[AArch64][SVE] Generate smull/umull instead of sve v2i64 mul
authorDavid Green <david.green@arm.com>
Wed, 26 Apr 2023 21:12:00 +0000 (22:12 +0100)
committerDavid Green <david.green@arm.com>
Wed, 26 Apr 2023 21:12:00 +0000 (22:12 +0100)
commitd340ef697d905f81ede9747cb64160177bc0c53c
tree567c304c0e23107a8388631cabda9ba64a3c5f8a
parent79e76211df9a6f9f583184f1c54e3a085ff1a6a1
[AArch64][SVE] Generate smull/umull instead of sve v2i64 mul

A neon smull/umull should be preferred over a sve v2i64 mul with two extends.
It will be both less instructions and a lower cost multiply instruction.

Differential Revision: https://reviews.llvm.org/D148248
llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
llvm/test/CodeGen/AArch64/aarch64-smull.ll