#include "mlir/Transforms/Utils.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/DenseSet.h"
-#include "llvm/ADT/SetVector.h"
+#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#define DEBUG_TYPE "licm"
-using llvm::SetVector;
-
using namespace mlir;
namespace {
};
} // end anonymous namespace
+static bool
+checkInvarianceOfNestedIfOps(Operation *op, Value *indVar,
+ SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist);
+static bool isOpLoopInvariant(Operation &op, Value *indVar,
+ SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist);
+
+static bool
+areAllOpsInTheBlockListInvariant(Region &blockList, Value *indVar,
+ SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist);
+
+static bool isMemRefDereferencingOp(Operation &op) {
+ // TODO(asabne): Support DMA Ops.
+ if (isa<LoadOp>(op) || isa<StoreOp>(op)) {
+ return true;
+ }
+ return false;
+}
+
FunctionPassBase *mlir::createLoopInvariantCodeMotionPass() {
return new LoopInvariantCodeMotion();
}
+// Returns true if the individual op is loop invariant.
+bool isOpLoopInvariant(Operation &op, Value *indVar,
+ SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist) {
+ LLVM_DEBUG(llvm::dbgs() << "iterating on op: " << op;);
+
+ if (isa<AffineIfOp>(op)) {
+ if (!checkInvarianceOfNestedIfOps(&op, indVar, definedOps, opsToHoist)) {
+ return false;
+ }
+ } else if (isa<AffineForOp>(op)) {
+ // If the body of a predicated region has a for loop, we don't hoist the
+ // 'affine.if'.
+ return false;
+ } else if (isa<DmaStartOp>(op) || isa<DmaWaitOp>(op)) {
+ // TODO(asabne): Support DMA ops.
+ return false;
+ } else if (!isa<ConstantOp>(op)) {
+ if (isMemRefDereferencingOp(op)) {
+ Value *memref = isa<LoadOp>(op) ? cast<LoadOp>(op).getMemRef()
+ : cast<StoreOp>(op).getMemRef();
+ for (auto *user : memref->getUsers()) {
+ // If this memref has a user that is a DMA, give up because these
+ // operations write to this memref.
+ if (isa<DmaStartOp>(op) || isa<DmaWaitOp>(op)) {
+ return false;
+ }
+ // If the memref used by the load/store is used in a store elsewhere in
+ // the loop nest, we do not hoist. Similarly, if the memref used in a
+ // load is also being stored too, we do not hoist the load.
+ if (isa<StoreOp>(user) || (isa<LoadOp>(user) && isa<StoreOp>(op))) {
+ if (&op != user) {
+ SmallVector<AffineForOp, 8> userIVs;
+ getLoopIVs(*user, &userIVs);
+ // Check that userIVs don't contain the for loop around the op.
+ if (llvm::is_contained(userIVs, getForInductionVarOwner(indVar))) {
+ return false;
+ }
+ }
+ }
+ }
+ }
+
+ // Insert this op in the defined ops list.
+ definedOps.insert(&op);
+
+ if (op.getNumOperands() == 0 && !isa<AffineTerminatorOp>(op)) {
+ LLVM_DEBUG(llvm::dbgs() << "\nNon-constant op with 0 operands\n");
+ return false;
+ }
+ for (unsigned int i = 0; i < op.getNumOperands(); ++i) {
+ auto *operandSrc = op.getOperand(i)->getDefiningOp();
+
+ LLVM_DEBUG(
+ op.getOperand(i)->print(llvm::dbgs() << "\nIterating on operand\n"));
+
+ // If the loop IV is the operand, this op isn't loop invariant.
+ if (indVar == op.getOperand(i)) {
+ LLVM_DEBUG(llvm::dbgs() << "\nLoop IV is the operand\n");
+ return false;
+ }
+
+ if (operandSrc != nullptr) {
+ LLVM_DEBUG(llvm::dbgs()
+ << *operandSrc << "\nIterating on operand src\n");
+
+ // If the value was defined in the loop (outside of the
+ // if/else region), and that operation itself wasn't meant to
+ // be hoisted, then mark this operation loop dependent.
+ if (definedOps.count(operandSrc) && opsToHoist.count(operandSrc) == 0) {
+ return false;
+ }
+ }
+ }
+ }
+
+ // If no operand was loop variant, mark this op for motion.
+ opsToHoist.insert(&op);
+ return true;
+}
+
+// Checks if all ops in a region (i.e. list of blocks) are loop invariant.
+bool areAllOpsInTheBlockListInvariant(
+ Region &blockList, Value *indVar, SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist) {
+
+ for (auto &b : blockList) {
+ for (auto &op : b) {
+ if (!isOpLoopInvariant(op, indVar, definedOps, opsToHoist)) {
+ return false;
+ }
+ }
+ }
+
+ return true;
+}
+
+// Returns true if the affine.if op can be hoisted.
+bool checkInvarianceOfNestedIfOps(Operation *op, Value *indVar,
+ SmallPtrSetImpl<Operation *> &definedOps,
+ SmallPtrSetImpl<Operation *> &opsToHoist) {
+ assert(isa<AffineIfOp>(op));
+ auto ifOp = cast<AffineIfOp>(op);
+
+ if (!areAllOpsInTheBlockListInvariant(ifOp.getThenBlocks(), indVar,
+ definedOps, opsToHoist)) {
+ return false;
+ }
+
+ if (!areAllOpsInTheBlockListInvariant(ifOp.getElseBlocks(), indVar,
+ definedOps, opsToHoist)) {
+ return false;
+ }
+
+ return true;
+}
+
void LoopInvariantCodeMotion::runOnAffineForOp(AffineForOp forOp) {
auto *loopBody = forOp.getBody();
+ auto *indVar = forOp.getInductionVar();
+ SmallPtrSet<Operation *, 8> definedOps;
// This is the place where hoisted instructions would reside.
FuncBuilder b(forOp.getOperation());
- // This vector is used to place loop invariant operations.
+ SmallPtrSet<Operation *, 8> opsToHoist;
SmallVector<Operation *, 8> opsToMove;
- SetVector<Operation *> loopDefinedOps;
- // Generate forward slice which contains ops that fall under the transitive
- // definition closure following the loop induction variable.
- getForwardSlice(forOp, &loopDefinedOps);
-
- LLVM_DEBUG(for (auto i
- : loopDefinedOps) {
- i->print(llvm::dbgs() << "\nLoop-dependent op\n");
- });
-
for (auto &op : *loopBody) {
- // If the operation is loop invariant, insert it into opsToMove.
- if (!isa<AffineForOp>(op) && !isa<AffineTerminatorOp>(op) &&
- loopDefinedOps.count(&op) != 1) {
- LLVM_DEBUG(op.print(llvm::dbgs() << "\nLICM'ing op\n"));
- opsToMove.push_back(&op);
+ // We don't hoist for loops.
+ if (!isa<AffineForOp>(op)) {
+ if (!isa<AffineTerminatorOp>(op)) {
+ if (isOpLoopInvariant(op, indVar, definedOps, opsToHoist)) {
+ opsToMove.push_back(&op);
+ }
+ }
}
}
- // For all instructions that we found to be invariant, place them sequentially
+ // For all instructions that we found to be invariant, place sequentially
// right before the for loop.
for (auto *op : opsToMove) {
op->moveBefore(forOp);
}
- LLVM_DEBUG(forOp.getOperation()->print(llvm::dbgs() << "\nModified loop\n"));
+ LLVM_DEBUG(forOp.getOperation()->print(llvm::dbgs() << "Modified loop\n"));
// If the for loop body has a single operation (the terminator), erase it.
if (forOp.getBody()->getOperations().size() == 1) {
}
void LoopInvariantCodeMotion::runOnFunction() {
-
// Walk through all loops in a function in innermost-loop-first order. This
// way, we first LICM from the inner loop, and place the ops in
// the outer loop, which in turn can be further LICM'ed.
}
-func @nested_loops_with_common_and_uncommon_invariant_code() {
+func @dependent_stores() {
%m = alloc() : memref<10xf32>
%cf7 = constant 7.0 : f32
%cf8 = constant 8.0 : f32
%v0 = addf %cf7, %cf8 : f32
affine.for %i1 = 0 to 10 {
%v1 = addf %cf7, %cf7 : f32
- store %v0, %m[%i1] : memref<10xf32>
+ store %v1, %m[%i1] : memref<10xf32>
store %v0, %m[%i0] : memref<10xf32>
}
}
// CHECK-NEXT: %1 = addf %cst, %cst_0 : f32
// CHECK-NEXT: %2 = addf %cst, %cst : f32
// CHECK-NEXT: affine.for %i0 = 0 to 10 {
- // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: store %2, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+ return
+}
+
+func @independent_stores() {
+ %m = alloc() : memref<10xf32>
+ %cf7 = constant 7.0 : f32
+ %cf8 = constant 8.0 : f32
+
+ affine.for %i0 = 0 to 10 {
+ %v0 = addf %cf7, %cf8 : f32
+ affine.for %i1 = 0 to 10 {
+ %v1 = addf %cf7, %cf7 : f32
+ store %v0, %m[%i0] : memref<10xf32>
+ store %v1, %m[%i1] : memref<10xf32>
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 7.000000e+00 : f32
+ // CHECK-NEXT: %cst_0 = constant 8.000000e+00 : f32
+ // CHECK-NEXT: %1 = addf %cst, %cst_0 : f32
+ // CHECK-NEXT: %2 = addf %cst, %cst : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: store %2, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+ return
+}
+
+func @load_dependent_store() {
+ %m = alloc() : memref<10xf32>
+ %cf7 = constant 7.0 : f32
+ %cf8 = constant 8.0 : f32
+
+ affine.for %i0 = 0 to 10 {
+ %v0 = addf %cf7, %cf8 : f32
+ affine.for %i1 = 0 to 10 {
+ %v1 = addf %cf7, %cf7 : f32
+ store %v0, %m[%i1] : memref<10xf32>
+ %v2 = load %m[%i0] : memref<10xf32>
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 7.000000e+00 : f32
+ // CHECK-NEXT: %cst_0 = constant 8.000000e+00 : f32
+ // CHECK-NEXT: %1 = addf %cst, %cst_0 : f32
+ // CHECK-NEXT: %2 = addf %cst, %cst : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
// CHECK-NEXT: affine.for %i1 = 0 to 10 {
// CHECK-NEXT: store %1, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: %3 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+ return
+}
+
+func @load_after_load() {
+ %m = alloc() : memref<10xf32>
+ %cf7 = constant 7.0 : f32
+ %cf8 = constant 8.0 : f32
+
+ affine.for %i0 = 0 to 10 {
+ %v0 = addf %cf7, %cf8 : f32
+ affine.for %i1 = 0 to 10 {
+ %v1 = addf %cf7, %cf7 : f32
+ %v3 = load %m[%i1] : memref<10xf32>
+ %v2 = load %m[%i0] : memref<10xf32>
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 7.000000e+00 : f32
+ // CHECK-NEXT: %cst_0 = constant 8.000000e+00 : f32
+ // CHECK-NEXT: %1 = addf %cst, %cst_0 : f32
+ // CHECK-NEXT: %2 = addf %cst, %cst : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: %3 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: %4 = load %0[%i1] : memref<10xf32>
// CHECK-NEXT: }
// CHECK-NEXT: }
// CHECK-NEXT: return
return
}
+func @invariant_affine_if2() {
+ %m = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ store %cf9, %m[%i1] : memref<10xf32>
+
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %1 = addf %cst, %cst : f32
+ // CHECK-NEXT: store %1, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+func @invariant_affine_nested_if() {
+ %m = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ store %cf9, %m[%i0] : memref<10xf32>
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ store %cf9, %m[%i1] : memref<10xf32>
+ }
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %1 = addf %cst, %cst : f32
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: store %1, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+func @invariant_affine_nested_if_else() {
+ %m = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ store %cf9, %m[%i0] : memref<10xf32>
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ store %cf9, %m[%i0] : memref<10xf32>
+ } else {
+ store %cf9, %m[%i1] : memref<10xf32>
+ }
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %1 = addf %cst, %cst : f32
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: } else {
+ // CHECK-NEXT: store %1, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+func @invariant_affine_nested_if_else2() {
+ %m = alloc() : memref<10xf32>
+ %m2 = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ %tload1 = load %m[%i0] : memref<10xf32>
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ store %cf9, %m2[%i0] : memref<10xf32>
+ } else {
+ %tload2 = load %m[%i0] : memref<10xf32>
+ }
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %1 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %2 = addf %cst, %cst : f32
+ // CHECK-NEXT: %3 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: store %2, %1[%i0] : memref<10xf32>
+ // CHECK-NEXT: } else {
+ // CHECK-NEXT: %4 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+
+func @invariant_affine_nested_if2() {
+ %m = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ %v1 = load %m[%i0] : memref<10xf32>
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %v2 = load %m[%i0] : memref<10xf32>
+ }
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %1 = addf %cst, %cst : f32
+ // CHECK-NEXT: %2 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %3 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+func @invariant_affine_for_inside_affine_if() {
+ %m = alloc() : memref<10xf32>
+ %cf8 = constant 8.0 : f32
+ affine.for %i0 = 0 to 10 {
+ affine.for %i1 = 0 to 10 {
+ affine.if (d0, d1) : (d1 - d0 >= 0) (%i0, %i0) {
+ %cf9 = addf %cf8, %cf8 : f32
+ store %cf9, %m[%i0] : memref<10xf32>
+ affine.for %i2 = 0 to 10 {
+ store %cf9, %m[%i2] : memref<10xf32>
+ }
+ }
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: affine.if #set0(%i0, %i0) {
+ // CHECK-NEXT: %1 = addf %cst, %cst : f32
+ // CHECK-NEXT: store %1, %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: affine.for %i2 = 0 to 10 {
+ // CHECK-NEXT: store %1, %0[%i2] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+
func @invariant_constant_and_load() {
%m = alloc() : memref<100xf32>
+ %m2 = alloc() : memref<100xf32>
affine.for %i0 = 0 to 5 {
%c0 = constant 0 : index
- %v = load %m[%c0] : memref<100xf32>
+ %v = load %m2[%c0] : memref<100xf32>
store %v, %m[%i0] : memref<100xf32>
}
// CHECK: %0 = alloc() : memref<100xf32>
+ // CHECK-NEXT: %1 = alloc() : memref<100xf32>
// CHECK-NEXT: %c0 = constant 0 : index
- // CHECK-NEXT: %1 = load %0[%c0] : memref<100xf32>
+ // CHECK-NEXT: %2 = load %1[%c0] : memref<100xf32>
// CHECK-NEXT: affine.for %i0 = 0 to 5 {
- // CHECK-NEXT: store %1, %0[%i0] : memref<100xf32>
+ // CHECK-NEXT: store %2, %0[%i0] : memref<100xf32>
// CHECK-NEXT: }
// CHECK-NEXT: return
return
}
+
+func @nested_load_store_same_memref() {
+ %m = alloc() : memref<10xf32>
+ %cst = constant 8.0 : f32
+ %c0 = constant 0 : index
+ affine.for %i0 = 0 to 10 {
+ %v0 = load %m[%c0] : memref<10xf32>
+ affine.for %i1 = 0 to 10 {
+ store %cst, %m[%i1] : memref<10xf32>
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: %c0 = constant 0 : index
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: %1 = load %0[%c0] : memref<10xf32>
+ // CHECK-NEXT: affine.for %i1 = 0 to 10 {
+ // CHECK-NEXT: store %cst, %0[%i1] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}
+
+
+func @nested_load_store_same_memref2() {
+ %m = alloc() : memref<10xf32>
+ %cst = constant 8.0 : f32
+ %c0 = constant 0 : index
+ affine.for %i0 = 0 to 10 {
+ store %cst, %m[%c0] : memref<10xf32>
+ affine.for %i1 = 0 to 10 {
+ %v0 = load %m[%i0] : memref<10xf32>
+ }
+ }
+
+ // CHECK: %0 = alloc() : memref<10xf32>
+ // CHECK-NEXT: %cst = constant 8.000000e+00 : f32
+ // CHECK-NEXT: %c0 = constant 0 : index
+ // CHECK-NEXT: affine.for %i0 = 0 to 10 {
+ // CHECK-NEXT: store %cst, %0[%c0] : memref<10xf32>
+ // CHECK-NEXT: %1 = load %0[%i0] : memref<10xf32>
+ // CHECK-NEXT: }
+ // CHECK-NEXT: return
+
+ return
+}