Commit 921bd299 authored by Johannes Doerfert's avatar Johannes Doerfert
Browse files

[OpenMP] Remove alignment for global <-> local reduction functions 

The alignment did likely not help much but increases the memory
requirement. Note that half of the affected accesses are all performed
by a single thread in each block. The reads are by consecutive threads
in a single block.
parent abe71b77

clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp

+5 −27
Original line number Diff line number Diff line
@@ -85,18 +85,6 @@ public: 
  ~ExecutionRuntimeModesRAII() { ExecMode = SavedExecMode; }

};



/// GPU Configuration:  This information can be derived from cuda registers,

/// however, providing compile time constants helps generate more efficient

/// code.  For all practical purposes this is fine because the configuration

/// is the same for all known NVPTX architectures.

enum MachineConfiguration : unsigned {

  /// See "llvm/Frontend/OpenMP/OMPGridValues.h" for various related target

  /// specific Grid Values like GV_Warp_Size, GV_Slot_Size



  /// Global memory alignment for performance.

  GlobalMemoryAlignment = 128,

};



static const ValueDecl *getPrivateItem(const Expr *RefExpr) {

  RefExpr = RefExpr->IgnoreParens();

  if (const auto *ASE = dyn_cast<ArraySubscriptExpr>(RefExpr)) {
@@ -119,31 +107,23 @@ static const ValueDecl *getPrivateItem(const Expr *RefExpr) { 
  return cast<ValueDecl>(ME->getMemberDecl()->getCanonicalDecl());

}





static RecordDecl *buildRecordForGlobalizedVars(

    ASTContext &C, ArrayRef<const ValueDecl *> EscapedDecls,

    ArrayRef<const ValueDecl *> EscapedDeclsForTeams,

    llvm::SmallDenseMap<const ValueDecl *, const FieldDecl *>

        &MappedDeclsFields, int BufSize) {

        &MappedDeclsFields,

    int BufSize) {

  using VarsDataTy = std::pair<CharUnits /*Align*/, const ValueDecl *>;

  if (EscapedDecls.empty() && EscapedDeclsForTeams.empty())

    return nullptr;

  SmallVector<VarsDataTy, 4> GlobalizedVars;

  for (const ValueDecl *D : EscapedDecls)

    GlobalizedVars.emplace_back(

        CharUnits::fromQuantity(std::max(

            C.getDeclAlign(D).getQuantity(),

            static_cast<CharUnits::QuantityType>(GlobalMemoryAlignment))),

        D);

    GlobalizedVars.emplace_back(C.getDeclAlign(D), D);

  for (const ValueDecl *D : EscapedDeclsForTeams)

    GlobalizedVars.emplace_back(C.getDeclAlign(D), D);

  llvm::stable_sort(GlobalizedVars, [](VarsDataTy L, VarsDataTy R) {

    return L.first > R.first;

  });



  // Build struct _globalized_locals_ty {

  //         /*  globalized vars  */[WarSize] align (max(decl_align,

  //         GlobalMemoryAlignment))

  //         /*  globalized vars  */[WarSize] align (decl_align)

  //         /*  globalized vars  */ for EscapedDeclsForTeams

  //       };

  RecordDecl *GlobalizedRD = C.buildImplicitRecord("_globalized_locals_ty");
@@ -182,9 +162,7 @@ static RecordDecl *buildRecordForGlobalizedVars( 
          /*BW=*/nullptr, /*Mutable=*/false,

          /*InitStyle=*/ICIS_NoInit);

      Field->setAccess(AS_public);

      llvm::APInt Align(32, std::max(C.getDeclAlign(VD).getQuantity(),

                                     static_cast<CharUnits::QuantityType>(

                                         GlobalMemoryAlignment)));

      llvm::APInt Align(32, Pair.first.getQuantity());

      Field->addAttr(AlignedAttr::CreateImplicit(

          C, /*IsAlignmentExpr=*/true,

          IntegerLiteral::Create(C, Align,

clang/test/OpenMP/nvptx_teams_reduction_codegen.cpp

+30 −30
Original line number Diff line number Diff line
@@ -253,7 +253,7 @@ int bar(int n){ 
// CHECK1-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP7]], align 8

// CHECK1-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 128

// CHECK1-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 8

// CHECK1-NEXT:    ret void

//

//
@@ -294,7 +294,7 @@ int bar(int n){ 
// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8

// CHECK1-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 128

// CHECK1-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 8

// CHECK1-NEXT:    store double [[TMP9]], ptr [[TMP7]], align 8

// CHECK1-NEXT:    ret void

//
@@ -583,13 +583,13 @@ int bar(int n){ 
// CHECK1-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP7]], align 1

// CHECK1-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 128

// CHECK1-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 4

// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i64 0, i64 1

// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 8

// CHECK1-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4

// CHECK1-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 128

// CHECK1-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 4

// CHECK1-NEXT:    ret void

//

//
@@ -634,13 +634,13 @@ int bar(int n){ 
// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8

// CHECK1-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 128

// CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 4

// CHECK1-NEXT:    store i8 [[TMP9]], ptr [[TMP7]], align 1

// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i64 0, i64 1

// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 8

// CHECK1-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 128

// CHECK1-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 4

// CHECK1-NEXT:    store float [[TMP13]], ptr [[TMP11]], align 4

// CHECK1-NEXT:    ret void

//
@@ -1156,13 +1156,13 @@ int bar(int n){ 
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4

// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 128

// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 4

// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i64 0, i64 1

// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 8

// CHECK1-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP11]], align 2

// CHECK1-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 128

// CHECK1-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 4

// CHECK1-NEXT:    ret void

//

//
@@ -1207,13 +1207,13 @@ int bar(int n){ 
// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8

// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 128

// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4

// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[TMP7]], align 4

// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i64 0, i64 1

// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 8

// CHECK1-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK1-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 128

// CHECK1-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 4

// CHECK1-NEXT:    store i16 [[TMP13]], ptr [[TMP11]], align 2

// CHECK1-NEXT:    ret void

//
@@ -1446,7 +1446,7 @@ int bar(int n){ 
// CHECK2-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP7]], align 8

// CHECK2-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 128

// CHECK2-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 8

// CHECK2-NEXT:    ret void

//

//
@@ -1487,7 +1487,7 @@ int bar(int n){ 
// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK2-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 128

// CHECK2-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 8

// CHECK2-NEXT:    store double [[TMP9]], ptr [[TMP7]], align 8

// CHECK2-NEXT:    ret void

//
@@ -1776,13 +1776,13 @@ int bar(int n){ 
// CHECK2-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP7]], align 1

// CHECK2-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 128

// CHECK2-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 4

// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK2-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4

// CHECK2-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 128

// CHECK2-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 4

// CHECK2-NEXT:    ret void

//

//
@@ -1827,13 +1827,13 @@ int bar(int n){ 
// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK2-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 128

// CHECK2-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 4

// CHECK2-NEXT:    store i8 [[TMP9]], ptr [[TMP7]], align 1

// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK2-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 128

// CHECK2-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 4

// CHECK2-NEXT:    store float [[TMP13]], ptr [[TMP11]], align 4

// CHECK2-NEXT:    ret void

//
@@ -2349,13 +2349,13 @@ int bar(int n){ 
// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4

// CHECK2-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 128

// CHECK2-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 4

// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK2-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP11]], align 2

// CHECK2-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 128

// CHECK2-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 4

// CHECK2-NEXT:    ret void

//

//
@@ -2400,13 +2400,13 @@ int bar(int n){ 
// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 128

// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4

// CHECK2-NEXT:    store i32 [[TMP9]], ptr [[TMP7]], align 4

// CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK2-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1024 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK2-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 128

// CHECK2-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 4

// CHECK2-NEXT:    store i16 [[TMP13]], ptr [[TMP11]], align 2

// CHECK2-NEXT:    ret void

//
@@ -2639,7 +2639,7 @@ int bar(int n){ 
// CHECK3-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP7]], align 8

// CHECK3-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 128

// CHECK3-NEXT:    store double [[TMP9]], ptr [[TMP8]], align 8

// CHECK3-NEXT:    ret void

//

//
@@ -2680,7 +2680,7 @@ int bar(int n){ 
// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK3-NEXT:    [[E:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x double], ptr [[E]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 128

// CHECK3-NEXT:    [[TMP9:%.*]] = load double, ptr [[TMP8]], align 8

// CHECK3-NEXT:    store double [[TMP9]], ptr [[TMP7]], align 8

// CHECK3-NEXT:    ret void

//
@@ -2969,13 +2969,13 @@ int bar(int n){ 
// CHECK3-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP7]], align 1

// CHECK3-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 128

// CHECK3-NEXT:    store i8 [[TMP9]], ptr [[TMP8]], align 4

// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK3-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2048 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4

// CHECK3-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 128

// CHECK3-NEXT:    store float [[TMP13]], ptr [[TMP12]], align 4

// CHECK3-NEXT:    ret void

//

//
@@ -3020,13 +3020,13 @@ int bar(int n){ 
// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK3-NEXT:    [[C:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x i8], ptr [[C]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 128

// CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP8]], align 4

// CHECK3-NEXT:    store i8 [[TMP9]], ptr [[TMP7]], align 1

// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK3-NEXT:    [[D:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_0]], ptr [[TMP4]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2048 x float], ptr [[D]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 128

// CHECK3-NEXT:    [[TMP13:%.*]] = load float, ptr [[TMP12]], align 4

// CHECK3-NEXT:    store float [[TMP13]], ptr [[TMP11]], align 4

// CHECK3-NEXT:    ret void

//
@@ -3542,13 +3542,13 @@ int bar(int n){ 
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP7]], align 4

// CHECK3-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 128

// CHECK3-NEXT:    store i32 [[TMP9]], ptr [[TMP8]], align 4

// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK3-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2048 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP11]], align 2

// CHECK3-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 128

// CHECK3-NEXT:    store i16 [[TMP13]], ptr [[TMP12]], align 4

// CHECK3-NEXT:    ret void

//

//
@@ -3593,13 +3593,13 @@ int bar(int n){ 
// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4

// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2:%.*]], ptr [[TMP4]], i32 0, i32 0

// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2048 x i32], ptr [[A]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 128

// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4

// CHECK3-NEXT:    store i32 [[TMP9]], ptr [[TMP7]], align 4

// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[TMP3]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 4

// CHECK3-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_2]], ptr [[TMP4]], i32 0, i32 1

// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2048 x i16], ptr [[B]], i32 0, i32 [[TMP5]]

// CHECK3-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 128

// CHECK3-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 4

// CHECK3-NEXT:    store i16 [[TMP13]], ptr [[TMP11]], align 2

// CHECK3-NEXT:    ret void

//

clang/test/OpenMP/target_teams_generic_loop_codegen.cpp

+10 −10
Original line number Diff line number Diff line
@@ -1815,7 +1815,7 @@ int foo() { 
// IR-GPU-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8

// IR-GPU-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// IR-GPU-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP5]]

// IR-GPU-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 128 [[TMP8]], ptr align 4 [[TMP7]], i64 400, i1 false)

// IR-GPU-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP8]], ptr align 4 [[TMP7]], i64 400, i1 false)

// IR-GPU-NEXT:    ret void

//

//
@@ -1863,7 +1863,7 @@ int foo() { 
// IR-GPU-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8

// IR-GPU-NEXT:    [[SUM:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 0, i32 0

// IR-GPU-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1024 x [10 x [10 x i32]]], ptr [[SUM]], i32 0, i32 [[TMP5]]

// IR-GPU-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP7]], ptr align 128 [[TMP8]], i64 400, i1 false)

// IR-GPU-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP7]], ptr align 4 [[TMP8]], i64 400, i1 false)

// IR-GPU-NEXT:    ret void

//

//

clang/test/OpenMP/reduction_implicit_map.cpp

+16 −16

File changed.

Contains only whitespace changes.