Skip to content

[flang][cuda] Avoid extra load in c_f_pointer lowering with c_devptr #108090

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.

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
Sep 11, 2024

Conversation

clementval
Copy link
Contributor

@clementval clementval commented Sep 10, 2024

Remove unnecessary load of the cptr component when getting the __address. fir.coordinate_of operation can be chained so the load is not needed.

@llvmbot llvmbot added flang Flang issues not falling into any other category flang:fir-hlfir labels Sep 10, 2024
@llvmbot
Copy link
Member

llvmbot commented Sep 10, 2024

@llvm/pr-subscribers-flang-fir-hlfir

Author: Valentin Clement (バレンタイン クレメン) (clementval)

Changes

Remove unnecessary load of the cptr component when getting the __address.


Full diff: https://github.com/llvm/llvm-project/pull/108090.diff

2 Files Affected:

  • (modified) flang/lib/Optimizer/Builder/FIRBuilder.cpp (+1-2)
  • (modified) flang/test/Lower/CUDA/cuda-devptr.cuf (+4-3)
diff --git a/flang/lib/Optimizer/Builder/FIRBuilder.cpp b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
index c5a135a189e8dc..d786d79ba8701b 100644
--- a/flang/lib/Optimizer/Builder/FIRBuilder.cpp
+++ b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
@@ -1594,8 +1594,7 @@ mlir::Value fir::factory::genCPtrOrCFunptrValue(fir::FirOpBuilder &builder,
       cPtrCoor = builder.create<fir::ExtractValueOp>(loc, addrFieldTy, cPtr,
                                                      arrayAttr);
     }
-    mlir::Value cptr = builder.create<fir::LoadOp>(loc, cPtrCoor);
-    return genCPtrOrCFunptrValue(builder, loc, cptr);
+    return genCPtrOrCFunptrValue(builder, loc, cPtrCoor);
   }
 
   if (fir::isa_ref_type(cPtr.getType())) {
diff --git a/flang/test/Lower/CUDA/cuda-devptr.cuf b/flang/test/Lower/CUDA/cuda-devptr.cuf
index 21c5088b640fc7..2eac890970d52b 100644
--- a/flang/test/Lower/CUDA/cuda-devptr.cuf
+++ b/flang/test/Lower/CUDA/cuda-devptr.cuf
@@ -40,8 +40,9 @@ end
 ! CHECK: %[[X:.*]]:2 = hlfir.declare %{{.*}} {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<pointer>, uniq_name = "_QFsub2Ex"} : (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>) -> (!fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>, !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>)
 ! CHECK: %[[CPTR:.*]] = fir.field_index cptr, !fir.type<_QM__fortran_builtinsT__builtin_c_devptr{cptr:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>
 ! CHECK: %[[CPTR_COORD:.*]] = fir.coordinate_of %{{.*}}#1, %[[CPTR]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_devptr{cptr:!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>}>>, !fir.field) -> !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>
-! CHECK: %[[CPTR_LOAD:.*]] = fir.load %[[CPTR_COORD]] : !fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>
-! CHECK: %[[ADDRESS:.*]] = fir.extract_value %[[CPTR_LOAD]], [0 : index] : (!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>) -> i64
-! CHECK: %[[ADDRESS_IDX:.*]] = fir.convert %[[ADDRESS]] : (i64) -> !fir.ptr<!fir.array<?xf32>>
+! CHECK: %[[ADDRESS:.*]] = fir.field_index __address, !fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>
+! CHECK: %[[ADDRESS_COORD:.*]] = fir.coordinate_of %[[CPTR_COORD]], %[[ADDRESS]] : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>, !fir.field) -> !fir.ref<i64>
+! CHECK: %[[ADDRESS_LOADED:.*]] = fir.load %[[ADDRESS_COORD]] : !fir.ref<i64>
+! CHECK: %[[ADDRESS_IDX:.*]] = fir.convert %[[ADDRESS_LOADED]] : (i64) -> !fir.ptr<!fir.array<?xf32>>
 ! CHECK: %[[EMBOX:.*]] = fir.embox %[[ADDRESS_IDX]](%{{.*}}) : (!fir.ptr<!fir.array<?xf32>>, !fir.shape<1>) -> !fir.box<!fir.ptr<!fir.array<?xf32>>>
 ! CHECK: fir.store %[[EMBOX]] to %[[X]]#1 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xf32>>>>

@clementval clementval merged commit e67a666 into llvm:main Sep 11, 2024
11 checks passed
@clementval clementval deleted the cuf_c_f_pointer_fix branch September 11, 2024 02:33
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
flang:fir-hlfir flang Flang issues not falling into any other category
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants