Skip to content

Commit 4ae8008

Browse files
farzonlfarzonl
committed
Revert "Adjust bit cast instruction filter for DXIL Prepare pass (llvm#142678)"
This reverts commit 1d6e8ec.
1 parent 16b0d2f commit 4ae8008

File tree

3 files changed

+40
-121
lines changed

3 files changed

+40
-121
lines changed

‎llvm/lib/Target/DirectX/DXILPrepare.cpp

Lines changed: 2 additions & 42 deletions
Original file line numberDiff line numberDiff line change
@@ -148,49 +148,9 @@ class DXILPrepareModule : public ModulePass {
148148
Type *Ty) {
149149
// Omit bitcasts if the incoming value matches the instruction type.
150150
auto It = PointerTypes.find(Operand);
151-
if (It != PointerTypes.end()) {
152-
auto *OpTy = cast<TypedPointerType>(It->second)->getElementType();
153-
if (OpTy == Ty)
151+
if (It != PointerTypes.end())
152+
if (cast<TypedPointerType>(It->second)->getElementType() == Ty)
154153
return nullptr;
155-
}
156-
157-
Type *ValTy = Operand->getType();
158-
// Also omit the bitcast for matching global array types
159-
if (auto *GlobalVar = dyn_cast<GlobalVariable>(Operand))
160-
ValTy = GlobalVar->getValueType();
161-
162-
if (auto *AI = dyn_cast<AllocaInst>(Operand))
163-
ValTy = AI->getAllocatedType();
164-
165-
if (auto *ArrTy = dyn_cast<ArrayType>(ValTy)) {
166-
Type *ElTy = ArrTy->getElementType();
167-
if (ElTy == Ty)
168-
return nullptr;
169-
}
170-
171-
// finally, drill down GEP instructions until we get the array
172-
// that is being accessed, and compare element types
173-
if (ConstantExpr *GEPInstr = dyn_cast<ConstantExpr>(Operand)) {
174-
while (GEPInstr->getOpcode() == Instruction::GetElementPtr) {
175-
Value *OpArg = GEPInstr->getOperand(0);
176-
if (ConstantExpr *NewGEPInstr = dyn_cast<ConstantExpr>(OpArg)) {
177-
GEPInstr = NewGEPInstr;
178-
continue;
179-
}
180-
181-
if (auto *GlobalVar = dyn_cast<GlobalVariable>(OpArg))
182-
ValTy = GlobalVar->getValueType();
183-
if (auto *AI = dyn_cast<AllocaInst>(Operand))
184-
ValTy = AI->getAllocatedType();
185-
if (auto *ArrTy = dyn_cast<ArrayType>(ValTy)) {
186-
Type *ElTy = ArrTy->getElementType();
187-
if (ElTy == Ty)
188-
return nullptr;
189-
}
190-
break;
191-
}
192-
}
193-
194154
// Insert bitcasts where we are removing the instruction.
195155
Builder.SetInsertPoint(&Inst);
196156
// This code only gets hit in opaque-pointer mode, so the type of the

‎llvm/test/CodeGen/DirectX/llc-vector-load-scalarize.ll

Lines changed: 38 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -60,15 +60,19 @@ define <4 x i32> @load_array_vec_test() #0 {
6060
define <4 x i32> @load_vec_test() #0 {
6161
; CHECK-LABEL: define <4 x i32> @load_vec_test(
6262
; CHECK-SAME: ) #[[ATTR0]] {
63-
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(3) @vecData.scalarized, align 4
64-
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 1), align 4
65-
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 2), align 4
66-
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 3), align 4
67-
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0
68-
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[TMP2]], i32 1
69-
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[TMP3]], i32 2
70-
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> [[DOTUPTO2]], i32 [[TMP4]], i32 3
71-
; CHECK-NEXT: ret <4 x i32> [[TMP5]]
63+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr addrspace(3) @vecData.scalarized to ptr addrspace(3)
64+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) [[TMP1]], align 4
65+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 1) to ptr addrspace(3)
66+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) [[TMP3]], align 4
67+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 2) to ptr addrspace(3)
68+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) [[TMP5]], align 4
69+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 3) to ptr addrspace(3)
70+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) [[TMP7]], align 4
71+
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i32 0
72+
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[TMP4]], i32 1
73+
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[TMP6]], i32 2
74+
; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x i32> [[DOTUPTO2]], i32 [[TMP8]], i32 3
75+
; CHECK-NEXT: ret <4 x i32> [[TMP9]]
7276
;
7377
%1 = load <4 x i32>, <4 x i32> addrspace(3)* @"vecData", align 4
7478
ret <4 x i32> %1
@@ -99,23 +103,31 @@ define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
99103
define <4 x i32> @multid_load_test() #0 {
100104
; CHECK-LABEL: define <4 x i32> @multid_load_test(
101105
; CHECK-SAME: ) #[[ATTR0]] {
102-
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, align 4
103-
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 1), align 4
104-
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 2), align 4
105-
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 3), align 4
106-
; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), align 4
107-
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 1), align 4
108-
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 2), align 4
109-
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 3), align 4
110-
; CHECK-NEXT: [[DOTI08:%.*]] = add i32 [[TMP1]], [[TMP5]]
111-
; CHECK-NEXT: [[DOTI19:%.*]] = add i32 [[TMP2]], [[DOTI13]]
112-
; CHECK-NEXT: [[DOTI210:%.*]] = add i32 [[TMP3]], [[DOTI25]]
113-
; CHECK-NEXT: [[DOTI311:%.*]] = add i32 [[TMP4]], [[DOTI37]]
114-
; CHECK-NEXT: [[DOTUPTO015:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI08]], i32 0
115-
; CHECK-NEXT: [[DOTUPTO116:%.*]] = insertelement <4 x i32> [[DOTUPTO015]], i32 [[DOTI19]], i32 1
116-
; CHECK-NEXT: [[DOTUPTO217:%.*]] = insertelement <4 x i32> [[DOTUPTO116]], i32 [[DOTI210]], i32 2
117-
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> [[DOTUPTO217]], i32 [[DOTI311]], i32 3
118-
; CHECK-NEXT: ret <4 x i32> [[TMP6]]
106+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim to ptr addrspace(3)
107+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) [[TMP1]], align 4
108+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 1) to ptr addrspace(3)
109+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) [[TMP3]], align 4
110+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 2) to ptr addrspace(3)
111+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) [[TMP5]], align 4
112+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 3) to ptr addrspace(3)
113+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) [[TMP7]], align 4
114+
; CHECK-NEXT: [[TMP9:%.*]] = bitcast ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1) to ptr addrspace(3)
115+
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr addrspace(3) [[TMP9]], align 4
116+
; CHECK-NEXT: [[TMP11:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 1) to ptr addrspace(3)
117+
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) [[TMP11]], align 4
118+
; CHECK-NEXT: [[TMP13:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 2) to ptr addrspace(3)
119+
; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr addrspace(3) [[TMP13]], align 4
120+
; CHECK-NEXT: [[TMP15:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), i32 3) to ptr addrspace(3)
121+
; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr addrspace(3) [[TMP15]], align 4
122+
; CHECK-NEXT: [[DOTI05:%.*]] = add i32 [[TMP2]], [[TMP10]]
123+
; CHECK-NEXT: [[DOTI16:%.*]] = add i32 [[TMP4]], [[TMP12]]
124+
; CHECK-NEXT: [[DOTI27:%.*]] = add i32 [[TMP6]], [[TMP14]]
125+
; CHECK-NEXT: [[DOTI38:%.*]] = add i32 [[TMP8]], [[TMP16]]
126+
; CHECK-NEXT: [[DOTUPTO01215:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI05]], i32 0
127+
; CHECK-NEXT: [[DOTUPTO11316:%.*]] = insertelement <4 x i32> [[DOTUPTO01215]], i32 [[DOTI16]], i32 1
128+
; CHECK-NEXT: [[DOTUPTO21417:%.*]] = insertelement <4 x i32> [[DOTUPTO11316]], i32 [[DOTI27]], i32 2
129+
; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> [[DOTUPTO21417]], i32 [[DOTI38]], i32 3
130+
; CHECK-NEXT: ret <4 x i32> [[TMP17]]
119131
;
120132
%1 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 0, i32 0), align 4
121133
%2 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 1, i32 1), align 4

‎llvm/test/CodeGen/DirectX/noop_bitcast_global_array_type.ll

Lines changed: 0 additions & 53 deletions
This file was deleted.

0 commit comments

Comments
 (0)