[llvm] [mlir] scan lowering changes (PR #133149)
Anchu Rajendran S via llvm-commits
llvm-commits at lists.llvm.org
Wed Apr 16 13:01:39 PDT 2025
https://github.com/anchuraj updated https://github.com/llvm/llvm-project/pull/133149
>From 5e859ca97803922de5de73b4e1d0cab240b1cbb3 Mon Sep 17 00:00:00 2001
From: Anchu Rajendran <asudhaku at amd.com>
Date: Wed, 16 Apr 2025 14:58:08 -0500
Subject: [PATCH] Scan Lowering Changes
---
.../llvm/Frontend/OpenMP/OMPIRBuilder.h | 152 ++++++-
llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp | 421 +++++++++++++++++-
.../Frontend/OpenMPIRBuilderTest.cpp | 66 +++
.../OpenMP/OpenMPToLLVMIRTranslation.cpp | 268 ++++++++---
.../Target/LLVMIR/openmp-reduction-scan.mlir | 120 +++++
mlir/test/Target/LLVMIR/openmp-todo.mlir | 31 --
6 files changed, 958 insertions(+), 100 deletions(-)
create mode 100644 mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 6b104708bdb0d..386de1fa51f18 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -503,6 +503,37 @@ class OpenMPIRBuilder {
return allocaInst;
}
};
+
+ /// Type used throughout for insertion points.
+ using InsertPointTy = IRBuilder<>::InsertPoint;
+
+ /// Type used to represent an insertion point or an error value.
+ using InsertPointOrErrorTy = Expected<InsertPointTy>;
+
+ struct ScanInformation {
+ public:
+ /// Dominates the body of the loop before scan directive
+ llvm::BasicBlock *OMPBeforeScanBlock = nullptr;
+ /// Dominates the body of the loop before scan directive
+ llvm::BasicBlock *OMPAfterScanBlock = nullptr;
+ /// Controls the flow to before or after scan blocks
+ llvm::BasicBlock *OMPScanDispatch = nullptr;
+ /// Exit block of loop body
+ llvm::BasicBlock *OMPScanLoopExit = nullptr;
+ /// Block before loop body where scan initializations are done
+ llvm::BasicBlock *OMPScanInit = nullptr;
+ /// Block after loop body where scan finalizations are done
+ llvm::BasicBlock *OMPScanFinish = nullptr;
+ /// If true, it indicates Input phase is lowered; else it indicates
+ /// ScanPhase is lowered
+ bool OMPFirstScanLoop = false;
+ // Maps the private reduction variable to the pointer of the temporary
+ // buffer
+ llvm::SmallDenseMap<llvm::Value *, llvm::Value *> ScanBuffPtrs;
+ llvm::Value *IV;
+ llvm::Value *Span;
+ } ScanInfo;
+
/// Initialize the internal state, this will put structures types and
/// potentially other helpers into the underlying module. Must be called
/// before any other method and only once! This internal state includes types
@@ -519,12 +550,6 @@ class OpenMPIRBuilder {
/// Add attributes known for \p FnID to \p Fn.
void addAttributes(omp::RuntimeFunction FnID, Function &Fn);
- /// Type used throughout for insertion points.
- using InsertPointTy = IRBuilder<>::InsertPoint;
-
- /// Type used to represent an insertion point or an error value.
- using InsertPointOrErrorTy = Expected<InsertPointTy>;
-
/// Get the create a name using the platform specific separators.
/// \param Parts parts of the final name that needs separation
/// The created name has a first separator between the first and second part
@@ -729,6 +754,35 @@ class OpenMPIRBuilder {
LoopBodyGenCallbackTy BodyGenCB, Value *TripCount,
const Twine &Name = "loop");
+ /// Generator for the control flow structure of an OpenMP canonical loops if
+ /// the parent directive has an `inscan` modifier specified.
+ /// If the `inscan` modifier is specified, the region of the parent is
+ /// expected to have a `scan` directive. Based on the clauses in
+ /// scan directive, the body of the loop is split into two loops: Input loop
+ /// and Scan Loop. Input loop contains the code generated for input phase of
+ /// scan and Scan loop contains the code generated for scan phase of scan.
+ ///
+ /// \param Loc The insert and source location description.
+ /// \param BodyGenCB Callback that will generate the loop body code.
+ /// \param Start Value of the loop counter for the first iterations.
+ /// \param Stop Loop counter values past this will stop the loop.
+ /// \param Step Loop counter increment after each iteration; negative
+ /// means counting down.
+ /// \param IsSigned Whether Start, Stop and Step are signed integers.
+ /// \param InclusiveStop Whether \p Stop itself is a valid value for the loop
+ /// counter.
+ /// \param ComputeIP Insertion point for instructions computing the trip
+ /// count. Can be used to ensure the trip count is available
+ /// at the outermost loop of a loop nest. If not set,
+ /// defaults to the preheader of the generated loop.
+ /// \param Name Base name used to derive BB and instruction names.
+ ///
+ /// \returns A vector containing Loop Info of Input Loop and Scan Loop.
+ Expected<SmallVector<llvm::CanonicalLoopInfo *>> createCanonicalScanLoops(
+ const LocationDescription &Loc, LoopBodyGenCallbackTy BodyGenCB,
+ Value *Start, Value *Stop, Value *Step, bool IsSigned, bool InclusiveStop,
+ InsertPointTy ComputeIP, const Twine &Name);
+
/// Calculate the trip count of a canonical loop.
///
/// This allows specifying user-defined loop counter values using increment,
@@ -798,13 +852,16 @@ class OpenMPIRBuilder {
/// at the outermost loop of a loop nest. If not set,
/// defaults to the preheader of the generated loop.
/// \param Name Base name used to derive BB and instruction names.
+ /// \param InScan Whether loop has a scan reduction specified.
///
/// \returns An object representing the created control flow structure which
/// can be used for loop-associated directives.
- Expected<CanonicalLoopInfo *> createCanonicalLoop(
- const LocationDescription &Loc, LoopBodyGenCallbackTy BodyGenCB,
- Value *Start, Value *Stop, Value *Step, bool IsSigned, bool InclusiveStop,
- InsertPointTy ComputeIP = {}, const Twine &Name = "loop");
+ Expected<CanonicalLoopInfo *>
+ createCanonicalLoop(const LocationDescription &Loc,
+ LoopBodyGenCallbackTy BodyGenCB, Value *Start,
+ Value *Stop, Value *Step, bool IsSigned,
+ bool InclusiveStop, InsertPointTy ComputeIP = {},
+ const Twine &Name = "loop", bool InScan = false);
/// Collapse a loop nest into a single loop.
///
@@ -1532,6 +1589,45 @@ class OpenMPIRBuilder {
ArrayRef<OpenMPIRBuilder::ReductionInfo> ReductionInfos,
Function *ReduceFn, AttributeList FuncAttrs);
+ /// Creates the runtime call specified
+ /// \param Callee Function Declaration Value
+ /// \param Args Arguments passed to the call
+ /// \param Name Optional param to specify the name of the call Instruction.
+ ///
+ /// \return The Runtime call instruction created.
+ llvm::CallInst *emitNoUnwindRuntimeCall(llvm::FunctionCallee Callee,
+ ArrayRef<llvm::Value *> Args,
+ const llvm::Twine &Name);
+
+ /// Helper function for CreateCanonicalScanLoops to create InputLoop
+ /// in the firstGen and Scan Loop in the SecondGen
+ /// \param InputLoopGen Callback for generating the loop for input phase
+ /// \param ScanLoopGen Callback for generating the loop for scan phase
+ ///
+ /// \return error if any produced, else return success.
+ Error emitScanBasedDirectiveIR(
+ llvm::function_ref<Error()> InputLoopGen,
+ llvm::function_ref<Error(LocationDescription Loc)> ScanLoopGen);
+
+ /// Creates the basic blocks required for scan reduction.
+ void createScanBBs();
+
+ /// Dynamically allocates the buffer needed for scan reduction.
+ /// \param AllocaIP The IP where possibly-shared pointer of buffer needs to be
+ /// declared. \param ScanVars Scan Variables.
+ ///
+ /// \return error if any produced, else return success.
+ Error emitScanBasedDirectiveDeclsIR(InsertPointTy AllocaIP,
+ ArrayRef<llvm::Value *> ScanVars,
+ ArrayRef<llvm::Type *> ScanVarsType);
+
+ /// Copies the result back to the reduction variable.
+ /// \param ReductionInfos Array type containing the ReductionOps.
+ ///
+ /// \return error if any produced, else return success.
+ Error emitScanBasedDirectiveFinalsIR(
+ SmallVector<llvm::OpenMPIRBuilder::ReductionInfo> ReductionInfos);
+
/// This function emits a helper that gathers Reduce lists from the first
/// lane of every active warp to lanes in the first warp.
///
@@ -2179,7 +2275,6 @@ class OpenMPIRBuilder {
// block, if possible, or else at the end of the function. Also add a branch
// from current block to BB if current block does not have a terminator.
void emitBlock(BasicBlock *BB, Function *CurFn, bool IsFinished = false);
-
/// Emits code for OpenMP 'if' clause using specified \a BodyGenCallbackTy
/// Here is the logic:
/// if (Cond) {
@@ -2607,6 +2702,41 @@ class OpenMPIRBuilder {
BodyGenCallbackTy BodyGenCB,
FinalizeCallbackTy FiniCB, Value *Filter);
+ /// This function performs the scan reduction of the values updated in
+ /// the input phase. The reduction logic needs to be emitted between input
+ /// and scan loop returned by `CreateCanonicalScanLoops`. The following
+ /// is the code that is generated, `buffer` and `span` are expected to be
+ /// populated before executing the generated code.
+ ///
+ /// for (int k = 0; k != ceil(log2(span)); ++k) {
+ /// i=pow(2,k)
+ /// for (size cnt = last_iter; cnt >= i; --cnt)
+ /// buffer[cnt] op= buffer[cnt-i];
+ /// }
+ /// \param Loc The insert and source location description.
+ /// \param ReductionInfos Array type containing the ReductionOps.
+ ///
+ /// \returns The insertion position *after* the masked.
+ InsertPointOrErrorTy emitScanReduction(
+ const LocationDescription &Loc,
+ SmallVector<llvm::OpenMPIRBuilder::ReductionInfo> ReductionInfos);
+
+ /// This directive split and directs the control flow to input phase
+ /// blocks or scan phase blocks based on 1. whether input loop or scan loop
+ /// is executed, 2. whether exclusive or inclusive scan is used.
+ ///
+ /// \param Loc The insert and source location description.
+ /// \param AllocaIP The IP where the temporary buffer for scan reduction
+ // needs to be allocated.
+ /// \param ScanVars Scan Variables.
+ /// \param IsInclusive Whether it is an inclusive or exclusive scan.
+ ///
+ /// \returns The insertion position *after* the scan.
+ InsertPointOrErrorTy createScan(const LocationDescription &Loc,
+ InsertPointTy AllocaIP,
+ ArrayRef<llvm::Value *> ScanVars,
+ ArrayRef<llvm::Type *> ScanVarsType,
+ bool IsInclusive);
/// Generator for '#omp critical'
///
/// \param Loc The insert and source location description.
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 28662efc02882..23a548a1a60d0 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -59,6 +59,8 @@
#include "llvm/Transforms/Utils/LoopPeel.h"
#include "llvm/Transforms/Utils/UnrollLoop.h"
+#include <cassert>
+#include <cstddef>
#include <cstdint>
#include <optional>
@@ -3981,6 +3983,333 @@ OpenMPIRBuilder::createMasked(const LocationDescription &Loc,
/*Conditional*/ true, /*hasFinalize*/ true);
}
+llvm::CallInst *
+OpenMPIRBuilder::emitNoUnwindRuntimeCall(llvm::FunctionCallee Callee,
+ ArrayRef<llvm::Value *> Args,
+ const llvm::Twine &Name) {
+ llvm::CallInst *Call = Builder.CreateCall(
+ Callee, Args, SmallVector<llvm::OperandBundleDef, 1>(), Name);
+ Call->setDoesNotThrow();
+ return Call;
+}
+
+// Expects input basic block is dominated by BeforeScanBB.
+// Once Scan directive is encountered, the code after scan directive should be
+// dominated by AfterScanBB. Scan directive splits the code sequence to
+// scan and input phase. Based on whether inclusive or exclusive
+// clause is used in the scan directive and whether input loop or scan loop
+// is lowered, it adds jumps to input and scan phase. First Scan loop is the
+// input loop and second is the scan loop. The code generated handles only
+// inclusive scans now.
+OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::createScan(
+ const LocationDescription &Loc, InsertPointTy AllocaIP,
+ ArrayRef<llvm::Value *> ScanVars, ArrayRef<llvm::Type *> ScanVarsType,
+ bool IsInclusive) {
+ if (ScanInfo.OMPFirstScanLoop) {
+ llvm::Error Err =
+ emitScanBasedDirectiveDeclsIR(AllocaIP, ScanVars, ScanVarsType);
+ if (Err) {
+ return Err;
+ }
+ }
+ if (!updateToLocation(Loc))
+ return Loc.IP;
+
+ llvm::Value *IV = ScanInfo.IV;
+
+ if (ScanInfo.OMPFirstScanLoop) {
+ // Emit buffer[i] = red; at the end of the input phase.
+ for (size_t i = 0; i < ScanVars.size(); i++) {
+ Value *BuffPtr = ScanInfo.ScanBuffPtrs[ScanVars[i]];
+ Value *Buff = Builder.CreateLoad(Builder.getPtrTy(), BuffPtr);
+ Type *DestTy = ScanVarsType[i];
+ Value *Val = Builder.CreateInBoundsGEP(DestTy, Buff, IV, "arrayOffset");
+ Value *Src = Builder.CreateLoad(DestTy, ScanVars[i]);
+
+ Builder.CreateStore(Src, Val);
+ }
+ }
+ Builder.CreateBr(ScanInfo.OMPScanLoopExit);
+ emitBlock(ScanInfo.OMPScanDispatch, Builder.GetInsertBlock()->getParent());
+
+ if (!ScanInfo.OMPFirstScanLoop) {
+ IV = ScanInfo.IV;
+ // Emit red = buffer[i]; at the entrance to the scan phase.
+ // TODO: if exclusive scan, the red = buffer[i-1] needs to be updated.
+ for (size_t i = 0; i < ScanVars.size(); i++) {
+ Value *BuffPtr = ScanInfo.ScanBuffPtrs[ScanVars[i]];
+ Value *Buff = Builder.CreateLoad(Builder.getPtrTy(), BuffPtr);
+ Type *DestTy = ScanVarsType[i];
+ Value *SrcPtr =
+ Builder.CreateInBoundsGEP(DestTy, Buff, IV, "arrayOffset");
+ Value *Src = Builder.CreateLoad(DestTy, SrcPtr);
+ Builder.CreateStore(Src, ScanVars[i]);
+ }
+ }
+
+ // TODO: Update it to CreateBr and remove dead blocks
+ llvm::Value *CmpI = Builder.getInt1(true);
+ if (ScanInfo.OMPFirstScanLoop == IsInclusive) {
+ Builder.CreateCondBr(CmpI, ScanInfo.OMPBeforeScanBlock,
+ ScanInfo.OMPAfterScanBlock);
+ } else {
+ Builder.CreateCondBr(CmpI, ScanInfo.OMPAfterScanBlock,
+ ScanInfo.OMPBeforeScanBlock);
+ }
+ emitBlock(ScanInfo.OMPAfterScanBlock, Builder.GetInsertBlock()->getParent());
+ Builder.SetInsertPoint(ScanInfo.OMPAfterScanBlock);
+ return Builder.saveIP();
+}
+
+Error OpenMPIRBuilder::emitScanBasedDirectiveDeclsIR(
+ InsertPointTy AllocaIP, ArrayRef<Value *> ScanVars,
+ ArrayRef<Type *> ScanVarsType) {
+
+ Builder.restoreIP(AllocaIP);
+ // Create the shared pointer at alloca IP.
+ for (size_t i = 0; i < ScanVars.size(); i++) {
+ llvm::Value *BuffPtr =
+ Builder.CreateAlloca(Builder.getPtrTy(), nullptr, "vla");
+ ScanInfo.ScanBuffPtrs[ScanVars[i]] = BuffPtr;
+ }
+
+ // Allocate temporary buffer by master thread
+ auto BodyGenCB = [&](InsertPointTy AllocaIP,
+ InsertPointTy CodeGenIP) -> Error {
+ Builder.restoreIP(CodeGenIP);
+ Value *AllocSpan = Builder.CreateAdd(ScanInfo.Span, Builder.getInt32(1));
+ for (size_t i = 0; i < ScanVars.size(); i++) {
+ Type *IntPtrTy = Builder.getInt32Ty();
+ Constant *Allocsize = ConstantExpr::getSizeOf(ScanVarsType[i]);
+ Allocsize = ConstantExpr::getTruncOrBitCast(Allocsize, IntPtrTy);
+ Value *Buff = Builder.CreateMalloc(IntPtrTy, ScanVarsType[i], Allocsize,
+ AllocSpan, nullptr, "arr");
+ Builder.CreateStore(Buff, ScanInfo.ScanBuffPtrs[ScanVars[i]]);
+ }
+ return Error::success();
+ };
+ // TODO: Perform finalization actions for variables. This has to be
+ // called for variables which have destructors/finalizers.
+ auto FiniCB = [&](InsertPointTy CodeGenIP) { return llvm::Error::success(); };
+
+ Builder.SetInsertPoint(ScanInfo.OMPScanInit->getTerminator());
+ llvm::Value *FilterVal = Builder.getInt32(0);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy AfterIP =
+ createMasked(Builder.saveIP(), BodyGenCB, FiniCB, FilterVal);
+
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+ BasicBlock *InputBB = Builder.GetInsertBlock();
+ if (InputBB->getTerminator())
+ Builder.SetInsertPoint(Builder.GetInsertBlock()->getTerminator());
+ AfterIP = createBarrier(Builder.saveIP(), llvm::omp::OMPD_barrier);
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+
+ return Error::success();
+}
+
+Error OpenMPIRBuilder::emitScanBasedDirectiveFinalsIR(
+ SmallVector<ReductionInfo> ReductionInfos) {
+ auto BodyGenCB = [&](InsertPointTy AllocaIP,
+ InsertPointTy CodeGenIP) -> Error {
+ Builder.restoreIP(CodeGenIP);
+ for (ReductionInfo RedInfo : ReductionInfos) {
+ Value *PrivateVar = RedInfo.PrivateVariable;
+ Value *OrigVar = RedInfo.Variable;
+ Value *BuffPtr = ScanInfo.ScanBuffPtrs[PrivateVar];
+ Value *Buff = Builder.CreateLoad(Builder.getPtrTy(), BuffPtr);
+
+ Type *SrcTy = RedInfo.ElementType;
+ Value *Val =
+ Builder.CreateInBoundsGEP(SrcTy, Buff, ScanInfo.Span, "arrayOffset");
+ Value *Src = Builder.CreateLoad(SrcTy, Val);
+
+ Builder.CreateStore(Src, OrigVar);
+ Builder.CreateFree(Buff);
+ }
+ return Error::success();
+ };
+ // TODO: Perform finalization actions for variables. This has to be
+ // called for variables which have destructors/finalizers.
+ auto FiniCB = [&](InsertPointTy CodeGenIP) { return llvm::Error::success(); };
+
+ Builder.SetInsertPoint(ScanInfo.OMPScanFinish->getTerminator());
+ llvm::Value *FilterVal = Builder.getInt32(0);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy AfterIP =
+ createMasked(Builder.saveIP(), BodyGenCB, FiniCB, FilterVal);
+
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+ BasicBlock *InputBB = Builder.GetInsertBlock();
+ if (InputBB->getTerminator())
+ Builder.SetInsertPoint(Builder.GetInsertBlock()->getTerminator());
+ AfterIP = createBarrier(Builder.saveIP(), llvm::omp::OMPD_barrier);
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+ return Error::success();
+}
+
+OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::emitScanReduction(
+ const LocationDescription &Loc,
+ SmallVector<llvm::OpenMPIRBuilder::ReductionInfo> ReductionInfos) {
+
+ if (!updateToLocation(Loc))
+ return Loc.IP;
+ auto BodyGenCB = [&](InsertPointTy AllocaIP,
+ InsertPointTy CodeGenIP) -> Error {
+ Builder.restoreIP(CodeGenIP);
+ Function *CurFn = Builder.GetInsertBlock()->getParent();
+ // for (int k = 0; k <= ceil(log2(n)); ++k)
+ llvm::BasicBlock *LoopBB =
+ BasicBlock::Create(CurFn->getContext(), "omp.outer.log.scan.body");
+ llvm::BasicBlock *ExitBB =
+ splitBB(Builder, false, "omp.outer.log.scan.exit");
+ llvm::Function *F = llvm::Intrinsic::getOrInsertDeclaration(
+ Builder.GetInsertBlock()->getModule(),
+ (llvm::Intrinsic::ID)llvm::Intrinsic::log2, Builder.getDoubleTy());
+ llvm::BasicBlock *InputBB = Builder.GetInsertBlock();
+ llvm::Value *Arg =
+ Builder.CreateUIToFP(ScanInfo.Span, Builder.getDoubleTy());
+ llvm::Value *LogVal = emitNoUnwindRuntimeCall(F, Arg, "");
+ F = llvm::Intrinsic::getOrInsertDeclaration(
+ Builder.GetInsertBlock()->getModule(),
+ (llvm::Intrinsic::ID)llvm::Intrinsic::ceil, Builder.getDoubleTy());
+ LogVal = emitNoUnwindRuntimeCall(F, LogVal, "");
+ LogVal = Builder.CreateFPToUI(LogVal, Builder.getInt32Ty());
+ llvm::Value *NMin1 = Builder.CreateNUWSub(
+ ScanInfo.Span, llvm::ConstantInt::get(ScanInfo.Span->getType(), 1));
+ Builder.SetInsertPoint(InputBB);
+ Builder.CreateBr(LoopBB);
+ emitBlock(LoopBB, Builder.GetInsertBlock()->getParent());
+ Builder.SetInsertPoint(LoopBB);
+
+ PHINode *Counter = Builder.CreatePHI(Builder.getInt32Ty(), 2);
+ //// size pow2k = 1;
+ PHINode *Pow2K = Builder.CreatePHI(Builder.getInt32Ty(), 2);
+ Counter->addIncoming(llvm::ConstantInt::get(Builder.getInt32Ty(), 0),
+ InputBB);
+ Pow2K->addIncoming(llvm::ConstantInt::get(Builder.getInt32Ty(), 1),
+ InputBB);
+ //// for (size i = n - 1; i >= 2 ^ k; --i)
+ //// tmp[i] op= tmp[i-pow2k];
+ llvm::BasicBlock *InnerLoopBB =
+ BasicBlock::Create(CurFn->getContext(), "omp.inner.log.scan.body");
+ llvm::BasicBlock *InnerExitBB =
+ BasicBlock::Create(CurFn->getContext(), "omp.inner.log.scan.exit");
+ llvm::Value *CmpI = Builder.CreateICmpUGE(NMin1, Pow2K);
+ Builder.CreateCondBr(CmpI, InnerLoopBB, InnerExitBB);
+ emitBlock(InnerLoopBB, Builder.GetInsertBlock()->getParent());
+ Builder.SetInsertPoint(InnerLoopBB);
+ auto *IVal = Builder.CreatePHI(Builder.getInt32Ty(), 2);
+ IVal->addIncoming(NMin1, LoopBB);
+ for (ReductionInfo RedInfo : ReductionInfos) {
+ Value *ReductionVal = RedInfo.PrivateVariable;
+ Value *BuffPtr = ScanInfo.ScanBuffPtrs[ReductionVal];
+ Value *Buff = Builder.CreateLoad(Builder.getPtrTy(), BuffPtr);
+ Type *DestTy = RedInfo.ElementType;
+ Value *IV = Builder.CreateAdd(IVal, Builder.getInt32(1));
+ Value *LHSPtr =
+ Builder.CreateInBoundsGEP(DestTy, Buff, IV, "arrayOffset");
+ Value *OffsetIval = Builder.CreateNUWSub(IV, Pow2K);
+ Value *RHSPtr =
+ Builder.CreateInBoundsGEP(DestTy, Buff, OffsetIval, "arrayOffset");
+ Value *LHS = Builder.CreateLoad(DestTy, LHSPtr);
+ Value *RHS = Builder.CreateLoad(DestTy, RHSPtr);
+ llvm::Value *Result;
+ InsertPointOrErrorTy AfterIP =
+ RedInfo.ReductionGen(Builder.saveIP(), LHS, RHS, Result);
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.CreateStore(Result, LHSPtr);
+ }
+ llvm::Value *NextIVal = Builder.CreateNUWSub(
+ IVal, llvm::ConstantInt::get(Builder.getInt32Ty(), 1));
+ IVal->addIncoming(NextIVal, Builder.GetInsertBlock());
+ CmpI = Builder.CreateICmpUGE(NextIVal, Pow2K);
+ Builder.CreateCondBr(CmpI, InnerLoopBB, InnerExitBB);
+ emitBlock(InnerExitBB, Builder.GetInsertBlock()->getParent());
+ llvm::Value *Next = Builder.CreateNUWAdd(
+ Counter, llvm::ConstantInt::get(Counter->getType(), 1));
+ Counter->addIncoming(Next, Builder.GetInsertBlock());
+ // pow2k <<= 1;
+ llvm::Value *NextPow2K = Builder.CreateShl(Pow2K, 1, "", /*HasNUW=*/true);
+ Pow2K->addIncoming(NextPow2K, Builder.GetInsertBlock());
+ llvm::Value *Cmp = Builder.CreateICmpNE(Next, LogVal);
+ Builder.CreateCondBr(Cmp, LoopBB, ExitBB);
+ Builder.SetInsertPoint(ExitBB->getFirstInsertionPt());
+ return Error::success();
+ };
+
+ // TODO: Perform finalization actions for variables. This has to be
+ // called for variables which have destructors/finalizers.
+ auto FiniCB = [&](InsertPointTy CodeGenIP) { return llvm::Error::success(); };
+
+ llvm::Value *FilterVal = Builder.getInt32(0);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy AfterIP =
+ createMasked(Builder.saveIP(), BodyGenCB, FiniCB, FilterVal);
+
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+ AfterIP = createBarrier(Builder.saveIP(), llvm::omp::OMPD_barrier);
+
+ if (!AfterIP)
+ return AfterIP.takeError();
+ Builder.restoreIP(*AfterIP);
+ Error Err = emitScanBasedDirectiveFinalsIR(ReductionInfos);
+ if (Err) {
+ return Err;
+ }
+
+ return AfterIP;
+}
+
+Error OpenMPIRBuilder::emitScanBasedDirectiveIR(
+ llvm::function_ref<Error()> InputLoopGen,
+ llvm::function_ref<Error(LocationDescription Loc)> ScanLoopGen) {
+
+ {
+ // Emit loop with input phase:
+ // for (i: 0..<num_iters>) {
+ // <input phase>;
+ // buffer[i] = red;
+ // }
+ ScanInfo.OMPFirstScanLoop = true;
+ auto Result = InputLoopGen();
+ if (Result)
+ return Result;
+ }
+ {
+ // Emit loop with scan phase:
+ // for (i: 0..<num_iters>) {
+ // red = buffer[i];
+ // <scan phase>;
+ // }
+ ScanInfo.OMPFirstScanLoop = false;
+ auto Result = ScanLoopGen(Builder.saveIP());
+ if (Result)
+ return Result;
+ }
+ return Error::success();
+}
+
+void OpenMPIRBuilder::createScanBBs() {
+ Function *Fun = Builder.GetInsertBlock()->getParent();
+ ScanInfo.OMPScanDispatch =
+ BasicBlock::Create(Fun->getContext(), "omp.inscan.dispatch");
+ ScanInfo.OMPAfterScanBlock =
+ BasicBlock::Create(Fun->getContext(), "omp.after.scan.bb");
+ ScanInfo.OMPBeforeScanBlock =
+ BasicBlock::Create(Fun->getContext(), "omp.before.scan.bb");
+ ScanInfo.OMPScanLoopExit =
+ BasicBlock::Create(Fun->getContext(), "omp.scan.loop.exit");
+}
+
CanonicalLoopInfo *OpenMPIRBuilder::createLoopSkeleton(
DebugLoc DL, Value *TripCount, Function *F, BasicBlock *PreInsertBefore,
BasicBlock *PostInsertBefore, const Twine &Name) {
@@ -4078,10 +4407,95 @@ OpenMPIRBuilder::createCanonicalLoop(const LocationDescription &Loc,
return CL;
}
+Expected<SmallVector<llvm::CanonicalLoopInfo *>>
+OpenMPIRBuilder::createCanonicalScanLoops(
+ const LocationDescription &Loc, LoopBodyGenCallbackTy BodyGenCB,
+ Value *Start, Value *Stop, Value *Step, bool IsSigned, bool InclusiveStop,
+ InsertPointTy ComputeIP, const Twine &Name) {
+ LocationDescription ComputeLoc =
+ ComputeIP.isSet() ? LocationDescription(ComputeIP, Loc.DL) : Loc;
+ updateToLocation(ComputeLoc);
+
+ Value *TripCount = calculateCanonicalLoopTripCount(
+ ComputeLoc, Start, Stop, Step, IsSigned, InclusiveStop, Name);
+ ScanInfo.Span = TripCount;
+ ScanInfo.OMPScanInit = splitBB(Builder, true, "scan.init");
+ Builder.SetInsertPoint(ScanInfo.OMPScanInit);
+
+ auto BodyGen = [=](InsertPointTy CodeGenIP, Value *IV) {
+ /// The control of the loopbody of following structure:
+ ///
+ /// InputBlock
+ /// |
+ /// ContinueBlock
+ ///
+ /// is transformed to:
+ ///
+ /// InputBlock
+ /// |
+ /// OMPScanDispatch
+ ///
+ /// OMPBeforeScanBlock
+ /// |
+ /// OMPScanLoopExit
+ /// |
+ /// ContinueBlock
+ ///
+ /// OMPBeforeScanBlock dominates the control flow of code generated until
+ /// scan directive is encountered and OMPAfterScanBlock dominates the
+ /// control flow of code generated after scan is encountered. The successor
+ /// of OMPScanDispatch can be OMPBeforeScanBlock or OMPAfterScanBlock based
+ /// on 1.whether it is in Input phase or Scan Phase , 2. whether it is an
+ /// exclusive or inclusive scan.
+ Builder.restoreIP(CodeGenIP);
+ ScanInfo.IV = IV;
+ createScanBBs();
+ BasicBlock *InputBlock = Builder.GetInsertBlock();
+ Instruction *Terminator = InputBlock->getTerminator();
+ assert(Terminator->getNumSuccessors() == 1);
+ BasicBlock *ContinueBlock = Terminator->getSuccessor(0);
+ Terminator->setSuccessor(0, ScanInfo.OMPScanDispatch);
+ emitBlock(ScanInfo.OMPBeforeScanBlock,
+ Builder.GetInsertBlock()->getParent());
+ Builder.CreateBr(ScanInfo.OMPScanLoopExit);
+ emitBlock(ScanInfo.OMPScanLoopExit, Builder.GetInsertBlock()->getParent());
+ Builder.CreateBr(ContinueBlock);
+ Builder.SetInsertPoint(ScanInfo.OMPBeforeScanBlock->getFirstInsertionPt());
+ return BodyGenCB(Builder.saveIP(), IV);
+ };
+
+ SmallVector<llvm::CanonicalLoopInfo *> Result;
+ const auto &&InputLoopGen = [&]() -> Error {
+ auto LoopInfo =
+ createCanonicalLoop(Builder.saveIP(), BodyGen, Start, Stop, Step,
+ IsSigned, InclusiveStop, ComputeIP, Name, true);
+ if (!LoopInfo)
+ return LoopInfo.takeError();
+ Result.push_back(*LoopInfo);
+ Builder.restoreIP((*LoopInfo)->getAfterIP());
+ return Error::success();
+ };
+ const auto &&ScanLoopGen = [&](LocationDescription Loc) -> Error {
+ auto LoopInfo =
+ createCanonicalLoop(Loc, BodyGen, Start, Stop, Step, IsSigned,
+ InclusiveStop, ComputeIP, Name, true);
+ if (!LoopInfo)
+ return LoopInfo.takeError();
+ Result.push_back(*LoopInfo);
+ Builder.restoreIP((*LoopInfo)->getAfterIP());
+ ScanInfo.OMPScanFinish = Builder.GetInsertBlock();
+ return Error::success();
+ };
+ Error Err = emitScanBasedDirectiveIR(InputLoopGen, ScanLoopGen);
+ if (Err) {
+ return Err;
+ }
+ return Result;
+}
+
Value *OpenMPIRBuilder::calculateCanonicalLoopTripCount(
const LocationDescription &Loc, Value *Start, Value *Stop, Value *Step,
bool IsSigned, bool InclusiveStop, const Twine &Name) {
-
// Consider the following difficulties (assuming 8-bit signed integers):
// * Adding \p Step to the loop counter which passes \p Stop may overflow:
// DO I = 1, 100, 50
@@ -4141,7 +4555,7 @@ Value *OpenMPIRBuilder::calculateCanonicalLoopTripCount(
Expected<CanonicalLoopInfo *> OpenMPIRBuilder::createCanonicalLoop(
const LocationDescription &Loc, LoopBodyGenCallbackTy BodyGenCB,
Value *Start, Value *Stop, Value *Step, bool IsSigned, bool InclusiveStop,
- InsertPointTy ComputeIP, const Twine &Name) {
+ InsertPointTy ComputeIP, const Twine &Name, bool InScan) {
LocationDescription ComputeLoc =
ComputeIP.isSet() ? LocationDescription(ComputeIP, Loc.DL) : Loc;
@@ -4152,6 +4566,9 @@ Expected<CanonicalLoopInfo *> OpenMPIRBuilder::createCanonicalLoop(
Builder.restoreIP(CodeGenIP);
Value *Span = Builder.CreateMul(IV, Step);
Value *IndVar = Builder.CreateAdd(Span, Start);
+ if (InScan) {
+ ScanInfo.IV = IndVar;
+ }
return BodyGenCB(Builder.saveIP(), IndVar);
};
LocationDescription LoopLoc = ComputeIP.isSet() ? Loc.IP : Builder.saveIP();
diff --git a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
index 2d3d318be7ff1..4cc312c1b0f1c 100644
--- a/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
+++ b/llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp
@@ -1440,6 +1440,16 @@ TEST_F(OpenMPIRBuilderTest, CanonicalLoopSimple) {
EXPECT_EQ(&Loop->getAfter()->front(), RetInst);
}
+void createScan(llvm::Value *scanVar, llvm::Type *scanType,
+ OpenMPIRBuilder &OMPBuilder, IRBuilder<> &Builder,
+ OpenMPIRBuilder::LocationDescription Loc,
+ OpenMPIRBuilder::InsertPointTy &allocaIP) {
+ using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
+ ASSERT_EXPECTED_INIT(
+ InsertPointTy, retIp,
+ OMPBuilder.createScan(Loc, allocaIP, {scanVar}, {scanType}, true));
+ Builder.restoreIP(retIp);
+}
TEST_F(OpenMPIRBuilderTest, CanonicalLoopTripCount) {
OpenMPIRBuilder OMPBuilder(*M);
@@ -5336,6 +5346,62 @@ TEST_F(OpenMPIRBuilderTest, CreateReductions) {
EXPECT_TRUE(findGEPZeroOne(ReductionFn->getArg(1), FirstRHS, SecondRHS));
}
+TEST_F(OpenMPIRBuilderTest, ScanReduction) {
+ using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
+ OpenMPIRBuilder OMPBuilder(*M);
+ OMPBuilder.initialize();
+ IRBuilder<> Builder(BB);
+ OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
+ Value *TripCount = F->getArg(0);
+ Type *LCTy = TripCount->getType();
+ Value *StartVal = ConstantInt::get(LCTy, 1);
+ Value *StopVal = ConstantInt::get(LCTy, 100);
+ Value *Step = ConstantInt::get(LCTy, 1);
+ auto allocaIP = Builder.saveIP();
+
+ llvm::Value *scanVar = Builder.CreateAlloca(Builder.getFloatTy());
+ llvm::Value *origVar = Builder.CreateAlloca(Builder.getFloatTy());
+ unsigned NumBodiesGenerated = 0;
+ auto LoopBodyGenCB = [&](InsertPointTy CodeGenIP, llvm::Value *LC) {
+ NumBodiesGenerated += 1;
+ Builder.restoreIP(CodeGenIP);
+ createScan(scanVar, Builder.getFloatTy(), OMPBuilder, Builder, Loc,
+ allocaIP);
+ return Error::success();
+ };
+ SmallVector<CanonicalLoopInfo *> Loops;
+ ASSERT_EXPECTED_INIT(SmallVector<CanonicalLoopInfo *>, loopsVec,
+ OMPBuilder.createCanonicalScanLoops(
+ Loc, LoopBodyGenCB, StartVal, StopVal, Step, false,
+ false, Builder.saveIP(), "scan"));
+ Loops = loopsVec;
+ EXPECT_EQ(Loops.size(), 2U);
+ CanonicalLoopInfo *InputLoop = Loops.front();
+ CanonicalLoopInfo *ScanLoop = Loops.back();
+ Builder.restoreIP(ScanLoop->getAfterIP());
+ InputLoop->assertOK();
+ ScanLoop->assertOK();
+
+ EXPECT_EQ(InputLoop->getPreheader()->getSinglePredecessor(),
+ &F->getEntryBlock());
+ EXPECT_EQ(ScanLoop->getAfter(), Builder.GetInsertBlock());
+ EXPECT_EQ(NumBodiesGenerated, 2U);
+ SmallVector<OpenMPIRBuilder::ReductionInfo> reductionInfos = {
+ {Builder.getFloatTy(), origVar, scanVar,
+ /*EvaluationKind=*/OpenMPIRBuilder::EvalKind::Scalar, sumReduction,
+ /*ReductionGenClang=*/nullptr, sumAtomicReduction}};
+ OpenMPIRBuilder::LocationDescription RedLoc({InputLoop->getAfterIP(), DL});
+ llvm::BasicBlock *Cont = splitBB(Builder, false, "omp.scan.loop.cont");
+ ASSERT_EXPECTED_INIT(InsertPointTy, retIp,
+ OMPBuilder.emitScanReduction(RedLoc, reductionInfos));
+ Builder.restoreIP(retIp);
+ Builder.CreateBr(Cont);
+ SmallVector<CallInst *> MaskedCalls;
+ findCalls(F, omp::RuntimeFunction::OMPRTL___kmpc_masked, OMPBuilder,
+ MaskedCalls);
+ ASSERT_EQ(MaskedCalls.size(), 1u);
+}
+
TEST_F(OpenMPIRBuilderTest, CreateTwoReductions) {
using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
OpenMPIRBuilder OMPBuilder(*M);
diff --git a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
index 8d1cc9b10a950..c84789e634101 100644
--- a/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp
@@ -47,6 +47,10 @@
using namespace mlir;
+llvm::SmallDenseMap<llvm::Value *, llvm::Type *> ReductionVarToType;
+llvm::OpenMPIRBuilder::InsertPointTy
+ parallelAllocaIP; // TODO: change this alloca IP to point to originalvar
+ // allocaIP. ReductionDecl need to be linked to scan var.
namespace {
static llvm::omp::ScheduleKind
convertToScheduleKind(std::optional<omp::ClauseScheduleKind> schedKind) {
@@ -86,7 +90,9 @@ class OpenMPLoopInfoStackFrame
: public LLVM::ModuleTranslation::StackFrameBase<OpenMPLoopInfoStackFrame> {
public:
MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(OpenMPLoopInfoStackFrame)
- llvm::CanonicalLoopInfo *loopInfo = nullptr;
+ // For constructs like scan, one Loop info frame can contain multiple
+ // Canonical Loops
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos;
};
/// Custom error class to signal translation errors that don't need reporting,
@@ -169,6 +175,10 @@ 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");
@@ -232,8 +242,8 @@ static LogicalResult checkImplementationStatus(Operation &op) {
op.getReductionSyms())
result = todo("reduction");
if (op.getReductionMod() &&
- op.getReductionMod().value() != omp::ReductionModifier::defaultmod)
- result = todo("reduction with modifier");
+ op.getReductionMod().value() == omp::ReductionModifier::task)
+ result = todo("reduction with task modifier");
};
auto checkTaskReduction = [&todo](auto op, LogicalResult &result) {
if (!op.getTaskReductionVars().empty() || op.getTaskReductionByref() ||
@@ -253,6 +263,7 @@ 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);
@@ -382,15 +393,15 @@ 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;
}
/// Converts the given region that appears within an OpenMP dialect operation to
@@ -1133,6 +1144,11 @@ initReductionVars(OP op, ArrayRef<BlockArgument> reductionArgs,
// 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) {
+
+ llvm::Type *reductionType =
+ moduleTranslation.convertType(reductionDecls[i].getType());
+ ReductionVarToType[privateReductionVariables[i]] = reductionType;
+
SmallVector<llvm::Value *, 1> phis;
// map block argument to initializer region
@@ -1206,9 +1222,11 @@ static void collectReductionInfo(
atomicGen = owningAtomicReductionGens[i];
llvm::Value *variable =
moduleTranslation.lookupValue(loop.getReductionVars()[i]);
+ llvm::Type *reductionType =
+ moduleTranslation.convertType(reductionDecls[i].getType());
+ 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});
@@ -2342,27 +2360,60 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
if (failed(handleError(regionBlock, opInst)))
return failure();
- builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
- 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 (failed(handleError(wsloopIP, opInst)))
- return failure();
-
- // Process the reductions if required.
- if (failed(createReductionsAndCleanup(
- wsloopOp, builder, moduleTranslation, allocaIP, reductionDecls,
- privateReductionVariables, isByRef, wsloopOp.getNowait(),
- /*isTeamsReduction=*/false)))
- return failure();
+ SmallVector<llvm::CanonicalLoopInfo *> loopInfos =
+ findCurrentLoopInfos(moduleTranslation);
+ auto inputLoopFinishIp = loopInfos.front()->getAfterIP();
+ bool isInScanRegion =
+ wsloopOp.getReductionMod() && (wsloopOp.getReductionMod().value() ==
+ mlir::omp::ReductionModifier::inscan);
+ if (isInScanRegion) {
+ 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::OpenMPIRBuilder::InsertPointOrErrorTy redIP =
+ ompBuilder->emitScanReduction(builder.saveIP(), reductionInfos);
+ if (failed(handleError(redIP, opInst)))
+ return failure();
+ builder.restoreIP(*redIP);
+ builder.CreateBr(cont);
+ }
+ 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 (failed(handleError(wsloopIP, opInst)))
+ return failure();
+ }
+ builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
+ 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,
privateVarsInfo.privatizers);
@@ -2528,6 +2579,7 @@ convertOmpParallel(omp::ParallelOp opInst, llvm::IRBuilderBase &builder,
llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
findAllocaInsertPoint(builder, moduleTranslation);
+ parallelAllocaIP = allocaIP;
llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP =
@@ -2553,6 +2605,64 @@ convertOrderKind(std::optional<omp::ClauseOrderKind> o) {
llvm_unreachable("Unknown ClauseOrderKind kind");
}
+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();
+ for (auto val : mlirScanVars) {
+ llvm::Value *llvmVal = moduleTranslation.lookupValue(val);
+ llvmScanVars.push_back(llvmVal);
+ llvmScanVarsType.push_back(ReductionVarToType[llvmVal]);
+ val.getDefiningOp();
+ }
+ auto parallelOp = scanOp->getParentOfType<omp::ParallelOp>();
+ if (!parallelOp) {
+ return failure();
+ }
+ llvm::OpenMPIRBuilder::InsertPointTy allocaIP = parallelAllocaIP;
+ llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
+ llvm::OpenMPIRBuilder::InsertPointOrErrorTy afterIP =
+ moduleTranslation.getOpenMPBuilder()->createScan(
+ ompLoc, allocaIP, llvmScanVars, llvmScanVarsType, isInclusive);
+ if (failed(handleError(afterIP, opInst)))
+ return failure();
+
+ builder.restoreIP(*afterIP);
+
+ // TODO: The argument of LoopnestOp is stored into the index variable and this
+ // variable is used across scan operation. However that makes the mlir
+ // invalid.(`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.`). The argument of LoopNestOp need to be loaded again
+ // after ScanOp again so mlir generated is valid.
+ auto parentOp = scanOp->getParentOp();
+ auto loopOp = cast<omp::LoopNestOp>(parentOp);
+ if (loopOp) {
+ auto &firstBlock = *(scanOp->getParentRegion()->getBlocks()).begin();
+ auto &ins = *(firstBlock.begin());
+ if (isa<LLVM::StoreOp>(ins)) {
+ LLVM::StoreOp storeOp = dyn_cast<LLVM::StoreOp>(ins);
+ auto src = moduleTranslation.lookupValue(storeOp->getOperand(0));
+ if (src == moduleTranslation.lookupValue(
+ (loopOp.getRegion().getArguments())[0])) {
+ auto dest = moduleTranslation.lookupValue(storeOp->getOperand(1));
+ builder.CreateStore(src, dest);
+ }
+ }
+ }
+ return success();
+}
+
/// Converts an OpenMP simd loop into LLVM IR using OpenMPIRBuilder.
static LogicalResult
convertOmpSimd(Operation &opInst, llvm::IRBuilderBase &builder,
@@ -2626,12 +2736,15 @@ 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);
+ }
return cleanupPrivateVars(builder, moduleTranslation, simdOp.getLoc(),
privateVarsInfo.llvmVars,
@@ -2698,16 +2811,53 @@ convertOmpLoopNest(Operation &opInst, llvm::IRBuilderBase &builder,
ompLoc.DL);
computeIP = loopInfos.front()->getPreheaderIP();
}
+ if (auto wsloopOp = loopOp->getParentOfType<omp::WsloopOp>()) {
+ bool 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);
+ llvm::Expected<SmallVector<llvm::CanonicalLoopInfo *>> loopResults =
+ ompBuilder->createCanonicalScanLoops(
+ loc, bodyGen, lowerBound, upperBound, step,
+ /*IsSigned=*/true, loopOp.getLoopInclusive(), computeIP,
+ "loop");
+
+ if (failed(handleError(loopResults, *loopOp)))
+ return failure();
+ auto inputLoop = loopResults->front();
+ auto scanLoop = loopResults->back();
+ moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
+ [&](OpenMPLoopInfoStackFrame &frame) {
+ frame.loopInfos.push_back(inputLoop);
+ frame.loopInfos.push_back(scanLoop);
+ return WalkResult::interrupt();
+ });
+ builder.restoreIP(scanLoop->getAfterIP());
+ return success();
+ } else {
+ llvm::Expected<llvm::CanonicalLoopInfo *> loopResult =
+ ompBuilder->createCanonicalLoop(
+ loc, bodyGen, lowerBound, upperBound, step,
+ /*IsSigned=*/true, loopOp.getLoopInclusive(), computeIP);
- llvm::Expected<llvm::CanonicalLoopInfo *> loopResult =
- ompBuilder->createCanonicalLoop(
- loc, bodyGen, lowerBound, upperBound, step,
- /*IsSigned=*/true, loopOp.getLoopInclusive(), computeIP);
+ if (failed(handleError(loopResult, *loopOp)))
+ return failure();
- if (failed(handleError(loopResult, *loopOp)))
- return failure();
+ loopInfos.push_back(*loopResult);
+ }
+ } else {
+ 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);
+ loopInfos.push_back(*loopResult);
+ }
}
// Collapse loops. Store the insertion point because LoopInfos may get
@@ -2719,7 +2869,8 @@ convertOmpLoopNest(Operation &opInst, llvm::IRBuilderBase &builder,
// after applying transformations.
moduleTranslation.stackWalk<OpenMPLoopInfoStackFrame>(
[&](OpenMPLoopInfoStackFrame &frame) {
- frame.loopInfo = ompBuilder->collapseLoops(ompLoc.DL, loopInfos, {});
+ frame.loopInfos.push_back(
+ ompBuilder->collapseLoops(ompLoc.DL, loopInfos, {}));
return WalkResult::interrupt();
});
@@ -4329,18 +4480,20 @@ 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,
@@ -5370,6 +5523,9 @@ convertHostOrTargetOperation(Operation *op, llvm::IRBuilderBase &builder,
.Case([&](omp::WsloopOp) {
return convertOmpWsloop(*op, builder, moduleTranslation);
})
+ .Case([&](omp::ScanOp) {
+ 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..a88c1993aebe1
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/openmp-reduction-scan.mlir
@@ -0,0 +1,120 @@
+// 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)
+ %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: %[[BUFF:.+]] = alloca i32, i32 100, align 4
+//CHECK: omp_loop.preheader{{.*}}: ; preds = %omp.wsloop.region
+//CHECK: omp_loop.after: ; preds = %omp_loop.exit
+//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: ; preds = %omp.inner.log.scan.exit, %omp_loop.after
+//CHECK: %[[K:.+]] = phi i32 [ 0, %omp_loop.after ], [ %[[NEXTK:.+]], %omp.inner.log.scan.exit ]
+//CHECK: %[[I:.+]] = phi i32 [ 1, %omp_loop.after ], [ %[[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: call void @__kmpc_barrier{{.*}}
+//CHECK: br label %omp.scan.loop.cont
+//CHECK: omp.scan.loop.cont: ; preds = %omp.outer.log.scan.exit
+//CHECK: br label %omp_loop.preheader{{.*}}
+//CHECK: omp_loop.after{{.*}}: ; preds = %omp_loop.exit{{.*}}
+//CHECK: %[[ARRLAST:.+]] = getelementptr inbounds i32, ptr %[[BUFF]], i32 100
+//CHECK: %[[RES:.+]] = load i32, ptr %[[ARRLAST]], align 4
+//CHECK: store i32 %[[RES]], ptr %loadgep{{.*}}, align 4
+//CHECK: omp.inscan.dispatch{{.*}}: ; preds = %omp_loop.body{{.*}}
+//CHECK: store i32 0, ptr %[[REDPRIV:.+]], align 4
+//CHECK: %[[arrayOffset1:.+]] = getelementptr inbounds i32, ptr %[[BUFF]], i32 %{{.*}}
+//CHECK: %[[BUFFVAL1:.+]] = load i32, ptr %[[arrayOffset1]], align 4
+//CHECK: store i32 %[[BUFFVAL1]], ptr %[[REDPRIV]], align 4
+//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: %[[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: store i32 0, ptr %[[REDPRIV]], align 4
+//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: %[[ARRAYOFFSET2:.+]] = getelementptr inbounds i32, ptr %[[BUFF]], i32 %{{.*}}
+//CHECK: %[[REDPRIVVAL:.+]] = load i32, ptr %[[REDPRIV]], 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 7eafe396082e4..7b8e8b509d72b 100644
--- a/mlir/test/Target/LLVMIR/openmp-todo.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-todo.mlir
@@ -212,37 +212,6 @@ llvm.func @simd_reduction(%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 @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) {
- omp.loop_nest (%iv) : i32 = (%lb) to (%ub) step (%step) {
- omp.scan inclusive(%prv : !llvm.ptr)
- omp.yield
- }
- }
- llvm.return
-}
-
-// -----
-
llvm.func @single_allocate(%x : !llvm.ptr) {
// expected-error at below {{not yet implemented: Unhandled clause allocate in omp.single operation}}
// expected-error at below {{LLVM Translation failed for operation: omp.single}}
More information about the llvm-commits
mailing list