Commit Graph

5312 Commits

Author SHA1 Message Date
Charles Schlosser
8ae3b1aaa5 Fix loongarch unsigned pabsdiff
libeigen/eigen!2071
2025-11-09 19:19:43 +00:00
Rasmus Munk Larsen
23a5482fc0 Misc. packet math cleanups.
libeigen/eigen!2070

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-08 21:57:20 +00:00
Charles Schlosser
8b85f5933a Fix realview
libeigen/eigen!2062
2025-11-08 13:36:43 +00:00
Rasmus Munk Larsen
ffcd7bdbd6 Avoid breaking the build on older compilers.
See merge request libeigen/eigen!2068

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-07 21:25:09 +00:00
Antonio Sánchez
da867c31c9 Fix defines in AVX512 custom TRSM kernel.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Broken at head by !2063.

See merge request libeigen/eigen!2066
2025-11-07 20:14:24 +00:00
Rasmus Munk Larsen
8a9bfb72d7 Rename preduce_half for HVX. 2025-11-07 16:52:07 +00:00
Antonio Sánchez
ed989c7504 Enable generic clang backend tests.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Enable generic clang backend tests.

Added an AVX512 job using the generic clang backend.

Also fixed up some guards in the custom AVX512 gemm/trsm kernels so they don't
start defining things if they aren't used.

See merge request libeigen/eigen!2063
2025-11-07 01:37:12 +00:00
Rasmus Munk Larsen
3368ac6c69 Don't set platform-specific vectorization macros for generic backend.
See merge request libeigen/eigen!2065

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-06 23:12:54 +00:00
Rasmus Munk Larsen
fecfa7f27e Fixes to make generic backend build with AVX512
See merge request libeigen/eigen!2064

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-06 22:50:32 +00:00
Rasmus Munk Larsen
ec93a6d098 Add a generic Eigen backend based on clang vector extensions
The goal of this MR is to implement a generic SIMD backend (packet ops) for Eigen that uses clang vector extensions instead of platform-dependent intrinsics. Ideally, this should make it possible to build Eigen and achieve reasonable speed on any platform that has a recent clang compiler, without having to write any inline assembly or intrinsics.

Caveats:

* The current implementation is a proof of concept and supports vectorization for float, double, int32_t, and int64_t using fixed-size 512-bit vectors (a somewhat arbitrary choice). I have not done much to tune this for speed yet.
* For now, there is no way to enable this other than setting -DEIGEN_VECTORIZE_GENERIC on the command line.
* This only compiles with newer versions of clang. I have tested that it compiles and all tests pass with clang 19.1.7.

https://clang.llvm.org/docs/LanguageExtensions.html#vectors-and-extended-vectors

Closes #2998 and #2997

See merge request libeigen/eigen!2051

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
Co-authored-by: Antonio Sánchez <cantonios@google.com>
2025-11-06 21:52:19 +00:00
Antonio Sánchez
142caf889c Fix MKL enum conversion warning.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Fix MKL enum conversion warning.

### Reference issue
<!--
You can link to a specific issue using the gitlab syntax #<issue number>. 
If the MR fixes an issue, write "Fixes #<issue number>" to have the issue automatically closed on merge.
-->
Fixes #2999.

Closes #2999

See merge request libeigen/eigen!2061
2025-11-05 18:03:22 +00:00
Tyler Veness
06f5cb4878 Use wrapper macro for multidimensional subscript feature test
See merge request libeigen/eigen!2059
2025-11-04 22:26:27 +00:00
Rasmus Munk Larsen
63fc0bc8c1 Make TernarySelectOp immune to const differences.
See merge request libeigen/eigen!2058

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-04 21:42:32 +00:00
Rasmus Munk Larsen
71703a9816 Make assume_aligned a no-op on ARM & ARM64 when msan is used, to work around a missing linker symbol. 2025-11-04 20:26:36 +00:00
Tyler Veness
f95b4698fc Add support for C++23 multidimensional subscript operator
I'm not sure where to put tests for this, assuming they're needed. They also wouldn't run in CI anyway since CI only exercises the C++17 codepaths.

See merge request libeigen/eigen!2053
2025-11-04 07:03:04 +00:00
Rasmus Munk Larsen
b6fcddccfc Get rid of pblend packet op.
There was only a single code path left in TensorEvaluator using pblend. We can replace that with a call to the more general TernarySelectOp and get rid of pblend entirely from Core.

Closes #2998

See merge request libeigen/eigen!2056

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-03 23:27:50 +00:00
Rasmus Munk Larsen
ed9a0e59ba Fix more bugs in !2052
Fixes #2998

Closes #2998

See merge request libeigen/eigen!2057

Co-authored-by: Rasmus Munk Larsen <rmlarsen@google.com>
2025-11-03 20:26:17 +00:00
Rasmus Munk Larsen
a20fc40e4e Revert "simplify squaredNorm"
This causes some subtle alignment-related bugs, which @chuckyschluz is currently investigating.

See merge request libeigen/eigen!2055
2025-11-03 18:59:51 +00:00
Antonio Sánchez
04eb06b354 Fix doc references for nullary expressions.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Fix doc references for nullary expressions.

### Reference issue
<!--
You can link to a specific issue using the gitlab syntax #<issue number>. 
If the MR fixes an issue, write "Fixes #<issue number>" to have the issue automatically closed on merge.
-->

Fixes #2997.

### Additional information
<!--Any additional information you think is important.-->

Closes #2997

See merge request libeigen/eigen!2054
2025-11-03 18:47:04 +00:00
Rasmus Munk Larsen
bfdbc031c2 Fixes #2998. 2025-11-02 23:58:16 +00:00
Rasmus Munk Larsen
8716f109e4 Implement assume_aligned using the standard API
This implements `Eigen::internal::assume_aligned` to match the API for C++20 standard as best as possible using either `std::assume_aligned` or `__builtin_assume_aligned` if available. If neither is available, the function is a no-op.

The override macro `EIGEN_ASSUME_ALIGNED` was changed to a `EIGEN_DONT_ASSUME_ALIGNED`, which now forces the function to be a no-op.

See merge request libeigen/eigen!2052
2025-11-01 12:04:19 +00:00
Rasmus Munk Larsen
ce70a507c0 Enable more generic packet ops for double.
See merge request libeigen/eigen!2050
2025-10-30 19:14:43 +00:00
Charles Schlosser
fb5bb3e98f simplify squaredNorm
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->
My review of https://gitlab.com/libeigen/eigen/-/merge_requests/2048 reminded me there was a much easier and better way of doing this for complex arrays.


### Reference issue
<!--
You can link to a specific issue using the gitlab syntax #<issue number>. 
If the MR fixes an issue, write "Fixes #<issue number>" to have the issue automatically closed on merge.
-->

### Additional information
<!--Any additional information you think is important.-->

See merge request libeigen/eigen!2049
2025-10-30 02:51:15 +00:00
Rasmus Munk Larsen
ece9a4c0b6 Always vectorize abs2() for non-complex types.
For several packet types, `abs2` was not vectorized even if it only requires `pmul`. Get rid of the confusing and redundant `HasAbs2` enum and instead check `HasMul` in addition to making sure the scalar type is not complex.

See merge request libeigen/eigen!2048
2025-10-30 02:42:56 +00:00
Antonio Sánchez
60122df698 Allow user to configure if free is allowed at runtime.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Allow user to configure if `free` is allowed at runtime.

Reverts to Eigen 3.4 behavior by default, where `free(...)` is allowed if `EIGEN_RUNTIME_NO_MALLOC` is defined but `set_is_malloc_allowed(true)`.  Adds a separate `set_is_free_allowed(...)` to explicitly control use of `std::free`.

### Reference issue
<!--
You can link to a specific issue using the gitlab syntax #<issue number>. 
If the MR fixes an issue, write "Fixes #<issue number>" to have the issue automatically closed on merge.
-->

Fixes #2983.

### Additional information
<!--Any additional information you think is important.-->

Closes #2983

See merge request libeigen/eigen!2047
2025-10-28 22:29:00 +00:00
Antonio Sánchez
8e60d4173c Support AVX for i686.
<!-- 
Thanks for contributing a merge request!

We recommend that first-time contributors read our [contribution guidelines](https://eigen.tuxfamily.org/index.php?title=Contributing_to_Eigen).

Before submitting the MR, please complete the following checks:
- Create one PR per feature or bugfix,
- Run the test suite to verify your changes.
  See our [test guidelines](https://eigen.tuxfamily.org/index.php?title=Tests).
- Add tests to cover the bug addressed or any new feature.
- Document new features.  If it is a substantial change, add it to the [Changelog](https://gitlab.com/libeigen/eigen/-/blob/master/CHANGELOG.md).
- Leave the following box checked when submitting: `Allow commits from members who can merge to the target branch`.
  This allows us to rebase and merge your change.

Note that we are a team of volunteers; we appreciate your patience during the review process.
-->

### Description
<!--Please explain your changes.-->

Support AVX for i686.

There was an existing work-around for windows.  Added the more generic
architecture comparison to also apply for linux.

### Reference issue
<!--
You can link to a specific issue using the gitlab syntax #<issue number>. 
If the MR fixes an issue, write "Fixes #<issue number>" to have the issue automatically closed on merge.
-->

Fixes #2991.

Closes #2991

See merge request libeigen/eigen!2037
2025-10-20 21:47:42 +00:00
Rasmus Munk Larsen
565db1c603 Optimize ApplyOnTheRight for Jacobi rotations when FMA is available 2025-10-12 00:04:19 +00:00
Charlie Schlosser
cd4f989f8f assume_aligned uses bytes not bits 2025-10-10 14:08:10 -04:00
Antonio Sánchez
ef3c5c1d1d Add workaround for using std::fma for scalar multiply-add. 2025-10-09 18:57:46 +00:00
Charles Schlosser
5996176b88 Fix alignment bug in avx pcast<Packet4l, Packet4d> 2025-10-09 02:50:42 +00:00
Laurenz
4bd382df56 Fix SSE PacketMath Compilation Error on QNX 2025-10-08 17:13:16 +00:00
Charles Schlosser
13bd14974d fix errors in windows builds and tests 2025-10-07 22:47:35 +00:00
Jeremy Nimmer
eea6587b0e Fix scalar_inner_product_op when binary ops return a different type 2025-10-05 22:51:50 +00:00
Sergiu Deitsch
32b0f386bc Eliminate possible -Wstringop-overflow warning in .setZero() 2025-10-04 00:03:03 +02:00
Antonio Sánchez
4916887f2c Update geo_homogeneous test, add eval() to PermutationMatrix. 2025-10-01 18:01:11 +00:00
Sergiu Deitsch
8d7ebac6ec Disambiguate multiplication of a permutation matrix and a homogeneous vector 2025-09-27 14:05:28 +02:00
Charles Schlosser
dbd25f632b Fix select: return typed comparisons if vectorized 2025-09-24 05:38:12 +00:00
Sergiu Deitsch
4df215785b Support matrix multiplication of homogeneous row vectors 2025-09-23 14:56:28 +00:00
Rasmus Munk Larsen
2d170aea11 Define pcmp_le generically in terms of pcmp_eq and pcmp_lt. 2025-09-23 14:34:57 +00:00
Sergiu Deitsch
ea869e183b Add missing bool SSE2 PacketMath comparison 2025-09-22 21:28:45 +02:00
Sergiu Deitsch
14477c5d43 Replace deprecated std::is_trivial by an internal definition 2025-09-22 16:59:10 +00:00
Sergiu Deitsch
62fbd276e0 Provide hints for deprecated functionality 2025-09-22 16:00:42 +00:00
Antonio Sánchez
a627f72cd6 Add "Version" file and update version. 2025-09-20 02:08:59 +00:00
Antonio Sanchez
f426eff949 Add inline/device-function attributes to fma. 2025-09-02 22:51:35 +00:00
Antonio Sánchez
8a8fbc8f5e Don't enable AVX for wasm. 2025-08-29 21:50:25 +00:00
Antonio Sánchez
7f0cb638c5 Specialize numext::madd for half/bfloat16. 2025-08-29 18:11:25 +00:00
Antonio Sánchez
1e9d7ed7d3 Add missing semicolon to has_fma definitions to fix GPU builds. 2025-08-29 17:19:28 +00:00
Antonio Sánchez
2e8cc042a1 Replace calls to numext::fma with numext:madd. 2025-08-28 21:40:19 +00:00
Charles Schlosser
38b51d5b7e Mitigate setConstant regression with custom scalars 2025-08-26 20:04:17 +00:00
Antonio Sanchez
d2a70fe4e2 Make permutation products aliasing by default. 2025-08-25 18:39:06 +00:00