http://github.com/tensorflow/runtime/commit/e03a192239c4f59c63db1b4d54b4c15f1d0a1ad7.

PiperOrigin-RevId: 481277525

PiperOrigin-RevId: 481275791

Updates LLVM usage to match
[d8415b02a519](https://github.com/llvm/llvm-project/commit/d8415b02a519)

PiperOrigin-RevId: 481275732

PiperOrigin-RevId: 481266358

PiperOrigin-RevId: 481265147

PiperOrigin-RevId: 481258105

PiperOrigin-RevId: 481253023

PiperOrigin-RevId: 481252019

[XLA:GPU] Check that the types used in the matmul are supported by cublasLt. If they are not supported, we fall back to legacy cublas.

This fixes an issue found by some JAX dot_general tests which depended on device (p100/v100). There are certain combinations of types which work on some devices but not on others. However, according to the official cublasLt documentation, these combinations of types are unsupported by cublasLt on all devices.

PiperOrigin-RevId: 481251576

PiperOrigin-RevId: 481251198

Guard an aarch64-specific TF-oneDNN code block with an ifdef to make sure it doesn't affect the x86 backend, per discussion here:
https://github.com/tensorflow/tensorflow/pull/57987#discussion_r993731524

PiperOrigin-RevId: 481247334

PiperOrigin-RevId: 481246613

Unify functions to check whether oneDNN is enabled: [Python test_util checker](https://github.com/tensorflow/tensorflow/blob/76b9fa22857148a562f3d9b5af6843402a93c15b/tensorflow/python/framework/test_util.py#L360-L362), [Pywrapped checker](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/util/port.cc#L62-L67) and [C++ checker](https://github.com/tensorflow/tensorflow/blob/76b9fa22857148a562f3d9b5af6843402a93c15b/tensorflow/core/util/util.cc#L130-L183).

Move the main function to port.cc to avoid cyclic dependency. Keeping the fix minimal to cherry-pick into TF 2.11.

Inconsistencies could cause bugs because Python layer may assume oneDNN is disabled when it's not. For example, `//tensorflow/python/framework:config_test` will fail when run on Cascade Lake or newer CPUs because:
* The test is [supposed to be skipped](https://github.com/tensorflow/tensorflow/blob/76b9fa22857148a562f3d9b5af6843402a93c15b/tensorflow/python/framework/config_test.py#L682-L684) when oneDNN is turned on.
* Python `test_util` `IsMklEnabled` only checks for [static defines](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/util/port.cc#L62-L67) and [`TF_ENABLE_ONEDNN_OPTS` environment variable](https://github.com/tensorflow/tensorflow/blob/76b9fa22857148a562f3d9b5af6843402a93c15b/tensorflow/python/framework/test_util.py#L362). When the var is unset, the Python checker will just think oneDNN is disabled. But oneDNN is turned on by default on Cascade Lake and newer Intel CPUs (even when the env var is unset).
* The test was not skipped when it should have been skipped.

PiperOrigin-RevId: 481244351

PiperOrigin-RevId: 481240161

[XLA:Python] Expose the Python 3.11 function PyCode_Addr2Location from the XLA:Python traceback library.

This allows us to get start column information and end line/column information under Python 3.11.

PiperOrigin-RevId: 481240029

This updates the behavior of flatbuffer_import to be consistent with
that of flatbuffer_export -- it's now possible to losslessly
round-trip between flatbuffer and MLIR formats in the presence of
model_control_dependencies metadata.

PiperOrigin-RevId: 481236206

Also, add more optimization/simplification passes.

PiperOrigin-RevId: 481229419

PiperOrigin-RevId: 481226973

PiperOrigin-RevId: 481223775

http://github.com/tensorflow/runtime/commit/b23dc948cfe0660dd889fb77844d95849876da23.

PiperOrigin-RevId: 481216888

- Use xla::runtime::DialectRegistry and xla::runtime::PassManager to wrap corresponding mlir types
- Restrict visibility of packages under
   /compiler/xla/mlir/transforms/runtime
   /compiler/xla/mlir/ir/runtime/
   /compiler/xla/mlir/utils/runtime/

PiperOrigin-RevId: 481214910

- HloConstantSplitter pass will ensure that each constant has a single use by
  duplicating multi-use constants.
- Doing this before sharding propagation prevents inadvertent and unintended
  propagation of sharding. As an example, a shared constant between automatic
  sharded region and manually sharded region can cause the manual sharding to
  leak into the automatic sharded code.

PiperOrigin-RevId: 481203013

PiperOrigin-RevId: 481195613

PiperOrigin-RevId: 481195475

PiperOrigin-RevId: 481193247

BEFORE

---------------------------------------------------------------
Benchmark                     Time             CPU   Iterations
---------------------------------------------------------------
BM_FlatMemrefX12All        81.3 ns         81.3 ns      8617874
BM_FlatMemrefX12None       69.1 ns         69.1 ns     10129964

AFTER

---------------------------------------------------------------
Benchmark                     Time             CPU   Iterations
---------------------------------------------------------------
BM_FlatMemrefX12All        54.8 ns         54.8 ns     12683098
BM_FlatMemrefX12None       47.7 ns         47.7 ns     14667555

PiperOrigin-RevId: 481192231

PiperOrigin-RevId: 481192084

PiperOrigin-RevId: 481190338

PiperOrigin-RevId: 481184853

PiperOrigin-RevId: 481184687

http://github.com/tensorflow/runtime/commit/0e71dd2f3edc157ac9ef78643618ef2bdad9718c.

PiperOrigin-RevId: 481184177

Updates LLVM usage to match
[06da9b94ae37](https://github.com/llvm/llvm-project/commit/06da9b94ae37)

PiperOrigin-RevId: 481181270

Log when TPUReshardVariables is inserted. This information is often needed for performance optimization.

PiperOrigin-RevId: 481181240

PiperOrigin-RevId: 481177778

PiperOrigin-RevId: 481176034

PiperOrigin-RevId: 481174166

This is something that I forgot about when reviewing cl/481105084.

Importer needed to be changed manually, but exports is autogenerated and already works properly.

PiperOrigin-RevId: 481164618

PiperOrigin-RevId: 481159038

This error was caused by shifting a signed integer:
```
third_party/tensorflow/compiler/mlir/lite/utils/low_bit_utils.cc:44:38: runtime error: left shift of negative value -98
```

This is fixed by more intentional casting.

PiperOrigin-RevId: 481157662

PiperOrigin-RevId: 481153499