[mlir][OpenMP] Lower device clause for target data/enter/exit/update #174665
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
mlir:llvm
mlir
flang
Member
|
@llvm/pr-subscribers-flang-fir-hlfir @llvm/pr-subscribers-mlir-llvm Author: Chi-Chun, Chen (chichunchen) ChangesAllow the OpenMP device clause on target data/enter/exit/update to be lowered from variables instead of constatns only. Patch is 22.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/174665.diff 4 Files Affected:
diff --git a/flang/docs/OpenMPSupport.md b/flang/docs/OpenMPSupport.md
index 21966c5489108..1b25fc4890847 100644
--- a/flang/docs/OpenMPSupport.md
+++ b/flang/docs/OpenMPSupport.md
@@ -37,9 +37,9 @@ Note : No distinction is made between the support in Parser/Semantics, MLIR, Low
| simd construct | P | Implicit linearization is skipped if iv is a pointer or allocatable|
| declare simd construct | N | |
| do simd construct | P | linear clause is not supported |
-| target data construct | P | device clause not supported |
+| target data construct | Y | |
| target construct | Y | |
-| target update construct | P | device clause not supported |
+| target update construct | Y | |
| declare target directive | Y | |
| teams construct | Y | |
| distribute construct | Y | |
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 614f06017a324..2698a4b4e89db 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -456,11 +456,7 @@ static LogicalResult checkImplementationStatus(Operation &op) {
omp::AtomicCaptureOp>([&](auto op) { checkHint(op, result); })
.Case<omp::TargetEnterDataOp, omp::TargetExitDataOp>(
[&](auto op) { checkDepend(op, result); })
- .Case<omp::TargetUpdateOp>([&](auto op) {
- checkDepend(op, result);
- checkDevice(op, result);
- })
- .Case<omp::TargetDataOp>([&](auto op) { checkDevice(op, result); })
+ .Case<omp::TargetUpdateOp>([&](auto op) { checkDepend(op, result); })
.Case([&](omp::TargetOp op) {
checkAllocate(op, result);
checkBare(op, result);
@@ -5051,7 +5047,7 @@ static LogicalResult
convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
LLVM::ModuleTranslation &moduleTranslation) {
llvm::Value *ifCond = nullptr;
- int64_t deviceID = llvm::omp::OMP_DEVICEID_UNDEF;
+ llvm::Value *deviceID = builder.getInt64(llvm::omp::OMP_DEVICEID_UNDEF);
SmallVector<Value> mapVars;
SmallVector<Value> useDevicePtrVars;
SmallVector<Value> useDeviceAddrVars;
@@ -5067,6 +5063,11 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
bool isOffloadEntry =
isTargetDevice || !ompBuilder->Config.TargetTriples.empty();
+ auto getDeviceID = [&](mlir::Value dev) -> llvm::Value * {
+ llvm::Value *v = moduleTranslation.lookupValue(dev);
+ return builder.CreateIntCast(v, builder.getInt64Ty(), /*isSigned=*/true);
+ };
+
LogicalResult result =
llvm::TypeSwitch<Operation *, LogicalResult>(op)
.Case([&](omp::TargetDataOp dataOp) {
@@ -5076,10 +5077,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = dataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = dataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = dataOp.getDevice())
+ deviceID = getDeviceID(devId);
mapVars = dataOp.getMapVars();
useDevicePtrVars = dataOp.getUseDevicePtrVars();
@@ -5093,10 +5092,9 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = enterDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = enterDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = enterDataOp.getDevice())
+ deviceID = getDeviceID(devId);
+
RTLFn =
enterDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_begin_nowait_mapper
@@ -5112,10 +5110,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = exitDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = exitDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = exitDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn = exitDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_end_nowait_mapper
@@ -5131,10 +5127,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = updateDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = updateDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = updateDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn =
updateDataOp.getNowait()
@@ -5287,13 +5281,13 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP = [&]() {
if (isa<omp::TargetDataOp>(op))
return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
- builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB,
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB,
/*MapperFunc=*/nullptr, bodyGenCB,
/*DeviceAddrCB=*/nullptr);
- return ompBuilder->createTargetData(
- ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB, &RTLFn);
+ return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB, &RTLFn);
}();
if (failed(handleError(afterIP, *op)))
diff --git a/mlir/test/Target/LLVMIR/omptarget-device.mlir b/mlir/test/Target/LLVMIR/omptarget-device.mlir
index b4c9744cc0c87..64f488a7bb44a 100644
--- a/mlir/test/Target/LLVMIR/omptarget-device.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-device.mlir
@@ -1,7 +1,7 @@
-// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
- llvm.func @foo(%d16 : i16, %d32 : i32, %d64 : i64) {
+ llvm.func @_QPopenmp_target(%d16 : i16, %d32 : i32, %d64 : i64) {
%x = llvm.mlir.constant(0 : i32) : i32
// Constant i16 -> i64 in the runtime call.
@@ -47,7 +47,7 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
}
}
-// CHECK-LABEL: define void @foo(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK-LABEL: define void @_QPopenmp_target(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
// CHECK: br label %entry
// CHECK: entry:
@@ -65,4 +65,210 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
// Variable i64
-// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
\ No newline at end of file
+// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPopenmp_target_data(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %one = llvm.mlir.constant(1 : i64) : i64
+ %buf = llvm.alloca %one x i32 : (i64) -> !llvm.ptr
+ %map = omp.map.info var_ptr(%buf : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = ""}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_data device(%c1_i16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_data device(%c2_i32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_data device(%c3_i64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i16 -> cast to i64.
+ omp.target_data device(%d16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i32 -> cast to i64.
+ omp.target_data device(%d32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i64 stays i64.
+ omp.target_data device(%d64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPopenmp_target_data(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPomp_target_enter_exit(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+
+ %m_to = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+ %m_from = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(from) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_enter_data device(%c1_i16 : i16) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_enter_data device(%c2_i32 : i32) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_enter_data device(%c3_i64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Variable cases (enter) ----
+ omp.target_enter_data device(%d16 : i16) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d32 : i32) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Constant cases (exit) ----
+ omp.target_exit_data device(%c1_i16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c2_i32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c3_i64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ // ---- Variable cases (exit) ----
+ omp.target_exit_data device(%d16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPomp_target_enter_exit(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant enter cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable enter cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_BEGIN:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_BEGIN:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Constant exit cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable exit cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_END:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_END:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @target_update_dev_clause(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+ %m = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // ---- Constant cases ----
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_update device(%c1_i16 : i16) map_entries(%m : !llvm.ptr)
+
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_update device(%c2_i32 : i32) map_entries(%m : !llvm.ptr)
+
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_update device(%c3_i64 : i64) map_entries(%m : !llvm.ptr)
+
+ // ---- Variable cases ----
+ omp.target_update device(%d16 : i16) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d32 : i32) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d64 : i64) map_entries(%m : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @target_update_dev_clause(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
\ No newline at end of file
diff --git a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
index e289d5d013eaa..0b4d63125f82f 100644
--- a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
@@ -156,6 +156,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: %[[VAL_8:.*]] = load i32, ptr %[[VAL_7]], align 4
// CHECK: %[[VAL_9:.*]] = icmp slt i32 %[[VAL_8]], 10
// CHECK: %[[VAL_10:.*]] = load i32, ptr %[[VAL_6]], align 4
+// CHECK: %[[DEV_I64_BEGIN:.*]] = sext i32 %[[VAL_10:.*]] to i64
// CHECK: br label %[[VAL_11:.*]]
// CHECK: entry: ; preds = %[[VAL_12:.*]]
// CHECK: br i1 %[[VAL_9]], label %[[VAL_13:.*]], label %[[VAL_14:.*]]
@@ -176,7 +177,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: ...
[truncated]
|
Member
|
@llvm/pr-subscribers-mlir-openmp Author: Chi-Chun, Chen (chichunchen) ChangesAllow the OpenMP device clause on target data/enter/exit/update to be lowered from variables instead of constatns only. Patch is 22.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/174665.diff 4 Files Affected:
diff --git a/flang/docs/OpenMPSupport.md b/flang/docs/OpenMPSupport.md
index 21966c5489108..1b25fc4890847 100644
--- a/flang/docs/OpenMPSupport.md
+++ b/flang/docs/OpenMPSupport.md
@@ -37,9 +37,9 @@ Note : No distinction is made between the support in Parser/Semantics, MLIR, Low
| simd construct | P | Implicit linearization is skipped if iv is a pointer or allocatable|
| declare simd construct | N | |
| do simd construct | P | linear clause is not supported |
-| target data construct | P | device clause not supported |
+| target data construct | Y | |
| target construct | Y | |
-| target update construct | P | device clause not supported |
+| target update construct | Y | |
| declare target directive | Y | |
| teams construct | Y | |
| distribute construct | Y | |
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 614f06017a324..2698a4b4e89db 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -456,11 +456,7 @@ static LogicalResult checkImplementationStatus(Operation &op) {
omp::AtomicCaptureOp>([&](auto op) { checkHint(op, result); })
.Case<omp::TargetEnterDataOp, omp::TargetExitDataOp>(
[&](auto op) { checkDepend(op, result); })
- .Case<omp::TargetUpdateOp>([&](auto op) {
- checkDepend(op, result);
- checkDevice(op, result);
- })
- .Case<omp::TargetDataOp>([&](auto op) { checkDevice(op, result); })
+ .Case<omp::TargetUpdateOp>([&](auto op) { checkDepend(op, result); })
.Case([&](omp::TargetOp op) {
checkAllocate(op, result);
checkBare(op, result);
@@ -5051,7 +5047,7 @@ static LogicalResult
convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
LLVM::ModuleTranslation &moduleTranslation) {
llvm::Value *ifCond = nullptr;
- int64_t deviceID = llvm::omp::OMP_DEVICEID_UNDEF;
+ llvm::Value *deviceID = builder.getInt64(llvm::omp::OMP_DEVICEID_UNDEF);
SmallVector<Value> mapVars;
SmallVector<Value> useDevicePtrVars;
SmallVector<Value> useDeviceAddrVars;
@@ -5067,6 +5063,11 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
bool isOffloadEntry =
isTargetDevice || !ompBuilder->Config.TargetTriples.empty();
+ auto getDeviceID = [&](mlir::Value dev) -> llvm::Value * {
+ llvm::Value *v = moduleTranslation.lookupValue(dev);
+ return builder.CreateIntCast(v, builder.getInt64Ty(), /*isSigned=*/true);
+ };
+
LogicalResult result =
llvm::TypeSwitch<Operation *, LogicalResult>(op)
.Case([&](omp::TargetDataOp dataOp) {
@@ -5076,10 +5077,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = dataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = dataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = dataOp.getDevice())
+ deviceID = getDeviceID(devId);
mapVars = dataOp.getMapVars();
useDevicePtrVars = dataOp.getUseDevicePtrVars();
@@ -5093,10 +5092,9 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = enterDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = enterDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = enterDataOp.getDevice())
+ deviceID = getDeviceID(devId);
+
RTLFn =
enterDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_begin_nowait_mapper
@@ -5112,10 +5110,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = exitDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = exitDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = exitDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn = exitDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_end_nowait_mapper
@@ -5131,10 +5127,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = updateDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = updateDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = updateDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn =
updateDataOp.getNowait()
@@ -5287,13 +5281,13 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP = [&]() {
if (isa<omp::TargetDataOp>(op))
return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
- builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB,
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB,
/*MapperFunc=*/nullptr, bodyGenCB,
/*DeviceAddrCB=*/nullptr);
- return ompBuilder->createTargetData(
- ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB, &RTLFn);
+ return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB, &RTLFn);
}();
if (failed(handleError(afterIP, *op)))
diff --git a/mlir/test/Target/LLVMIR/omptarget-device.mlir b/mlir/test/Target/LLVMIR/omptarget-device.mlir
index b4c9744cc0c87..64f488a7bb44a 100644
--- a/mlir/test/Target/LLVMIR/omptarget-device.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-device.mlir
@@ -1,7 +1,7 @@
-// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
- llvm.func @foo(%d16 : i16, %d32 : i32, %d64 : i64) {
+ llvm.func @_QPopenmp_target(%d16 : i16, %d32 : i32, %d64 : i64) {
%x = llvm.mlir.constant(0 : i32) : i32
// Constant i16 -> i64 in the runtime call.
@@ -47,7 +47,7 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
}
}
-// CHECK-LABEL: define void @foo(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK-LABEL: define void @_QPopenmp_target(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
// CHECK: br label %entry
// CHECK: entry:
@@ -65,4 +65,210 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
// Variable i64
-// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
\ No newline at end of file
+// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPopenmp_target_data(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %one = llvm.mlir.constant(1 : i64) : i64
+ %buf = llvm.alloca %one x i32 : (i64) -> !llvm.ptr
+ %map = omp.map.info var_ptr(%buf : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = ""}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_data device(%c1_i16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_data device(%c2_i32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_data device(%c3_i64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i16 -> cast to i64.
+ omp.target_data device(%d16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i32 -> cast to i64.
+ omp.target_data device(%d32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i64 stays i64.
+ omp.target_data device(%d64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPopenmp_target_data(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPomp_target_enter_exit(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+
+ %m_to = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+ %m_from = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(from) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_enter_data device(%c1_i16 : i16) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_enter_data device(%c2_i32 : i32) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_enter_data device(%c3_i64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Variable cases (enter) ----
+ omp.target_enter_data device(%d16 : i16) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d32 : i32) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Constant cases (exit) ----
+ omp.target_exit_data device(%c1_i16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c2_i32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c3_i64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ // ---- Variable cases (exit) ----
+ omp.target_exit_data device(%d16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPomp_target_enter_exit(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant enter cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable enter cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_BEGIN:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_BEGIN:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Constant exit cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable exit cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_END:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_END:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @target_update_dev_clause(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+ %m = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // ---- Constant cases ----
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_update device(%c1_i16 : i16) map_entries(%m : !llvm.ptr)
+
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_update device(%c2_i32 : i32) map_entries(%m : !llvm.ptr)
+
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_update device(%c3_i64 : i64) map_entries(%m : !llvm.ptr)
+
+ // ---- Variable cases ----
+ omp.target_update device(%d16 : i16) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d32 : i32) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d64 : i64) map_entries(%m : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @target_update_dev_clause(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
\ No newline at end of file
diff --git a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
index e289d5d013eaa..0b4d63125f82f 100644
--- a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
@@ -156,6 +156,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: %[[VAL_8:.*]] = load i32, ptr %[[VAL_7]], align 4
// CHECK: %[[VAL_9:.*]] = icmp slt i32 %[[VAL_8]], 10
// CHECK: %[[VAL_10:.*]] = load i32, ptr %[[VAL_6]], align 4
+// CHECK: %[[DEV_I64_BEGIN:.*]] = sext i32 %[[VAL_10:.*]] to i64
// CHECK: br label %[[VAL_11:.*]]
// CHECK: entry: ; preds = %[[VAL_12:.*]]
// CHECK: br i1 %[[VAL_9]], label %[[VAL_13:.*]], label %[[VAL_14:.*]]
@@ -176,7 +177,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: ...
[truncated]
|
Member
|
@llvm/pr-subscribers-mlir Author: Chi-Chun, Chen (chichunchen) ChangesAllow the OpenMP device clause on target data/enter/exit/update to be lowered from variables instead of constatns only. Patch is 22.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/174665.diff 4 Files Affected:
diff --git a/flang/docs/OpenMPSupport.md b/flang/docs/OpenMPSupport.md
index 21966c5489108..1b25fc4890847 100644
--- a/flang/docs/OpenMPSupport.md
+++ b/flang/docs/OpenMPSupport.md
@@ -37,9 +37,9 @@ Note : No distinction is made between the support in Parser/Semantics, MLIR, Low
| simd construct | P | Implicit linearization is skipped if iv is a pointer or allocatable|
| declare simd construct | N | |
| do simd construct | P | linear clause is not supported |
-| target data construct | P | device clause not supported |
+| target data construct | Y | |
| target construct | Y | |
-| target update construct | P | device clause not supported |
+| target update construct | Y | |
| declare target directive | Y | |
| teams construct | Y | |
| distribute construct | Y | |
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 614f06017a324..2698a4b4e89db 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -456,11 +456,7 @@ static LogicalResult checkImplementationStatus(Operation &op) {
omp::AtomicCaptureOp>([&](auto op) { checkHint(op, result); })
.Case<omp::TargetEnterDataOp, omp::TargetExitDataOp>(
[&](auto op) { checkDepend(op, result); })
- .Case<omp::TargetUpdateOp>([&](auto op) {
- checkDepend(op, result);
- checkDevice(op, result);
- })
- .Case<omp::TargetDataOp>([&](auto op) { checkDevice(op, result); })
+ .Case<omp::TargetUpdateOp>([&](auto op) { checkDepend(op, result); })
.Case([&](omp::TargetOp op) {
checkAllocate(op, result);
checkBare(op, result);
@@ -5051,7 +5047,7 @@ static LogicalResult
convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
LLVM::ModuleTranslation &moduleTranslation) {
llvm::Value *ifCond = nullptr;
- int64_t deviceID = llvm::omp::OMP_DEVICEID_UNDEF;
+ llvm::Value *deviceID = builder.getInt64(llvm::omp::OMP_DEVICEID_UNDEF);
SmallVector<Value> mapVars;
SmallVector<Value> useDevicePtrVars;
SmallVector<Value> useDeviceAddrVars;
@@ -5067,6 +5063,11 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
bool isOffloadEntry =
isTargetDevice || !ompBuilder->Config.TargetTriples.empty();
+ auto getDeviceID = [&](mlir::Value dev) -> llvm::Value * {
+ llvm::Value *v = moduleTranslation.lookupValue(dev);
+ return builder.CreateIntCast(v, builder.getInt64Ty(), /*isSigned=*/true);
+ };
+
LogicalResult result =
llvm::TypeSwitch<Operation *, LogicalResult>(op)
.Case([&](omp::TargetDataOp dataOp) {
@@ -5076,10 +5077,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = dataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = dataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = dataOp.getDevice())
+ deviceID = getDeviceID(devId);
mapVars = dataOp.getMapVars();
useDevicePtrVars = dataOp.getUseDevicePtrVars();
@@ -5093,10 +5092,9 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = enterDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = enterDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = enterDataOp.getDevice())
+ deviceID = getDeviceID(devId);
+
RTLFn =
enterDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_begin_nowait_mapper
@@ -5112,10 +5110,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = exitDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = exitDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = exitDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn = exitDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_end_nowait_mapper
@@ -5131,10 +5127,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = updateDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = updateDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = updateDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn =
updateDataOp.getNowait()
@@ -5287,13 +5281,13 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP = [&]() {
if (isa<omp::TargetDataOp>(op))
return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
- builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB,
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB,
/*MapperFunc=*/nullptr, bodyGenCB,
/*DeviceAddrCB=*/nullptr);
- return ompBuilder->createTargetData(
- ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB, &RTLFn);
+ return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB, &RTLFn);
}();
if (failed(handleError(afterIP, *op)))
diff --git a/mlir/test/Target/LLVMIR/omptarget-device.mlir b/mlir/test/Target/LLVMIR/omptarget-device.mlir
index b4c9744cc0c87..64f488a7bb44a 100644
--- a/mlir/test/Target/LLVMIR/omptarget-device.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-device.mlir
@@ -1,7 +1,7 @@
-// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
- llvm.func @foo(%d16 : i16, %d32 : i32, %d64 : i64) {
+ llvm.func @_QPopenmp_target(%d16 : i16, %d32 : i32, %d64 : i64) {
%x = llvm.mlir.constant(0 : i32) : i32
// Constant i16 -> i64 in the runtime call.
@@ -47,7 +47,7 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
}
}
-// CHECK-LABEL: define void @foo(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK-LABEL: define void @_QPopenmp_target(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
// CHECK: br label %entry
// CHECK: entry:
@@ -65,4 +65,210 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
// Variable i64
-// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
\ No newline at end of file
+// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPopenmp_target_data(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %one = llvm.mlir.constant(1 : i64) : i64
+ %buf = llvm.alloca %one x i32 : (i64) -> !llvm.ptr
+ %map = omp.map.info var_ptr(%buf : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = ""}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_data device(%c1_i16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_data device(%c2_i32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_data device(%c3_i64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i16 -> cast to i64.
+ omp.target_data device(%d16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i32 -> cast to i64.
+ omp.target_data device(%d32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i64 stays i64.
+ omp.target_data device(%d64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPopenmp_target_data(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPomp_target_enter_exit(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+
+ %m_to = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+ %m_from = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(from) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_enter_data device(%c1_i16 : i16) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_enter_data device(%c2_i32 : i32) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_enter_data device(%c3_i64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Variable cases (enter) ----
+ omp.target_enter_data device(%d16 : i16) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d32 : i32) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Constant cases (exit) ----
+ omp.target_exit_data device(%c1_i16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c2_i32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c3_i64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ // ---- Variable cases (exit) ----
+ omp.target_exit_data device(%d16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPomp_target_enter_exit(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant enter cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable enter cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_BEGIN:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_BEGIN:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Constant exit cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable exit cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_END:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_END:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @target_update_dev_clause(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+ %m = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // ---- Constant cases ----
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_update device(%c1_i16 : i16) map_entries(%m : !llvm.ptr)
+
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_update device(%c2_i32 : i32) map_entries(%m : !llvm.ptr)
+
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_update device(%c3_i64 : i64) map_entries(%m : !llvm.ptr)
+
+ // ---- Variable cases ----
+ omp.target_update device(%d16 : i16) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d32 : i32) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d64 : i64) map_entries(%m : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @target_update_dev_clause(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
\ No newline at end of file
diff --git a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
index e289d5d013eaa..0b4d63125f82f 100644
--- a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
@@ -156,6 +156,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: %[[VAL_8:.*]] = load i32, ptr %[[VAL_7]], align 4
// CHECK: %[[VAL_9:.*]] = icmp slt i32 %[[VAL_8]], 10
// CHECK: %[[VAL_10:.*]] = load i32, ptr %[[VAL_6]], align 4
+// CHECK: %[[DEV_I64_BEGIN:.*]] = sext i32 %[[VAL_10:.*]] to i64
// CHECK: br label %[[VAL_11:.*]]
// CHECK: entry: ; preds = %[[VAL_12:.*]]
// CHECK: br i1 %[[VAL_9]], label %[[VAL_13:.*]], label %[[VAL_14:.*]]
@@ -176,7 +177,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: ...
[truncated]
|
Member
|
@llvm/pr-subscribers-flang-openmp Author: Chi-Chun, Chen (chichunchen) ChangesAllow the OpenMP device clause on target data/enter/exit/update to be lowered from variables instead of constatns only. Patch is 22.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/174665.diff 4 Files Affected:
diff --git a/flang/docs/OpenMPSupport.md b/flang/docs/OpenMPSupport.md
index 21966c5489108..1b25fc4890847 100644
--- a/flang/docs/OpenMPSupport.md
+++ b/flang/docs/OpenMPSupport.md
@@ -37,9 +37,9 @@ Note : No distinction is made between the support in Parser/Semantics, MLIR, Low
| simd construct | P | Implicit linearization is skipped if iv is a pointer or allocatable|
| declare simd construct | N | |
| do simd construct | P | linear clause is not supported |
-| target data construct | P | device clause not supported |
+| target data construct | Y | |
| target construct | Y | |
-| target update construct | P | device clause not supported |
+| target update construct | Y | |
| declare target directive | Y | |
| teams construct | Y | |
| distribute construct | Y | |
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 614f06017a324..2698a4b4e89db 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -456,11 +456,7 @@ static LogicalResult checkImplementationStatus(Operation &op) {
omp::AtomicCaptureOp>([&](auto op) { checkHint(op, result); })
.Case<omp::TargetEnterDataOp, omp::TargetExitDataOp>(
[&](auto op) { checkDepend(op, result); })
- .Case<omp::TargetUpdateOp>([&](auto op) {
- checkDepend(op, result);
- checkDevice(op, result);
- })
- .Case<omp::TargetDataOp>([&](auto op) { checkDevice(op, result); })
+ .Case<omp::TargetUpdateOp>([&](auto op) { checkDepend(op, result); })
.Case([&](omp::TargetOp op) {
checkAllocate(op, result);
checkBare(op, result);
@@ -5051,7 +5047,7 @@ static LogicalResult
convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
LLVM::ModuleTranslation &moduleTranslation) {
llvm::Value *ifCond = nullptr;
- int64_t deviceID = llvm::omp::OMP_DEVICEID_UNDEF;
+ llvm::Value *deviceID = builder.getInt64(llvm::omp::OMP_DEVICEID_UNDEF);
SmallVector<Value> mapVars;
SmallVector<Value> useDevicePtrVars;
SmallVector<Value> useDeviceAddrVars;
@@ -5067,6 +5063,11 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
bool isOffloadEntry =
isTargetDevice || !ompBuilder->Config.TargetTriples.empty();
+ auto getDeviceID = [&](mlir::Value dev) -> llvm::Value * {
+ llvm::Value *v = moduleTranslation.lookupValue(dev);
+ return builder.CreateIntCast(v, builder.getInt64Ty(), /*isSigned=*/true);
+ };
+
LogicalResult result =
llvm::TypeSwitch<Operation *, LogicalResult>(op)
.Case([&](omp::TargetDataOp dataOp) {
@@ -5076,10 +5077,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = dataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = dataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = dataOp.getDevice())
+ deviceID = getDeviceID(devId);
mapVars = dataOp.getMapVars();
useDevicePtrVars = dataOp.getUseDevicePtrVars();
@@ -5093,10 +5092,9 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = enterDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = enterDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = enterDataOp.getDevice())
+ deviceID = getDeviceID(devId);
+
RTLFn =
enterDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_begin_nowait_mapper
@@ -5112,10 +5110,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = exitDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = exitDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = exitDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn = exitDataOp.getNowait()
? llvm::omp::OMPRTL___tgt_target_data_end_nowait_mapper
@@ -5131,10 +5127,8 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
if (auto ifVar = updateDataOp.getIfExpr())
ifCond = moduleTranslation.lookupValue(ifVar);
- if (auto devId = updateDataOp.getDevice())
- if (auto constOp = devId.getDefiningOp<LLVM::ConstantOp>())
- if (auto intAttr = dyn_cast<IntegerAttr>(constOp.getValue()))
- deviceID = intAttr.getInt();
+ if (mlir::Value devId = updateDataOp.getDevice())
+ deviceID = getDeviceID(devId);
RTLFn =
updateDataOp.getNowait()
@@ -5287,13 +5281,13 @@ convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP = [&]() {
if (isa<omp::TargetDataOp>(op))
return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
- builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB,
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB,
/*MapperFunc=*/nullptr, bodyGenCB,
/*DeviceAddrCB=*/nullptr);
- return ompBuilder->createTargetData(
- ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
- info, genMapInfoCB, customMapperCB, &RTLFn);
+ return ompBuilder->createTargetData(ompLoc, allocaIP, builder.saveIP(),
+ deviceID, ifCond, info, genMapInfoCB,
+ customMapperCB, &RTLFn);
}();
if (failed(handleError(afterIP, *op)))
diff --git a/mlir/test/Target/LLVMIR/omptarget-device.mlir b/mlir/test/Target/LLVMIR/omptarget-device.mlir
index b4c9744cc0c87..64f488a7bb44a 100644
--- a/mlir/test/Target/LLVMIR/omptarget-device.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-device.mlir
@@ -1,7 +1,7 @@
-// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+// RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
- llvm.func @foo(%d16 : i16, %d32 : i32, %d64 : i64) {
+ llvm.func @_QPopenmp_target(%d16 : i16, %d32 : i32, %d64 : i64) {
%x = llvm.mlir.constant(0 : i32) : i32
// Constant i16 -> i64 in the runtime call.
@@ -47,7 +47,7 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
}
}
-// CHECK-LABEL: define void @foo(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK-LABEL: define void @_QPopenmp_target(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
// CHECK: br label %entry
// CHECK: entry:
@@ -65,4 +65,210 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-
// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
// Variable i64
-// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
\ No newline at end of file
+// CHECK: call i32 @__tgt_target_kernel(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPopenmp_target_data(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %one = llvm.mlir.constant(1 : i64) : i64
+ %buf = llvm.alloca %one x i32 : (i64) -> !llvm.ptr
+ %map = omp.map.info var_ptr(%buf : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = ""}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_data device(%c1_i16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_data device(%c2_i32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_data device(%c3_i64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i16 -> cast to i64.
+ omp.target_data device(%d16 : i16) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i32 -> cast to i64.
+ omp.target_data device(%d32 : i32) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ // Variable i64 stays i64.
+ omp.target_data device(%d64 : i64) map_entries(%map : !llvm.ptr) {
+ omp.terminator
+ }
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPopenmp_target_data(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// Variable i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 1, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr null)
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @_QPomp_target_enter_exit(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+
+ %m_to = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+ %m_from = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(from) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // Constant i16 -> i64 in the runtime call.
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_enter_data device(%c1_i16 : i16) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i32 -> i64 in the runtime call.
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_enter_data device(%c2_i32 : i32) map_entries(%m_to : !llvm.ptr)
+
+ // Constant i64 stays i64 in the runtime call.
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_enter_data device(%c3_i64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Variable cases (enter) ----
+ omp.target_enter_data device(%d16 : i16) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d32 : i32) map_entries(%m_to : !llvm.ptr)
+ omp.target_enter_data device(%d64 : i64) map_entries(%m_to : !llvm.ptr)
+
+ // ---- Constant cases (exit) ----
+ omp.target_exit_data device(%c1_i16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c2_i32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%c3_i64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ // ---- Variable cases (exit) ----
+ omp.target_exit_data device(%d16 : i16) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d32 : i32) map_entries(%m_from : !llvm.ptr)
+ omp.target_exit_data device(%d64 : i64) map_entries(%m_from : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @_QPomp_target_enter_exit(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant enter cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable enter cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_BEGIN:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D16_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_BEGIN:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %[[D32_I64_BEGIN]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_begin_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Constant exit cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable exit cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64_END:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D16_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64_END:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %[[D32_I64_END]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_end_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
+
+// -----
+
+module attributes {omp.is_target_device = false, omp.target_triples = ["nvptx64-nvidia-cuda"]} {
+ llvm.func @target_update_dev_clause(%d16 : i16, %d32 : i32, %d64 : i64) {
+ %c1 = llvm.mlir.constant(1 : i64) : i64
+ %var = llvm.alloca %c1 x i32 : (i64) -> !llvm.ptr
+ %m = omp.map.info var_ptr(%var : !llvm.ptr, i32) map_clauses(to) capture(ByRef) -> !llvm.ptr {name = "var"}
+
+ // ---- Constant cases ----
+ %c1_i16 = llvm.mlir.constant(1 : i16) : i16
+ omp.target_update device(%c1_i16 : i16) map_entries(%m : !llvm.ptr)
+
+ %c2_i32 = llvm.mlir.constant(2 : i32) : i32
+ omp.target_update device(%c2_i32 : i32) map_entries(%m : !llvm.ptr)
+
+ %c3_i64 = llvm.mlir.constant(3 : i64) : i64
+ omp.target_update device(%c3_i64 : i64) map_entries(%m : !llvm.ptr)
+
+ // ---- Variable cases ----
+ omp.target_update device(%d16 : i16) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d32 : i32) map_entries(%m : !llvm.ptr)
+ omp.target_update device(%d64 : i64) map_entries(%m : !llvm.ptr)
+
+ llvm.return
+ }
+}
+
+// CHECK-LABEL: define void @target_update_dev_clause(i16 %{{.*}}, i32 %{{.*}}, i64 %{{.*}}) {
+// CHECK: br label %entry
+// CHECK: entry:
+
+// ---- Constant cases (device id is 2nd argument) ----
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 1, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 2, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+// CHECK-DAG: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 3, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// ---- Variable cases ----
+// Variable i16 -> i64
+// CHECK: %[[D16_I64:.*]] = sext i16 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D16_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i32 -> i64
+// CHECK: %[[D32_I64:.*]] = sext i32 %{{.*}} to i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %[[D32_I64]], i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// Variable i64 stays i64
+// CHECK: call void @__tgt_target_data_update_mapper(ptr {{.*}}, i64 %{{.*}}, i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}})
+
+// CHECK: ret void
+// CHECK: }
\ No newline at end of file
diff --git a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
index e289d5d013eaa..0b4d63125f82f 100644
--- a/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/omptarget-llvm.mlir
@@ -156,6 +156,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: %[[VAL_8:.*]] = load i32, ptr %[[VAL_7]], align 4
// CHECK: %[[VAL_9:.*]] = icmp slt i32 %[[VAL_8]], 10
// CHECK: %[[VAL_10:.*]] = load i32, ptr %[[VAL_6]], align 4
+// CHECK: %[[DEV_I64_BEGIN:.*]] = sext i32 %[[VAL_10:.*]] to i64
// CHECK: br label %[[VAL_11:.*]]
// CHECK: entry: ; preds = %[[VAL_12:.*]]
// CHECK: br i1 %[[VAL_9]], label %[[VAL_13:.*]], label %[[VAL_14:.*]]
@@ -176,7 +177,7 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
// CHECK: ...
[truncated]
|
🐧 Linux x64 Test Results
✅ The build succeeded and all tests passed. |
chichunchen
force-pushed
the
cchen/flang/enter_exit_device
branch
from
7bb1214 to
c844398
Compare
Extend OpenMP device clause lowering for target data, target enter data, target exit data, and target update to accept non-constant values. Previously, only constant device IDs could be lowered to LLVM IR. Add Flang tests to validate device clause handling and mark the feature as supported in the OpenMPSupport documentation. New tests cover: - target teams - target teams distribute - target teams distribute parallel do - target teams distribute parallel do simd - target data Tests for target update and target enter/exit were already present in Flang.
chichunchen
force-pushed
the
cchen/flang/enter_exit_device
branch
from
c844398 to
e02a664
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Extend OpenMP device clause lowering for target data, target enter data,
target exit data, and target update to accept non-constant values.
Previously, only constant device IDs could be lowered to LLVM IR.
Add Flang tests to validate device clause handling and mark the feature
as supported in the OpenMPSupport documentation. New tests cover:
Tests for target update and target enter/exit were
already present in Flang.