[Mlir-commits] [flang] [mlir] [MLIR][LLVMIR] Adding scan lowering to llvm on the mlir side (PR #165788)
Anchu Rajendran S
llvmlistbot at llvm.org
Thu Oct 30 15:09:01 PDT 2025
https://github.com/anchuraj created https://github.com/llvm/llvm-project/pull/165788
Scan reductions are supported in OpenMP with the the help of scan directive. Reduction clause of the for workshare loop/simd directive takes an `inscan` modifier if scan reduction is specified. With an `inscan` modifier, the body of the directive should specify a `scan` directive. This PR implements the lowering logic for scan reductions in workshare loops of OpenMP. OpenMPIRBuilder support can be found in https://github.com/llvm/llvm-project/pull/136035. Support for nested loops/ exclusive clause is not done in this PR
>From cb38aae937f89b0c77aed5f147998f02db7c2d44 Mon Sep 17 00:00:00 2001
From: Anchu Rajendran <asudhaku at amd.com>
Date: Thu, 30 Oct 2025 17:01:59 -0500
Subject: [PATCH] [MLIR][LLVMIR] Adding scan lowering to llvm on the mlir side
---
flang/lib/Lower/OpenMP/OpenMP.cpp | 39 +-
.../OpenMP/OpenMPToLLVMIRTranslation.cpp | 380 ++++++++++++++----
.../Target/LLVMIR/openmp-reduction-scan.mlir | 130 ++++++
mlir/test/Target/LLVMIR/openmp-todo.mlir | 38 +-
4 files changed, 503 insertions(+), 84 deletions(-)
create mode 100644 mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index f86ee01355104..5d82466889b1e 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -2326,12 +2326,41 @@ genParallelOp(lower::AbstractConverter &converter, lower::SymMap &symTable,
static mlir::omp::ScanOp
genScanOp(lower::AbstractConverter &converter, lower::SymMap &symTable,
- semantics::SemanticsContext &semaCtx, mlir::Location loc,
- const ConstructQueue &queue, ConstructQueue::const_iterator item) {
+ semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
+ mlir::Location loc, const ConstructQueue &queue,
+ ConstructQueue::const_iterator item) {
mlir::omp::ScanOperands clauseOps;
genScanClauses(converter, semaCtx, item->clauses, loc, clauseOps);
- return mlir::omp::ScanOp::create(converter.getFirOpBuilder(),
- converter.getCurrentLocation(), clauseOps);
+ mlir::omp::ScanOp scanOp = mlir::omp::ScanOp::create(
+ converter.getFirOpBuilder(), converter.getCurrentLocation(), clauseOps);
+ // If there are nested loops all indices should be loaded after
+ // the scan construct as otherwise, it would result in using the index
+ // variable across scan directive.
+ // (`Intra-iteration dependences from a statement in the structured
+ // block sequence that precede a scan directive to a statement in the
+ // structured block sequence that follows a scan directive must not exist,
+ // except for dependences for the list items specified in an inclusive or
+ // exclusive clause.`).
+ // TODO: If there are nested loops, it is not handled.
+ mlir::omp::LoopNestOp loopNestOp =
+ scanOp->getParentOfType<mlir::omp::LoopNestOp>();
+ assert(loopNestOp.getNumLoops() == 1 &&
+ "Scan directive inside nested do loops is not handled yet.");
+ mlir::Region ®ion = loopNestOp->getRegion(0);
+ mlir::Value indexVal = fir::getBase(region.getArgument(0));
+ lower::pft::Evaluation *doConstructEval = eval.parentConstruct;
+ fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+ lower::pft::Evaluation *doLoop = &doConstructEval->getFirstNestedEvaluation();
+ auto *doStmt = doLoop->getIf<parser::NonLabelDoStmt>();
+ assert(doStmt && "Expected do loop to be in the nested evaluation");
+ const auto &loopControl =
+ std::get<std::optional<parser::LoopControl>>(doStmt->t);
+ const parser::LoopControl::Bounds *bounds =
+ std::get_if<parser::LoopControl::Bounds>(&loopControl->u);
+ mlir::Operation *storeOp =
+ setLoopVar(converter, loc, indexVal, bounds->name.thing.symbol);
+ firOpBuilder.setInsertionPointAfter(storeOp);
+ return scanOp;
}
static mlir::omp::SectionsOp
@@ -3416,7 +3445,7 @@ static void genOMPDispatch(lower::AbstractConverter &converter,
loc, queue, item);
break;
case llvm::omp::Directive::OMPD_scan:
- newOp = genScanOp(converter, symTable, semaCtx, loc, queue, item);
+ newOp = genScanOp(converter, symTable, semaCtx, eval, loc, queue, item);
break;
case llvm::omp::Directive::OMPD_section:
llvm_unreachable("genOMPDispatch: OMPD_section");
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 1e2099d6cc1b2..eb4378dd6f719 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -37,6 +37,7 @@
#include "llvm/TargetParser/Triple.h"
#include "llvm/Transforms/Utils/ModuleUtils.h"
+#include <cassert>
#include <cstdint>
#include <iterator>
#include <numeric>
@@ -77,6 +78,22 @@ class OpenMPAllocaStackFrame
llvm::OpenMPIRBuilder::InsertPointTy allocaInsertPoint;
};
+/// ModuleTranslation stack frame for OpenMP operations. This keeps track of the
+/// insertion points for allocas of parent of the current parallel region. The
+/// insertion point is used to allocate variables to be share by the threads
+/// executing the parallel region. Lowering of scan reduction requires declaring
+/// shared pointers to the temporary buffer to perform scan reduction.
+class OpenMPParallelAllocaStackFrame
+ : public StateStackFrameBase<OpenMPParallelAllocaStackFrame> {
+public:
+ MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(OpenMPParallelAllocaStackFrame)
+
+ explicit OpenMPParallelAllocaStackFrame(
+ llvm::OpenMPIRBuilder::InsertPointTy allocaIP)
+ : allocaInsertPoint(allocaIP) {}
+ llvm::OpenMPIRBuilder::InsertPointTy allocaInsertPoint;
+};
+
/// Stack frame to hold a \see llvm::CanonicalLoopInfo representing the
/// collapsed canonical loop information corresponding to an \c omp.loop_nest
/// operation.
@@ -84,7 +101,13 @@ class OpenMPLoopInfoStackFrame
: public StateStackFrameBase<OpenMPLoopInfoStackFrame> {
public:
MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(OpenMPLoopInfoStackFrame)
- llvm::CanonicalLoopInfo *loopInfo = nullptr;
+ /// For constructs like scan, one LoopInfo frame can contain multiple
+ /// Canonical Loops as a single openmpLoopNestOp will be split into input
+ /// loop and scan loop.
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos;
+ llvm::ScanInfo *scanInfo;
+ llvm::DenseMap<llvm::Value *, llvm::Type *> *reductionVarToType =
+ new llvm::DenseMap<llvm::Value *, llvm::Type *>();
};
/// Custom error class to signal translation errors that don't need reporting,
@@ -323,6 +346,10 @@ static LogicalResult checkImplementationStatus(Operation &op) {
if (op.getDistScheduleChunkSize())
result = todo("dist_schedule with chunk_size");
};
+ auto checkExclusive = [&todo](auto op, LogicalResult &result) {
+ if (!op.getExclusiveVars().empty())
+ result = todo("exclusive");
+ };
auto checkHint = [](auto op, LogicalResult &) {
if (op.getHint())
op.emitWarning("hint clause discarded");
@@ -371,9 +398,14 @@ static LogicalResult checkImplementationStatus(Operation &op) {
if (!op.getReductionVars().empty() || op.getReductionByref() ||
op.getReductionSyms())
result = todo("reduction");
- if (op.getReductionMod() &&
- op.getReductionMod().value() != omp::ReductionModifier::defaultmod)
- result = todo("reduction with modifier");
+ if (op.getReductionMod()) {
+ if (isa<omp::WsloopOp>(op)) {
+ if (op.getReductionMod().value() == omp::ReductionModifier::task)
+ result = todo("reduction with task modifier");
+ } else {
+ result = todo("reduction with modifier");
+ }
+ }
};
auto checkTaskReduction = [&todo](auto op, LogicalResult &result) {
if (!op.getTaskReductionVars().empty() || op.getTaskReductionByref() ||
@@ -397,6 +429,7 @@ static LogicalResult checkImplementationStatus(Operation &op) {
checkOrder(op, result);
})
.Case([&](omp::OrderedRegionOp op) { checkParLevelSimd(op, result); })
+ .Case([&](omp::ScanOp op) { checkExclusive(op, result); })
.Case([&](omp::SectionsOp op) {
checkAllocate(op, result);
checkPrivate(op, result);
@@ -531,15 +564,59 @@ findAllocaInsertPoint(llvm::IRBuilderBase &builder,
/// Find the loop information structure for the loop nest being translated. It
/// will return a `null` value unless called from the translation function for
/// a loop wrapper operation after successfully translating its body.
-static llvm::CanonicalLoopInfo *
-findCurrentLoopInfo(LLVM::ModuleTranslation &moduleTranslation) {
- llvm::CanonicalLoopInfo *loopInfo = nullptr;
+static SmallVector<llvm::CanonicalLoopInfo *>
+findCurrentLoopInfos(LLVM::ModuleTranslation &moduleTranslation) {
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos;
moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
[&](OpenMPLoopInfoStackFrame &frame) {
- loopInfo = frame.loopInfo;
+ loopInfos = frame.loopInfos;
return WalkResult::interrupt();
});
- return loopInfo;
+ return loopInfos;
+}
+
+// LoopFrame stores the scaninfo which is used for scan reduction.
+// Upon encountering an `inscan` reduction modifier, `scanInfoInitialize`
+// initializes the ScanInfo and is used when scan directive is encountered
+// in the body of the loop nest.
+static llvm::ScanInfo *
+findScanInfo(LLVM::ModuleTranslation &moduleTranslation) {
+ llvm::ScanInfo *scanInfo;
+ moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
+ [&](OpenMPLoopInfoStackFrame &frame) {
+ scanInfo = frame.scanInfo;
+ return WalkResult::interrupt();
+ });
+ return scanInfo;
+}
+
+// The types of reduction vars are used for lowering scan directive which
+// appears in the body of the loop. The types are stored in loop frame when
+// reduction clause is encountered and is used when scan directive is
+// encountered.
+static llvm::DenseMap<llvm::Value *, llvm::Type *> *
+findReductionVarTypes(LLVM::ModuleTranslation &moduleTranslation) {
+ llvm::DenseMap<llvm::Value *, llvm::Type *> *reductionVarToType = nullptr;
+ moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
+ [&](OpenMPLoopInfoStackFrame &frame) {
+ reductionVarToType = frame.reductionVarToType;
+ return WalkResult::interrupt();
+ });
+ return reductionVarToType;
+}
+
+// Scan reduction requires a shared buffer to be allocated to perform reduction.
+// ParallelAllocaStackFrame holds the allocaIP where shared allocation can be
+// done.
+static llvm::OpenMPIRBuilder::InsertPointTy
+findParallelAllocaIP(LLVM::ModuleTranslation &moduleTranslation) {
+ llvm::OpenMPIRBuilder::InsertPointTy parallelAllocaIP;
+ moduleTranslation.stackWalk<OpenMPParallelAllocaStackFrame>(
+ [&](OpenMPParallelAllocaStackFrame &frame) {
+ parallelAllocaIP = frame.allocaInsertPoint;
+ return WalkResult::interrupt();
+ });
+ return parallelAllocaIP;
}
/// Converts the given region that appears within an OpenMP dialect operation to
@@ -1254,11 +1331,17 @@ initReductionVars(OP op, ArrayRef<BlockArgument> reductionArgs,
for (auto [data, addr] : deferredStores)
builder.CreateStore(data, addr);
+ llvm::DenseMap<llvm::Value *, llvm::Type *> *reductionVarToType =
+ findReductionVarTypes(moduleTranslation);
// Before the loop, store the initial values of reductions into reduction
// variables. Although this could be done after allocas, we don't want to mess
// up with the alloca insertion point.
for (unsigned i = 0; i < op.getNumReductionVars(); ++i) {
SmallVector<llvm::Value *, 1> phis;
+ llvm::Type *reductionType =
+ moduleTranslation.convertType(reductionDecls[i].getType());
+ if (reductionVarToType != nullptr)
+ (*reductionVarToType)[privateReductionVariables[i]] = reductionType;
// map block argument to initializer region
mapInitializationArgs(op, moduleTranslation, reductionDecls,
@@ -1330,15 +1413,20 @@ static void collectReductionInfo(
// Collect the reduction information.
reductionInfos.reserve(numReductions);
+ llvm::DenseMap<llvm::Value *, llvm::Type *> *reductionVarToType =
+ findReductionVarTypes(moduleTranslation);
for (unsigned i = 0; i < numReductions; ++i) {
llvm::OpenMPIRBuilder::ReductionGenAtomicCBTy atomicGen = nullptr;
if (owningAtomicReductionGens[i])
atomicGen = owningAtomicReductionGens[i];
llvm::Value *variable =
moduleTranslation.lookupValue(loop.getReductionVars()[i]);
+ llvm::Type *reductionType =
+ moduleTranslation.convertType(reductionDecls[i].getType());
+ if (reductionVarToType != nullptr)
+ (*reductionVarToType)[privateReductionVariables[i]] = reductionType;
reductionInfos.push_back(
- {moduleTranslation.convertType(reductionDecls[i].getType()), variable,
- privateReductionVariables[i],
+ {reductionType, variable, privateReductionVariables[i],
/*EvaluationKind=*/llvm::OpenMPIRBuilder::EvalKind::Scalar,
owningReductionGens[i],
/*ReductionGenClang=*/nullptr, atomicGen});
@@ -2543,6 +2631,9 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
std::optional<omp::ScheduleModifier> scheduleMod = wsloopOp.getScheduleMod();
bool isSimd = wsloopOp.getScheduleSimd();
bool loopNeedsBarrier = !wsloopOp.getNowait();
+ bool isInScanRegion =
+ wsloopOp.getReductionMod() && (wsloopOp.getReductionMod().value() ==
+ mlir::omp::ReductionModifier::inscan);
// The only legal way for the direct parent to be omp.distribute is that this
// represents 'distribute parallel do'. Otherwise, this is a regular
@@ -2574,20 +2665,81 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
if (failed(handleError(regionBlock, opInst)))
return failure();
- llvm::CanonicalLoopInfo *loopInfo = findCurrentLoopInfo(moduleTranslation);
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos =
+ findCurrentLoopInfos(moduleTranslation);
+
+ const auto &&wsloopCodeGen = [&](llvm::CanonicalLoopInfo *loopInfo,
+ bool noLoopMode, bool inputScanLoop) {
+ bool emitLinearVarInit = !isInScanRegion || inputScanLoop;
+ // Emit Initialization and Update IR for linear variables
+ if (emitLinearVarInit && !wsloopOp.getLinearVars().empty()) {
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterBarrierIP =
+ linearClauseProcessor.initLinearVar(builder, moduleTranslation,
+ loopInfo->getPreheader());
+ if (failed(handleError(afterBarrierIP, *loopOp)))
+ return failure();
+ builder.restoreIP(*afterBarrierIP);
+ linearClauseProcessor.updateLinearVar(builder, loopInfo->getBody(),
+ loopInfo->getIndVar());
+ linearClauseProcessor.outlineLinearFinalizationBB(builder,
+ loopInfo->getExit());
+ }
- // Emit Initialization and Update IR for linear variables
- if (!wsloopOp.getLinearVars().empty()) {
- llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterBarrierIP =
- linearClauseProcessor.initLinearVar(builder, moduleTranslation,
- loopInfo->getPreheader());
- if (failed(handleError(afterBarrierIP, *loopOp)))
+ builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy wsloopIP =
+ ompBuilder->applyWorkshareLoop(
+ ompLoc.DL, loopInfo, allocaIP, loopNeedsBarrier,
+ convertToScheduleKind(schedule), chunk, isSimd,
+ scheduleMod == omp::ScheduleModifier::monotonic,
+ scheduleMod == omp::ScheduleModifier::nonmonotonic, isOrdered,
+ workshareLoopType, noLoopMode);
+
+ if (failed(handleError(wsloopIP, opInst)))
+ return failure();
+
+ bool emitLinearVarFinalize = !isInScanRegion || !inputScanLoop;
+ // Emit finalization and in-place rewrites for linear vars.
+ if (emitLinearVarFinalize && !wsloopOp.getLinearVars().empty()) {
+ llvm::OpenMPIRBuilder::InsertPointTy oldIP = builder.saveIP();
+ if (loopInfo->getLastIter())
+ return failure();
+ // assert(loopInfo->getLastIter() &&
+ // "`lastiter` in CanonicalLoopInfo is nullptr");
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterBarrierIP =
+ linearClauseProcessor.finalizeLinearVar(builder, moduleTranslation,
+ loopInfo->getLastIter());
+ if (failed(handleError(afterBarrierIP, *loopOp)))
+ return failure();
+ for (size_t index = 0; index < wsloopOp.getLinearVars().size(); index++)
+ linearClauseProcessor.rewriteInPlace(builder, "omp.loop_nest.region",
+ index);
+ builder.restoreIP(oldIP);
+ }
+ if (!inputScanLoop || !isInScanRegion)
+ popCancelFinalizationCB(cancelTerminators, *ompBuilder, wsloopIP.get());
+
+ return llvm::success();
+ };
+
+ if (isInScanRegion) {
+ auto inputLoopFinishIp = loopInfos.front()->getAfterIP();
+ builder.restoreIP(inputLoopFinishIp);
+ SmallVector<OwningReductionGen> owningReductionGens;
+ SmallVector<OwningAtomicReductionGen> owningAtomicReductionGens;
+ SmallVector<llvm::OpenMPIRBuilder::ReductionInfo> reductionInfos;
+ collectReductionInfo(wsloopOp, builder, moduleTranslation, reductionDecls,
+ owningReductionGens, owningAtomicReductionGens,
+ privateReductionVariables, reductionInfos);
+ llvm::BasicBlock *cont = splitBB(builder, false, "omp.scan.loop.cont");
+ llvm::ScanInfo *scanInfo = findScanInfo(moduleTranslation);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy redIP =
+ ompBuilder->emitScanReduction(builder.saveIP(), reductionInfos,
+ scanInfo);
+ if (failed(handleError(redIP, opInst)))
return failure();
- builder.restoreIP(*afterBarrierIP);
- linearClauseProcessor.updateLinearVar(builder, loopInfo->getBody(),
- loopInfo->getIndVar());
- linearClauseProcessor.outlineLinearFinalizationBB(builder,
- loopInfo->getExit());
+
+ builder.restoreIP(*redIP);
+ builder.CreateBr(cont);
}
builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
@@ -2612,42 +2764,34 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
}
}
- llvm::OpenMPIRBuilder::InsertPointOrErrorTy wsloopIP =
- ompBuilder->applyWorkshareLoop(
- ompLoc.DL, loopInfo, allocaIP, loopNeedsBarrier,
- convertToScheduleKind(schedule), chunk, isSimd,
- scheduleMod == omp::ScheduleModifier::monotonic,
- scheduleMod == omp::ScheduleModifier::nonmonotonic, isOrdered,
- workshareLoopType, noLoopMode);
-
- if (failed(handleError(wsloopIP, opInst)))
- return failure();
-
- // Emit finalization and in-place rewrites for linear vars.
- if (!wsloopOp.getLinearVars().empty()) {
- llvm::OpenMPIRBuilder::InsertPointTy oldIP = builder.saveIP();
- assert(loopInfo->getLastIter() &&
- "`lastiter` in CanonicalLoopInfo is nullptr");
- llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterBarrierIP =
- linearClauseProcessor.finalizeLinearVar(builder, moduleTranslation,
- loopInfo->getLastIter());
- if (failed(handleError(afterBarrierIP, *loopOp)))
+ // For Scan loops input loop need not pop cancellation CB and hence, it is set
+ // false for the first loop
+ bool inputScanLoop = isInScanRegion;
+ for (llvm::CanonicalLoopInfo *loopInfo : loopInfos) {
+ if (failed(wsloopCodeGen(loopInfo, noLoopMode, inputScanLoop)))
return failure();
- for (size_t index = 0; index < wsloopOp.getLinearVars().size(); index++)
- linearClauseProcessor.rewriteInPlace(builder, "omp.loop_nest.region",
- index);
- builder.restoreIP(oldIP);
+ inputScanLoop = false;
}
- // Set the correct branch target for task cancellation
- popCancelFinalizationCB(cancelTerminators, *ompBuilder, wsloopIP.get());
-
- // Process the reductions if required.
- if (failed(createReductionsAndCleanup(
- wsloopOp, builder, moduleTranslation, allocaIP, reductionDecls,
- privateReductionVariables, isByRef, wsloopOp.getNowait(),
- /*isTeamsReduction=*/false)))
- return failure();
+ // todo: change builder.saveIP to wsLoopIP
+ if (isInScanRegion) {
+ SmallVector<Region *> reductionRegions;
+ llvm::transform(reductionDecls, std::back_inserter(reductionRegions),
+ [](omp::DeclareReductionOp reductionDecl) {
+ return &reductionDecl.getCleanupRegion();
+ });
+ if (failed(inlineOmpRegionCleanup(
+ reductionRegions, privateReductionVariables, moduleTranslation,
+ builder, "omp.reduction.cleanup")))
+ return failure();
+ } else {
+ // Process the reductions if required.
+ if (failed(createReductionsAndCleanup(
+ wsloopOp, builder, moduleTranslation, allocaIP, reductionDecls,
+ privateReductionVariables, isByRef, wsloopOp.getNowait(),
+ /*isTeamsReduction=*/false)))
+ return failure();
+ }
return cleanupPrivateVars(builder, moduleTranslation, wsloopOp.getLoc(),
privateVarsInfo.llvmVars,
@@ -2815,6 +2959,8 @@ convertOmpParallel(omp::ParallelOp opInst, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
findAllocaInsertPoint(builder, moduleTranslation);
llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
+ LLVM::ModuleTranslation::SaveStack<OpenMPParallelAllocaStackFrame> frame(
+ moduleTranslation, allocaIP);
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP =
ompBuilder->createParallel(ompLoc, allocaIP, bodyGenCB, privCB, finiCB,
@@ -2935,12 +3081,15 @@ convertOmpSimd(Operation &opInst, llvm::IRBuilderBase &builder,
return failure();
builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
- llvm::CanonicalLoopInfo *loopInfo = findCurrentLoopInfo(moduleTranslation);
- ompBuilder->applySimd(loopInfo, alignedVars,
- simdOp.getIfExpr()
- ? moduleTranslation.lookupValue(simdOp.getIfExpr())
- : nullptr,
- order, simdlen, safelen);
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos =
+ findCurrentLoopInfos(moduleTranslation);
+ for (llvm::CanonicalLoopInfo *loopInfo : loopInfos) {
+ ompBuilder->applySimd(
+ loopInfo, alignedVars,
+ simdOp.getIfExpr() ? moduleTranslation.lookupValue(simdOp.getIfExpr())
+ : nullptr,
+ order, simdlen, safelen);
+ }
// We now need to reduce the per-simd-lane reduction variable into the
// original variable. This works a bit differently to other reductions (e.g.
@@ -2991,6 +3140,40 @@ convertOmpSimd(Operation &opInst, llvm::IRBuilderBase &builder,
privateVarsInfo.privatizers);
}
+static LogicalResult
+convertOmpScan(Operation &opInst, llvm::IRBuilderBase &builder,
+ LLVM::ModuleTranslation &moduleTranslation) {
+ if (failed(checkImplementationStatus(opInst)))
+ return failure();
+ auto scanOp = cast<omp::ScanOp>(opInst);
+ bool isInclusive = scanOp.hasInclusiveVars();
+ SmallVector<llvm::Value *> llvmScanVars;
+ SmallVector<llvm::Type *> llvmScanVarsType;
+ mlir::OperandRange mlirScanVars = scanOp.getInclusiveVars();
+ if (!isInclusive)
+ mlirScanVars = scanOp.getExclusiveVars();
+
+ llvm::DenseMap<llvm::Value *, llvm::Type *> *reductionVarToType =
+ findReductionVarTypes(moduleTranslation);
+ for (auto val : mlirScanVars) {
+ llvm::Value *llvmVal = moduleTranslation.lookupValue(val);
+ llvmScanVars.push_back(llvmVal);
+ llvmScanVarsType.push_back((*reductionVarToType)[llvmVal]);
+ }
+ llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
+ findParallelAllocaIP(moduleTranslation);
+ llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
+ llvm::ScanInfo *scanInfo = findScanInfo(moduleTranslation);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP =
+ moduleTranslation.getOpenMPBuilder()->createScan(
+ ompLoc, allocaIP, llvmScanVars, llvmScanVarsType, isInclusive,
+ scanInfo);
+ if (failed(handleError(afterIP, opInst)))
+ return failure();
+ builder.restoreIP(*afterIP);
+ return success();
+}
+
/// Converts an OpenMP loop nest into LLVM IR using OpenMPIRBuilder.
static LogicalResult
convertOmpLoopNest(Operation &opInst, llvm::IRBuilderBase &builder,
@@ -3052,14 +3235,50 @@ convertOmpLoopNest(Operation &opInst, llvm::IRBuilderBase &builder,
computeIP = loopInfos.front()->getPreheaderIP();
}
+ bool isInScanRegion = false;
+ if (auto wsloopOp = loopOp->getParentOfType<omp::WsloopOp>())
+ isInScanRegion =
+ wsloopOp.getReductionMod() && (wsloopOp.getReductionMod().value() ==
+ mlir::omp::ReductionModifier::inscan);
+ if (isInScanRegion) {
+ // TODO: Handle nesting if Scan loop is nested in a loop
+ assert(loopOp.getNumLoops() == 1 &&
+ "Scan directive inside nested do loops is not handled yet.");
+ llvm::Expected<llvm::ScanInfo *> res = ompBuilder->scanInfoInitialize();
+ if (failed(handleError(res, *loopOp)))
+ return failure();
+ llvm::ScanInfo *scanInfo = res.get();
+ moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
+ [&](OpenMPLoopInfoStackFrame &frame) {
+ frame.scanInfo = scanInfo;
+ return WalkResult::interrupt();
+ });
+ llvm::Expected<llvm::SmallVector<llvm::CanonicalLoopInfo *>> loopResults =
+ ompBuilder->createCanonicalScanLoops(
+ loc, bodyGen, lowerBound, upperBound, step,
+ /*IsSigned=*/true, loopOp.getLoopInclusive(), computeIP, "loop",
+ scanInfo);
+
+ if (failed(handleError(loopResults, *loopOp)))
+ return failure();
+ llvm::CanonicalLoopInfo *inputLoop = loopResults.get().front();
+ llvm::CanonicalLoopInfo *scanLoop = loopResults.get().back();
+ moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
+ [&](OpenMPLoopInfoStackFrame &frame) {
+ frame.loopInfos.push_back(inputLoop);
+ frame.loopInfos.push_back(scanLoop);
+ return WalkResult::interrupt();
+ });
+ builder.restoreIP(scanLoop->getAfterIP());
+ // TODO: tiling and collapse are not yet implemented for scan reduction
+ return success();
+ }
llvm::Expected<llvm::CanonicalLoopInfo *> loopResult =
ompBuilder->createCanonicalLoop(
loc, bodyGen, lowerBound, upperBound, step,
/*IsSigned=*/true, loopOp.getLoopInclusive(), computeIP);
-
if (failed(handleError(loopResult, *loopOp)))
return failure();
-
loopInfos.push_back(*loopResult);
}
@@ -3102,7 +3321,7 @@ convertOmpLoopNest(Operation &opInst, llvm::IRBuilderBase &builder,
assert(newTopLoopInfo && "New top loop information is missing");
moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
[&](OpenMPLoopInfoStackFrame &frame) {
- frame.loopInfo = newTopLoopInfo;
+ frame.loopInfos.push_back(newTopLoopInfo);
return WalkResult::interrupt();
});
@@ -4965,18 +5184,20 @@ convertOmpDistribute(Operation &opInst, llvm::IRBuilderBase &builder,
bool loopNeedsBarrier = false;
llvm::Value *chunk = nullptr;
- llvm::CanonicalLoopInfo *loopInfo =
- findCurrentLoopInfo(moduleTranslation);
- llvm::OpenMPIRBuilder::InsertPointOrErrorTy wsloopIP =
- ompBuilder->applyWorkshareLoop(
- ompLoc.DL, loopInfo, allocaIP, loopNeedsBarrier,
- convertToScheduleKind(schedule), chunk, isSimd,
- scheduleMod == omp::ScheduleModifier::monotonic,
- scheduleMod == omp::ScheduleModifier::nonmonotonic, isOrdered,
- workshareLoopType);
-
- if (!wsloopIP)
- return wsloopIP.takeError();
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos =
+ findCurrentLoopInfos(moduleTranslation);
+ for (llvm::CanonicalLoopInfo *loopInfo : loopInfos) {
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy wsloopIP =
+ ompBuilder->applyWorkshareLoop(
+ ompLoc.DL, loopInfo, allocaIP, loopNeedsBarrier,
+ convertToScheduleKind(schedule), chunk, isSimd,
+ scheduleMod == omp::ScheduleModifier::monotonic,
+ scheduleMod == omp::ScheduleModifier::nonmonotonic, isOrdered,
+ workshareLoopType);
+
+ if (!wsloopIP)
+ return wsloopIP.takeError();
+ }
}
if (failed(cleanupPrivateVars(builder, moduleTranslation,
@@ -6167,6 +6388,11 @@ convertHostOrTargetOperation(Operation *op, llvm::IRBuilderBase &builder,
.Case([&](omp::WsloopOp) {
return convertOmpWsloop(*op, builder, moduleTranslation);
})
+ .Case([&](omp::ScanOp) {
+ if (failed(checkImplementationStatus(*op)))
+ return failure();
+ return convertOmpScan(*op, builder, moduleTranslation);
+ })
.Case([&](omp::SimdOp) {
return convertOmpSimd(*op, builder, moduleTranslation);
})
diff --git a/mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir b/mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir
new file mode 100644
index 0000000000000..ed04a069b998f
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir
@@ -0,0 +1,130 @@
+// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+omp.declare_reduction @add_reduction_i32 : i32 init {
+^bb0(%arg0: i32):
+ %0 = llvm.mlir.constant(0 : i32) : i32
+ omp.yield(%0 : i32)
+} combiner {
+^bb0(%arg0: i32, %arg1: i32):
+ %0 = llvm.add %arg0, %arg1 : i32
+ omp.yield(%0 : i32)
+}
+// CHECK-LABEL: @scan_reduction
+llvm.func @scan_reduction() {
+ %0 = llvm.mlir.constant(1 : i64) : i64
+ %1 = llvm.alloca %0 x i32 {bindc_name = "z"} : (i64) -> !llvm.ptr
+ %2 = llvm.mlir.constant(1 : i64) : i64
+ %3 = llvm.alloca %2 x i32 {bindc_name = "y"} : (i64) -> !llvm.ptr
+ %4 = llvm.mlir.constant(1 : i64) : i64
+ %5 = llvm.alloca %4 x i32 {bindc_name = "x"} : (i64) -> !llvm.ptr
+ %6 = llvm.mlir.constant(1 : i64) : i64
+ %7 = llvm.alloca %6 x i32 {bindc_name = "k"} : (i64) -> !llvm.ptr
+ %8 = llvm.mlir.constant(0 : index) : i64
+ %9 = llvm.mlir.constant(1 : index) : i64
+ %10 = llvm.mlir.constant(100 : i32) : i32
+ %11 = llvm.mlir.constant(1 : i32) : i32
+ %12 = llvm.mlir.constant(0 : i32) : i32
+ %13 = llvm.mlir.constant(100 : index) : i64
+ %14 = llvm.mlir.addressof @_QFEa : !llvm.ptr
+ %15 = llvm.mlir.addressof @_QFEb : !llvm.ptr
+ omp.parallel {
+ %37 = llvm.mlir.constant(1 : i64) : i64
+ %38 = llvm.alloca %37 x i32 {bindc_name = "k", pinned} : (i64) -> !llvm.ptr
+ %39 = llvm.mlir.constant(1 : i64) : i64
+ omp.wsloop reduction(mod: inscan, @add_reduction_i32 %5 -> %arg0 : !llvm.ptr) {
+ omp.loop_nest (%arg1) : i32 = (%11) to (%10) inclusive step (%11) {
+ llvm.store %arg1, %38 : i32, !llvm.ptr
+ %40 = llvm.load %arg0 : !llvm.ptr -> i32
+ %41 = llvm.load %38 : !llvm.ptr -> i32
+ %42 = llvm.sext %41 : i32 to i64
+ %50 = llvm.getelementptr %14[%42] : (!llvm.ptr, i64) -> !llvm.ptr, i32
+ %51 = llvm.load %50 : !llvm.ptr -> i32
+ %52 = llvm.add %40, %51 : i32
+ llvm.store %52, %arg0 : i32, !llvm.ptr
+ omp.scan inclusive(%arg0 : !llvm.ptr)
+ llvm.store %arg1, %38 : i32, !llvm.ptr
+ %53 = llvm.load %arg0 : !llvm.ptr -> i32
+ %54 = llvm.load %38 : !llvm.ptr -> i32
+ %55 = llvm.sext %54 : i32 to i64
+ %63 = llvm.getelementptr %15[%55] : (!llvm.ptr, i64) -> !llvm.ptr, i32
+ llvm.store %53, %63 : i32, !llvm.ptr
+ omp.yield
+ }
+ }
+ omp.terminator
+ }
+ llvm.return
+}
+llvm.mlir.global internal @_QFEa() {addr_space = 0 : i32} : !llvm.array<100 x i32> {
+ %0 = llvm.mlir.zero : !llvm.array<100 x i32>
+ llvm.return %0 : !llvm.array<100 x i32>
+}
+llvm.mlir.global internal @_QFEb() {addr_space = 0 : i32} : !llvm.array<100 x i32> {
+ %0 = llvm.mlir.zero : !llvm.array<100 x i32>
+ llvm.return %0 : !llvm.array<100 x i32>
+}
+llvm.mlir.global internal constant @_QFECn() {addr_space = 0 : i32} : i32 {
+ %0 = llvm.mlir.constant(100 : i32) : i32
+ llvm.return %0 : i32
+}
+//CHECK: %vla = alloca ptr, align 8
+//CHECK: omp_parallel
+//CHECK: store ptr %vla, ptr %gep_vla, align 8
+//CHECK: @__kmpc_fork_call
+//CHECK: void @scan_reduction..omp_par
+//CHECK: %[[BUFF_PTR:.+]] = load ptr, ptr %gep_vla
+//CHECK: @__kmpc_masked
+//CHECK: @__kmpc_barrier
+//CHECK: @__kmpc_masked
+//CHECK: @__kmpc_barrier
+//CHECK: omp.scan.loop.cont:
+//CHECK: @__kmpc_masked
+//CHECK: @__kmpc_barrier
+//CHECK: %[[FREE_VAR:.+]] = load ptr, ptr %[[BUFF_PTR]], align 8
+//CHECK: %[[ARRLAST:.+]] = getelementptr inbounds i32, ptr %[[FREE_VAR]], i32 100
+//CHECK: %[[RES:.+]] = load i32, ptr %[[ARRLAST]], align 4
+//CHECK: store i32 %[[RES]], ptr %loadgep{{.*}}, align 4
+//CHECK: tail call void @free(ptr %[[FREE_VAR]])
+//CHECK: @__kmpc_end_masked
+//CHECK: omp.inscan.dispatch{{.*}}: ; preds = %omp_loop.body{{.*}}
+//CHECK: %[[BUFFVAR:.+]] = load ptr, ptr %[[BUFF_PTR]], align 8
+//CHECK: %[[arrayOffset1:.+]] = getelementptr inbounds i32, ptr %[[BUFFVAR]], i32 %{{.*}}
+//CHECK: %[[BUFFVAL1:.+]] = load i32, ptr %[[arrayOffset1]], align 4
+//CHECK: store i32 %[[BUFFVAL1]], ptr %{{.*}}, align 4
+//CHECK: %[[LOG:.+]] = call double @llvm.log2.f64(double 1.000000e+02) #0
+//CHECK: %[[CEIL:.+]] = call double @llvm.ceil.f64(double %[[LOG]]) #0
+//CHECK: %[[UB:.+]] = fptoui double %[[CEIL]] to i32
+//CHECK: br label %omp.outer.log.scan.body
+//CHECK: omp.outer.log.scan.body:
+//CHECK: %[[K:.+]] = phi i32 [ 0, %{{.*}} ], [ %[[NEXTK:.+]], %omp.inner.log.scan.exit ]
+//CHECK: %[[I:.+]] = phi i32 [ 1, %{{.*}} ], [ %[[NEXTI:.+]], %omp.inner.log.scan.exit ]
+//CHECK: %[[CMP1:.+]] = icmp uge i32 99, %[[I]]
+//CHECK: br i1 %[[CMP1]], label %omp.inner.log.scan.body, label %omp.inner.log.scan.exit
+//CHECK: omp.inner.log.scan.exit: ; preds = %omp.inner.log.scan.body, %omp.outer.log.scan.body
+//CHECK: %[[NEXTK]] = add nuw i32 %[[K]], 1
+//CHECK: %[[NEXTI]] = shl nuw i32 %[[I]], 1
+//CHECK: %[[CMP2:.+]] = icmp ne i32 %[[NEXTK]], %[[UB]]
+//CHECK: br i1 %[[CMP2]], label %omp.outer.log.scan.body, label %omp.outer.log.scan.exit
+//CHECK: omp.outer.log.scan.exit: ; preds = %omp.inner.log.scan.exit
+//CHECK: @__kmpc_end_masked
+//CHECK: omp.inner.log.scan.body: ; preds = %omp.inner.log.scan.body, %omp.outer.log.scan.body
+//CHECK: %[[CNT:.+]] = phi i32 [ 99, %omp.outer.log.scan.body ], [ %[[CNTNXT:.+]], %omp.inner.log.scan.body ]
+//CHECK: %[[BUFF:.+]] = load ptr, ptr %[[BUFF_PTR]]
+//CHECK: %[[IND1:.+]] = add i32 %[[CNT]], 1
+//CHECK: %[[IND1PTR:.+]] = getelementptr inbounds i32, ptr %[[BUFF]], i32 %[[IND1]]
+//CHECK: %[[IND2:.+]] = sub nuw i32 %[[IND1]], %[[I]]
+//CHECK: %[[IND2PTR:.+]] = getelementptr inbounds i32, ptr %[[BUFF]], i32 %[[IND2]]
+//CHECK: %[[IND1VAL:.+]] = load i32, ptr %[[IND1PTR]], align 4
+//CHECK: %[[IND2VAL:.+]] = load i32, ptr %[[IND2PTR]], align 4
+//CHECK: %[[REDVAL:.+]] = add i32 %[[IND1VAL]], %[[IND2VAL]]
+//CHECK: store i32 %[[REDVAL]], ptr %[[IND1PTR]], align 4
+//CHECK: %[[CNTNXT]] = sub nuw i32 %[[CNT]], 1
+//CHECK: %[[CMP3:.+]] = icmp uge i32 %[[CNTNXT]], %[[I]]
+//CHECK: br i1 %[[CMP3]], label %omp.inner.log.scan.body, label %omp.inner.log.scan.exit
+//CHECK: omp.inscan.dispatch: ; preds = %omp_loop.body
+//CHECK: br i1 true, label %omp.before.scan.bb, label %omp.after.scan.bb
+//CHECK: omp.loop_nest.region: ; preds = %omp.before.scan.bb
+//CHECK: %[[BUFFER:.+]] = load ptr, ptr %loadgep_vla, align 8
+//CHECK: %[[ARRAYOFFSET2:.+]] = getelementptr inbounds i32, ptr %[[BUFFER]], i32 %{{.*}}
+//CHECK-NEXT: %[[REDPRIVVAL:.+]] = load i32, ptr %{{.*}}, align 4
+//CHECK: store i32 %[[REDPRIVVAL]], ptr %[[ARRAYOFFSET2]], align 4
+//CHECK: br label %omp.scan.loop.exit
diff --git a/mlir/test/Target/LLVMIR/openmp-todo.mlir b/mlir/test/Target/LLVMIR/openmp-todo.mlir
index 2fa4470bb8300..32784b14f5302 100644
--- a/mlir/test/Target/LLVMIR/openmp-todo.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-todo.mlir
@@ -129,6 +129,37 @@ llvm.func @simd_linear(%lb : i32, %ub : i32, %step : i32, %x : !llvm.ptr) {
// -----
+omp.declare_reduction @add_f32 : f32
+init {
+^bb0(%arg: f32):
+ %0 = llvm.mlir.constant(0.0 : f32) : f32
+ omp.yield (%0 : f32)
+}
+combiner {
+^bb1(%arg0: f32, %arg1: f32):
+ %1 = llvm.fadd %arg0, %arg1 : f32
+ omp.yield (%1 : f32)
+}
+atomic {
+^bb2(%arg2: !llvm.ptr, %arg3: !llvm.ptr):
+ %2 = llvm.load %arg3 : !llvm.ptr -> f32
+ llvm.atomicrmw fadd %arg2, %2 monotonic : !llvm.ptr, f32
+ omp.yield
+}
+llvm.func @simd_reduction(%lb : i32, %ub : i32, %step : i32, %x : !llvm.ptr) {
+ // expected-error at below {{not yet implemented: Unhandled clause reduction with modifier in omp.simd operation}}
+ // expected-error at below {{LLVM Translation failed for operation: omp.simd}}
+ omp.simd reduction(mod:inscan, @add_f32 %x -> %prv : !llvm.ptr) {
+ omp.loop_nest (%iv) : i32 = (%lb) to (%ub) step (%step) {
+ omp.scan inclusive(%prv : !llvm.ptr)
+ omp.yield
+ }
+ }
+ llvm.return
+}
+
+// -----
+
omp.declare_reduction @add_f32 : f32
init {
^bb0(%arg: f32):
@@ -147,17 +178,20 @@ atomic {
omp.yield
}
llvm.func @scan_reduction(%lb : i32, %ub : i32, %step : i32, %x : !llvm.ptr) {
- // expected-error at below {{not yet implemented: Unhandled clause reduction with modifier in omp.wsloop operation}}
// expected-error at below {{LLVM Translation failed for operation: omp.wsloop}}
omp.wsloop reduction(mod:inscan, @add_f32 %x -> %prv : !llvm.ptr) {
+ // expected-error at below {{LLVM Translation failed for operation: omp.loop_nest}}
omp.loop_nest (%iv) : i32 = (%lb) to (%ub) step (%step) {
- omp.scan inclusive(%prv : !llvm.ptr)
+ // expected-error at below {{not yet implemented: Unhandled clause exclusive in omp.scan operation}}
+ // expected-error at below {{LLVM Translation failed for operation: omp.scan}}
+ omp.scan exclusive(%prv : !llvm.ptr)
omp.yield
}
}
llvm.return
}
+
// -----
llvm.func @single_allocate(%x : !llvm.ptr) {
More information about the Mlir-commits
mailing list