diff options
author | David Sherwood <david.sherwood@arm.com> | 2025-02-10 16:04:43 +0000 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-02-10 16:04:43 +0000 |
commit | 0010a3c97ef4df11aa50b381ea801c9ba8dd516f (patch) | |
tree | f28a10f85878f38ff6a56313c062eadbf83f889a /clang/lib/CodeGen/CodeGenModule.cpp | |
parent | 1c583c19bb7914a2686e245b7e1d14f82fe454eb (diff) | |
download | llvm-0010a3c97ef4df11aa50b381ea801c9ba8dd516f.zip llvm-0010a3c97ef4df11aa50b381ea801c9ba8dd516f.tar.gz llvm-0010a3c97ef4df11aa50b381ea801c9ba8dd516f.tar.bz2 |
[NFC][LoopVectorize] Add more partial reduction tests (#126525)
* Adds variants of dotp (dotp_i8_to_i64_has_neon_dotprod,
dotp_i16_to_i64_has_neon_dotprod) that show how the loop
vectoriser has generated fixed-width partial reductions
without any matching NEON udot instruction.
* Adds loops that could also benefit from partial
reductions once the work is done to recognise patterns
such as
%zext = zext i8 %load to i32
%acc.next = add i32 %acc, %zext
See zext_add_reduc_i8_i32, etc. I intend to follow up with
a patch to add support for vectorising such patterns.
Diffstat (limited to 'clang/lib/CodeGen/CodeGenModule.cpp')
0 files changed, 0 insertions, 0 deletions