[flang][cuda] Lower attribute for module variables #81226
Merged
Conversation
This file contains 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
flang
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesPropagate the CUDA attribute to fir.global operation for simple module variables. Full diff: https://github.com/llvm/llvm-project/pull/81226.diff 5 Files Affected:
diff --git a/flang/include/flang/Optimizer/Builder/FIRBuilder.h b/flang/include/flang/Optimizer/Builder/FIRBuilder.h
index 5384f6e8121ec6..f50dacd327a7c0 100644
--- a/flang/include/flang/Optimizer/Builder/FIRBuilder.h
+++ b/flang/include/flang/Optimizer/Builder/FIRBuilder.h
@@ -230,12 +230,14 @@ class FirOpBuilder : public mlir::OpBuilder, public mlir::OpBuilder::Listener {
llvm::StringRef name,
mlir::StringAttr linkage = {},
mlir::Attribute value = {}, bool isConst = false,
- bool isTarget = false);
+ bool isTarget = false,
+ fir::CUDAAttributeAttr cudaAttr = {});
fir::GlobalOp createGlobal(mlir::Location loc, mlir::Type type,
llvm::StringRef name, bool isConst, bool isTarget,
std::function<void(FirOpBuilder &)> bodyBuilder,
- mlir::StringAttr linkage = {});
+ mlir::StringAttr linkage = {},
+ fir::CUDAAttributeAttr cudaAttr = {});
/// Create a global constant (read-only) value.
fir::GlobalOp createGlobalConstant(mlir::Location loc, mlir::Type type,
diff --git a/flang/include/flang/Optimizer/Dialect/FIROps.td b/flang/include/flang/Optimizer/Dialect/FIROps.td
index b954a0cc74d0e1..d505fedd6e6415 100644
--- a/flang/include/flang/Optimizer/Dialect/FIROps.td
+++ b/flang/include/flang/Optimizer/Dialect/FIROps.td
@@ -2737,7 +2737,8 @@ def fir_GlobalOp : fir_Op<"global", [IsolatedFromAbove, Symbol]> {
OptionalAttr<AnyAttr>:$initVal,
OptionalAttr<UnitAttr>:$constant,
OptionalAttr<UnitAttr>:$target,
- OptionalAttr<StrAttr>:$linkName
+ OptionalAttr<StrAttr>:$linkName,
+ OptionalAttr<fir_CUDAAttributeAttr>:$cuda_attr
);
let regions = (region AtMostRegion<1>:$region);
diff --git a/flang/lib/Lower/ConvertVariable.cpp b/flang/lib/Lower/ConvertVariable.cpp
index f14267f1234217..2f23757f497ea5 100644
--- a/flang/lib/Lower/ConvertVariable.cpp
+++ b/flang/lib/Lower/ConvertVariable.cpp
@@ -138,7 +138,8 @@ static bool isConstant(const Fortran::semantics::Symbol &sym) {
static fir::GlobalOp defineGlobal(Fortran::lower::AbstractConverter &converter,
const Fortran::lower::pft::Variable &var,
llvm::StringRef globalName,
- mlir::StringAttr linkage);
+ mlir::StringAttr linkage,
+ fir::CUDAAttributeAttr cudaAttr = {});
static mlir::Location genLocation(Fortran::lower::AbstractConverter &converter,
const Fortran::semantics::Symbol &sym) {
@@ -462,7 +463,8 @@ void Fortran::lower::createGlobalInitialization(
static fir::GlobalOp defineGlobal(Fortran::lower::AbstractConverter &converter,
const Fortran::lower::pft::Variable &var,
llvm::StringRef globalName,
- mlir::StringAttr linkage) {
+ mlir::StringAttr linkage,
+ fir::CUDAAttributeAttr cudaAttr) {
fir::FirOpBuilder &builder = converter.getFirOpBuilder();
const Fortran::semantics::Symbol &sym = var.getSymbol();
mlir::Location loc = genLocation(converter, sym);
@@ -500,8 +502,9 @@ static fir::GlobalOp defineGlobal(Fortran::lower::AbstractConverter &converter,
}
}
if (!global)
- global = builder.createGlobal(loc, symTy, globalName, linkage,
- mlir::Attribute{}, isConst, var.isTarget());
+ global =
+ builder.createGlobal(loc, symTy, globalName, linkage, mlir::Attribute{},
+ isConst, var.isTarget(), cudaAttr);
if (Fortran::semantics::IsAllocatableOrPointer(sym) &&
!Fortran::semantics::IsProcedure(sym)) {
const auto *details =
@@ -2219,7 +2222,10 @@ void Fortran::lower::defineModuleVariable(
// Do nothing. Mapping will be done on user side.
} else {
std::string globalName = converter.mangleName(sym);
- defineGlobal(converter, var, globalName, linkage);
+ fir::CUDAAttributeAttr cudaAttr =
+ Fortran::lower::translateSymbolCUDAAttribute(
+ converter.getFirOpBuilder().getContext(), sym);
+ defineGlobal(converter, var, globalName, linkage, cudaAttr);
}
}
diff --git a/flang/lib/Optimizer/Builder/FIRBuilder.cpp b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
index 141f8fcd3ab5fc..cce120c1f872c4 100644
--- a/flang/lib/Optimizer/Builder/FIRBuilder.cpp
+++ b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
@@ -271,19 +271,21 @@ mlir::Value fir::FirOpBuilder::createHeapTemporary(
/// Create a global variable in the (read-only) data section. A global variable
/// must have a unique name to identify and reference it.
-fir::GlobalOp fir::FirOpBuilder::createGlobal(mlir::Location loc,
- mlir::Type type,
- llvm::StringRef name,
- mlir::StringAttr linkage,
- mlir::Attribute value,
- bool isConst, bool isTarget) {
+fir::GlobalOp fir::FirOpBuilder::createGlobal(
+ mlir::Location loc, mlir::Type type, llvm::StringRef name,
+ mlir::StringAttr linkage, mlir::Attribute value, bool isConst,
+ bool isTarget, fir::CUDAAttributeAttr cudaAttr) {
auto module = getModule();
auto insertPt = saveInsertionPoint();
if (auto glob = module.lookupSymbol<fir::GlobalOp>(name))
return glob;
setInsertionPoint(module.getBody(), module.getBody()->end());
- auto glob =
- create<fir::GlobalOp>(loc, name, isConst, isTarget, type, value, linkage);
+ llvm::SmallVector<mlir::NamedAttribute> attrs;
+ if (cudaAttr)
+ attrs.push_back(mlir::NamedAttribute(
+ mlir::StringAttr::get(module.getContext(), "cuda_attr"), cudaAttr));
+ auto glob = create<fir::GlobalOp>(loc, name, isConst, isTarget, type, value,
+ linkage, attrs);
restoreInsertionPoint(insertPt);
return glob;
}
@@ -291,7 +293,7 @@ fir::GlobalOp fir::FirOpBuilder::createGlobal(mlir::Location loc,
fir::GlobalOp fir::FirOpBuilder::createGlobal(
mlir::Location loc, mlir::Type type, llvm::StringRef name, bool isConst,
bool isTarget, std::function<void(FirOpBuilder &)> bodyBuilder,
- mlir::StringAttr linkage) {
+ mlir::StringAttr linkage, fir::CUDAAttributeAttr cudaAttr) {
auto module = getModule();
auto insertPt = saveInsertionPoint();
if (auto glob = module.lookupSymbol<fir::GlobalOp>(name))
diff --git a/flang/test/Lower/CUDA/cuda-data-attribute.cuf b/flang/test/Lower/CUDA/cuda-data-attribute.cuf
index b02701bf3aea5a..7596c6b21efb0d 100644
--- a/flang/test/Lower/CUDA/cuda-data-attribute.cuf
+++ b/flang/test/Lower/CUDA/cuda-data-attribute.cuf
@@ -3,6 +3,18 @@
! Test lowering of CUDA attribute on variables.
+module cuda_var
+ real, constant :: mod_a_rc
+! CHECK: fir.global @_QMcuda_varEmod_a_rc {cuda_attr = #fir.cuda<constant>} : f32
+ real, device :: mod_b_ra
+! CHECK: fir.global @_QMcuda_varEmod_b_ra {cuda_attr = #fir.cuda<device>} : f32
+ real, allocatable, managed :: mod_c_rm
+! CHECK: fir.global @_QMcuda_varEmod_c_rm {cuda_attr = #fir.cuda<managed>} : !fir.box<!fir.heap<f32>>
+ real, allocatable, pinned :: mod_d_rp
+! CHECK: fir.global @_QMcuda_varEmod_d_rp {cuda_attr = #fir.cuda<pinned>} : !fir.box<!fir.heap<f32>>
+
+contains
+
subroutine local_var_attrs
real, constant :: rc
real, device :: rd
@@ -10,46 +22,43 @@ subroutine local_var_attrs
real, allocatable, pinned :: rp
end subroutine
-! CHECK-LABEL: func.func @_QPlocal_var_attrs()
-! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<constant>, uniq_name = "_QFlocal_var_attrsErc"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
-! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<device>, uniq_name = "_QFlocal_var_attrsErd"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
-! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFlocal_var_attrsErm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
-! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFlocal_var_attrsErp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
+! CHECK-LABEL: func.func @_QMcuda_varPlocal_var_attrs()
+! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<constant>, uniq_name = "_QMcuda_varFlocal_var_attrsErc"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<device>, uniq_name = "_QMcuda_varFlocal_var_attrsErd"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFlocal_var_attrsErm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
+! CHECK: %{{.*}}:2 = hlfir.declare %{{.*}} {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFlocal_var_attrsErp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
-! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<constant>, uniq_name = "_QFlocal_var_attrsErc"} : (!fir.ref<f32>) -> !fir.ref<f32>
-! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<device>, uniq_name = "_QFlocal_var_attrsErd"} : (!fir.ref<f32>) -> !fir.ref<f32>
-! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFlocal_var_attrsErm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> !fir.ref<!fir.box<!fir.heap<f32>>>
-! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFlocal_var_attrsErp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> !fir.ref<!fir.box<!fir.heap<f32>>>
+! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<constant>, uniq_name = "_QMcuda_varFlocal_var_attrsErc"} : (!fir.ref<f32>) -> !fir.ref<f32>
+! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<device>, uniq_name = "_QMcuda_varFlocal_var_attrsErd"} : (!fir.ref<f32>) -> !fir.ref<f32>
+! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFlocal_var_attrsErm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> !fir.ref<!fir.box<!fir.heap<f32>>>
+! FIR: %{{.*}} = fir.declare %{{.*}} {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFlocal_var_attrsErp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> !fir.ref<!fir.box<!fir.heap<f32>>>
subroutine dummy_arg_constant(dc)
real, constant :: dc
end subroutine
-! CHECK-LABEL: func.func @_QPdummy_arg_constant(
+! CHECK-LABEL: func.func @_QMcuda_varPdummy_arg_constant(
! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<f32> {fir.bindc_name = "dc", fir.cuda_attr = #fir.cuda<constant>}
-! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<constant>, uniq_name = "_QFdummy_arg_constantEdc"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<constant>, uniq_name = "_QMcuda_varFdummy_arg_constantEdc"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
subroutine dummy_arg_device(dd)
real, device :: dd
end subroutine
-! CHECK-LABEL: func.func @_QPdummy_arg_device(
+! CHECK-LABEL: func.func @_QMcuda_varPdummy_arg_device(
! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<f32> {fir.bindc_name = "dd", fir.cuda_attr = #fir.cuda<device>}) {
-! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<device>, uniq_name = "_QFdummy_arg_deviceEdd"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<device>, uniq_name = "_QMcuda_varFdummy_arg_deviceEdd"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
subroutine dummy_arg_managed(dm)
real, allocatable, managed :: dm
end subroutine
-! CHECK-LABEL: func.func @_QPdummy_arg_managed(
+! CHECK-LABEL: func.func @_QMcuda_varPdummy_arg_managed(
! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<!fir.box<!fir.heap<f32>>> {fir.bindc_name = "dm", fir.cuda_attr = #fir.cuda<managed>}) {
-! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFdummy_arg_managedEdm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
+! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<managed>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFdummy_arg_managedEdm"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
subroutine dummy_arg_pinned(dp)
real, allocatable, pinned :: dp
end subroutine
-! CHECK-LABEL: func.func @_QPdummy_arg_pinned(
+! CHECK-LABEL: func.func @_QMcuda_varPdummy_arg_pinned(
! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<!fir.box<!fir.heap<f32>>> {fir.bindc_name = "dp", fir.cuda_attr = #fir.cuda<pinned>}) {
-! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFdummy_arg_pinnedEdp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
-
-
-
-
+! CHECK: %{{.*}}:2 = hlfir.declare %[[ARG0]] {cuda_attr = #fir.cuda<pinned>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QMcuda_varFdummy_arg_pinnedEdp"} : (!fir.ref<!fir.box<!fir.heap<f32>>>) -> (!fir.ref<!fir.box<!fir.heap<f32>>>, !fir.ref<!fir.box<!fir.heap<f32>>>)
+end module
|
jeanPerier
approved these changes
Feb 9, 2024
Thanks for the pointer to the other PR. I was looking to avoid this constexpr in the extra declaration for a long time and it was indeed pretty easy :-) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Propagate the CUDA attribute to fir.global operation for simple module variables.