[PowerPC][LoopVectorize] Extend getRegisterClassForType to consider double and other...
authorJinsong Ji <jji@us.ibm.com>
Fri, 3 Jan 2020 20:25:19 +0000 (20:25 +0000)
committerJinsong Ji <jji@us.ibm.com>
Mon, 6 Jan 2020 18:44:59 +0000 (18:44 +0000)
commite29a2e6be4e114b4233a2e0feedb74b2f34cf782
treef8fd479d04b670935e2420fa700db92ccb81857a
parent7ae3d335467a24faa80ebd9b31446c649570ca0c
[PowerPC][LoopVectorize] Extend getRegisterClassForType to consider double and other floating point type

In https://reviews.llvm.org/D67148, we use isFloatTy to test floating
point type, otherwise we return GPRRC.
So 'double' will be classified as GPRRC, which is not accurate.

This patch covers other floating point types.

Reviewed By: #powerpc, nemanjai

Differential Revision: https://reviews.llvm.org/D71946
llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
llvm/test/Transforms/LoopVectorize/PowerPC/reg-usage.ll