ARM: don't mangle DAG constant if it has more than one use

The basic optimisation was to convert (mul $LHS, $complex_constant) into
roughly "(shl (mul $LHS, $simple_constant), $simple_amt)" when it was expected
to be cheaper. The original logic checks that the mul only has one use (since
we're mangling $complex_constant), but when used in even more complex
addressing modes there may be an outer addition that can pick up the wrong
value too.

I *think* the ARM addressing-mode problem is actually unreachable at the
moment, but that depends on complex assessments of the profitability of
pre-increment addressing modes so I've put a real check in there instead of an
assertion.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@259228 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Tim Northover 2016-01-29 19:18:46 +00:00
parent a42f385f92
commit b226159d1c
2 changed files with 19 additions and 2 deletions

View File

@ -747,7 +747,7 @@ bool ARMDAGToDAGISel::SelectLdStSOReg(SDValue N, SDValue &Base, SDValue &Offset,
// If Offset is a multiply-by-constant and it's profitable to extract a shift
// and use it in a shifted operand do so.
if (Offset.getOpcode() == ISD::MUL) {
if (Offset.getOpcode() == ISD::MUL && N.hasOneUse()) {
unsigned PowerOfTwo = 0;
SDValue NewMulConst;
if (canExtractShiftFromMul(Offset, 31, PowerOfTwo, NewMulConst)) {
@ -1422,7 +1422,7 @@ bool ARMDAGToDAGISel::SelectT2AddrModeSoReg(SDValue N,
// If OffReg is a multiply-by-constant and it's profitable to extract a shift
// and use it in a shifted operand do so.
if (OffReg.getOpcode() == ISD::MUL) {
if (OffReg.getOpcode() == ISD::MUL && N.hasOneUse()) {
unsigned PowerOfTwo = 0;
SDValue NewMulConst;
if (canExtractShiftFromMul(OffReg, 3, PowerOfTwo, NewMulConst)) {

View File

@ -239,3 +239,20 @@ define void @test_well_formed_dag(i32 %in1, i32 %in2, i32* %addr) {
store i32 %add, i32* %addr
ret void
}
define { i32, i32 } @test_multi_use_add(i32 %base, i32 %offset) {
; CHECK-LABEL: test_multi_use_add:
; CHECK-THUMB: movs [[CONST:r[0-9]+]], #28
; CHECK-THUMB: movt [[CONST]], #1
%prod = mul i32 %offset, 65564
%sum = add i32 %base, %prod
%ptr = inttoptr i32 %sum to i32*
%loaded = load i32, i32* %ptr
%ret.tmp = insertvalue { i32, i32 } undef, i32 %sum, 0
%ret = insertvalue { i32, i32 } %ret.tmp, i32 %loaded, 1
ret { i32, i32 } %ret
}