projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 8c8aae8) | patch
[CodeGen] Use SPLAT_VECTOR for zeroinitialiser with scalable types
authorDavid Sherwood <david.sherwood@arm.com>
Mon, 27 Apr 2020 14:25:23 +0000 (15:25 +0100)
committerDavid Sherwood <david.sherwood@arm.com>
Mon, 27 Apr 2020 14:57:59 +0000 (15:57 +0100)
commit096b25a8d8ed8a5162dfead59ed9cd9c2b4825aa
treea6a14ad4e8fed8987ac851fe163881c57f2a27f3tree | snapshot
parent8c8aae852b5e60929156054808af941fc0745d46commit | diff
[CodeGen] Use SPLAT_VECTOR for zeroinitialiser with scalable types

Summary:
When generating code for the LLVM IR zeroinitialiser operation, if
the vector type is scalable we should be using SPLAT_VECTOR instead
of BUILD_VECTOR.

Differential Revision: https://reviews.llvm.org/D78636
llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp diff | blob | history
llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp diff | blob | history
Domain: System / Toolchain;