Skip to content

[flang][cuda] Support derived type in cuf.data_transfer conversion #115557

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 12, 2024
+30 −5
Merged

[flang][cuda] Support derived type in cuf.data_transfer conversion #115557

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 17 additions & 5 deletions flang/lib/Optimizer/Transforms/CUFOpConversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -507,8 +507,11 @@ struct CUFDataTransferOpConversion
using OpRewritePattern::OpRewritePattern;

CUFDataTransferOpConversion(mlir::MLIRContext *context,
const mlir::SymbolTable &symtab)
: OpRewritePattern(context), symtab{symtab} {}
const mlir::SymbolTable &symtab,
mlir::DataLayout *dl,
const fir::LLVMTypeConverter *typeConverter)
: OpRewritePattern(context), symtab{symtab}, dl{dl},
typeConverter{typeConverter} {}

mlir::LogicalResult
matchAndRewrite(cuf::DataTransferOp op,
Expand Down Expand Up @@ -576,7 +579,13 @@ struct CUFDataTransferOpConversion
nbElement = builder.createIntegerConstant(
loc, i64Ty, seqTy.getConstantArraySize());
}
int width = computeWidth(loc, dstTy, kindMap);
unsigned width = 0;
if (fir::isa_derived(dstTy)) {
mlir::Type structTy = typeConverter->convertType(dstTy);
width = dl->getTypeSizeInBits(structTy) / 8;
} else {
width = computeWidth(loc, dstTy, kindMap);
}
mlir::Value widthValue = rewriter.create<mlir::arith::ConstantOp>(
loc, i64Ty, rewriter.getIntegerAttr(i64Ty, width));
mlir::Value bytes =
Expand Down Expand Up @@ -647,6 +656,8 @@ struct CUFDataTransferOpConversion

private:
const mlir::SymbolTable &symtab;
mlir::DataLayout *dl;
const fir::LLVMTypeConverter *typeConverter;
};

struct CUFLaunchOpConversion
Expand Down Expand Up @@ -749,6 +760,7 @@ void cuf::populateCUFToFIRConversionPatterns(
patterns.insert<CUFAllocOpConversion>(patterns.getContext(), &dl, &converter);
patterns.insert<CUFAllocateOpConversion, CUFDeallocateOpConversion,
CUFFreeOpConversion>(patterns.getContext());
patterns.insert<CUFDataTransferOpConversion, CUFLaunchOpConversion>(
patterns.getContext(), symtab);
patterns.insert<CUFDataTransferOpConversion>(patterns.getContext(), symtab,
&dl, &converter);
patterns.insert<CUFLaunchOpConversion>(patterns.getContext(), symtab);
}
13 changes: 13 additions & 0 deletions flang/test/Fir/CUDA/cuda-data-transfer.fir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -295,4 +295,17 @@ func.func @_QPscalar_to_array() {
// CHECK-LABEL: func.func @_QPscalar_to_array()
// CHECK: _FortranACUFDataTransferDescDescNoRealloc

func.func @_QPtest_type() {
%0 = cuf.alloc !fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}> {bindc_name = "a", data_attr = #cuf.cuda<device>, uniq_name = "_QFtest_typeEa"} -> !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>
%1 = fir.declare %0 {data_attr = #cuf.cuda<device>, uniq_name = "_QFtest_typeEa"} : (!fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>) -> !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>
%2 = fir.alloca !fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}> {bindc_name = "b", uniq_name = "_QFtest_typeEb"}
%3 = fir.declare %2 {uniq_name = "_QFtest_typeEb"} : (!fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>) -> !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>
cuf.data_transfer %3 to %1 {transfer_kind = #cuf.cuda_transfer<host_device>} : !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>, !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>>
cuf.free %1 : !fir.ref<!fir.type<_QMbarTcmplx{id:i32,c:complex<f32>}>> {data_attr = #cuf.cuda<device>}
return
}

// CHECK-LABEL: func.func @_QPtest_type()
// CHECK: fir.call @_FortranACUFDataTransferPtrPtr(%{{.*}}, %{{.*}}, %c12{{.*}}, %c0{{.*}}, %{{.*}}, %{{.*}}) : (!fir.llvm_ptr<i8>, !fir.llvm_ptr<i8>, i64, i32, !fir.ref<i8>, i32) -> none

} // end of module
Loading