PiperOrigin-RevId: 480928553

PiperOrigin-RevId: 480927625

PiperOrigin-RevId: 480927276

PiperOrigin-RevId: 480926841

PiperOrigin-RevId: 480926009

PiperOrigin-RevId: 480922000

Tiling `linalg.matmul` with any (< 3) number of tile sizes.

PiperOrigin-RevId: 480921681

PiperOrigin-RevId: 480919007

So that it can be included from header files without bringing
in cpu_executable as a dependence.

PiperOrigin-RevId: 480917177

The two files typically need to be consumed jointly, as gpu_delegate_native_jni.cc contains the native code needed for (successfully) initializing the native part of GpuDelegate.

PiperOrigin-RevId: 480914567

PiperOrigin-RevId: 480909268

- when the asset source and destination paths are the same, prevent copying.

PiperOrigin-RevId: 480907795

PiperOrigin-RevId: 480904957

Seems like the pattern already supports shrink axis mask and new axis mask.

PiperOrigin-RevId: 480904459

PiperOrigin-RevId: 480899914

PiperOrigin-RevId: 480885011

Currently, the type base branching logic for same scale is not mature yet.
It would be better to add this later when int16 MLIR quantizer becomes
concretized.

PiperOrigin-RevId: 480878972

PiperOrigin-RevId: 480875344

http://github.com/tensorflow/runtime/commit/00734af980b920f9f963badf85fbeb12d576fde5.

PiperOrigin-RevId: 480873402

Currently, programs compiled through the MLIR pipeline that use non-standard
layouts for results see transposed values.

This normalization is only correct if none of the instructions inside the
computation have non-standard layouts (which is the case for programs
compiled through the MLIR pipeline, but not true in general).

This fixes tensorflow/compiler/xla/tests:client_test_cpu with XLA_FLAGS=--xla_cpu_enable_mlir_lowering=true

PiperOrigin-RevId: 480870452

Updates LLVM usage to match
[7ac10e5b347f](https://github.com/llvm/llvm-project/commit/7ac10e5b347f)

PiperOrigin-RevId: 480868167

Also extract some common utilities for scatter and gather into
mhlo_scatter_utils, and rename it to mhlo_scatter_gather_utils.

The functionality is the same as the XLA:Current simplifier.
The test cases were copied as well, and a test case for slice
sizes containing 0 was added.

PiperOrigin-RevId: 480864143

This fixes //third_party/tensorflow/compiler/xla/tests:reduce_precision_test_cpu
with XLA_FLAGS=--xla_cpu_enable_mlir_lowering=true.

PiperOrigin-RevId: 480863541

Updating existing object without creation of new one.

PiperOrigin-RevId: 480862684

PiperOrigin-RevId: 480856355

Currently, programs compiled through the MLIR pipeline that use non-standard
layouts for results see transposed values.

This fixes tensorflow/compiler/xla/tests:client_test_cpu with XLA_FLAGS=--xla_cpu_enable_mlir_lowering=true

PiperOrigin-RevId: 480845821

Currently, programs compiled through the MLIR pipeline that use non-standard
layouts for results see transposed values.

This fixes tensorflow/compiler/xla/tests:client_test_cpu with XLA_FLAGS=--xla_cpu_enable_mlir_lowering=true

PiperOrigin-RevId: 480838013

PiperOrigin-RevId: 480834358

PiperOrigin-RevId: 480833633

PiperOrigin-RevId: 480833632

PiperOrigin-RevId: 480822135

PiperOrigin-RevId: 480821629

This introduces a dependency from XLA to TensorFlow, the test should be rewritten to avoid this and be made a raw py_test  instead of a tf_py_test

PiperOrigin-RevId: 480815635

PiperOrigin-RevId: 480792155

The only purpose of this macros over the native cc_test is to reintroduce the
protobuf library on tests targets when it is disabled from the TSL libraries.

PiperOrigin-RevId: 480786829

This change is needed for the upcoming Bazel 6.0 release.
Context: https://github.com/bazelbuild/bazel/issues/12932

Breakage: https://buildkite.com/bazel/bazel-at-head-plus-downstream/builds/2670#0183c9f5-c730-4e6d-9374-2bc9ace37761
PiperOrigin-RevId: 480784917

Adds logging messages for TF quantization for better logging & debuggability.

PiperOrigin-RevId: 480772029

PiperOrigin-RevId: 480771570

PiperOrigin-RevId: 480770359

PiperOrigin-RevId: 480763101