| 5dd585abb84c5d13af0017f78741e29505f7b5f7 |
|
15-Feb-2018 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Make conversions from ShapedBuffer <-> ScopedShapedBuffer efficient by
moving memory ownership instead of copying.
PiperOrigin-RevId: 185871648
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| e4532d20973c4c00854492362665317551661c18 |
|
22-Dec-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Merge changes from github.
PiperOrigin-RevId: 179953488
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 9648f8040a559f6cf9bbe0501ba96f2b2c2864b1 |
|
16-Dec-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Automated g4 rollback of changelist 179258973
PiperOrigin-RevId: 179260538
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| d55f532867a3670d66460c5ee3b774519542adc1 |
|
16-Dec-2017 |
Dandelion Man? <dandelion@google.com> |
Merge changes from github.
PiperOrigin-RevId: 179258973
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| bfcc0970d61952ee894eaa4ef3256033239359b7 |
|
10-Nov-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Change HloSharding to allow getting a ShapeTree for non-tuple types.
Add reverse iteration to ShapeTree.
PiperOrigin-RevId: 175341255
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| c38781f6ec9710c0102bdc9d95bf6176fd96d1ce |
|
09-Nov-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
When sharding a tuple, we typically want to describe the data sharding
of each individual subtensor individually. Tuples are essentially just
containers - the tensors they contain should be able to be sharded
differently.
Tuples are hierarchically structured, but shardings were designed to
not contain the sharded type (the sharded type is inferred from the
output type of the instruction the sharding is applied to). Therefore,
shardings for tuples contain shardings for each subtensor as a
non-structured list.
This list is ordered as a preorder walk of the tuple shape, and of
course only the leaf nodes of the tuple shape are stored. The
structure is reapplied when the sharded instruction's shape is known.
PiperOrigin-RevId: 175132692
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 5ead76420dee762a5f710fda6893075f1292d5d3 |
|
19-Aug-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Reduce XLA compile time by ~7% for a convolutional image model:
* Added CompactPointerSet<T>, which is optimized for set size <= 1.
* Changed expensive CHECKs to DCHECKS in buffer_assignment.cc
* Reserve space in DFS state array before starting DFS.
* Use unsigned arithmetic in DFS state maintenance.
* HloInstruction:
- Moved frequently used fields to start for better cache locality.
- Use InlinedVector instead of vector for operand array.
- Use InlinedVector instead of vector for DFS stack.
* Pre-compute "is array" and "is tuple" for LogicalBuffer.
* PointsToSet:
- Combine two ShapeTrees into one.
- Use CompactPointerSet instead of std::set to hold sources.
- Use CompactPointerSet instead of std::set to hold flattened buffers.
* ShapeTree: use unique_ptr instead of optional for shape storage
(reduces size and destruction overhead).
* Add proper const qualifiers to some FlatSet iterator methods.
Co-author=jeff
PiperOrigin-RevId: 165759117
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| f3cd80f1f84fedbd2b9ccc064c2c5c28eb9711b8 |
|
04-Aug-2017 |
Mark Heffernan <meheff@google.com> |
Fix const_iterator in ShapeTree. The existing implementation didn't work
because const Foo<T>* and Foo<const T>* are not convertible in C++ which broke
the internal machinery of the iterator in the const case.
PiperOrigin-RevId: 164276236
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 1e236617d308bcfcf030ea363854f02c3c7c64f2 |
|
02-Aug-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Add an iterator implementing std::forward_iterator_tag to ShapeTree.
This allows it to be iterated over by a for-loop.
PiperOrigin-RevId: 163950729
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 181816fe27684585bface6e2260a0ff1c890e3e9 |
|
29-Jun-2017 |
Justin Lebar <jlebar@google.com> |
Speed up TuplePointsToAnalysis.
This analysis is one of the most expensive parts of the HLO optimization
pipeline.
- Avoid one or two unnecessary hashtable lookups in
PopulateDefinedBuffersAndAliases.
- Add a mode to ShapeTree wherein we avoid copying Shapes.
- Use templated functors rather than std::function in ShapeTree's
iterators, thus avoiding the overhead of std::function.
PiperOrigin-RevId: 160487485
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 05412bd367198ec491ca034b4bc634784c03125c |
|
07-Jun-2017 |
Mark Heffernan <meheff@google.com> |
[XLA] Simplify Shape traversal visitors.
Simplify shape traversal visitors in ShapeUtil and ShapeTree. Add a non-Status form because most uses of the traversal methods do not use it, and remove is_leaf parameter from ShapeTree.ForEach* as it is not frequently used.
PiperOrigin-RevId: 158201574
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| 2ee09b873a7f658fba151d8e39d2a8bc67e136a6 |
|
02-Jun-2017 |
Mark Heffernan <meheff@google.com> |
[XLA] Various improvements to ShapeTree.
Add support for holding non-copyable types, operator==, and a
CopySubtreeFrom method for copying a subtree from one ShapeTree to
another.
PiperOrigin-RevId: 157777636
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| ea910532bc21e265d5af7bdcd63d836e4204acf2 |
|
17-Apr-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
Fix ShapeTree operator= to actually work. Also added some tests.
Change: 153385325
/external/tensorflow/tensorflow/compiler/xla/shape_tree.h
|
| a241fe8ec5bc46e06c48c3ad609f00bc7a5030d9 |
|
17-Apr-2017 |
A. Unique TensorFlower <gardener@tensorflow.org> |
[XLA] Improve ShapeTree implementation.
1) Ensure PODs are zero-initialized via the shape-constructor; i.e.
ShapeTree<bool>(shape) has all data items initialized to false.
2) Add a default constructor, which makes it possible to use operator[] in a
FlatMap<_, ShapeTree<_>>. This creates a tree where !ShapeTree::IsValid.
3) Change the representation so that the root node holds the shape by value,
while non-root nodes don't hold a shape. This is "cleaner" than holding an
unused shape in non-root nodes, and also simplifies the implementation.
There is now a struct internal::ShapeTreeNode only used by ShapeTree.
4) Simplify constructors to delegate to a single implementation.
5) Fix bug in ShapeTree(Shape shape, T init_value) constructor, which caused
the existing code to only work for T=bool.
6) Fix documentation to indicate that a T data value is held for every node in
the tree, not just leaf nodes. That was the original behavior, which is
unchanged in this CL; it's just the documentation that was wrong.
Change: 153305130
/external/tensorflow/tensorflow/compiler/xla/shape_tree.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/shape_tree.h
|