This is an archive of the discontinued LLVM Phabricator instance.

Differential D75161

[ReleaseNotes] Mention new matrix intrinsics.
ClosedPublic

Authored by fhahn on Feb 26 2020, 3:05 AM.

Details

Reviewers
anemet
Gerolf
Commits
rG456e9c2e14bc: [ReleaseNotes] Mention new matrix intrinsics.

Diff Detail

Event Timeline

fhahn created this revision.Feb 26 2020, 3:05 AM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 26 2020, 3:05 AM
Harbormaster completed remote builds in B47281: Diff 246657.Feb 26 2020, 4:25 AM
anemet accepted this revision.Feb 26 2020, 8:27 AM

LGTM

This revision is now accepted and ready to land.Feb 26 2020, 8:27 AM
fhahn closed this revision.Feb 28 2020, 1:44 PM

Landed as 456e9c2e14bc on the 10.x release branch

Revision Contents

PathSize
llvm/
docs/
3 lines
6 lines

Diff 246657

llvm/docs/LangRef.rst

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 14,547 Lines • ▼ Show 20 Lines
If the intrinsic call has the ``nnan`` fast-math flag then the operation can If the intrinsic call has the ``nnan`` fast-math flag then the operation can
assume that NaNs are not present in the input vector. assume that NaNs are not present in the input vector.
Arguments: Arguments:
"""""""""" """"""""""
The argument to this intrinsic must be a vector of floating-point values. The argument to this intrinsic must be a vector of floating-point values.
.. _i_matrixintrinsics:
Matrix Intrinsics Matrix Intrinsics
----------------- -----------------
Operations on matrixes requiring shape information (like number of rows/columns Operations on matrixes requiring shape information (like number of rows/columns
or the memory layout) can be expressed using the matrix intrinsics. Matrixes are or the memory layout) can be expressed using the matrix intrinsics. Matrixes are
embedded in a flat vector and the intrinsics take the dimensions as arguments. embedded in a flat vector and the intrinsics take the dimensions as arguments.
Currently column-major layout is assumed. The intrinsics support both integer Currently column-major layout is assumed. The intrinsics support both integer
and floating point matrixes. and floating point matrixes.
▲ Show 20 Lines • Show All 3,915 Lines • Show Last 20 Lines

llvm/docs/ReleaseNotes.rst

Show First 20 Lines • Show All 100 Lines • ▼ Show 20 Lines
* The inter-procedural analysis and optimization capabilities in the Attributor * The inter-procedural analysis and optimization capabilities in the Attributor
framework and pass have been substantially advanced (initial commit framework and pass have been substantially advanced (initial commit
`D59918 <https://reviews.llvm.org/D59918>`_, `LLVM-Dev talk <https://youtu.be/CzWkc_JcfS0>`_). `D59918 <https://reviews.llvm.org/D59918>`_, `LLVM-Dev talk <https://youtu.be/CzWkc_JcfS0>`_).
In this release, 19 different attributes are inferred, including 12 LLVM IR In this release, 19 different attributes are inferred, including 12 LLVM IR
attributes and 7 "abstract" attributes, such as liveness. The Attributor is attributes and 7 "abstract" attributes, such as liveness. The Attributor is
still under heavy development and disabled by default, to enable an early run still under heavy development and disabled by default, to enable an early run
pass ``-mllvm -attributor-disable=false`` to an invocation of clang. pass ``-mllvm -attributor-disable=false`` to an invocation of clang.
* New matrix math intrinsics have been added to LLVM
(see :ref:`LLVM Language Reference Manual <i_matrixintrinsics>`), together
with the LowerMatrixIntrinsics pass. The pass lowers matrix intrinsics
to a set of efficient vector instructions. The lowering pass is off
by default and can be enabled by passing ``-mllvm -enable-matrix`` to an
invocation of clang.
Changes to the LLVM IR Changes to the LLVM IR
---------------------- ----------------------
* Unnamed function arguments now get printed with their automatically * Unnamed function arguments now get printed with their automatically
generated name (e.g. "i32 %0") in definitions. This may require front-ends generated name (e.g. "i32 %0") in definitions. This may require front-ends
to update their tests; if so there is a script utils/add_argument_names.py to update their tests; if so there is a script utils/add_argument_names.py
▲ Show 20 Lines • Show All 201 Lines • Show Last 20 Lines