[flang][cuda] Update implicit data transfer for device component #147882
+43
−20
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
flang
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesUpdate the detection of implicit data transfer when a device resident allocatable derived-type component is involved and remove the TODOs. Full diff: https://github.com/llvm/llvm-project/pull/147882.diff 4 Files Affected:
diff --git a/flang/include/flang/Evaluate/tools.h b/flang/include/flang/Evaluate/tools.h
index 18c244f6f450f..96ed86f468350 100644
--- a/flang/include/flang/Evaluate/tools.h
+++ b/flang/include/flang/Evaluate/tools.h
@@ -1359,24 +1359,7 @@ inline bool IsCUDADataTransfer(const A &lhs, const B &rhs) {
/// Check if the expression is a mix of host and device variables that require
/// implicit data transfer.
-inline bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr) {
- unsigned hostSymbols{0};
- unsigned deviceSymbols{0};
- for (const Symbol &sym : CollectCudaSymbols(expr)) {
- if (IsCUDADeviceSymbol(sym)) {
- ++deviceSymbols;
- } else {
- if (sym.owner().IsDerivedType()) {
- if (IsCUDADeviceSymbol(sym.owner().GetSymbol()->GetUltimate())) {
- ++deviceSymbols;
- }
- }
- ++hostSymbols;
- }
- }
- bool hasConstant{HasConstant(expr)};
- return (hasConstant || (hostSymbols > 0)) && deviceSymbols > 0;
-}
+bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr);
// Checks whether the symbol on the LHS is present in the RHS expression.
bool CheckForSymbolMatch(const Expr<SomeType> *lhs, const Expr<SomeType> *rhs);
diff --git a/flang/lib/Evaluate/tools.cpp b/flang/lib/Evaluate/tools.cpp
index 6a57d87a30e93..3d9f06308d8c1 100644
--- a/flang/lib/Evaluate/tools.cpp
+++ b/flang/lib/Evaluate/tools.cpp
@@ -1154,6 +1154,31 @@ template semantics::UnorderedSymbolSet CollectCudaSymbols(
template semantics::UnorderedSymbolSet CollectCudaSymbols(
const Expr<SubscriptInteger> &);
+bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr) {
+ semantics::UnorderedSymbolSet hostSymbols;
+ semantics::UnorderedSymbolSet deviceSymbols;
+
+ SymbolVector symbols{GetSymbolVector(expr)};
+ std::reverse(symbols.begin(), symbols.end());
+ bool skipNext{false};
+ for (const Symbol &sym : symbols) {
+ bool isComponent{sym.owner().IsDerivedType()};
+ bool skipComponent{false};
+ if (!skipNext) {
+ if (IsCUDADeviceSymbol(sym)) {
+ deviceSymbols.insert(sym);
+ } else if (isComponent) {
+ skipComponent = true; // Component is not device. Look on the base.
+ } else {
+ hostSymbols.insert(sym);
+ }
+ }
+ skipNext = isComponent && !skipComponent;
+ }
+ bool hasConstant{HasConstant(expr)};
+ return (hasConstant || (hostSymbols.size() > 0)) && deviceSymbols.size() > 0;
+}
+
// HasVectorSubscript()
struct HasVectorSubscriptHelper
: public AnyTraverse<HasVectorSubscriptHelper, bool,
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 696473605a4e0..ff26aa8e33649 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -4842,8 +4842,6 @@ class FirConverter : public Fortran::lower::AbstractConverter {
.detailsIf<Fortran::semantics::ObjectEntityDetails>()) {
if (details->cudaDataAttr() &&
*details->cudaDataAttr() != Fortran::common::CUDADataAttr::Pinned) {
- if (sym.owner().IsDerivedType() && IsAllocatable(sym.GetUltimate()))
- TODO(loc, "Device resident allocatable derived-type component");
// TODO: This should probably being checked in semantic and give a
// proper error.
assert(
diff --git a/flang/test/Lower/CUDA/cuda-data-transfer.cuf b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
index 68a0202f951fe..3a9b55996d9b1 100644
--- a/flang/test/Lower/CUDA/cuda-data-transfer.cuf
+++ b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
@@ -7,6 +7,10 @@ module mod1
integer :: i
end type
+ type :: t2
+ integer, device, allocatable, dimension(:) :: x
+ end type
+
integer, device, dimension(11:20) :: cdev
contains
@@ -419,3 +423,16 @@ end subroutine
! CHECK: fir.do_concurrent.loop
! CHECK-NOT: cuf.data_transfer
! CHECK: hlfir.assign
+
+
+subroutine sub22()
+ use mod1
+ type(t2) :: a
+ integer :: b(100)
+ allocate(a%x(100))
+
+ b = a%x
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub22()
+! CHECK: cuf.data_transfer
|
|
@llvm/pr-subscribers-flang-semantics Author: Valentin Clement (バレンタイン クレメン) (clementval) ChangesUpdate the detection of implicit data transfer when a device resident allocatable derived-type component is involved and remove the TODOs. Full diff: https://github.com/llvm/llvm-project/pull/147882.diff 4 Files Affected:
diff --git a/flang/include/flang/Evaluate/tools.h b/flang/include/flang/Evaluate/tools.h
index 18c244f6f450f..96ed86f468350 100644
--- a/flang/include/flang/Evaluate/tools.h
+++ b/flang/include/flang/Evaluate/tools.h
@@ -1359,24 +1359,7 @@ inline bool IsCUDADataTransfer(const A &lhs, const B &rhs) {
/// Check if the expression is a mix of host and device variables that require
/// implicit data transfer.
-inline bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr) {
- unsigned hostSymbols{0};
- unsigned deviceSymbols{0};
- for (const Symbol &sym : CollectCudaSymbols(expr)) {
- if (IsCUDADeviceSymbol(sym)) {
- ++deviceSymbols;
- } else {
- if (sym.owner().IsDerivedType()) {
- if (IsCUDADeviceSymbol(sym.owner().GetSymbol()->GetUltimate())) {
- ++deviceSymbols;
- }
- }
- ++hostSymbols;
- }
- }
- bool hasConstant{HasConstant(expr)};
- return (hasConstant || (hostSymbols > 0)) && deviceSymbols > 0;
-}
+bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr);
// Checks whether the symbol on the LHS is present in the RHS expression.
bool CheckForSymbolMatch(const Expr<SomeType> *lhs, const Expr<SomeType> *rhs);
diff --git a/flang/lib/Evaluate/tools.cpp b/flang/lib/Evaluate/tools.cpp
index 6a57d87a30e93..3d9f06308d8c1 100644
--- a/flang/lib/Evaluate/tools.cpp
+++ b/flang/lib/Evaluate/tools.cpp
@@ -1154,6 +1154,31 @@ template semantics::UnorderedSymbolSet CollectCudaSymbols(
template semantics::UnorderedSymbolSet CollectCudaSymbols(
const Expr<SubscriptInteger> &);
+bool HasCUDAImplicitTransfer(const Expr<SomeType> &expr) {
+ semantics::UnorderedSymbolSet hostSymbols;
+ semantics::UnorderedSymbolSet deviceSymbols;
+
+ SymbolVector symbols{GetSymbolVector(expr)};
+ std::reverse(symbols.begin(), symbols.end());
+ bool skipNext{false};
+ for (const Symbol &sym : symbols) {
+ bool isComponent{sym.owner().IsDerivedType()};
+ bool skipComponent{false};
+ if (!skipNext) {
+ if (IsCUDADeviceSymbol(sym)) {
+ deviceSymbols.insert(sym);
+ } else if (isComponent) {
+ skipComponent = true; // Component is not device. Look on the base.
+ } else {
+ hostSymbols.insert(sym);
+ }
+ }
+ skipNext = isComponent && !skipComponent;
+ }
+ bool hasConstant{HasConstant(expr)};
+ return (hasConstant || (hostSymbols.size() > 0)) && deviceSymbols.size() > 0;
+}
+
// HasVectorSubscript()
struct HasVectorSubscriptHelper
: public AnyTraverse<HasVectorSubscriptHelper, bool,
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 696473605a4e0..ff26aa8e33649 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -4842,8 +4842,6 @@ class FirConverter : public Fortran::lower::AbstractConverter {
.detailsIf<Fortran::semantics::ObjectEntityDetails>()) {
if (details->cudaDataAttr() &&
*details->cudaDataAttr() != Fortran::common::CUDADataAttr::Pinned) {
- if (sym.owner().IsDerivedType() && IsAllocatable(sym.GetUltimate()))
- TODO(loc, "Device resident allocatable derived-type component");
// TODO: This should probably being checked in semantic and give a
// proper error.
assert(
diff --git a/flang/test/Lower/CUDA/cuda-data-transfer.cuf b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
index 68a0202f951fe..3a9b55996d9b1 100644
--- a/flang/test/Lower/CUDA/cuda-data-transfer.cuf
+++ b/flang/test/Lower/CUDA/cuda-data-transfer.cuf
@@ -7,6 +7,10 @@ module mod1
integer :: i
end type
+ type :: t2
+ integer, device, allocatable, dimension(:) :: x
+ end type
+
integer, device, dimension(11:20) :: cdev
contains
@@ -419,3 +423,16 @@ end subroutine
! CHECK: fir.do_concurrent.loop
! CHECK-NOT: cuf.data_transfer
! CHECK: hlfir.assign
+
+
+subroutine sub22()
+ use mod1
+ type(t2) :: a
+ integer :: b(100)
+ allocate(a%x(100))
+
+ b = a%x
+end subroutine
+
+! CHECK-LABEL: func.func @_QPsub22()
+! CHECK: cuf.data_transfer
|
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.
Update the detection of implicit data transfer when a device resident allocatable derived-type component is involved and remove the TODOs.