[ARM] Combine fadd into fcmla
authorDavid Green <david.green@arm.com>
Wed, 5 Apr 2023 09:31:19 +0000 (10:31 +0100)
committerDavid Green <david.green@arm.com>
Wed, 5 Apr 2023 09:31:19 +0000 (10:31 +0100)
commitb4df2b2c6c75f02e2c26c3c947718816ca0be56a
treef6b4a0550b1843bffe55f678ea6103d64cc6ca05
parenta5e1a93ea10ffc06a32df6d9f410e9b297ed136d
[ARM] Combine fadd into fcmla

This is the MVE equivalent of https://reviews.llvm.org/D146407. It adds a
target combine for fadd(a, vcmla(b, c, d)) -> vcmla(fadd(a, b), c, d), pushing
the fadd into the operands of the fcmla, which can help simplify away some
additions.

Differential Revision: https://reviews.llvm.org/D147200
llvm/lib/Target/ARM/ARMISelLowering.cpp
llvm/test/CodeGen/Thumb2/mve-complex-deinterleaving-mixed-cases.ll
llvm/test/CodeGen/Thumb2/mve-vcmla.ll