[flang][NFC] use llvm.intr.stacksave/restore instead of opaque calls #108562
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
The new LLVM stack save/restore intrinsic operations are more convenient than function calls because they do not add function declarations to the module and therefore do not block the parallelisation of passes. Furthermore they could be much more easily marked with memory effects than function calls if that ever proved useful. This builds on top of llvm#107879. Resolves llvm#108016
llvmbot
added
flang
|
@llvm/pr-subscribers-flang-fir-hlfir @llvm/pr-subscribers-flang-codegen Author: Tom Eccles (tblah) ChangesThe new LLVM stack save/restore intrinsic operations are more convenient than function calls because they do not add function declarations to the module and therefore do not block the parallelisation of passes. Furthermore they could be much more easily marked with memory effects than function calls if that ever proved useful. This builds on top of #107879. Resolves #108016 Patch is 44.59 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/108562.diff 25 Files Affected:
diff --git a/flang/include/flang/Optimizer/Builder/FIRBuilder.h b/flang/include/flang/Optimizer/Builder/FIRBuilder.h
index f7151f26f09cb3..379199dfc710d2 100644
--- a/flang/include/flang/Optimizer/Builder/FIRBuilder.h
+++ b/flang/include/flang/Optimizer/Builder/FIRBuilder.h
@@ -253,6 +253,15 @@ class FirOpBuilder : public mlir::OpBuilder, public mlir::OpBuilder::Listener {
mlir::ValueRange lenParams = {},
llvm::ArrayRef<mlir::NamedAttribute> attrs = {});
+ /// Create an LLVM stack save intrinsic op. Returns the saved stack pointer.
+ /// The stack address space is fetched from the data layout of the current
+ /// module.
+ mlir::Value genStackSave(mlir::Location loc);
+
+ /// Create an LLVM stack restore intrinsic op. stackPointer should be a value
+ /// previously returned from genStackSave.
+ void genStackRestore(mlir::Location loc, mlir::Value stackPointer);
+
/// Create a global value.
fir::GlobalOp createGlobal(mlir::Location loc, mlir::Type type,
llvm::StringRef name,
@@ -729,6 +738,9 @@ elideExtentsAlreadyInType(mlir::Type type, mlir::ValueRange shape);
llvm::SmallVector<mlir::Value>
elideLengthsAlreadyInType(mlir::Type type, mlir::ValueRange lenParams);
+/// Get the address space which should be used for allocas
+uint64_t getAllocaAddressSpace(mlir::DataLayout *dataLayout);
+
} // namespace fir::factory
#endif // FORTRAN_OPTIMIZER_BUILDER_FIRBUILDER_H
diff --git a/flang/include/flang/Optimizer/Builder/LowLevelIntrinsics.h b/flang/include/flang/Optimizer/Builder/LowLevelIntrinsics.h
index e5a7113149346c..9be051632f93d9 100644
--- a/flang/include/flang/Optimizer/Builder/LowLevelIntrinsics.h
+++ b/flang/include/flang/Optimizer/Builder/LowLevelIntrinsics.h
@@ -42,12 +42,6 @@ mlir::func::FuncOp getLlvmGetRounding(FirOpBuilder &builder);
/// Get the `llvm.set.rounding` intrinsic.
mlir::func::FuncOp getLlvmSetRounding(FirOpBuilder &builder);
-/// Get the `llvm.stacksave` intrinsic.
-mlir::func::FuncOp getLlvmStackSave(FirOpBuilder &builder);
-
-/// Get the `llvm.stackrestore` intrinsic.
-mlir::func::FuncOp getLlvmStackRestore(FirOpBuilder &builder);
-
/// Get the `llvm.init.trampoline` intrinsic.
mlir::func::FuncOp getLlvmInitTrampoline(FirOpBuilder &builder);
diff --git a/flang/include/flang/Optimizer/Support/DataLayout.h b/flang/include/flang/Optimizer/Support/DataLayout.h
index d21576bb95f795..6072425b7d637f 100644
--- a/flang/include/flang/Optimizer/Support/DataLayout.h
+++ b/flang/include/flang/Optimizer/Support/DataLayout.h
@@ -45,7 +45,6 @@ void setMLIRDataLayoutFromAttributes(mlir::ModuleOp mlirModule,
/// std::nullopt.
std::optional<mlir::DataLayout>
getOrSetDataLayout(mlir::ModuleOp mlirModule, bool allowDefaultLayout = false);
-
} // namespace fir::support
#endif // FORTRAN_OPTIMIZER_SUPPORT_DATALAYOUT_H
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 3e9db06b61d502..6724d93e0c5524 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -3256,15 +3256,10 @@ class FirConverter : public Fortran::lower::AbstractConverter {
const Fortran::parser::CharBlock &endPosition =
eval.getLastNestedEvaluation().position;
localSymbols.pushScope();
- mlir::func::FuncOp stackSave = fir::factory::getLlvmStackSave(*builder);
- mlir::func::FuncOp stackRestore =
- fir::factory::getLlvmStackRestore(*builder);
- mlir::Value stackPtr =
- builder->create<fir::CallOp>(toLocation(), stackSave).getResult(0);
+ mlir::Value stackPtr = builder->genStackSave(toLocation());
mlir::Location endLoc = genLocation(endPosition);
- stmtCtx.attachCleanup([=]() {
- builder->create<fir::CallOp>(endLoc, stackRestore, stackPtr);
- });
+ stmtCtx.attachCleanup(
+ [=]() { builder->genStackRestore(endLoc, stackPtr); });
Fortran::semantics::Scope &scope =
bridge.getSemanticsContext().FindScope(endPosition);
scopeBlockIdMap.try_emplace(&scope, ++blockId);
diff --git a/flang/lib/Lower/ConvertCall.cpp b/flang/lib/Lower/ConvertCall.cpp
index f445a21e560bc9..a085affd6c7126 100644
--- a/flang/lib/Lower/ConvertCall.cpp
+++ b/flang/lib/Lower/ConvertCall.cpp
@@ -368,22 +368,9 @@ std::pair<fir::ExtendedValue, bool> Fortran::lower::genCallOpAndResult(
if (!extents.empty() || !lengths.empty()) {
auto *bldr = &converter.getFirOpBuilder();
- auto stackSaveFn = fir::factory::getLlvmStackSave(builder);
- auto stackSaveSymbol = bldr->getSymbolRefAttr(stackSaveFn.getName());
- mlir::Value sp;
- fir::CallOp call = bldr->create<fir::CallOp>(
- loc, stackSaveSymbol, stackSaveFn.getFunctionType().getResults(),
- mlir::ValueRange{});
- if (call.getNumResults() != 0)
- sp = call.getResult(0);
- stmtCtx.attachCleanup([bldr, loc, sp]() {
- auto stackRestoreFn = fir::factory::getLlvmStackRestore(*bldr);
- auto stackRestoreSymbol =
- bldr->getSymbolRefAttr(stackRestoreFn.getName());
- bldr->create<fir::CallOp>(loc, stackRestoreSymbol,
- stackRestoreFn.getFunctionType().getResults(),
- mlir::ValueRange{sp});
- });
+ mlir::Value sp = bldr->genStackSave(loc);
+ stmtCtx.attachCleanup(
+ [bldr, loc, sp]() { bldr->genStackRestore(loc, sp); });
}
mlir::Value temp =
builder.createTemporary(loc, type, ".result", extents, resultLengths);
diff --git a/flang/lib/Optimizer/Builder/FIRBuilder.cpp b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
index d786d79ba8701b..ddc97cacda2aea 100644
--- a/flang/lib/Optimizer/Builder/FIRBuilder.cpp
+++ b/flang/lib/Optimizer/Builder/FIRBuilder.cpp
@@ -18,6 +18,7 @@
#include "flang/Optimizer/Dialect/FIRAttr.h"
#include "flang/Optimizer/Dialect/FIROpsSupport.h"
#include "flang/Optimizer/Dialect/FIRType.h"
+#include "flang/Optimizer/Support/DataLayout.h"
#include "flang/Optimizer/Support/FatalError.h"
#include "flang/Optimizer/Support/InternalNames.h"
#include "flang/Optimizer/Support/Utils.h"
@@ -328,6 +329,18 @@ mlir::Value fir::FirOpBuilder::createHeapTemporary(
name, dynamicLength, dynamicShape, attrs);
}
+mlir::Value fir::FirOpBuilder::genStackSave(mlir::Location loc) {
+ mlir::DataLayout dataLayout(getModule());
+ mlir::Type voidPtr = mlir::LLVM::LLVMPointerType::get(
+ getContext(), fir::factory::getAllocaAddressSpace(&dataLayout));
+ return create<mlir::LLVM::StackSaveOp>(loc, voidPtr);
+}
+
+void fir::FirOpBuilder::genStackRestore(mlir::Location loc,
+ mlir::Value stackPointer) {
+ create<mlir::LLVM::StackRestoreOp>(loc, stackPointer);
+}
+
/// 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(
@@ -1664,3 +1677,10 @@ void fir::factory::setInternalLinkage(mlir::func::FuncOp func) {
mlir::LLVM::LinkageAttr::get(func->getContext(), internalLinkage);
func->setAttr("llvm.linkage", linkage);
}
+
+uint64_t fir::factory::getAllocaAddressSpace(mlir::DataLayout *dataLayout) {
+ if (dataLayout)
+ if (mlir::Attribute addrSpace = dataLayout->getAllocaMemorySpace())
+ return mlir::cast<mlir::IntegerAttr>(addrSpace).getUInt();
+ return 0;
+}
diff --git a/flang/lib/Optimizer/Builder/LowLevelIntrinsics.cpp b/flang/lib/Optimizer/Builder/LowLevelIntrinsics.cpp
index bb5f77d5d4d1de..411a48614af6c8 100644
--- a/flang/lib/Optimizer/Builder/LowLevelIntrinsics.cpp
+++ b/flang/lib/Optimizer/Builder/LowLevelIntrinsics.cpp
@@ -76,25 +76,6 @@ fir::factory::getLlvmSetRounding(fir::FirOpBuilder &builder) {
funcTy);
}
-mlir::func::FuncOp fir::factory::getLlvmStackSave(fir::FirOpBuilder &builder) {
- // FIXME: This should query the target alloca address space
- auto ptrTy = builder.getRefType(builder.getIntegerType(8));
- auto funcTy =
- mlir::FunctionType::get(builder.getContext(), std::nullopt, {ptrTy});
- return builder.createFunction(builder.getUnknownLoc(), "llvm.stacksave.p0",
- funcTy);
-}
-
-mlir::func::FuncOp
-fir::factory::getLlvmStackRestore(fir::FirOpBuilder &builder) {
- // FIXME: This should query the target alloca address space
- auto ptrTy = builder.getRefType(builder.getIntegerType(8));
- auto funcTy =
- mlir::FunctionType::get(builder.getContext(), {ptrTy}, std::nullopt);
- return builder.createFunction(builder.getUnknownLoc(), "llvm.stackrestore.p0",
- funcTy);
-}
-
mlir::func::FuncOp
fir::factory::getLlvmInitTrampoline(fir::FirOpBuilder &builder) {
auto ptrTy = builder.getRefType(builder.getIntegerType(8));
diff --git a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
index a2a9cff4c4977e..f6cb26ff9613f4 100644
--- a/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
+++ b/flang/lib/Optimizer/CodeGen/TargetRewrite.cpp
@@ -1236,25 +1236,18 @@ class TargetRewrite : public fir::impl::TargetRewritePassBase<TargetRewrite> {
inline void clearMembers() { setMembers(nullptr, nullptr, nullptr); }
- uint64_t getAllocaAddressSpace() const {
- if (dataLayout)
- if (mlir::Attribute addrSpace = dataLayout->getAllocaMemorySpace())
- return llvm::cast<mlir::IntegerAttr>(addrSpace).getUInt();
- return 0;
- }
-
// Inserts a call to llvm.stacksave at the current insertion
// point and the given location. Returns the call's result Value.
inline mlir::Value genStackSave(mlir::Location loc) {
- mlir::Type voidPtr = mlir::LLVM::LLVMPointerType::get(
- rewriter->getContext(), getAllocaAddressSpace());
- return rewriter->create<mlir::LLVM::StackSaveOp>(loc, voidPtr);
+ fir::FirOpBuilder builder(*rewriter, getModule());
+ return builder.genStackSave(loc);
}
// Inserts a call to llvm.stackrestore at the current insertion
// point and the given location and argument.
inline void genStackRestore(mlir::Location loc, mlir::Value sp) {
- rewriter->create<mlir::LLVM::StackRestoreOp>(loc, sp);
+ fir::FirOpBuilder builder(*rewriter, getModule());
+ return builder.genStackRestore(loc, sp);
}
fir::CodeGenSpecifics *specifics = nullptr;
diff --git a/flang/lib/Optimizer/Transforms/StackArrays.cpp b/flang/lib/Optimizer/Transforms/StackArrays.cpp
index a8f1a744cda5fe..1b929928dc7164 100644
--- a/flang/lib/Optimizer/Transforms/StackArrays.cpp
+++ b/flang/lib/Optimizer/Transforms/StackArrays.cpp
@@ -734,28 +734,12 @@ void AllocMemConversion::insertStackSaveRestore(
auto mod = oldAlloc->getParentOfType<mlir::ModuleOp>();
fir::FirOpBuilder builder{rewriter, mod};
- mlir::func::FuncOp stackSaveFn = fir::factory::getLlvmStackSave(builder);
- mlir::SymbolRefAttr stackSaveSym =
- builder.getSymbolRefAttr(stackSaveFn.getName());
-
builder.setInsertionPoint(oldAlloc);
- mlir::Value sp =
- builder
- .create<fir::CallOp>(oldAlloc.getLoc(), stackSaveSym,
- stackSaveFn.getFunctionType().getResults(),
- mlir::ValueRange{})
- .getResult(0);
-
- mlir::func::FuncOp stackRestoreFn =
- fir::factory::getLlvmStackRestore(builder);
- mlir::SymbolRefAttr stackRestoreSym =
- builder.getSymbolRefAttr(stackRestoreFn.getName());
+ mlir::Value sp = builder.genStackSave(oldAlloc.getLoc());
auto createStackRestoreCall = [&](mlir::Operation *user) {
builder.setInsertionPoint(user);
- builder.create<fir::CallOp>(user->getLoc(), stackRestoreSym,
- stackRestoreFn.getFunctionType().getResults(),
- mlir::ValueRange{sp});
+ builder.genStackRestore(user->getLoc(), sp);
};
for (mlir::Operation *user : oldAlloc->getUsers()) {
diff --git a/flang/lib/Optimizer/Transforms/StackReclaim.cpp b/flang/lib/Optimizer/Transforms/StackReclaim.cpp
index 8a60a9e64f704b..bd3e49a47bc399 100644
--- a/flang/lib/Optimizer/Transforms/StackReclaim.cpp
+++ b/flang/lib/Optimizer/Transforms/StackReclaim.cpp
@@ -7,6 +7,7 @@
//===----------------------------------------------------------------------===//
#include "flang/Common/Fortran.h"
+#include "flang/Optimizer/Builder/FIRBuilder.h"
#include "flang/Optimizer/Dialect/FIRDialect.h"
#include "flang/Optimizer/Dialect/FIROps.h"
#include "flang/Optimizer/Transforms/Passes.h"
@@ -31,34 +32,20 @@ class StackReclaimPass : public fir::impl::StackReclaimBase<StackReclaimPass> {
};
} // namespace
-uint64_t getAllocaAddressSpace(Operation *op) {
- mlir::ModuleOp module = mlir::dyn_cast_or_null<mlir::ModuleOp>(op);
- if (!module)
- module = op->getParentOfType<mlir::ModuleOp>();
-
- if (mlir::Attribute addrSpace =
- mlir::DataLayout(module).getAllocaMemorySpace())
- return llvm::cast<mlir::IntegerAttr>(addrSpace).getUInt();
- return 0;
-}
-
void StackReclaimPass::runOnOperation() {
auto *op = getOperation();
- auto *context = &getContext();
- mlir::OpBuilder builder(context);
- mlir::Type voidPtr =
- mlir::LLVM::LLVMPointerType::get(context, getAllocaAddressSpace(op));
+ fir::FirOpBuilder builder(op, fir::getKindMapping(op));
op->walk([&](fir::DoLoopOp loopOp) {
mlir::Location loc = loopOp.getLoc();
if (!loopOp.getRegion().getOps<fir::AllocaOp>().empty()) {
builder.setInsertionPointToStart(&loopOp.getRegion().front());
- auto stackSaveOp = builder.create<LLVM::StackSaveOp>(loc, voidPtr);
+ mlir::Value sp = builder.genStackSave(loc);
auto *terminator = loopOp.getRegion().back().getTerminator();
builder.setInsertionPoint(terminator);
- builder.create<LLVM::StackRestoreOp>(loc, stackSaveOp);
+ builder.genStackRestore(loc, sp);
}
});
}
diff --git a/flang/test/HLFIR/order_assignments/where-scheduling.f90 b/flang/test/HLFIR/order_assignments/where-scheduling.f90
index d3665d234a7125..ab87ae92de5799 100644
--- a/flang/test/HLFIR/order_assignments/where-scheduling.f90
+++ b/flang/test/HLFIR/order_assignments/where-scheduling.f90
@@ -134,7 +134,7 @@ end function f
!CHECK-NEXT: run 1 save : where/mask
!CHECK-NEXT: run 2 evaluate: where/region_assign1
!CHECK-LABEL: ------------ scheduling where in _QPonly_once ------------
-!CHECK-NEXT: unknown effect: %{{[0-9]+}} = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+!CHECK-NEXT: unknown effect: %{{[0-9]+}} = llvm.intr.stacksave : !llvm.ptr
!CHECK-NEXT: run 1 save (w): where/mask
!CHECK-NEXT: run 2 evaluate: where/region_assign1
!CHECK-NEXT: run 3 evaluate: where/region_assign2
diff --git a/flang/test/Lower/HLFIR/block_bindc_pocs.f90 b/flang/test/Lower/HLFIR/block_bindc_pocs.f90
index 090eeb35ea88b0..ed07d88c53a606 100644
--- a/flang/test/Lower/HLFIR/block_bindc_pocs.f90
+++ b/flang/test/Lower/HLFIR/block_bindc_pocs.f90
@@ -8,9 +8,9 @@ subroutine test_proc() bind(C)
end subroutine test_proc
end interface
end module m
-!CHECK-DAG: %[[S0:.*]] = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+!CHECK-DAG: %[[S0:.*]] = llvm.intr.stacksave : !llvm.ptr
!CHECK-DAG: fir.call @test_proc() proc_attrs<bind_c> fastmath<contract> : () -> ()
-!CHECK-DAG: fir.call @llvm.stackrestore.p0(%[[S0]]) fastmath<contract> : (!fir.ref<i8>) -> ()
+!CHECK-DAG: llvm.intr.stackrestore %[[S0]] : !llvm.ptr
!CHECK-DAG: func.func private @test_proc() attributes {fir.bindc_name = "test_proc"}
subroutine test
BLOCK
diff --git a/flang/test/Lower/HLFIR/elemental-array-ops.f90 b/flang/test/Lower/HLFIR/elemental-array-ops.f90
index 80801fdde0d729..9929c17ec33994 100644
--- a/flang/test/Lower/HLFIR/elemental-array-ops.f90
+++ b/flang/test/Lower/HLFIR/elemental-array-ops.f90
@@ -182,12 +182,12 @@ end subroutine char_return
! CHECK: %[[VAL_23:.*]] = arith.constant 0 : index
! CHECK: %[[VAL_24:.*]] = arith.cmpi sgt, %[[VAL_22]], %[[VAL_23]] : index
! CHECK: %[[VAL_25:.*]] = arith.select %[[VAL_24]], %[[VAL_22]], %[[VAL_23]] : index
-! CHECK: %[[VAL_26:.*]] = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+! CHECK: %[[VAL_26:.*]] = llvm.intr.stacksave : !llvm.ptr
! CHECK: %[[VAL_27:.*]] = fir.call @_QPcallee(%[[VAL_2]], %[[VAL_25]], %[[VAL_20]]) fastmath<contract> : (!fir.ref<!fir.char<1,3>>, index, !fir.boxchar<1>) -> !fir.boxchar<1>
! CHECK: %[[VAL_28:.*]]:2 = hlfir.declare %[[VAL_2]] typeparams %[[VAL_25]] {uniq_name = ".tmp.func_result"} : (!fir.ref<!fir.char<1,3>>, index) -> (!fir.ref<!fir.char<1,3>>, !fir.ref<!fir.char<1,3>>)
! CHECK: %[[MustFree:.*]] = arith.constant false
! CHECK: %[[ResultTemp:.*]] = hlfir.as_expr %[[VAL_28]]#0 move %[[MustFree]] : (!fir.ref<!fir.char<1,3>>, i1) -> !hlfir.expr<!fir.char<1,3>>
-! CHECK: fir.call @llvm.stackrestore.p0(%[[VAL_26]]) fastmath<contract> : (!fir.ref<i8>) -> ()
+! CHECK: llvm.intr.stackrestore %[[VAL_26]] : !llvm.ptr
! CHECK: hlfir.yield_element %[[ResultTemp]] : !hlfir.expr<!fir.char<1,3>>
! CHECK: }
! CHECK: %[[VAL_29:.*]] = arith.constant 0 : index
diff --git a/flang/test/Lower/HLFIR/proc-pointer-comp-pass.f90 b/flang/test/Lower/HLFIR/proc-pointer-comp-pass.f90
index 247008e3a93df2..c1a827fe36ab1d 100644
--- a/flang/test/Lower/HLFIR/proc-pointer-comp-pass.f90
+++ b/flang/test/Lower/HLFIR/proc-pointer-comp-pass.f90
@@ -105,6 +105,6 @@ subroutine test5(x)
! CHECK: %[[VAL_7:.*]] = arith.constant 0 : index
! CHECK: %[[VAL_8:.*]] = arith.cmpi sgt, %[[VAL_6]], %[[VAL_7]] : index
! CHECK: %[[VAL_9:.*]] = arith.select %[[VAL_8]], %[[VAL_6]], %[[VAL_7]] : index
-! CHECK: %[[VAL_10:.*]] = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+! CHECK: %[[VAL_10:.*]] = llvm.intr.stacksave : !llvm.ptr
! CHECK: %[[VAL_11:.*]] = fir.box_addr %[[VAL_4]] : (!fir.boxproc<(!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3{c:!fir.char<1,4>,p:!fir.boxproc<(!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3>>) -> !fir.boxchar<1>>}>>) -> !fir.boxchar<1>>) -> ((!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3{c:!fir.char<1,4>,p:!fir.boxproc<(!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3>>) -> !fir.boxchar<1>>}>>) -> !fir.boxchar<1>)
! CHECK: %[[VAL_12:.*]] = fir.call %[[VAL_11]](%[[VAL_1]], %[[VAL_9]], %[[VAL_2]]#1) fastmath<contract> : (!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3{c:!fir.char<1,4>,p:!fir.boxproc<(!fir.ref<!fir.char<1,4>>, index, !fir.ref<!fir.type<_QMmTt3>>) -> !fir.boxchar<1>>}>>) -> !fir.boxchar<1>
diff --git a/flang/test/Lower/HLFIR/where-nonelemental.f90 b/flang/test/Lower/HLFIR/where-nonelemental.f90
index f0a6857f0f4b99..15a281b0ba6813 100644
--- a/flang/test/Lower/HLFIR/where-nonelemental.f90
+++ b/flang/test/Lower/HLFIR/where-nonelemental.f90
@@ -26,10 +26,10 @@ real elemental function elem_func(x)
! CHECK-LABEL: func.func @_QPtest_where(
! CHECK: hlfir.where {
! CHECK-NOT: hlfir.exactly_once
-! CHECK: %[[VAL_17:.*]] = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+! CHECK: %[[VAL_17:.*]] = llvm.intr.stacksave : !llvm.ptr
! CHECK: %[[VAL_19:.*]] = fir.call @_QPlogical_func1() fastmath<contract> : () -> !fir.array<100x!fir.logical<4>>
! CHECK: hlfir.yield %{{.*}} : !hlfir.expr<100x!fir.logical<4>> cleanup {
-! CHECK: fir.call @llvm.stackrestore.p0(%[[VAL_17]]) fastmath<contract> : (!fir.ref<i8>) -> ()
+! CHECK: llvm.intr.stackrestore %[[VAL_17]] : !llvm.ptr
! CHECK: }
! CHECK: } do {
! CHECK: hlfir.region_assign {
@@ -70,10 +70,10 @@ real elemental function elem_func(x)
! CHECK: }
! CHECK: hlfir.elsewhere mask {
! CHECK: %[[VAL_62:.*]] = hlfir.exactly_once : !hlfir.expr<100x!fir.logical<4>> {
-! CHECK: %[[VAL_72:.*]] = fir.call @llvm.stacksave.p0() fastmath<contract> : () -> !fir.ref<i8>
+! CHECK: %[[VAL_72:.*]] = llvm.intr.stacksave : !llvm.ptr
! CHECK: fir.call @_QPlogical_func2() fast...
[truncated]
|
jeanPerier
approved these changes
Sep 16, 2024
I decided to use std::optional instead of std::unique_ptr because we don't need heap allocation here.
tblah
added a commit
to tblah/llvm-project
that referenced
this pull request
Sep 16, 2024
Since llvm#108562, StackArrays no longer has to create function declarations at the module level to use stacksave/stackrestore LLVM intrinsics. This will allow it to run in parallel on multiple functions at the same time.
tblah
added a commit
that referenced
this pull request
Sep 17, 2024
Since #108562, StackArrays no longer has to create function declarations at the module level to use stacksave/stackrestore LLVM intrinsics. This will allow it to run in parallel on multiple functions at the same time.
tmsri
pushed a commit
to tmsri/llvm-project
that referenced
this pull request
Sep 19, 2024
) Since llvm#108562, StackArrays no longer has to create function declarations at the module level to use stacksave/stackrestore LLVM intrinsics. This will allow it to run in parallel on multiple functions at the same time.
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.
The new LLVM stack save/restore intrinsic operations are more convenient than function calls because they do not add function declarations to the module and therefore do not block the parallelisation of passes. Furthermore they could be much more easily marked with memory effects than function calls if that ever proved useful.
This builds on top of #107879.
Resolves #108016