Commit Graph

186511 Commits

Author SHA1 Message Date
Marcin Radomski
689bf5ef28 [XLA:GPU] Add Thunk::TransformAllNestedThunks
A function similar to ForAllThunksMutable, but capable of wrapping or replacing nested thunks with new ones. In case the thunk has some specific requirements (e.g. assumes the directly nested thunk is a `SequentialThunk`), the implementation needs to ensure the assumptions still hold. The transform function must be infallible, but it may be a no-op and return the original `unique_ptr<Thunk>`.

The use for this is in buffer checksumming: to cover all thunks, we need to recursively insert pre-/post-execution checksum thunks around execution of nested thunks. Unfortunately, ForAllThunksMutable doesn't allow mutating the thunk's *container*, which would create the need to switch() over thunk kind, special-case any thunk that may contain nested thunks, and expose the internals to the caller to reimplement the recursion.

PiperOrigin-RevId: 826014413
2025-10-30 08:00:41 -07:00
Alex
2f1ef8b1a0 PR #33188: [ROCm] Fix hermetic tests when executing on rbe worker
Imported from GitHub PR https://github.com/openxla/xla/pull/33188

📝 Summary of Changes
Fix tests when executing on rbe

🎯 Justification
Fix hermetic dependencies so right libs are loaded.
Libs are put as a data dependency as many of the rocm libs may have their own
RPATH=$ORIGIN/..., they will try to load the libs from the place where they are.
This leads to a situation that one lib located in _solid_data dir will try to find its
dependencies in the same dir. However dependency might be located in the data
directory. Then either test can't load the lib or tries to load one from the system libs.

🚀 Kind of Contribution
Please remove what does not apply: 🐛 Bug Fix

📊 Benchmark (for Performance Improvements)
Build fix

🧪 Unit Tests:
Build fix

🧪 Execution Tests:
Build fix

Copybara import of the project:

--
43913bfbacc042c7a81e692cfa28e0a0733c3989 by Alexandros Theodoridis <atheodor@amd.com>:

Add minimal hip runtime deps to kernel_headers

--
ffb62f536608b00445fda4fce0a3323c5bf6fe9d by Alexandros Theodoridis <atheodor@amd.com>:

Fix hermetic build invalid local rpath

Merging this change closes #33188

PiperOrigin-RevId: 826014075
2025-10-30 07:50:50 -07:00
A. Unique TensorFlower
ca45a1e4bb [XLA:GPU] Allow to map slice of memory with multicast object.
PiperOrigin-RevId: 826013717
2025-10-30 07:37:58 -07:00
A. Unique TensorFlower
181ff64d18 Automated Code Change
PiperOrigin-RevId: 826012733
2025-10-30 07:22:32 -07:00
Ilya Tikhonovskiy
398fefb520 [XLA:GPU] Enable fusion of broadcast and reshape into Triton scaled-dot.
This change allows broadcast and reshape operations on the scale operands of `scaled-dot` to be fused into the Triton kernel. It generalizes the operand fusion logic to handle all four operands of `scaled-dot` and adds support for `BroadcastOp` and `ExpandDimsOp` in the Triton MLIR conversion. A new test case is added to verify this fusion.

PiperOrigin-RevId: 826012089
2025-10-30 07:11:55 -07:00
Adrian Kuegel
3bd63fafc5 Restore setting the default value for xla_gpu_enable_dynamic_slice_fusion flag.
The line was accidentally removed, this change adds it back with the previous
default (false).

PiperOrigin-RevId: 826008957
2025-10-30 06:58:33 -07:00
Thomas Joerg
fbaeea227b [XLA:GPU] Test that DotDecomposer canonicalizes batch dims.
Existing tests do not cover this.

PiperOrigin-RevId: 826006833
2025-10-30 06:45:23 -07:00
A. Unique TensorFlower
0a1309a2e5 Automated Code Change
PiperOrigin-RevId: 825996910
2025-10-30 06:11:26 -07:00
Nikita Putikhin
3449303622 Handle PyCUtensorMapObject in extractTmaDesc in the launcher
Reenables failing tests

PiperOrigin-RevId: 825983235
2025-10-30 05:15:23 -07:00
A. Unique TensorFlower
82adb21aad Automated Code Change
PiperOrigin-RevId: 825978577
2025-10-30 05:00:30 -07:00
A. Unique TensorFlower
930d6052e2 Automated Code Change
PiperOrigin-RevId: 825969551
2025-10-30 04:26:33 -07:00
A. Unique TensorFlower
8a9bec96a6 Automated Code Change
PiperOrigin-RevId: 825961742
2025-10-30 03:56:21 -07:00
Alexander Shaposhnikov
d29d0f8635 Minor cleanup in YnnDimensions.
PiperOrigin-RevId: 825955545
2025-10-30 03:37:49 -07:00
A. Unique TensorFlower
cfa7953589 Automated Code Change
PiperOrigin-RevId: 825940381
2025-10-30 02:48:52 -07:00
Adrian Kuegel
1fa646265a [XLA:GPU] Avoid a segfault in StreamAttributeAnnotator
Currently it is assumed that GetTupleElement is never the root of a
computation. That assumption is not necessarily true, e.g. during autotuning of
Cublas Gemm calls we can have a GetTupleElement op as root.

PiperOrigin-RevId: 825932301
2025-10-30 02:36:46 -07:00
A. Unique TensorFlower
50d82c5b2d compat: Update forward compatibility horizon to 2025-10-30
PiperOrigin-RevId: 825930743
2025-10-30 02:28:11 -07:00
A. Unique TensorFlower
cdd331ca3f Update GraphDef version to 2396.
PiperOrigin-RevId: 825930720
2025-10-30 02:15:19 -07:00
Adrian Kuegel
ce8015c614 [XLA:CPU] Remove obsolete IndexedArrayAnalysisPrinterPass.
It is not being used anymore.

PiperOrigin-RevId: 825927601
2025-10-30 02:04:18 -07:00
A. Unique TensorFlower
6f1d4574bd Automated Code Change
PiperOrigin-RevId: 825902854
2025-10-30 00:58:39 -07:00
Bhatu
a36834c399 Update rules_ml_toolchain to version with nvcc wrapper fixes .
PiperOrigin-RevId: 825832143
2025-10-29 20:42:44 -07:00
Zixuan Jiang
ba10feaa24 Add an overload for SpmdPartitioner::SetPartitionedHlo to avoid unnecessary lambda functions.
PiperOrigin-RevId: 825819367
2025-10-29 20:10:18 -07:00
A. Unique TensorFlower
512f1e48cb Use reuse semantics instead of copy semantics when calling DisassembleIntoSingleDeviceArrays, since disassemble always aliases.
Use copy semantics instead of reuse semantics when calling CopyToHostBuffer, since copy to host always copies.

PiperOrigin-RevId: 825815935
2025-10-29 19:53:06 -07:00
Subhankar Shah
28e9e5ea27 [XLA:MSA] Allow block prefetching for custom call prefetches that have aliased uses.
* Extend alternate memory chunk reservations for aliased uses.
* Add pinned allocations in alternate memory for aliased uses.
* Mark all aliased allocations as colocated.

Pin all values aliased with the prefetched source value to default memory.

PiperOrigin-RevId: 825801181
2025-10-29 19:02:42 -07:00
Eugene Zhulenev
e99aad85fe [xla:codegen] Cleanup MlirKernelSource APIs
PiperOrigin-RevId: 825795569
2025-10-29 18:41:47 -07:00
Will Froom
860f543d1c [XLA:CPU][XTile] Create lowering for Iota.
PiperOrigin-RevId: 825789498
2025-10-29 18:30:39 -07:00
A. Unique TensorFlower
c7055c2e5b Reverts 0b0ff7c8ac
PiperOrigin-RevId: 825779916
2025-10-29 18:10:47 -07:00
Niklas Vangerow
fe2a783077 Migrate broadcast_simple_test to use PjRt.
PiperOrigin-RevId: 825775803
2025-10-29 17:58:13 -07:00
Felix Wang
152b2338d9 Refactor xla codebase to avoid the dynamic_cast, use ClassOf or DynCast instead.
PiperOrigin-RevId: 825773229
2025-10-29 17:49:47 -07:00
A. Unique TensorFlower
9511b51e61 Remove dependency on private absl/base:endian.
PiperOrigin-RevId: 825772784
2025-10-29 17:37:56 -07:00
Eugene Zhulenev
d4b7f15aee [xla:ffi] NFC: Use AttrTag<T> for tagging regular arguments
PiperOrigin-RevId: 825769996
2025-10-29 17:19:10 -07:00
Hyeontaek Lim
b1d5462115 [IFRT] Remove -DIFRT_REQUIRE_USER_CONTEXT that is no longer used for detecting missing user contexts.
IFRT users such as JAX perform checks the presence of user contexts in IFRT objects in their layer.

PiperOrigin-RevId: 825760230
2025-10-29 16:46:05 -07:00
Parker Schuh
0b0ff7c8ac Change RawSEDeviceMemory to be AsyncValueRef.
PiperOrigin-RevId: 825735739
2025-10-29 15:49:26 -07:00
Eugene Zhulenev
0f559dec93 [xla:cpu] Move buffer allocation info encoding to tf2xla
PiperOrigin-RevId: 825732652
2025-10-29 15:40:10 -07:00
Eugene Zhulenev
756a72760a [xla:codegen] Remove MlirKernelDefinition alias
PiperOrigin-RevId: 825724819
2025-10-29 15:22:52 -07:00
Kanish Anand
ff2b8b600d Refactor: Move method definitions from mesh_and_axis.h to .cc file
PiperOrigin-RevId: 825722377
2025-10-29 15:10:16 -07:00
Jake Harmon
83051de423 Add option to tag PJRT wheels with nightly timestamp
PiperOrigin-RevId: 825706994
2025-10-29 14:33:36 -07:00
Felix Wang
cecce70fb2 Rename rail-aligned into world-level in collective_ops_utils.h
Network rail usually refers to a set of  NICs connected by the same fabric/switch, e.g. [Rail-optimized topology](https://developer.nvidia.com/blog/doubling-all2all-performance-with-nvidia-collective-communication-library-2-12/).

PiperOrigin-RevId: 825696577
2025-10-29 14:11:20 -07:00
A. Unique TensorFlower
ca3d7d6305 Integrate LLVM at llvm/llvm-project@028bfa255e
Updates LLVM usage to match
[028bfa255e90](https://github.com/llvm/llvm-project/commit/028bfa255e90)

PiperOrigin-RevId: 825670183
2025-10-29 13:04:28 -07:00
Henning Becker
757f0ac980 Add proto serialization for GpuComputeCapability
PiperOrigin-RevId: 825657032
2025-10-29 12:31:09 -07:00
A. Unique TensorFlower
9cbe7bd184 Enable reduced fingerprint for text-proto models
PiperOrigin-RevId: 825618056
2025-10-29 11:02:17 -07:00
Will Froom
09d56a9643 [XLA:CPU][XTile] Add lowering for reshape.
PiperOrigin-RevId: 825605674
2025-10-29 10:47:02 -07:00
Ilya Tikhonovskiy
82dc95c293 [XLA:GPU] rename thunk_checksum_tracing_pass to thunk_buffer_debug_pass
It is pure mechanical move cl.

The goal is to use the pass for all the buffer debug checks. We have checksum and nan_counter kernels at the moment.

PiperOrigin-RevId: 825602375
2025-10-29 10:27:29 -07:00
Will Froom
d717d76122 [XLA:CPU][XTile] Add lowering for broadcast.
PiperOrigin-RevId: 825578568
2025-10-29 09:28:41 -07:00
Fergus Henderson
98a24eb07e Fix test breakage when address/memory/thread sanitizer was enabled.
Update the production code to not set the `*_allocated_bytes` fields
when using address/memory/thread sanitizer, since those sanitizers don't
support `mallinfo()` -- `mallinfo()` may return invalid results when sanitizers
are enabled .

Update test to not expect the `*_allocated_bytes` fields to be valid when using
address/memory/thread sanitizer.

Also update docs in `memory_info.h` to make it clearer that even if `isSupported()`
returns true, it is possible that only _some_ of the fields are supported.

Also document that `memory_info.h` supports Windows; this was already clear from
other parts of the header file that described the behaviour on Windows.

PiperOrigin-RevId: 825505016
2025-10-29 05:47:04 -07:00
Will Froom
684717efe0 [XLA][XTile] Add pass to verify that a module conforms to XTile specification.
PiperOrigin-RevId: 825488424
2025-10-29 04:49:44 -07:00
Alex
8dc7ce7547 PR #33085: [ROCm] Fix too strict default spanw strategy for rbe builds
Imported from GitHub PR https://github.com/openxla/xla/pull/33085

📝 Summary of Changes
Fix too strict spawn strategy for rbe builds

🎯 Justification
remote only execution is not possible for all the tests

🚀 Kind of Contribution
Please remove what does not apply: 🐛 Bug Fix

📊 Benchmark (for Performance Improvements)
Not relevant

🧪 Unit Tests:
Not relevant

🧪 Execution Tests:
Not relevant

Copybara import of the project:

--
df73e6e006c47d5ada1e14ced8f2ae94c0df7dd8 by Alexandros Theodoridis <atheodor@amd.com>:

Fix too strict default spanw strategy for rbe builds

Merging this change closes #33085

PiperOrigin-RevId: 825463234
2025-10-29 03:33:03 -07:00
Marcin Radomski
8b47f52ef7 [XLA:GPU] Add BufferDebugLogEntryMetadataStore
Encoding extra metadata about an debug log entry within its ID limits how much
information we can pass. To remove the limitation without the need to pass
extra data between host and device, introduce a metadata store that provides a
opaque ID -> metadata mapping.

Follow up patches will make checksum/NaN tracing use
BufferDebugLogEntryMetadataStore shared between all thunks that operate on
BufferDebugLog:

- BuffersChecksumThunks put the metadata into the store and use the returned
  entry_ids to identify the checksums from BufferDebugLog,
- xla_gpu_buffer_debug_log_dump reads the BufferDebugLog and uses the store to
  resolve the entry_ids into the metadata.

PiperOrigin-RevId: 825462635
2025-10-29 03:18:55 -07:00
A. Unique TensorFlower
7b7a64f3c8 Automated Code Change
PiperOrigin-RevId: 825457902
2025-10-29 03:03:30 -07:00
dependabot[bot]
bbd2fb5cf8 PR #33141: Bump github/codeql-action from 4.30.9 to 4.31.0
Imported from GitHub PR https://github.com/openxla/xla/pull/33141

Bumps [github/codeql-action](https://github.com/github/codeql-action) from 4.30.9 to 4.31.0.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/github/codeql-action/releases">github/codeql-action's releases</a>.</em></p>
<blockquote>
<h2>v4.31.0</h2>
<h1>CodeQL Action Changelog</h1>
<p>See the <a href="https://github.com/github/codeql-action/releases">releases page</a> for the relevant changes to the CodeQL CLI and language packs.</p>
<h2>4.31.0 - 24 Oct 2025</h2>
<ul>
<li>Bump minimum CodeQL bundle version to 2.17.6. <a href="https://redirect.github.com/github/codeql-action/pull/3223">#3223</a></li>
<li>When SARIF files are uploaded by the <code>analyze</code> or <code>upload-sarif</code> actions, the CodeQL Action automatically performs post-processing steps to prepare the data for the upload. Previously, these post-processing steps were only performed before an upload took place. We are now changing this so that the post-processing steps will always be performed, even when the SARIF files are not uploaded. This does not change anything for the <code>upload-sarif</code> action. For <code>analyze</code>, this may affect Advanced Setup for CodeQL users who specify a value other than <code>always</code> for the <code>upload</code> input. <a href="https://redirect.github.com/github/codeql-action/pull/3222">#3222</a></li>
</ul>
<p>See the full <a href="https://github.com/github/codeql-action/blob/v4.31.0/CHANGELOG.md">CHANGELOG.md</a> for more information.</p>
</blockquote>
</details>
<details>
<summary>Changelog</summary>
<p><em>Sourced from <a href="https://github.com/github/codeql-action/blob/main/CHANGELOG.md">github/codeql-action's changelog</a>.</em></p>
<blockquote>
<h1>CodeQL Action Changelog</h1>
<p>See the <a href="https://github.com/github/codeql-action/releases">releases page</a> for the relevant changes to the CodeQL CLI and language packs.</p>
<h2>[UNRELEASED]</h2>
<p>No user facing changes.</p>
<h2>4.31.0 - 24 Oct 2025</h2>
<ul>
<li>Bump minimum CodeQL bundle version to 2.17.6. <a href="https://redirect.github.com/github/codeql-action/pull/3223">#3223</a></li>
<li>When SARIF files are uploaded by the <code>analyze</code> or <code>upload-sarif</code> actions, the CodeQL Action automatically performs post-processing steps to prepare the data for the upload. Previously, these post-processing steps were only performed before an upload took place. We are now changing this so that the post-processing steps will always be performed, even when the SARIF files are not uploaded. This does not change anything for the <code>upload-sarif</code> action. For <code>analyze</code>, this may affect Advanced Setup for CodeQL users who specify a value other than <code>always</code> for the <code>upload</code> input. <a href="https://redirect.github.com/github/codeql-action/pull/3222">#3222</a></li>
</ul>
<h2>4.30.9 - 17 Oct 2025</h2>
<ul>
<li>Update default CodeQL bundle version to 2.23.3. <a href="https://redirect.github.com/github/codeql-action/pull/3205">#3205</a></li>
<li>Experimental: A new <code>setup-codeql</code> action has been added which is similar to <code>init</code>, except it only installs the CodeQL CLI and does not initialize a database. Do not use this in production as it is part of an internal experiment and subject to change at any time. <a href="https://redirect.github.com/github/codeql-action/pull/3204">#3204</a></li>
</ul>
<h2>4.30.8 - 10 Oct 2025</h2>
<p>No user facing changes.</p>
<h2>4.30.7 - 06 Oct 2025</h2>
<ul>
<li>[v4+ only] The CodeQL Action now runs on Node.js v24. <a href="https://redirect.github.com/github/codeql-action/pull/3169">#3169</a></li>
</ul>
<h2>3.30.6 - 02 Oct 2025</h2>
<ul>
<li>Update default CodeQL bundle version to 2.23.2. <a href="https://redirect.github.com/github/codeql-action/pull/3168">#3168</a></li>
</ul>
<h2>3.30.5 - 26 Sep 2025</h2>
<ul>
<li>We fixed a bug that was introduced in <code>3.30.4</code> with <code>upload-sarif</code> which resulted in files without a <code>.sarif</code> extension not getting uploaded. <a href="https://redirect.github.com/github/codeql-action/pull/3160">#3160</a></li>
</ul>
<h2>3.30.4 - 25 Sep 2025</h2>
<ul>
<li>We have improved the CodeQL Action's ability to validate that the workflow it is used in does not use different versions of the CodeQL Action for different workflow steps. Mixing different versions of the CodeQL Action in the same workflow is unsupported and can lead to unpredictable results. A warning will now be emitted from the <code>codeql-action/init</code> step if different versions of the CodeQL Action are detected in the workflow file. Additionally, an error will now be thrown by the other CodeQL Action steps if they load a configuration file that was generated by a different version of the <code>codeql-action/init</code> step. <a href="https://redirect.github.com/github/codeql-action/pull/3099">#3099</a> and <a href="https://redirect.github.com/github/codeql-action/pull/3100">#3100</a></li>
<li>We added support for reducing the size of dependency caches for Java analyses, which will reduce cache usage and speed up workflows. This will be enabled automatically at a later time. <a href="https://redirect.github.com/github/codeql-action/pull/3107">#3107</a></li>
<li>You can now run the latest CodeQL nightly bundle by passing <code>tools: nightly</code> to the <code>init</code> action. In general, the nightly bundle is unstable and we only recommend running it when directed by GitHub staff. <a href="https://redirect.github.com/github/codeql-action/pull/3130">#3130</a></li>
<li>Update default CodeQL bundle version to 2.23.1. <a href="https://redirect.github.com/github/codeql-action/pull/3118">#3118</a></li>
</ul>
<h2>3.30.3 - 10 Sep 2025</h2>
<p>No user facing changes.</p>
<h2>3.30.2 - 09 Sep 2025</h2>
<ul>
<li>Fixed a bug which could cause language autodetection to fail. <a href="https://redirect.github.com/github/codeql-action/pull/3084">#3084</a></li>
<li>Experimental: The <code>quality-queries</code> input that was added in <code>3.29.2</code> as part of an internal experiment is now deprecated and will be removed in an upcoming version of the CodeQL Action. It has been superseded by a new <code>analysis-kinds</code> input, which is part of the same internal experiment. Do not use this in production as it is subject to change at any time. <a href="https://redirect.github.com/github/codeql-action/pull/3064">#3064</a></li>
</ul>
<!-- raw HTML omitted -->
</blockquote>
<p>... (truncated)</p>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a href="4e94bd11f7"><code>4e94bd1</code></a> Merge pull request <a href="https://redirect.github.com/github/codeql-action/issues/3235">#3235</a> from github/update-v4.31.0-1d36546c1</li>
<li><a href="8f11182164"><code>8f11182</code></a> Update changelog for v4.31.0</li>
<li><a href="1d36546c14"><code>1d36546</code></a> Merge pull request <a href="https://redirect.github.com/github/codeql-action/issues/3234">#3234</a> from github/mbg/changelog/post-processing</li>
<li><a href="08ada26e6a"><code>08ada26</code></a> Add changelog entry for post-processing change</li>
<li><a href="b843cbeed0"><code>b843cbe</code></a> Merge pull request <a href="https://redirect.github.com/github/codeql-action/issues/3233">#3233</a> from github/mbg/getOptionalEnvVar</li>
<li><a href="1ecd563919"><code>1ecd563</code></a> Use <code>getOptionalEnvVar</code> in <code>writePostProcessedFiles</code></li>
<li><a href="e576807920"><code>e576807</code></a> Merge pull request <a href="https://redirect.github.com/github/codeql-action/issues/3223">#3223</a> from github/henrymercer/bump-minimum</li>
<li><a href="ad35676669"><code>ad35676</code></a> Add <code>getOptionalEnvVar</code> function</li>
<li><a href="d75645b13f"><code>d75645b</code></a> Merge pull request <a href="https://redirect.github.com/github/codeql-action/issues/3222">#3222</a> from github/mbg/upload-lib/post-process</li>
<li><a href="710606cc35"><code>710606c</code></a> Check that <code>outputPath</code> is non-empty</li>
<li>Additional commits viewable in <a href="16140ae1a1...4e94bd11f7">compare view</a></li>
</ul>
</details>
<br />

[![Dependabot compatibility score](https://dependabot-badges.githubapp.com/badges/compatibility_score?dependency-name=github/codeql-action&package-manager=github_actions&previous-version=4.30.9&new-version=4.31.0)](https://docs.github.com/en/github/managing-security-vulnerabilities/about-dependabot-security-updates#about-compatibility-scores)

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.

[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)

---

<details>
<summary>Dependabot commands and options</summary>
<br />

You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
- `@dependabot show <dependency name> ignore conditions` will show all of the ignore conditions of the specified dependency
- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

</details>
Copybara import of the project:

--
cbe7908eed34d441708d7360f23dad04e5b48ee1 by dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>:

Bump github/codeql-action from 4.30.9 to 4.31.0

Bumps [github/codeql-action](https://github.com/github/codeql-action) from 4.30.9 to 4.31.0.
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](16140ae1a1...4e94bd11f7)

---
updated-dependencies:
- dependency-name: github/codeql-action
  dependency-version: 4.31.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

Merging this change closes #33141

PiperOrigin-RevId: 825449574
2025-10-29 02:52:45 -07:00
Ilia Sergachev
2e53225273 PR #33205: [GPU] Fix reduce-precision simplification.
Imported from GitHub PR https://github.com/openxla/xla/pull/33205

📝 Summary of Changes
The simplification was unintentionally disabled in
2accf052cb.

🎯 Justification
Bug fix.

🚀 Kind of Contribution
🐛 Bug Fix, ️ Performance Improvement

📊 Benchmark (for Performance Improvements)
No

🧪 Unit Tests:
Yes.

🧪 Execution Tests:
No.
Copybara import of the project:

--
2fb682c10ff49212044dd995ba97aa329e52bb71 by Ilia Sergachev <isergachev@nvidia.com>:

[GPU] Fix reduce-precision simplification.

Was unintentionally disabled in
2accf052cb.

Merging this change closes #33205

PiperOrigin-RevId: 825449067
2025-10-29 02:40:39 -07:00