maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[FastISel][AArch64] Use the proper FMOV instruction to materialize a +0.0. 

Use FMOVWSr/FMOVXDr instead of FMOVSr/FMOVDr, which have the proper register
class to be used with the zero register. This makes the MachineInstruction
verifier happy again.

This is related to <rdar://problem/18027157>.

llvm-svn: 216040
This commit is contained in:
Juergen Ributzka 2014-08-20 01:10:36 +00:00
parent a0b988cb63
commit 0781b860e4
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llvm/lib/Target/AArch64/AArch64FastISel.cpp
View File

@ -291,7 +291,7 @@ unsigned AArch64FastISel::AArch64MaterializeFP(const ConstantFP *CFP, MVT VT) {
// register, because the immediate version of fmov cannot encode zero.
if (Val.isPosZero()) {
unsigned ZReg = Is64Bit ? AArch64::XZR : AArch64::WZR;
unsigned Opc = Is64Bit ? AArch64::FMOVDr : AArch64::FMOVSr;
unsigned Opc = Is64Bit ? AArch64::FMOVXDr : AArch64::FMOVWSr;
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
.addReg(ZReg, getKillRegState(true));
return ResultReg;

2
llvm/test/CodeGen/AArch64/arm64-fast-isel-materialize.ll
View File

@ -1,4 +1,4 @@
; RUN: llc -O0 -fast-isel-abort -mtriple=arm64-apple-darwin < %s | FileCheck %s
; RUN: llc -O0 -fast-isel-abort -verify-machineinstrs -mtriple=arm64-apple-darwin < %s | FileCheck %s
; Materialize using fmov
define float @fmov_float1() {