | | This adds a field to the proto so that we may serialize it. On TPUs, we can simulate higher precision by splitting a float32 number into several bfloat16 numbers such that their sum closely approximates the original number. A tensor contraction operation like convolution or a dot product can be computed by forming several partial products which approximate the correct answer to a closer margin. PiperOrigin-RevId: 209720948
* Automated g4 rollback of changelist 193451839Gravatar A. Unique TensorFlower2018-06-12
| | | | PiperOrigin-RevId: 200275406
* [XLA] s/tensorflow::Status/Status/.Gravatar Justin Lebar2018-05-11
| | | | | | | These are type aliases of one another; we'd like to be consistent and use the shorter one. PiperOrigin-RevId: 196322955
* Remove uses of the kTransposeDot fusionGravatar Sanjoy Das2018-05-07
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | I didn't remove the enum itself, but after this change removing the enum should be a simple NFC change (famous last words!). This will make it easier to implement BatchDot on CPU. The change removes usages of kTransposeDot by: - Teaching TransposeFolding to "fuse" transposes into dots by flipping the lhs_contracting_dims/rhs_contracting_dims fields. - Replacing the notion of transpose_lhs/transpose_rhs in the IR emitters with "has a non-canonical LHS contraction dimension"/"has a non-canonical RHS contraction dimension" where the canonical LHS and RHS contraction dims [0] are 1 and 0. Some tests were getting away with creating Dot instructions with their dimensions numbers unset. I've fixed these to create canonical dot operations instead. It is possible (but hard to tell without trying) that some of the IR emission logic and Eigen runtime calls can now be simplified further. For instance, instead of passing in a `transpose_lhs` and `transpose_rhs` to the Eigen GEMM routines, we could instead pass in the LHS and RHS contraction dimensions directly. [0] See HloInstruction::CreateCanonicalDot. PiperOrigin-RevId: 195514907
* Teach transpose folding about sharding.Gravatar A. Unique TensorFlower2018-04-18
| | | | PiperOrigin-RevId: 193451839
* [XLA] Always fold transposes into convs or dots regardless of use countGravatar David Majnemer2017-12-12
| | | | PiperOrigin-RevId: 178790193
* [XLA] Support transposing the spatial dimensions of a convolution's activationsGravatar David Majnemer2017-11-28
| | | | PiperOrigin-RevId: 177260886
* [XLA] Separate input and output spatial dimensions for convolutionGravatar David Majnemer2017-11-27
| | | | | | This lets us reason about input spatial dimensions as distinct from output spatial dimensions. By doing this, it opens up more opportunities for assigning more interesting, different, layouts for the activations and the output. PiperOrigin-RevId: 177117140
* [XLA] Teach transpose folding how to transpose the LHS of convolutionsGravatar David Majnemer2017-10-19
| | | | | | | This is now possible now that we have added the required fields to ConvolutionDimensionNumbers. PiperOrigin-RevId: 172807540
* [XLA] Make HloModule::computations() return raw pointers.Gravatar Justin Lebar2017-09-29
| | | | | | | | | | | | | | | | | | Like HloComputation::instructions(), HloModule::computations() used to return a list of unique_ptrs. But this is an implementation detail that shouldn't be leaked into the public API. This patch also adds HloModule::MakeNonFusionComputations(), because many of the callers of computations() went on to filter out all the fusion computations. It would be possible to implement MakeNonFusionComputations() "in place" using a filtering iterator, but I don't think it's necessary -- we never have *that* many computations, and since many callers go on to copy the list of non-fusion computations, making it unconditionally a copy is simpler and avoids a footgun. PiperOrigin-RevId: 170529051
* Cut dependencies between XLA client/service and CPU/GPU backends.Gravatar Justin Lebar2017-09-15
| | | | PiperOrigin-RevId: 168922817
* [XLA] Finish normalizing fusion computations into standard computationsGravatar A. Unique TensorFlower2017-07-26
| | | | PiperOrigin-RevId: 163210327
* [XLA] Run transpose_folding on nested computationsGravatar David Majnemer2017-04-26
| | | | | | | We only ran the pass on the entry computation which would make us lose out on optimization opportunities. Visit all computations to find any potential transpose folding opportunities. Change: 154343660
* [XLA] Add support for folding transpose into convolutionGravatar David Majnemer2017-04-08
| | | | Change: 152581336
* [XLA] Make `HloPass` an interface, NFCGravatar A. Unique TensorFlower2017-01-24
| | | | | This will allow inheritance from both `HloPassInterface` and `DfsHloVisitor`, so various passes which include a visitor can have handler methods overridden per backend. Change: 145477041
* Initial open-source release of XLA: Accelerated Linear Algebra.Gravatar Peter Hawkins2017-01-09
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
generated by cgit v1.2.3 (git 2.39.1) at 2024-06-27 23:06:18 +0000