1 files changed, 85 insertions, 0 deletions
diff --git a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
index 84fc77a1420f..bbbacd25bca6 100644
--- a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
+++ b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
@@ -646,6 +646,10 @@ static constexpr IntrinsicHandler handlers[]{
        {"dim", asValue},
        {"mask", asBox, handleDynamicOptional}}},
      /*isElemental=*/false},
+    {"syncthreads", &I::genSyncThreads, {}, /*isElemental=*/false},
+    {"syncthreads_and", &I::genSyncThreadsAnd, {}, /*isElemental=*/false},
+    {"syncthreads_count", &I::genSyncThreadsCount, {}, /*isElemental=*/false},
+    {"syncthreads_or", &I::genSyncThreadsOr, {}, /*isElemental=*/false},
     {"system",
      &I::genSystem,
      {{{"command", asBox}, {"exitstat", asBox, handleDynamicOptional}}},
@@ -655,6 +659,9 @@ static constexpr IntrinsicHandler handlers[]{
      {{{"count", asAddr}, {"count_rate", asAddr}, {"count_max", asAddr}}},
      /*isElemental=*/false},
     {"tand", &I::genTand},
+    {"threadfence", &I::genThreadFence, {}, /*isElemental=*/false},
+    {"threadfence_block", &I::genThreadFenceBlock, {}, /*isElemental=*/false},
+    {"threadfence_system", &I::genThreadFenceSystem, {}, /*isElemental=*/false},
     {"trailz", &I::genTrailz},
     {"transfer",
      &I::genTransfer,
@@ -7437,6 +7444,52 @@ IntrinsicLibrary::genSum(mlir::Type resultType,
                       resultType, args);
 }
 
+// SYNCTHREADS
+void IntrinsicLibrary::genSyncThreads(llvm::ArrayRef<fir::ExtendedValue> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.barrier0";
+  mlir::FunctionType funcType =
+      mlir::FunctionType::get(builder.getContext(), {}, {});
+  auto funcOp = builder.createFunction(loc, funcName, funcType);
+  llvm::SmallVector<mlir::Value> noArgs;
+  builder.create<fir::CallOp>(loc, funcOp, noArgs);
+}
+
+// SYNCTHREADS_AND
+mlir::Value
+IntrinsicLibrary::genSyncThreadsAnd(mlir::Type resultType,
+                                    llvm::ArrayRef<mlir::Value> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.barrier0.and";
+  mlir::MLIRContext *context = builder.getContext();
+  mlir::FunctionType ftype =
+      mlir::FunctionType::get(context, {resultType}, {args[0].getType()});
+  auto funcOp = builder.createFunction(loc, funcName, ftype);
+  return builder.create<fir::CallOp>(loc, funcOp, args).getResult(0);
+}
+
+// SYNCTHREADS_COUNT
+mlir::Value
+IntrinsicLibrary::genSyncThreadsCount(mlir::Type resultType,
+                                      llvm::ArrayRef<mlir::Value> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.barrier0.popc";
+  mlir::MLIRContext *context = builder.getContext();
+  mlir::FunctionType ftype =
+      mlir::FunctionType::get(context, {resultType}, {args[0].getType()});
+  auto funcOp = builder.createFunction(loc, funcName, ftype);
+  return builder.create<fir::CallOp>(loc, funcOp, args).getResult(0);
+}
+
+// SYNCTHREADS_OR
+mlir::Value
+IntrinsicLibrary::genSyncThreadsOr(mlir::Type resultType,
+                                   llvm::ArrayRef<mlir::Value> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.barrier0.or";
+  mlir::MLIRContext *context = builder.getContext();
+  mlir::FunctionType ftype =
+      mlir::FunctionType::get(context, {resultType}, {args[0].getType()});
+  auto funcOp = builder.createFunction(loc, funcName, ftype);
+  return builder.create<fir::CallOp>(loc, funcOp, args).getResult(0);
+}
+
 // SYSTEM
 fir::ExtendedValue
 IntrinsicLibrary::genSystem(std::optional<mlir::Type> resultType,
@@ -7567,6 +7620,38 @@ IntrinsicLibrary::genTranspose(mlir::Type resultType,
   return readAndAddCleanUp(resultMutableBox, resultType, "TRANSPOSE");
 }
 
+// THREADFENCE
+void IntrinsicLibrary::genThreadFence(llvm::ArrayRef<fir::ExtendedValue> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.membar.gl";
+  mlir::FunctionType funcType =
+      mlir::FunctionType::get(builder.getContext(), {}, {});
+  auto funcOp = builder.createFunction(loc, funcName, funcType);
+  llvm::SmallVector<mlir::Value> noArgs;
+  builder.create<fir::CallOp>(loc, funcOp, noArgs);
+}
+
+// THREADFENCE_BLOCK
+void IntrinsicLibrary::genThreadFenceBlock(
+    llvm::ArrayRef<fir::ExtendedValue> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.membar.cta";
+  mlir::FunctionType funcType =
+      mlir::FunctionType::get(builder.getContext(), {}, {});
+  auto funcOp = builder.createFunction(loc, funcName, funcType);
+  llvm::SmallVector<mlir::Value> noArgs;
+  builder.create<fir::CallOp>(loc, funcOp, noArgs);
+}
+
+// THREADFENCE_SYSTEM
+void IntrinsicLibrary::genThreadFenceSystem(
+    llvm::ArrayRef<fir::ExtendedValue> args) {
+  constexpr llvm::StringLiteral funcName = "llvm.nvvm.membar.sys";
+  mlir::FunctionType funcType =
+      mlir::FunctionType::get(builder.getContext(), {}, {});
+  auto funcOp = builder.createFunction(loc, funcName, funcType);
+  llvm::SmallVector<mlir::Value> noArgs;
+  builder.create<fir::CallOp>(loc, funcOp, noArgs);
+}
+
 // TRIM
 fir::ExtendedValue
 IntrinsicLibrary::genTrim(mlir::Type resultType,