Skip to content

Commit e37882e

Browse files
againullagainull
authored
[SYCL] Handle address space casts in the LowerWGScope (#3120)
1 parent 2b521f3 commit e37882e

File tree

3 files changed

+124
-32
lines changed

3 files changed

+124
-32
lines changed

‎llvm/lib/SYCLLowerIR/LowerWGScope.cpp‎

Lines changed: 14 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -237,6 +237,8 @@ static bool mayHaveSideEffects(const Instruction *I) {
237237
case Instruction::Call:
238238
assert(!isPFWICall(I) && "pfwi must have been handled separately");
239239
return true;
240+
case Instruction::AddrSpaceCast:
241+
return false;
240242
default:
241243
return true;
242244
}
@@ -630,6 +632,10 @@ static void fixupPrivateMemoryPFWILambdaCaptures(CallInst *PFWICall) {
630632
// whether it is an alloca with "work_item_scope"
631633
SmallVector<CaptureDesc, 4> PrivMemCaptures;
632634

635+
// Look through cast
636+
if (auto *Cast = dyn_cast<AddrSpaceCastInst>(LambdaObj))
637+
LambdaObj = Cast->getOperand(0);
638+
633639
for (auto *U : LambdaObj->users()) {
634640
GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(U);
635641

@@ -773,19 +779,21 @@ PreservedAnalyses SYCLLowerWGScopePass::run(Function &F, const llvm::Triple &TT,
773779
Instruction *First = nullptr;
774780
Instruction *Last = nullptr;
775781

776-
// Skip PHIs and allocas, as they don't have side effects and must never be
777-
// guarded with the WG leader test. Note that there should be no allocas in
778-
// local address space at this point - they must have been converted to
779-
// globals.
782+
// Skip PHIs, allocas and addrspacecasts associated with allocas, as they
783+
// don't have side effects and must never be guarded with the WG leader
784+
// test. Note that there should be no allocas in local address space at this
785+
// point - they must have been converted to globals.
780786
Instruction *I = BB.getFirstNonPHI();
781787

782-
for (; I->getOpcode() == Instruction::Alloca; I = I->getNextNode()) {
788+
for (; I->getOpcode() == Instruction::Alloca ||
789+
I->getOpcode() == Instruction::AddrSpaceCast;
790+
I = I->getNextNode()) {
783791
auto *AllocaI = dyn_cast<AllocaInst>(I);
784792
// Allocas marked with "work_item_scope" are those originating from
785793
// cl::sycl::private_memory<T> variables, which must be in private memory.
786794
// No shadows/materialization is needed for them because they can be
787795
// updated only within PFWIs
788-
if (!AllocaI->getMetadata(WI_SCOPE_MD))
796+
if (AllocaI && !AllocaI->getMetadata(WI_SCOPE_MD))
789797
Allocas.insert(AllocaI);
790798
}
791799
for (; I && (I != BB.getTerminator()); I = I->getNextNode()) {

‎llvm/test/SYCLLowerIR/addrspacecast_handling.ll‎

Lines changed: 87 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,87 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2+
; RUN: opt < %s -LowerWGScope -S | FileCheck %s
3+
4+
; Test checks that LowerWGScope pass can deal with addrpacecasts
5+
; associated with allocas and look through addrspacecasts when handling
6+
; private_memory<T> captures.
7+
8+
%struct.ham = type { i64, i64, i32, i32 }
9+
%struct.bar = type { i64 }
10+
%struct.spam = type { i64, i64, i64, i64, i32 }
11+
12+
; CHECK: @[[SHADOW4:.*]] = internal unnamed_addr addrspace(3) global %struct.ham addrspace(4)*
13+
; CHECK: @[[SHADOW3:.*]] = internal unnamed_addr addrspace(3) global %struct.spam
14+
; CHECK: @[[SHADOW2:.*]] = internal unnamed_addr addrspace(3) global %struct.ham
15+
; CHECK: @[[SHADOW1:.*]] = internal unnamed_addr addrspace(3) global %struct.bar
16+
17+
define linkonce_odr dso_local spir_func void @foo(%struct.ham addrspace(4)* dereferenceable_or_null(56) %arg, %struct.bar* byval(%struct.bar) align 8 %arg1) !work_group_scope !0 {
18+
; CHECK-LABEL: @foo(
19+
; CHECK-NEXT: bb:
20+
; CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_HAM:%.*]] addrspace(4)*, align 8
21+
; CHECK-NEXT: [[TMP0:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
22+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0:#.*]]
23+
; CHECK-NEXT: [[CMPZ3:%.*]] = icmp eq i64 [[TMP0]], 0
24+
; CHECK-NEXT: br i1 [[CMPZ3]], label [[LEADER:%.*]], label [[MERGE:%.*]]
25+
; CHECK: leader:
26+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast %struct.bar* [[ARG1:%.*]] to i8*
27+
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p0i8.i64(i8 addrspace(3)* align 8 bitcast (%struct.bar addrspace(3)* @[[SHADOW1]] to i8 addrspace(3)*), i8* align 8 [[TMP1]], i64 8, i1 false)
28+
; CHECK-NEXT: [[TMP2:%.*]] = bitcast [[STRUCT_HAM]] addrspace(4)* [[ARG:%.*]] to i8 addrspace(4)*
29+
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p4i8.i64(i8 addrspace(3)* align 16 bitcast (%struct.ham addrspace(3)* @[[SHADOW2]] to i8 addrspace(3)*), i8 addrspace(4)* align 8 [[TMP2]], i64 24, i1 false)
30+
; CHECK-NEXT: br label [[MERGE]]
31+
; CHECK: merge:
32+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
33+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast %struct.bar* [[ARG1]] to i8*
34+
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p3i8.i64(i8* align 8 [[TMP3]], i8 addrspace(3)* align 8 bitcast (%struct.bar addrspace(3)* @[[SHADOW1]] to i8 addrspace(3)*), i64 8, i1 false)
35+
; CHECK-NEXT: [[TMP4:%.*]] = bitcast [[STRUCT_HAM]] addrspace(4)* [[ARG]] to i8 addrspace(4)*
36+
; CHECK-NEXT: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* align 8 [[TMP4]], i8 addrspace(3)* align 16 bitcast (%struct.ham addrspace(3)* @[[SHADOW2]] to i8 addrspace(3)*), i64 24, i1 false)
37+
; CHECK-NEXT: [[TMP2:%.*]] = addrspacecast [[STRUCT_HAM]] addrspace(4)** [[TMP]] to [[STRUCT_HAM]] addrspace(4)* addrspace(4)*
38+
; CHECK-NEXT: [[TMP3:%.*]] = alloca [[STRUCT_SPAM:%.*]], align 8
39+
; CHECK-NEXT: [[TMP4:%.*]] = addrspacecast %struct.spam* [[TMP3]] to [[STRUCT_SPAM]] addrspace(4)*
40+
; CHECK-NEXT: [[TMP5:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
41+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
42+
; CHECK-NEXT: [[CMPZ:%.*]] = icmp eq i64 [[TMP5]], 0
43+
; CHECK-NEXT: br i1 [[CMPZ]], label [[WG_LEADER:%.*]], label [[WG_CF:%.*]]
44+
; CHECK: wg_leader:
45+
; CHECK-NEXT: store [[STRUCT_HAM]] addrspace(4)* [[ARG]], [[STRUCT_HAM]] addrspace(4)* addrspace(4)* [[TMP2]], align 8
46+
; CHECK-NEXT: br label [[WG_CF]]
47+
; CHECK: wg_cf:
48+
; CHECK-NEXT: [[TMP6:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
49+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
50+
; CHECK-NEXT: [[CMPZ2:%.*]] = icmp eq i64 [[TMP6]], 0
51+
; CHECK-NEXT: br i1 [[CMPZ2]], label [[TESTMAT:%.*]], label [[LEADERMAT:%.*]]
52+
; CHECK: TestMat:
53+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast %struct.spam* [[TMP3]] to i8*
54+
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p0i8.i64(i8 addrspace(3)* align 16 bitcast (%struct.spam addrspace(3)* @[[SHADOW3]] to i8 addrspace(3)*), i8* align 8 [[TMP7]], i64 36, i1 false)
55+
; CHECK-NEXT: [[MAT_LD:%.*]] = load [[STRUCT_HAM]] addrspace(4)*, [[STRUCT_HAM]] addrspace(4)** [[TMP]], align 8
56+
; CHECK-NEXT: store [[STRUCT_HAM]] addrspace(4)* [[MAT_LD]], [[STRUCT_HAM]] addrspace(4)* addrspace(3)* @[[SHADOW4]], align 8
57+
; CHECK-NEXT: br label [[LEADERMAT]]
58+
; CHECK: LeaderMat:
59+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
60+
; CHECK-NEXT: [[MAT_LD1:%.*]] = load [[STRUCT_HAM]] addrspace(4)*, [[STRUCT_HAM]] addrspace(4)* addrspace(3)* @[[SHADOW4]], align 8
61+
; CHECK-NEXT: store [[STRUCT_HAM]] addrspace(4)* [[MAT_LD1]], [[STRUCT_HAM]] addrspace(4)** [[TMP]], align 8
62+
; CHECK-NEXT: [[TMP8:%.*]] = bitcast %struct.spam* [[TMP3]] to i8*
63+
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p3i8.i64(i8* align 8 [[TMP8]], i8 addrspace(3)* align 16 bitcast (%struct.spam addrspace(3)* @[[SHADOW3]] to i8 addrspace(3)*), i64 36, i1 false)
64+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
65+
; CHECK-NEXT: [[TMP5:%.*]] = addrspacecast %struct.bar* [[ARG1]] to [[STRUCT_BAR:%.*]] addrspace(4)*
66+
; CHECK-NEXT: [[TMP6:%.*]] = addrspacecast [[STRUCT_SPAM]] addrspace(4)* [[TMP4]] to %struct.spam*
67+
; CHECK-NEXT: call spir_func void @widget(%struct.bar addrspace(4)* dereferenceable_or_null(32) [[TMP5]], %struct.spam* byval(%struct.spam) align 8 [[TMP6]])
68+
; CHECK-NEXT: ret void
69+
;
70+
bb:
71+
%tmp = alloca %struct.ham addrspace(4)*, align 8
72+
%tmp2 = addrspacecast %struct.ham addrspace(4)** %tmp to %struct.ham addrspace(4)* addrspace(4)*
73+
%tmp3 = alloca %struct.spam, align 8
74+
%tmp4 = addrspacecast %struct.spam* %tmp3 to %struct.spam addrspace(4)*
75+
store %struct.ham addrspace(4)* %arg, %struct.ham addrspace(4)* addrspace(4)* %tmp2, align 8
76+
%tmp5 = addrspacecast %struct.bar* %arg1 to %struct.bar addrspace(4)*
77+
%tmp6 = addrspacecast %struct.spam addrspace(4)* %tmp4 to %struct.spam*
78+
call spir_func void @widget(%struct.bar addrspace(4)* dereferenceable_or_null(32) %tmp5, %struct.spam* byval(%struct.spam) align 8 %tmp6)
79+
ret void
80+
}
81+
82+
define linkonce_odr dso_local spir_func void @widget(%struct.bar addrspace(4)* dereferenceable_or_null(32) %arg, %struct.spam* byval(%struct.spam) align 8 %arg1) !work_item_scope !0 !parallel_for_work_item !0 {
83+
bb:
84+
ret void
85+
}
86+
87+
!0 = !{}

‎llvm/test/SYCLLowerIR/pfwg_and_pfwi.ll‎

Lines changed: 23 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,6 @@
1313
%struct.foo = type { %struct.barney }
1414
%struct.foo.0 = type { i8 }
1515

16-
; CHECK: @[[GROUP_SHADOW_PTR:.*]] = internal unnamed_addr addrspace(3) global %struct.zot addrspace(4)*
1716
; CHECK: @[[PFWG_SHADOW_PTR:.*]] = internal unnamed_addr addrspace(3) global %struct.bar addrspace(4)*
1817
; CHECK: @[[PFWI_SHADOW:.*]] = internal unnamed_addr addrspace(3) global %struct.foo.0
1918
; CHECK: @[[PFWG_SHADOW:.*]] = internal unnamed_addr addrspace(3) global %struct.bar
@@ -23,53 +22,51 @@ define internal spir_func void @wibble(%struct.bar addrspace(4)* %arg, %struct.z
2322
; CHECK-LABEL: @wibble(
2423
; CHECK-NEXT: bb:
2524
; CHECK-NEXT: [[TMP:%.*]] = alloca [[STRUCT_BAR:%.*]] addrspace(4)*, align 8
26-
; CHECK-NEXT: [[TMP2:%.*]] = alloca [[STRUCT_FOO_0:%.*]], align 1
27-
; CHECK-NEXT: [[TMP0:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex
28-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272)
25+
; CHECK-NEXT: [[TMP_FOO:%.*]] = alloca [[STRUCT_FOO_0:%.*]], align 1
26+
; CHECK-NEXT: [[TMP0:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
27+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0:#.*]]
2928
; CHECK-NEXT: [[CMPZ3:%.*]] = icmp eq i64 [[TMP0]], 0
3029
; CHECK-NEXT: br i1 [[CMPZ3]], label [[LEADER:%.*]], label [[MERGE:%.*]]
3130
; CHECK: leader:
3231
; CHECK-NEXT: [[TMP1:%.*]] = bitcast %struct.zot* [[ARG1:%.*]] to i8*
3332
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p0i8.i64(i8 addrspace(3)* align 16 bitcast (%struct.zot addrspace(3)* @[[GROUP_SHADOW]] to i8 addrspace(3)*), i8* align 8 [[TMP1]], i64 96, i1 false)
34-
; CHECK-NEXT: [[ARG_CAST:%.*]] = bitcast [[STRUCT_BAR]] addrspace(4)* [[ARG:%.*]] to i8 addrspace(4)*
35-
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p4i8.i64(i8 addrspace(3)* align 8 getelementptr inbounds (%struct.bar, [[STRUCT_BAR]] addrspace(3)* @[[PFWG_SHADOW]], i32 0, i32 0), i8 addrspace(4)* align 8 [[ARG_CAST]], i64 1, i1 false)
33+
; CHECK-NEXT: [[TMP2:%.*]] = bitcast [[STRUCT_BAR]] addrspace(4)* [[ARG:%.*]] to i8 addrspace(4)*
34+
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p4i8.i64(i8 addrspace(3)* align 8 getelementptr inbounds (%struct.bar, [[STRUCT_BAR]] addrspace(3)* @[[PFWG_SHADOW]], i32 0, i32 0), i8 addrspace(4)* align 8 [[TMP2]], i64 1, i1 false)
3635
; CHECK-NEXT: br label [[MERGE]]
3736
; CHECK: merge:
38-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) #0
37+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
3938
; CHECK-NEXT: [[TMP3:%.*]] = bitcast %struct.zot* [[ARG1]] to i8*
4039
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p3i8.i64(i8* align 8 [[TMP3]], i8 addrspace(3)* align 16 bitcast (%struct.zot addrspace(3)* @[[GROUP_SHADOW]] to i8 addrspace(3)*), i64 96, i1 false)
4140
; CHECK-NEXT: [[TMP4:%.*]] = bitcast [[STRUCT_BAR]] addrspace(4)* [[ARG]] to i8 addrspace(4)*
4241
; CHECK-NEXT: call void @llvm.memcpy.p4i8.p3i8.i64(i8 addrspace(4)* align 8 [[TMP4]], i8 addrspace(3)* align 8 getelementptr inbounds (%struct.bar, [[STRUCT_BAR]] addrspace(3)* @[[PFWG_SHADOW]], i32 0, i32 0), i64 1, i1 false)
43-
; CHECK-NEXT: [[TMP5:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex
44-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272)
42+
; CHECK-NEXT: [[TMP5:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
43+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
4544
; CHECK-NEXT: [[CMPZ:%.*]] = icmp eq i64 [[TMP5]], 0
4645
; CHECK-NEXT: br i1 [[CMPZ]], label [[WG_LEADER:%.*]], label [[WG_CF:%.*]]
4746
; CHECK: wg_leader:
4847
; CHECK-NEXT: store [[STRUCT_BAR]] addrspace(4)* [[ARG]], [[STRUCT_BAR]] addrspace(4)** [[TMP]], align 8
4948
; CHECK-NEXT: [[TMP3:%.*]] = load [[STRUCT_BAR]] addrspace(4)*, [[STRUCT_BAR]] addrspace(4)** [[TMP]], align 8
50-
; CHECK-NEXT: [[TMP4:%.*]] = addrspacecast %struct.zot* [[ARG1]] to [[STRUCT_ZOT:%.*]] addrspace(4)*
51-
; CHECK-NEXT: store [[STRUCT_ZOT]] addrspace(4)* [[TMP4]], [[STRUCT_ZOT]] addrspace(4)* addrspace(3)* @[[GROUP_SHADOW_PTR]]
5249
; CHECK-NEXT: br label [[WG_CF]]
5350
; CHECK: wg_cf:
54-
; CHECK-NEXT: [[TMP4:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex
55-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272)
56-
; CHECK-NEXT: [[CMPZ2:%.*]] = icmp eq i64 [[TMP4]], 0
51+
; CHECK-NEXT: [[TMP6:%.*]] = load i64, i64 addrspace(1)* @__spirv_BuiltInLocalInvocationIndex, align 4
52+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
53+
; CHECK-NEXT: [[CMPZ2:%.*]] = icmp eq i64 [[TMP6]], 0
5754
; CHECK-NEXT: br i1 [[CMPZ2]], label [[TESTMAT:%.*]], label [[LEADERMAT:%.*]]
5855
; CHECK: TestMat:
59-
; CHECK-NEXT: [[TMP5:%.*]] = bitcast %struct.foo.0* [[TMP2]] to i8*
60-
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p0i8.i64(i8 addrspace(3)* align 8 getelementptr inbounds (%struct.foo.0, [[STRUCT_FOO_0]] addrspace(3)* @[[PFWI_SHADOW]], i32 0, i32 0), i8* align 1 [[TMP5]], i64 1, i1 false)
61-
; CHECK-NEXT: [[MAT_LD:%.*]] = load [[STRUCT_BAR]] addrspace(4)*, [[STRUCT_BAR]] addrspace(4)** [[TMP]]
62-
; CHECK-NEXT: store [[STRUCT_BAR]] addrspace(4)* [[MAT_LD]], [[STRUCT_BAR]] addrspace(4)* addrspace(3)* @[[PFWG_SHADOW_PTR]]
56+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast %struct.foo.0* [[TMP_FOO]] to i8*
57+
; CHECK-NEXT: call void @llvm.memcpy.p3i8.p0i8.i64(i8 addrspace(3)* align 8 getelementptr inbounds (%struct.foo.0, [[STRUCT_FOO_0]] addrspace(3)* @[[PFWI_SHADOW]], i32 0, i32 0), i8* align 1 [[TMP7]], i64 1, i1 false)
58+
; CHECK-NEXT: [[MAT_LD:%.*]] = load [[STRUCT_BAR]] addrspace(4)*, [[STRUCT_BAR]] addrspace(4)** [[TMP]], align 8
59+
; CHECK-NEXT: store [[STRUCT_BAR]] addrspace(4)* [[MAT_LD]], [[STRUCT_BAR]] addrspace(4)* addrspace(3)* @[[PFWG_SHADOW_PTR]], align 8
6360
; CHECK-NEXT: br label [[LEADERMAT]]
6461
; CHECK: LeaderMat:
65-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) #0
66-
; CHECK-NEXT: [[MAT_LD1:%.*]] = load [[STRUCT_BAR]] addrspace(4)*, [[STRUCT_BAR]] addrspace(4)* addrspace(3)* @[[PFWG_SHADOW_PTR]]
67-
; CHECK-NEXT: store [[STRUCT_BAR]] addrspace(4)* [[MAT_LD1]], [[STRUCT_BAR]] addrspace(4)** [[TMP]]
68-
; CHECK-NEXT: [[TMP6:%.*]] = bitcast %struct.foo.0* [[TMP2]] to i8*
69-
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p3i8.i64(i8* align 1 [[TMP6]], i8 addrspace(3)* align 8 getelementptr inbounds (%struct.foo.0, [[STRUCT_FOO_0]] addrspace(3)* @[[PFWI_SHADOW]], i32 0, i32 0), i64 1, i1 false)
70-
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) #0
71-
; CHECK-NEXT: [[WG_VAL_TMP4:%.*]] = load [[STRUCT_ZOT]] addrspace(4)*, [[STRUCT_ZOT]] addrspace(4)* addrspace(3)* @[[GROUP_SHADOW_PTR]]
72-
; CHECK-NEXT: call spir_func void @bar(%struct.zot addrspace(4)* [[WG_VAL_TMP4]], %struct.foo.0* byval(%struct.foo.0) align 1 [[TMP2]])
62+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
63+
; CHECK-NEXT: [[MAT_LD1:%.*]] = load [[STRUCT_BAR]] addrspace(4)*, [[STRUCT_BAR]] addrspace(4)* addrspace(3)* @[[PFWG_SHADOW_PTR]], align 8
64+
; CHECK-NEXT: store [[STRUCT_BAR]] addrspace(4)* [[MAT_LD1]], [[STRUCT_BAR]] addrspace(4)** [[TMP]], align 8
65+
; CHECK-NEXT: [[TMP8:%.*]] = bitcast %struct.foo.0* [[TMP_FOO]] to i8*
66+
; CHECK-NEXT: call void @llvm.memcpy.p0i8.p3i8.i64(i8* align 1 [[TMP8]], i8 addrspace(3)* align 8 getelementptr inbounds (%struct.foo.0, [[STRUCT_FOO_0]] addrspace(3)* @[[PFWI_SHADOW]], i32 0, i32 0), i64 1, i1 false)
67+
; CHECK-NEXT: call void @_Z22__spirv_ControlBarrierjjj(i32 2, i32 2, i32 272) [[ATTR0]]
68+
; CHECK-NEXT: [[TMP4:%.*]] = addrspacecast %struct.zot* [[ARG1]] to [[STRUCT_ZOT:%.*]] addrspace(4)*
69+
; CHECK-NEXT: call spir_func void @bar(%struct.zot addrspace(4)* [[TMP4]], %struct.foo.0* byval(%struct.foo.0) align 1 [[TMP_FOO]])
7370
; CHECK-NEXT: ret void
7471
;
7572
bb:

0 commit comments

Comments
 (0)