PiperOrigin-RevId: 549658605

PiperOrigin-RevId: 549654509

PiperOrigin-RevId: 549652138

PiperOrigin-RevId: 549648455

http://github.com/tensorflow/runtime/commit/6a9b1806562fcf3de8baf37f9c1cf0b475a2487f.

PiperOrigin-RevId: 549648158

PiperOrigin-RevId: 549646998

The RESOURCE_EXHASTED status contains payload that generates many pages of output when logging Python exceptions, obscuring the error message. The original structured information is still preserved and can be accessed using the original instance of status stored in the exception.

PiperOrigin-RevId: 549643312

It has no users left.

PiperOrigin-RevId: 549636361

Updates LLVM usage to match
[6c30b7886919](https://github.com/llvm/llvm-project/commit/6c30b7886919)

PiperOrigin-RevId: 549627514

PiperOrigin-RevId: 549624665

PiperOrigin-RevId: 549620455

1. Add global stack index builder which is used for mhlo-to-hlo translator.
See all changes under mhlo_to_hlo folder.

2. Add stack index frames reader which is used for hlo-to-mhlo translator.
See all changes under hlo_to_mhlo folder.

PiperOrigin-RevId: 549619471

Imported from GitHub PR https://github.com/openxla/xla/pull/4379

cc @akuegel

Thanks!
Copybara import of the project:

--
f8777b19371b87572b288604df558fa8ce57b75f by Chao Chen <cchen104@amd.com>:

correct amd gpu device info

Merging this change closes #4379

PiperOrigin-RevId: 549609765

PiperOrigin-RevId: 549604114

ir_emitter_unnested depends on fusions, which needs to depend on
ir_emitter to be able to emit calls.

PiperOrigin-RevId: 549600029

This will be good when we want to analyze all Triton fusions of an HLO for example with ncu.

PiperOrigin-RevId: 549579171

- Also give them clearer names (IMO, let me know if you disagree)
- Also get rid of the indirection through `GetNestedComputer()`.

PiperOrigin-RevId: 549568613

PiperOrigin-RevId: 549562986

PiperOrigin-RevId: 549562877

Also merge it with IrEmitter::EmitCallToNestedComputation.

PiperOrigin-RevId: 549553449

PiperOrigin-RevId: 549538997

I need to emit calls to nested computations outside of the scope of an
ir_emitter (and it's better to have these things be stateless whenever
possible anyway).

PiperOrigin-RevId: 549538403

PiperOrigin-RevId: 549534080

PiperOrigin-RevId: 549534066

PiperOrigin-RevId: 549533669

PiperOrigin-RevId: 549520448

PiperOrigin-RevId: 549509169

PiperOrigin-RevId: 549508753

Many TF API users depend on the two methods above to parse TF_Status. But the implementation directly accesses the internal of TF_Status. This breaks the ABI between TF and other libraries.

This change modifies `Set_TF_Status_from_Status` and `StatusFromTF_Status` to go through the API.

There's a small overhead in this conversion. Because a new absl::Status will be constructed. And strings will be copied. But I believe we should always bias towards correctness. For TF internal users who can see the layout of `TF_Status`, `TF_Status::status` should be used directly to avoid these limitations.

PiperOrigin-RevId: 549498469

Add passes for lowering from LMHLO to IREEInput dialect to be able to run XLA:GPU executables on top of IREE runtime (VM + HAL).

These passes currently are not enabled in the open source build because they require setting up XLA->IREE dependency in Bazel. This will be done separately. Currently we only build and test it using internal Google infrastructure.

PiperOrigin-RevId: 549487443

PiperOrigin-RevId: 549486441

PiperOrigin-RevId: 549483740

[AutoSharding] Add an option to allow generating fully replicated strategies for dot and convolution HLO ops.

PiperOrigin-RevId: 549483733

DTensor api.relayout and collectives need to recognize the new layout type.

PiperOrigin-RevId: 549477658

Add unused parameter enable_stub_generation to open-source version of pybind_extension to be in sync with the internal version.

PiperOrigin-RevId: 549476112

This is useful when we need to pass status through API boundary.

PiperOrigin-RevId: 549475209

Use MultipleIterationsAutoScaler inside the data service dispatcher implementation as follows:
- UpdateOptimalNumberOfWorkersMetric() in the maintenance thread.
- ReportProcessingTime() when receiving processing times from a WorkerHeartbeat.
- ReportTargetProcessingTime() when receiving a target processing time from a ClientHeartbeat.
- RemoveWorker() when detecting missing workers or executing MaybeRemoveTask.
- RemoveConsumer() when releasing missing clients.
- RegisterIteration() when creating a new Iteration.
- UnregisterIteration() when garbage-collecting old Iterations.

PiperOrigin-RevId: 549469531

PiperOrigin-RevId: 549468160

PiperOrigin-RevId: 549464529