diff options
Diffstat (limited to 'clang/test/CodeGenHLSL/BasicFeatures')
5 files changed, 468 insertions, 66 deletions
diff --git a/clang/test/CodeGenHLSL/BasicFeatures/AggregateSplatCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/AggregateSplatCast.hlsl index 512fcd4..9524f02 100644 --- a/clang/test/CodeGenHLSL/BasicFeatures/AggregateSplatCast.hlsl +++ b/clang/test/CodeGenHLSL/BasicFeatures/AggregateSplatCast.hlsl @@ -54,18 +54,16 @@ struct S { // struct splats // CHECK-LABEL: define void {{.*}}call3 -// CHECK: [[A:%.*]] = alloca <1 x i32>, align 4 +// CHECK: [[AA:%.*]] = alloca i32, align 4 // CHECK: [[s:%.*]] = alloca %struct.S, align 1 -// CHECK-NEXT: store <1 x i32> splat (i32 1), ptr [[A]], align 4 -// CHECK-NEXT: [[L:%.*]] = load <1 x i32>, ptr [[A]], align 4 -// CHECK-NEXT: [[VL:%.*]] = extractelement <1 x i32> [[L]], i32 0 +// CHECK-NEXT: store i32 %A, ptr [[AA]], align 4 +// CHECK-NEXT: [[L:%.*]] = load i32, ptr [[AA]], align 4 // CHECK-NEXT: [[G1:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 0 // CHECK-NEXT: [[G2:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 1 -// CHECK-NEXT: store i32 [[VL]], ptr [[G1]], align 4 -// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[VL]] to float +// CHECK-NEXT: store i32 [[L]], ptr [[G1]], align 4 +// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[L]] to float // CHECK-NEXT: store float [[C]], ptr [[G2]], align 4 -export void call3() { - int1 A = {1}; +export void call3(int A) { S s = (S)A; } @@ -85,3 +83,41 @@ export void call5() { int1 A = {1}; S s = (S)A; } + +struct BFields { + double DF; + int E: 15; + int : 8; + float F; +}; + +struct Derived : BFields { + int G; +}; + +// derived struct with bitfields splat from scalar +// CHECK-LABEL: call6 +// CHECK: [[AAddr:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[D:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: store i32 %A, ptr [[AAddr]], align 4 +// CHECK-NEXT: [[B:%.*]] = load i32, ptr [[AAddr]], align 4 +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 1 +// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[B]] to double +// CHECK-NEXT: store double [[C]], ptr [[Gep1]], align 8 +// CHECK-NEXT: [[H:%.*]] = trunc i32 [[B]] to i24 +// CHECK-NEXT: [[BFL:%.*]] = load i24, ptr [[E]], align 1 +// CHECK-NEXT: [[BFV:%.*]] = and i24 [[H]], 32767 +// CHECK-NEXT: [[BFC:%.*]] = and i24 [[BFL]], -32768 +// CHECK-NEXT: [[BFS:%.*]] = or i24 [[BFC]], [[BFV]] +// CHECK-NEXT: store i24 [[BFS]], ptr [[E]], align 1 +// CHECK-NEXT: [[C4:%.*]] = sitofp i32 [[B]] to float +// CHECK-NEXT: store float [[C4]], ptr [[Gep2]], align 4 +// CHECK-NEXT: store i32 [[B]], ptr [[Gep3]], align 4 +// CHECK-NEXT: ret void +export void call6(int A) { + Derived D = (Derived)A; +} diff --git a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl index ac02ddf..5f2182e 100644 --- a/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl +++ b/clang/test/CodeGenHLSL/BasicFeatures/ArrayElementwiseCast.hlsl @@ -10,7 +10,8 @@ // CHECK-NEXT: [[G1:%.*]] = getelementptr inbounds [2 x i32], ptr [[Tmp]], i32 0, i32 0 // CHECK-NEXT: [[G2:%.*]] = getelementptr inbounds [2 x i32], ptr [[Tmp]], i32 0, i32 1 // CHECK-NEXT: [[L:%.*]] = load i32, ptr [[G1]], align 4 -// CHECK-NEXT: store i32 [[L]], ptr [[B]], align 4 +// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[L]] to float +// CHECK-NEXT: store float [[C]], ptr [[B]], align 4 export void call0() { int A[2] = {0,1}; float B = (float)A; @@ -141,3 +142,46 @@ export void call7() { int A[1] = {1}; A = (int[1])s; } + +struct BFields { + double D; + int E: 15; + int : 8; + float F; +}; + +struct Derived : BFields { + int G; +}; + +// flatten from a derived struct with bitfields +// CHECK-LABEL: call8 +// CHECK: [[A:%.*]] = alloca [4 x i32], align 4 +// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false) +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 0 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 1 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds [4 x i32], ptr [[A]], i32 0, i32 3 +// CHECK-NEXT: [[Gep4:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep4]], i32 0, i32 1 +// CHECK-NEXT: [[Gep5:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep6:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep7:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 1 +// CHECK-NEXT: [[Z:%.*]] = load double, ptr [[Gep5]], align 8 +// CHECK-NEXT: [[C:%.*]] = fptosi double [[Z]] to i32 +// CHECK-NEXT: store i32 [[C]], ptr [[Gep]], align 4 +// CHECK-NEXT: [[BFL:%.*]] = load i24, ptr [[E]], align 1 +// CHECK-NEXT: [[BFShl:%.*]] = shl i24 [[BFL]], 9 +// CHECK-NEXT: [[BFAshr:%.*]] = ashr i24 [[BFShl]], 9 +// CHECK-NEXT: [[BFC:%.*]] = sext i24 [[BFAshr]] to i32 +// CHECK-NEXT: store i32 [[BFC]], ptr [[Gep1]], align 4 +// CHECK-NEXT: [[Y:%.*]] = load float, ptr [[Gep6]], align 4 +// CHECK-NEXT: [[C8:%.*]] = fptosi float [[Y]] to i32 +// CHECK-NEXT: store i32 [[C8]], ptr [[Gep2]], align 4 +// CHECK-NEXT: [[X:%.*]] = load i32, ptr [[Gep7]], align 4 +// CHECK-NEXT: store i32 [[X]], ptr [[Gep3]], align 4 +// CHECK-NEXT: ret void +export void call8(Derived D) { + int A[4] = (int[4])D; +} diff --git a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl index 7e83e5f..82ed754 100644 --- a/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl +++ b/clang/test/CodeGenHLSL/BasicFeatures/InitLists.hlsl @@ -61,10 +61,6 @@ struct EmptyDerived : Empty {}; struct UnnamedDerived : UnnamedOnly {}; -// CHECK-DAG: [[ConstE:@.*]] = private unnamed_addr constant %struct.Empty undef, align 1 -// CHECK-DAG: [[ConstUO:@.*]] = private unnamed_addr constant %struct.UnnamedOnly undef, align 1 -// CHECK-DAG: [[ConstED:@.*]] = private unnamed_addr constant %struct.EmptyDerived undef, align 1 -// CHECK-DAG: [[ConstUD:@.*]] = private unnamed_addr constant %struct.UnnamedDerived undef, align 1 // Case 1: Extraneous braces get ignored in literal instantiation. // CHECK-LABEL: define hidden void @_Z5case1v( @@ -911,15 +907,15 @@ TwoFloats case15(SlicyBits SB) { // CHECK-NEXT: [[X_ADDR:%.*]] = alloca ptr, align 4 // CHECK-NEXT: store ptr [[X]], ptr [[X_ADDR]], align 4 // CHECK-NEXT: [[X1:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 0 -// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[X_ADDR]], align 4, !nonnull [[META3:![0-9]+]], !align [[META4:![0-9]+]] // CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[TMP0]], align 4 // CHECK-NEXT: store float [[TMP1]], ptr [[X1]], align 1 // CHECK-NEXT: [[Y:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOFLOATS]], ptr [[AGG_RESULT]], i32 0, i32 1 -// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[X_ADDR]], align 4 +// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[X_ADDR]], align 4, !nonnull [[META3]], !align [[META4]] // CHECK-NEXT: [[TMP3:%.*]] = load float, ptr [[TMP2]], align 4 // CHECK-NEXT: [[MUL:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP3]], 1.500000e+00 // CHECK-NEXT: store float [[MUL]], ptr [[Y]], align 1 -// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[X_ADDR]], align 4 +// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[X_ADDR]], align 4, !nonnull [[META3]], !align [[META4]] // CHECK-NEXT: [[TMP5:%.*]] = load float, ptr [[TMP4]], align 4 // CHECK-NEXT: [[MUL2:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP5]], 2.000000e+00 // CHECK-NEXT: store float [[MUL2]], ptr [[TMP4]], align 4 @@ -964,94 +960,173 @@ FourFloats case16() { } +// CHECK-LABEL: define hidden noundef i32 @_Z12case17Helperi( +// CHECK-SAME: i32 noundef [[X:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4 +// CHECK-NEXT: store i32 [[X]], ptr [[X_ADDR]], align 4 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[X_ADDR]], align 4 +// CHECK-NEXT: ret i32 [[TMP0]] +// int case17Helper(int x) { return x; } // InitList with OpaqueValueExpr -// CHECK-LABEL: define hidden void {{.*}}case17 -// CHECK: [[X:%.*]] = alloca <2 x i32>, align 8 -// CHECK-NEXT: [[C:%.*]] = call noundef i32 {{.*}}case17Helper{{.*}}(i32 noundef 0) -// CHECK-NEXT: [[C1:%.*]] = call noundef i32 {{.*}}case17Helper{{.*}}(i32 noundef 1) -// CHECK-NEXT: [[VI:%.*]] = insertelement <2 x i32> poison, i32 [[C]], i32 0 -// CHECK-NEXT: [[VI2:%.*]] = insertelement <2 x i32> [[VI]], i32 [[C1]], i32 1 -// CHECK-NEXT: store <2 x i32> [[VI2]], ptr [[X]], align 8 -// CHECK-NEXT: ret void +// CHECK-LABEL: define hidden void @_Z6case17v( +// CHECK-SAME: ) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[X:%.*]] = alloca <2 x i32>, align 8 +// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_Z12case17Helperi(i32 noundef 0) #[[ATTR2]] +// CHECK-NEXT: [[CALL1:%.*]] = call noundef i32 @_Z12case17Helperi(i32 noundef 1) #[[ATTR2]] +// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <2 x i32> poison, i32 [[CALL]], i32 0 +// CHECK-NEXT: [[VECINIT2:%.*]] = insertelement <2 x i32> [[VECINIT]], i32 [[CALL1]], i32 1 +// CHECK-NEXT: store <2 x i32> [[VECINIT2]], ptr [[X]], align 8 +// CHECK-NEXT: ret void +// void case17() { int2 X = {case17Helper(0), case17Helper(1)}; } // InitList with Struct with unnamed bitfield on LHS -// CHECK-LABEL: case18 -// CHECK: [[U:%.*]] = alloca %struct.Unnamed, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[U]], ptr align 1 {{.*}}, i32 5, i1 false) +// CHECK-LABEL: define hidden void @_Z6case18v( +// CHECK-SAME: ) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[U:%.*]] = alloca [[STRUCT_UNNAMED:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[U]], ptr align 1 @__const._Z6case18v.U, i32 5, i1 false) +// CHECK-NEXT: ret void +// void case18() { Unnamed U = {1}; } // InitList with Struct with unnamed bitfield on RHS -// CHECK-LABEL: case19 -// CHECK: [[TI:%.*]] = alloca %struct.TwoInts, align 1 -// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw %struct.TwoInts, ptr [[TI]], i32 0, i32 0 -// CHECK-NEXT: [[A:%.*]] = getelementptr inbounds nuw %struct.Unnamed, ptr %U, i32 0, i32 0 -// CHECK-NEXT: [[L:%.*]] = load i32, ptr [[A]], align 1 -// CHECK-NEXT: store i32 [[L]], ptr [[Z]], align 1 -// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw %struct.TwoInts, ptr [[TI]], i32 0, i32 1 -// CHECK-NEXT: store i32 1, ptr [[W]], align 1 +// CHECK-LABEL: define hidden void @_Z6case197Unnamed( +// CHECK-SAME: ptr noundef byval([[STRUCT_UNNAMED:%.*]]) align 1 [[U:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1 +// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0 +// CHECK-NEXT: [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_UNNAMED]], ptr [[U]], i32 0, i32 0 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A]], align 1 +// CHECK-NEXT: store i32 [[TMP0]], ptr [[Z]], align 1 +// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1 +// CHECK-NEXT: store i32 1, ptr [[W]], align 1 +// CHECK-NEXT: ret void +// void case19(Unnamed U) { TwoInts TI = {U, 1}; } // InitList with Empty Struct on LHS -// CHECK-LABEL: case20 -// CHECK: [[E:%.*]] = alloca %struct.Empty, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[E]], ptr align 1 [[ConstE]], i32 1, i1 false) +// CHECK-LABEL: define hidden void @_Z6case20v( +// CHECK-SAME: ) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[E:%.*]] = alloca [[STRUCT_EMPTY:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[E]], ptr align 1 @__const._Z6case20v.E, i32 1, i1 false) +// CHECK-NEXT: ret void +// void case20() { Empty E = {}; } // InitList with Empty Struct on RHS -// CHECK-LABEL: case21 -// CHECK: [[TI:%.*]] = alloca %struct.TwoInts, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %TI, ptr align 1 {{.*}}, i32 8, i1 false) +// CHECK-LABEL: define hidden void @_Z6case215Empty( +// CHECK-SAME: ptr noundef byval([[STRUCT_EMPTY:%.*]]) align 1 [[E:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI]], ptr align 1 @__const._Z6case215Empty.TI, i32 8, i1 false) +// CHECK-NEXT: ret void +// void case21(Empty E) { TwoInts TI = {E, 1, 2}; } // InitList with Struct with only unnamed bitfield on LHS -// CHECK-LABEL: case22 -// CHECK: [[UO:%.*]] = alloca %struct.UnnamedOnly, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[UO]], ptr align 1 [[ConstUO]], i32 1, i1 false) +// CHECK-LABEL: define hidden void @_Z6case22v( +// CHECK-SAME: ) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[UO:%.*]] = alloca [[STRUCT_UNNAMEDONLY:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[UO]], ptr align 1 @__const._Z6case22v.UO, i32 1, i1 false) +// CHECK-NEXT: ret void +// void case22() { - UnnamedOnly UO = {}; + UnnamedOnly UO = {}; } // InitList with Struct with only unnamed bitfield on RHS -// CHECK-LABEL: case23 -// CHECK: [[TI:%.*]] = alloca %struct.TwoInts, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI]], ptr align 1 {{.*}}, i32 8, i1 false) +// CHECK-LABEL: define hidden void @_Z6case2311UnnamedOnly( +// CHECK-SAME: ptr noundef byval([[STRUCT_UNNAMEDONLY:%.*]]) align 1 [[UO:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[TI:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI]], ptr align 1 @__const._Z6case2311UnnamedOnly.TI, i32 8, i1 false) +// CHECK-NEXT: ret void +// void case23(UnnamedOnly UO) { TwoInts TI = {UO, 1, 2}; } // InitList with Derived empty struct on LHS // InitList with Derived unnamed bitfield on LHS -// CHECK-LABEL: case24 -// CHECK: [[ED:%.*]] = alloca %struct.EmptyDerived, align 1 -// CHECK-NEXT: [[UD:%.*]] = alloca %struct.UnnamedDerived, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %ED, ptr align 1 [[ConstED]], i32 1, i1 false) -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %UD, ptr align 1 [[ConstUD]], i32 1, i1 false) +// CHECK-LABEL: define hidden void @_Z6case24v( +// CHECK-SAME: ) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[ED:%.*]] = alloca [[STRUCT_EMPTYDERIVED:%.*]], align 1 +// CHECK-NEXT: [[UD:%.*]] = alloca [[STRUCT_UNNAMEDDERIVED:%.*]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[ED]], ptr align 1 @__const._Z6case24v.ED, i32 1, i1 false) +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[UD]], ptr align 1 @__const._Z6case24v.UD, i32 1, i1 false) +// CHECK-NEXT: ret void +// void case24() { EmptyDerived ED = {}; UnnamedDerived UD = {}; } -// CHECK-LABEL: case25 -// CHECK: [[TI1:%.*]] = alloca %struct.TwoInts, align 1 -// CHECK-NEXT: [[TI2:%.*]] = alloca %struct.TwoInts, align 1 -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %TI1, ptr align 1 {{.*}}, i32 8, i1 false) -// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 %TI2, ptr align 1 {{.*}}, i32 8, i1 false) +// CHECK-LABEL: define hidden void @_Z6case2512EmptyDerived14UnnamedDerived( +// CHECK-SAME: ptr noundef byval([[STRUCT_EMPTYDERIVED:%.*]]) align 1 [[ED:%.*]], ptr noundef byval([[STRUCT_UNNAMEDDERIVED:%.*]]) align 1 [[UD:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[TI1:%.*]] = alloca [[STRUCT_TWOINTS:%.*]], align 1 +// CHECK-NEXT: [[TI2:%.*]] = alloca [[STRUCT_TWOINTS]], align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI1]], ptr align 1 @__const._Z6case2512EmptyDerived14UnnamedDerived.TI1, i32 8, i1 false) +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[TI2]], ptr align 1 @__const._Z6case2512EmptyDerived14UnnamedDerived.TI2, i32 8, i1 false) +// CHECK-NEXT: ret void +// void case25(EmptyDerived ED, UnnamedDerived UD) { TwoInts TI1 = {ED, 1, 2}; TwoInts TI2 = {UD, 1, 2}; } + +// CHECK-LABEL: define hidden void @_Z6case267TwoInts( +// CHECK-SAME: ptr noundef byval([[STRUCT_TWOINTS:%.*]]) align 1 [[TI:%.*]]) #[[ATTR0]] { +// CHECK-NEXT: [[ENTRY:.*:]] +// CHECK-NEXT: [[F:%.*]] = alloca <4 x float>, align 16 +// CHECK-NEXT: [[F2:%.*]] = alloca <3 x float>, align 16 +// CHECK-NEXT: [[Z:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0 +// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[Z]], align 1 +// CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP0]] to float +// CHECK-NEXT: [[VECINIT:%.*]] = insertelement <4 x float> poison, float [[CONV]], i32 0 +// CHECK-NEXT: [[W:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1 +// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[W]], align 1 +// CHECK-NEXT: [[CONV1:%.*]] = sitofp i32 [[TMP1]] to float +// CHECK-NEXT: [[VECINIT2:%.*]] = insertelement <4 x float> [[VECINIT]], float [[CONV1]], i32 1 +// CHECK-NEXT: [[VECINIT3:%.*]] = insertelement <4 x float> [[VECINIT2]], float 1.000000e+00, i32 2 +// CHECK-NEXT: [[VECINIT4:%.*]] = insertelement <4 x float> [[VECINIT3]], float 2.000000e+00, i32 3 +// CHECK-NEXT: store <4 x float> [[VECINIT4]], ptr [[F]], align 16 +// CHECK-NEXT: [[Z5:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 0 +// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[Z5]], align 1 +// CHECK-NEXT: [[CONV6:%.*]] = sitofp i32 [[TMP2]] to float +// CHECK-NEXT: [[VECINIT7:%.*]] = insertelement <3 x float> <float 3.000000e+00, float poison, float poison>, float [[CONV6]], i32 1 +// CHECK-NEXT: [[W8:%.*]] = getelementptr inbounds nuw [[STRUCT_TWOINTS]], ptr [[TI]], i32 0, i32 1 +// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[W8]], align 1 +// CHECK-NEXT: [[CONV9:%.*]] = sitofp i32 [[TMP3]] to float +// CHECK-NEXT: [[VECINIT10:%.*]] = insertelement <3 x float> [[VECINIT7]], float [[CONV9]], i32 2 +// CHECK-NEXT: store <3 x float> [[VECINIT10]], ptr [[F2]], align 16 +// CHECK-NEXT: ret void +// +void case26(TwoInts TI) { + float4 F = float4(TI, 1, 2); + float3 F2 = float3(3, TI); +} +//. +// CHECK: [[META3]] = !{} +// CHECK: [[META4]] = !{i64 4} +//. diff --git a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl index 81b9f5b..4e29994 100644 --- a/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl +++ b/clang/test/CodeGenHLSL/BasicFeatures/StructElementwiseCast.hlsl @@ -1,4 +1,4 @@ -// RUN: %clang_cc1 -finclude-default-header -triple dxil-pc-shadermodel6.3-library -x hlsl -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s +// RUN: %clang_cc1 -finclude-default-header -fnative-half-type -triple dxil-pc-shadermodel6.3-library -x hlsl -emit-llvm -disable-llvm-passes -o - %s | FileCheck %s struct S { int X; @@ -127,14 +127,219 @@ struct T { // CHECK-NEXT: [[G1:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 0 // CHECK-NEXT: [[G2:%.*]] = getelementptr inbounds %struct.S, ptr [[s]], i32 0, i32 1 // CHECK-NEXT: [[G3:%.*]] = getelementptr inbounds %struct.T, ptr [[Tmp]], i32 0, i32 0 -// CHECK-NEXT: %gep3 = getelementptr inbounds %struct.T, ptr %agg-temp, i32 0, i32 1 -// CHECK-NEXT: %gep4 = getelementptr inbounds %struct.T, ptr %agg-temp, i32 0, i32 2 -// CHECK-NEXT: %load = load i32, ptr %gep2, align 4 -// CHECK-NEXT: store i32 %load, ptr %gep, align 4 -// CHECK-NEXT: %load5 = load i32, ptr %gep3, align 4 -// CHECK-NEXT: %conv = sitofp i32 %load5 to float -// CHECK-NEXT: store float %conv, ptr %gep1, align 4 +// CHECK-NEXT: [[G4:%.*]] = getelementptr inbounds %struct.T, ptr %agg-temp, i32 0, i32 1 +// CHECK-NEXT: [[G5:%.*]] = getelementptr inbounds %struct.T, ptr %agg-temp, i32 0, i32 2 +// CHECK-NEXT: [[L1:%.*]] = load i32, ptr [[G3]], align 4 +// CHECK-NEXT: store i32 [[L1]], ptr [[G1]], align 4 +// CHECK-NEXT: [[L2:%.*]] = load i32, ptr [[G4]], align 4 +// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[L2]] to float +// CHECK-NEXT: store float [[C]], ptr [[G2]], align 4 export void call8() { T t = {1,2,3}; S s = (S)t; } + +struct BFields { + double D; + int E: 15; + int : 8; + float F; +}; + +struct Derived : BFields { + int G; +}; + +// Derived Struct truncate to scalar +// CHECK-LABEL: call9 +// CHECK: [[D2:%.*]] = alloca double, align 8 +// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false) +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 1 +// CHECK-NEXT: [[A:%.*]] = load double, ptr [[Gep1]], align 8 +// CHECK-NEXT: store double [[A]], ptr [[D2]], align 8 +// CHECK-NEXT: ret void +export void call9(Derived D) { + double D2 = (double)D; +} + +// Derived struct from vector +// CHECK-LABEL: call10 +// CHECK: [[IAddr:%.*]] = alloca <4 x i32>, align 16 +// CHECK-NEXT: [[D:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: store <4 x i32> %I, ptr [[IAddr]], align 16 +// CHECK-NEXT: [[A:%.*]] = load <4 x i32>, ptr [[IAddr]], align 16 +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds %struct.Derived, ptr [[D]], i32 0, i32 1 +// CHECK-NEXT: [[VL:%.*]] = extractelement <4 x i32> [[A]], i64 0 +// CHECK-NEXT: [[C:%.*]] = sitofp i32 [[VL]] to double +// CHECK-NEXT: store double [[C]], ptr [[Gep1]], align 8 +// CHECK-NEXT: [[VL4:%.*]] = extractelement <4 x i32> [[A]], i64 1 +// CHECK-NEXT: [[B:%.*]] = trunc i32 [[VL4]] to i24 +// CHECK-NEXT: [[BFL:%.*]] = load i24, ptr [[E]], align 1 +// CHECK-NEXT: [[BFV:%.*]] = and i24 [[B]], 32767 +// CHECK-NEXT: [[BFC:%.*]] = and i24 [[BFL]], -32768 +// CHECK-NEXT: [[BFSet:%.*]] = or i24 [[BFC]], [[BFV]] +// CHECK-NEXT: store i24 [[BFSet]], ptr [[E]], align 1 +// CHECK-NEXT: [[VL5:%.*]] = extractelement <4 x i32> [[A]], i64 2 +// CHECK-NEXT: [[C6:%.*]] = sitofp i32 [[VL5]] to float +// CHECK-NEXT: store float [[C6]], ptr [[Gep2]], align 4 +// CHECK-NEXT: [[VL7:%.*]] = extractelement <4 x i32> [[A]], i64 3 +// CHECK-NEXT: store i32 [[VL7]], ptr [[Gep3]], align 4 +// CHECK-NEXT: ret void +export void call10(int4 I) { + Derived D = (Derived)I; +} + +// truncate derived struct +// CHECK-LABEL: call11 +// CHECK: [[B:%.*]] = alloca %struct.BFields, align 1 +// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 [[D]], i32 19, i1 false) +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.BFields, ptr [[B]], i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds %struct.BFields, ptr [[B]], i32 0, i32 0 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds %struct.BFields, ptr [[B]], i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0 +// CHECK-NEXT: [[E4:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep3]], i32 0, i32 1 +// CHECK-NEXT: [[Gep5:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep6:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep7:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 1 +// CHECK-NEXT: [[A:%.*]] = load double, ptr [[Gep5]], align 8 +// CHECK-NEXT: store double [[A]], ptr [[Gep1]], align 8 +// CHECK-NEXT: [[BFl:%.*]] = load i24, ptr [[E4]], align 1 +// CHECK-NEXT: [[Shl:%.*]] = shl i24 [[BFL]], 9 +// CHECK-NEXT: [[Ashr:%.*]] = ashr i24 [[Shl]], 9 +// CHECK-NEXT: [[BFC:%.*]] = sext i24 [[Ashr]] to i32 +// CHECK-NEXT: [[B:%.*]] = trunc i32 [[BFC]] to i24 +// CHECK-NEXT: [[BFL8:%.*]] = load i24, ptr [[E]], align 1 +// CHECK-NEXT: [[BFV:%.*]] = and i24 [[B]], 32767 +// CHECK-NEXT: [[BFC:%.*]] = and i24 [[BFL8]], -32768 +// CHECK-NEXT: [[BFSet:%.*]] = or i24 [[BFC]], [[BFV]] +// CHECK-NEXT: store i24 [[BFSet]], ptr [[E]], align 1 +// CHECK-NEXT: [[C:%.*]] = load float, ptr [[Gep6]], align 4 +// CHECK-NEXT: store float [[C]], ptr [[Gep2]], align 4 +// CHECK-NEXT: ret void +export void call11(Derived D) { + BFields B = (BFields)D; +} + +struct Empty { +}; + +// cast to an empty struct +// CHECK-LABEL: call12 +// CHECK: [[I:%.*]] = alloca <4 x i32>, align 16 +// CHECK-NEXT: [[E:%.*]] = alloca %struct.Empty, align 1 +// CHECK-NEXT: store <4 x i32> <i32 1, i32 2, i32 3, i32 4>, ptr [[I]], align 16 +// CHECK-NEXT: [[A:%.*]] = load <4 x i32>, ptr [[I]], align 16 +// CHECK-NEXt: ret void +export void call12() { + int4 I = {1,2,3,4}; + Empty E = (Empty)I; +} + +struct MoreBFields { + int A; + uint64_t B: 60; + float C; + uint16_t D: 10; + uint16_t E: 6; + int : 32; + double F; + int : 8; + uint G; +}; + +// more complicated bitfield case +// CHECK-LABEL: call13 +// CHECK: [[AA:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[MBF:%.*]] = alloca %struct.MoreBFields, align 1 +// CHECK-NEXT: store i32 %A, ptr [[AA]], align 4 +// CHECK-NEXT: [[Z:%.*]] = load i32, ptr [[AA]], align 4 +// get the gep for the struct. +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.MoreBFields, ptr [[MBF]], i32 0 +// CHECK-NEXT: [[FieldB:%.*]] = getelementptr inbounds nuw %struct.MoreBFields, ptr [[Gep]], i32 0, i32 1 +// D and E share the same field index +// CHECK-NEXT: [[FieldD:%.*]] = getelementptr inbounds nuw %struct.MoreBFields, ptr [[Gep]], i32 0, i32 3 +// CHECK-NEXT: [[FieldE:%.*]] = getelementptr inbounds nuw %struct.MoreBFields, ptr [[Gep]], i32 0, i32 3 +// CHECK-NEXT: [[FieldA:%.*]] = getelementptr inbounds %struct.MoreBFields, ptr [[MBF]], i32 0, i32 0 +// CHECK-NEXT: [[FieldC:%.*]] = getelementptr inbounds %struct.MoreBFields, ptr [[MBF]], i32 0, i32 2 +// CHECK-NEXT: [[FieldF:%.*]] = getelementptr inbounds %struct.MoreBFields, ptr [[MBF]], i32 0, i32 5 +// CHECK-NEXT: [[FieldG:%.*]] = getelementptr inbounds %struct.MoreBFields, ptr [[MBF]], i32 0, i32 7 +// store int A into field A +// CHECK-NEXT: store i32 [[Z]], ptr [[FieldA]], align 4 +// store int A in bitField B, do necessary conversions +// CHECK-NEXT: [[Conv:%.*]] = sext i32 [[Z]] to i64 +// CHECK-NEXT: [[BFL:%.*]] = load i64, ptr [[FieldB]], align 1 +// CHECK-NEXT: [[BFV:%.*]] = and i64 [[Conv]], 1152921504606846975 +// CHECK-NEXT: [[BFC:%.*]] = and i64 [[BFL]], -1152921504606846976 +// CHECK-NEXT: [[BFS:%.*]] = or i64 [[BFC]], [[BFV]] +// CHECK-NEXT: store i64 [[BFS]], ptr [[FieldB]], align 1 +// store int A into field C +// CHECK-NEXT: [[Conv5:%.*]] = sitofp i32 [[Z]] to float +// CHECK-NEXT: store float [[Conv5]], ptr [[FieldC]], align 4 +// store int A into bitfield D +// CHECK-NEXT: [[Conv6:%.*]] = trunc i32 [[Z]] to i16 +// CHECK-NEXT: [[FDL:%.*]] = load i16, ptr [[FieldD]], align 1 +// CHECK-NEXT: [[FDV:%.*]] = and i16 [[Conv6]], 1023 +// CHECK-NEXT: [[FDC:%.*]] = and i16 [[FDL]], -1024 +// CHECK-NEXT: [[FDS:%.*]] = or i16 [[FDC]], [[FDV]] +// CHECK-NEXT: store i16 [[FDS]], ptr [[FieldD]], align 1 +// store int A into bitfield E; +// CHECK-NEXT: [[Conv11:%.*]] = trunc i32 [[Z]] to i16 +// CHECK-NEXT: [[FEL:%.*]] = load i16, ptr [[FieldE]], align 1 +// CHECK-NEXT: [[FEV:%.*]] = and i16 [[Conv11]], 63 +// CHECK-NEXT: [[FESHL:%.*]] = shl i16 [[FEV]], 10 +// CHECK-NEXT: [[FEC:%.*]] = and i16 [[FEL]], 1023 +// CHECK-NEXT: [[FES:%.*]] = or i16 [[FEC]], [[FESHL]] +// CHECK-NEXT: store i16 [[FES]], ptr [[FieldE]], align 1 +// store int A into field F +// CHECK-NEXT: [[Conv16:%.*]] = sitofp i32 [[Z]] to double +// CHECK-NEXT: store double [[Conv16]], ptr [[FieldF]], align 8 +// store int A into field G +// CHECK-NEXT: store i32 [[Z]], ptr [[FieldG]], align 4 +// CHECK-NEXT: ret void +export void call13(int A) { + MoreBFields MBF = (MoreBFields)A; +} + +struct Inner { + int Z; + int Y : 25; +}; + +struct Outer { + int A; + Inner I; +}; + +// show usage of "extra" gep for struct containing bitfield +// CHECK-LABEL: call14 +// CHECK: [[AA:%.*]] = alloca i32, align 4 +// CHECK-NEXT: [[O:%.*]] = alloca %struct.Outer, align 1 +// CHECK-NEXT: store i32 %A, ptr [[AA]], align 4 +// CHECK-NEXT: [[Z:%.*]] = load i32, ptr [[AA]], align 4 +// CHECK-NEXT: [[FieldA:%.*]] = getelementptr inbounds %struct.Outer, ptr [[O]], i32 0, i32 0 +// showing real usage of "extra gep". need Inner struct to generate access of its bitfield. +// CHECK-NEXT: [[FieldI:%.*]] = getelementptr inbounds %struct.Outer, ptr [[O]], i32 0, i32 1 +// CHECK-NEXT: [[FieldY:%.*]] = getelementptr inbounds nuw %struct.Inner, ptr [[FieldI]], i32 0, i32 1 +// CHECK-NEXT: [[FieldZ:%.*]] = getelementptr inbounds %struct.Outer, ptr [[O]], i32 0, i32 1, i32 0 +// CHECK-NEXT: store i32 [[Z]], ptr [[FieldA]], align 4 +// CHECK-NEXT: store i32 [[Z]], ptr [[FieldZ]], align 4 +// CHECK-NEXT: [[BFL:%.*]] = load i32, ptr [[FieldY]], align 1 +// CHECK-NEXT: [[BFV:%.*]] = and i32 [[Z]], 33554431 +// CHECK-NEXT: [[BFC:%.*]] = and i32 [[BFL]], -33554432 +// CHECK-NEXT: [[BFS:%.*]] = or i32 [[BFC]], [[BFV]] +// CHECK-NEXT: store i32 [[BFS]], ptr [[FieldY]], align 1 +// CHECK-NEXT: ret void +export void call14(int A) { + Outer O = (Outer)A; +} diff --git a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl index 253b38a7c..26aa41a 100644 --- a/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl +++ b/clang/test/CodeGenHLSL/BasicFeatures/VectorElementwiseCast.hlsl @@ -79,3 +79,45 @@ export void call5() { S s = {1, 2.0}; int A = (int)s; } + +struct BFields { + double D; + int E: 15; + int : 8; + float F; +}; + +struct Derived : BFields { + int G; +}; + +// vector flat cast from derived struct with bitfield +// CHECK-LABEL: call6 +// CHECK: [[A:%.*]] = alloca <4 x i32>, align 16 +// CHECK-NEXT: [[Tmp:%.*]] = alloca %struct.Derived, align 1 +// CHECK-NEXT: [[FlatTmp:%.*]] = alloca <4 x i32>, align 16 +// CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 1 [[Tmp]], ptr align 1 %D, i32 19, i1 false) +// CHECK-NEXT: [[Gep:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0 +// CHECK-NEXT: [[E:%.*]] = getelementptr inbounds nuw %struct.BFields, ptr [[Gep]], i32 0, i32 1 +// CHECK-NEXT: [[Gep1:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 0 +// CHECK-NEXT: [[Gep2:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 0, i32 2 +// CHECK-NEXT: [[Gep3:%.*]] = getelementptr inbounds %struct.Derived, ptr [[Tmp]], i32 0, i32 1 +// CHECK-NEXT: [[Z:%.*]] = load <4 x i32>, ptr [[FlatTmp]], align 16 +// CHECK-NEXT: [[Y:%.*]] = load double, ptr [[Gep1]], align 8 +// CHECK-NEXT: [[C:%.*]] = fptosi double [[Y]] to i32 +// CHECK-NEXT: [[X:%.*]] = insertelement <4 x i32> [[Z]], i32 [[C]], i64 0 +// CHECK-NEXT: [[BFL:%.*]] = load i24, ptr [[E]], align 1 +// CHECK-NEXT: [[BFShl:%.*]] = shl i24 [[BFL]], 9 +// CHECK-NEXT: [[BFAshr:%.*]] = ashr i24 [[BFShl]], 9 +// CHECK-NEXT: [[BFC:%.*]] = sext i24 [[BFAshr]] to i32 +// CHECK-NEXT: [[W:%.*]] = insertelement <4 x i32> [[X]], i32 [[BFC]], i64 1 +// CHECK-NEXT: [[V:%.*]] = load float, ptr [[Gep2]], align 4 +// CHECK-NEXT: [[C4:%.*]] = fptosi float [[V]] to i32 +// CHECK-NEXT: [[U:%.*]] = insertelement <4 x i32> [[W]], i32 [[C4]], i64 2 +// CHECK-NEXT: [[T:%.*]] = load i32, ptr [[Gep3]], align 4 +// CHECK-NEXT: [[S:%.*]] = insertelement <4 x i32> [[U]], i32 [[T]], i64 3 +// CHECK-NEXT: store <4 x i32> [[S]], ptr [[A]], align 16 +// CHECK-NEXT: ret void +export void call6(Derived D) { + int4 A = (int4)D; +} |