[X86] getTargetVShiftNode - remove shift-by-constant handling.
authorSimon Pilgrim <llvm-dev@redking.me.uk>
Tue, 21 Dec 2021 13:16:41 +0000 (13:16 +0000)
committerSimon Pilgrim <llvm-dev@redking.me.uk>
Tue, 21 Dec 2021 13:16:48 +0000 (13:16 +0000)
commitdfa2ad1ad858a44d7d73de1b042fbe78256b74f7
treedc0792d33845a70d1cf5c64e003ff518fd75c5ad
parente751d97863fb48b7dd844e48c0ba564f6970b726
[X86] getTargetVShiftNode - remove shift-by-constant handling.

Move shift-by-constant handling and move it into its only user (VSHIFT intrinsics lowering).

This is some prep-work for getTargetVShiftNode to no longer take a scalar shift amount - we're introducing temporary ISD::EXTRACT_VECTOR_ELT nodes via SelectionDAG::getSplatValue to accommodate this which can cause various issues, including unnecessary scalarization and xmm->gpr->xmm transfers, and causes problems for 32-bit codegen if we fail to remove an (illegal) i64 scalar extracted from a (legal) vXi64 vector.
llvm/lib/Target/X86/X86ISelLowering.cpp