This adds support for sext in foldLogicCastConstant. This is a prerequisite for D36214.
Differential Revision: https://reviews.llvm.org/D36234
llvm-svn: 309880
Type *DestTy = Logic.getType();
Type *SrcTy = Cast->getSrcTy();
- // Move the logic operation ahead of a zext if the constant is unchanged in
- // the smaller source type. Performing the logic in a smaller type may provide
- // more information to later folds, and the smaller logic instruction may be
- // cheaper (particularly in the case of vectors).
+ // Move the logic operation ahead of a zext or sext if the constant is
+ // unchanged in the smaller source type. Performing the logic in a smaller
+ // type may provide more information to later folds, and the smaller logic
+ // instruction may be cheaper (particularly in the case of vectors).
Value *X;
if (match(Cast, m_OneUse(m_ZExt(m_Value(X))))) {
Constant *TruncC = ConstantExpr::getTrunc(C, SrcTy);
}
}
+ if (match(Cast, m_OneUse(m_SExt(m_Value(X))))) {
+ Constant *TruncC = ConstantExpr::getTrunc(C, SrcTy);
+ Constant *SextTruncC = ConstantExpr::getSExt(TruncC, DestTy);
+ if (SextTruncC == C) {
+ // LogicOpc (sext X), C --> sext (LogicOpc X, C)
+ Value *NewOp = Builder.CreateBinOp(LogicOpc, X, TruncC);
+ return new SExtInst(NewOp, DestTy);
+ }
+ }
+
return nullptr;
}
define i64 @test47(i8 %A) {
; CHECK-LABEL: @test47(
-; CHECK-NEXT: [[B:%.*]] = sext i8 %A to i64
-; CHECK-NEXT: [[C:%.*]] = and i64 [[B]], 4294967253
-; CHECK-NEXT: [[E:%.*]] = or i64 [[C]], 42
+; CHECK-NEXT: [[TMP1:%.*]] = or i8 [[A:%.*]], 42
+; CHECK-NEXT: [[C:%.*]] = sext i8 [[TMP1]] to i64
+; CHECK-NEXT: [[E:%.*]] = and i64 [[C]], 4294967295
; CHECK-NEXT: ret i64 [[E]]
;
%B = sext i8 %A to i32
define <4 x i32> @vec_sel_xor_multi_use(<4 x i32> %a, <4 x i32> %b, <4 x i1> %c) {
; CHECK-LABEL: @vec_sel_xor_multi_use(
-; CHECK-NEXT: [[MASK:%.*]] = sext <4 x i1> %c to <4 x i32>
-; CHECK-NEXT: [[MASK_FLIP1:%.*]] = xor <4 x i32> [[MASK]], <i32 -1, i32 0, i32 0, i32 0>
-; CHECK-NEXT: [[TMP1:%.*]] = xor <4 x i1> %c, <i1 false, i1 true, i1 true, i1 true>
-; CHECK-NEXT: [[TMP2:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> %a, <4 x i32> %b
-; CHECK-NEXT: [[ADD:%.*]] = add <4 x i32> [[TMP2]], [[MASK_FLIP1]]
+; CHECK-NEXT: [[TMP1:%.*]] = xor <4 x i1> [[C:%.*]], <i1 true, i1 false, i1 false, i1 false>
+; CHECK-NEXT: [[MASK_FLIP1:%.*]] = sext <4 x i1> [[TMP1]] to <4 x i32>
+; CHECK-NEXT: [[TMP2:%.*]] = xor <4 x i1> [[C]], <i1 false, i1 true, i1 true, i1 true>
+; CHECK-NEXT: [[TMP3:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]]
+; CHECK-NEXT: [[ADD:%.*]] = add <4 x i32> [[TMP3]], [[MASK_FLIP1]]
; CHECK-NEXT: ret <4 x i32> [[ADD]]
;
%mask = sext <4 x i1> %c to <4 x i32>
; optimization, we make the proposed smaller type (i8) larger (i16) to
; ensure correctness.
;
-; CHECK: %[[S0:.+]] = sext <2 x i8> {{.*}} to <2 x i16>
-; CHECK: %[[OR:.+]] = or <2 x i16> %[[S0]], <i16 1, i16 1>
-; CHECK: %[[E0:.+]] = extractelement <2 x i16> %[[OR]], i32 0
+; CHECK: %[[OR:.+]] = or <2 x i8> {{.*}}, <i8 1, i8 1>
+; CHECK: %[[S0:.+]] = sext <2 x i8> %[[OR]] to <2 x i16>
+; CHECK: %[[E0:.+]] = extractelement <2 x i16> %[[S0]], i32 0
; CHECK: %[[S1:.+]] = sext i16 %[[E0]] to i64
; CHECK: getelementptr inbounds i8, i8* %ptr, i64 %[[S1]]
-; CHECK: %[[E1:.+]] = extractelement <2 x i16> %[[OR]], i32 1
+; CHECK: %[[E1:.+]] = extractelement <2 x i16> %[[S0]], i32 1
; CHECK: %[[S2:.+]] = sext i16 %[[E1]] to i64
; CHECK: getelementptr inbounds i8, i8* %ptr, i64 %[[S2]]
;