Skip to content

Commit 132de3a

Browse files
authored
[MLIR] Fix arbitrary checks in affine LICM (#116469)
Fix arbitrary checks and hardcoding/specialcasing in affine LICM. Drop unnecessary (too much) debug logging. This pass is still unsound due to not handling aliases. This will have to be handled later.
1 parent be75020 commit 132de3a

File tree

2 files changed

+53
-94
lines changed

2 files changed

+53
-94
lines changed

mlir/lib/Dialect/Affine/Transforms/AffineLoopInvariantCodeMotion.cpp

Lines changed: 50 additions & 92 deletions
Original file line numberDiff line numberDiff line change
@@ -12,25 +12,9 @@
1212

1313
#include "mlir/Dialect/Affine/Passes.h"
1414

15-
#include "mlir/Analysis/SliceAnalysis.h"
16-
#include "mlir/Dialect/Affine/Analysis/AffineAnalysis.h"
17-
#include "mlir/Dialect/Affine/Analysis/AffineStructures.h"
18-
#include "mlir/Dialect/Affine/Analysis/LoopAnalysis.h"
1915
#include "mlir/Dialect/Affine/Analysis/Utils.h"
20-
#include "mlir/Dialect/Affine/IR/AffineOps.h"
21-
#include "mlir/Dialect/Affine/LoopUtils.h"
22-
#include "mlir/Dialect/Affine/Utils.h"
23-
#include "mlir/Dialect/Arith/IR/Arith.h"
2416
#include "mlir/Dialect/Func/IR/FuncOps.h"
25-
#include "mlir/IR/AffineExpr.h"
26-
#include "mlir/IR/AffineMap.h"
27-
#include "mlir/IR/Builders.h"
28-
#include "mlir/IR/Matchers.h"
2917
#include "mlir/Interfaces/SideEffectInterfaces.h"
30-
#include "llvm/ADT/DenseMap.h"
31-
#include "llvm/ADT/DenseSet.h"
32-
#include "llvm/ADT/SmallPtrSet.h"
33-
#include "llvm/Support/CommandLine.h"
3418
#include "llvm/Support/Debug.h"
3519
#include "llvm/Support/raw_ostream.h"
3620

@@ -41,17 +25,21 @@ namespace affine {
4125
} // namespace affine
4226
} // namespace mlir
4327

44-
#define DEBUG_TYPE "licm"
28+
#define DEBUG_TYPE "affine-licm"
4529

4630
using namespace mlir;
4731
using namespace mlir::affine;
4832

4933
namespace {
5034

5135
/// Affine loop invariant code motion (LICM) pass.
52-
/// TODO: The pass is missing zero-trip tests.
53-
/// TODO: This code should be removed once the new LICM pass can handle its
54-
/// uses.
36+
/// TODO: The pass is missing zero tripcount tests.
37+
/// TODO: When compared to the other standard LICM pass, this pass
38+
/// has some special handling for affine read/write ops but such handling
39+
/// requires aliasing to be sound, and as such this pass is unsound. In
40+
/// addition, this handling is nothing particular to affine memory ops but would
41+
/// apply to any memory read/write effect ops. Either aliasing should be handled
42+
/// or this pass can be removed and the standard LICM can be used.
5543
struct LoopInvariantCodeMotion
5644
: public affine::impl::AffineLoopInvariantCodeMotionBase<
5745
LoopInvariantCodeMotion> {
@@ -61,100 +49,80 @@ struct LoopInvariantCodeMotion
6149
} // namespace
6250

6351
static bool
64-
checkInvarianceOfNestedIfOps(AffineIfOp ifOp, Value indVar, ValueRange iterArgs,
52+
checkInvarianceOfNestedIfOps(AffineIfOp ifOp, AffineForOp loop,
6553
SmallPtrSetImpl<Operation *> &opsWithUsers,
6654
SmallPtrSetImpl<Operation *> &opsToHoist);
67-
static bool isOpLoopInvariant(Operation &op, Value indVar, ValueRange iterArgs,
55+
static bool isOpLoopInvariant(Operation &op, AffineForOp loop,
6856
SmallPtrSetImpl<Operation *> &opsWithUsers,
6957
SmallPtrSetImpl<Operation *> &opsToHoist);
7058

7159
static bool
72-
areAllOpsInTheBlockListInvariant(Region &blockList, Value indVar,
73-
ValueRange iterArgs,
60+
areAllOpsInTheBlockListInvariant(Region &blockList, AffineForOp loop,
7461
SmallPtrSetImpl<Operation *> &opsWithUsers,
7562
SmallPtrSetImpl<Operation *> &opsToHoist);
7663

77-
// Returns true if the individual op is loop invariant.
78-
static bool isOpLoopInvariant(Operation &op, Value indVar, ValueRange iterArgs,
64+
/// Returns true if `op` is invariant on `loop`.
65+
static bool isOpLoopInvariant(Operation &op, AffineForOp loop,
7966
SmallPtrSetImpl<Operation *> &opsWithUsers,
8067
SmallPtrSetImpl<Operation *> &opsToHoist) {
81-
LLVM_DEBUG(llvm::dbgs() << "iterating on op: " << op;);
68+
Value iv = loop.getInductionVar();
8269

8370
if (auto ifOp = dyn_cast<AffineIfOp>(op)) {
84-
if (!checkInvarianceOfNestedIfOps(ifOp, indVar, iterArgs, opsWithUsers,
85-
opsToHoist))
71+
if (!checkInvarianceOfNestedIfOps(ifOp, loop, opsWithUsers, opsToHoist))
8672
return false;
8773
} else if (auto forOp = dyn_cast<AffineForOp>(op)) {
88-
if (!areAllOpsInTheBlockListInvariant(forOp.getRegion(), indVar, iterArgs,
89-
opsWithUsers, opsToHoist))
74+
if (!areAllOpsInTheBlockListInvariant(forOp.getRegion(), loop, opsWithUsers,
75+
opsToHoist))
9076
return false;
9177
} else if (auto parOp = dyn_cast<AffineParallelOp>(op)) {
92-
if (!areAllOpsInTheBlockListInvariant(parOp.getRegion(), indVar, iterArgs,
93-
opsWithUsers, opsToHoist))
78+
if (!areAllOpsInTheBlockListInvariant(parOp.getRegion(), loop, opsWithUsers,
79+
opsToHoist))
9480
return false;
9581
} else if (!isMemoryEffectFree(&op) &&
96-
!isa<AffineReadOpInterface, AffineWriteOpInterface,
97-
AffinePrefetchOp>(&op)) {
82+
!isa<AffineReadOpInterface, AffineWriteOpInterface>(&op)) {
9883
// Check for side-effecting ops. Affine read/write ops are handled
9984
// separately below.
10085
return false;
101-
} else if (!matchPattern(&op, m_Constant())) {
86+
} else if (isa<AffineReadOpInterface, AffineWriteOpInterface>(op)) {
10287
// Register op in the set of ops that have users.
10388
opsWithUsers.insert(&op);
104-
if (isa<AffineReadOpInterface, AffineWriteOpInterface>(op)) {
105-
auto read = dyn_cast<AffineReadOpInterface>(op);
106-
Value memref = read ? read.getMemRef()
107-
: cast<AffineWriteOpInterface>(op).getMemRef();
108-
for (auto *user : memref.getUsers()) {
109-
// If this memref has a user that is a DMA, give up because these
110-
// operations write to this memref.
111-
if (isa<AffineDmaStartOp, AffineDmaWaitOp>(user))
89+
SmallVector<AffineForOp, 8> userIVs;
90+
auto read = dyn_cast<AffineReadOpInterface>(op);
91+
Value memref =
92+
read ? read.getMemRef() : cast<AffineWriteOpInterface>(op).getMemRef();
93+
for (auto *user : memref.getUsers()) {
94+
// If the memref used by the load/store is used in a store elsewhere in
95+
// the loop nest, we do not hoist. Similarly, if the memref used in a
96+
// load is also being stored too, we do not hoist the load.
97+
// FIXME: This is missing checking aliases.
98+
if (&op == user)
99+
continue;
100+
if (hasEffect<MemoryEffects::Write>(user, memref) ||
101+
(hasEffect<MemoryEffects::Read>(user, memref) &&
102+
isa<AffineWriteOpInterface>(op))) {
103+
userIVs.clear();
104+
getAffineForIVs(*user, &userIVs);
105+
// Check that userIVs don't contain the for loop around the op.
106+
if (llvm::is_contained(userIVs, loop))
112107
return false;
113-
// If the memref used by the load/store is used in a store elsewhere in
114-
// the loop nest, we do not hoist. Similarly, if the memref used in a
115-
// load is also being stored too, we do not hoist the load.
116-
if (isa<AffineWriteOpInterface>(user) ||
117-
(isa<AffineReadOpInterface>(user) &&
118-
isa<AffineWriteOpInterface>(op))) {
119-
if (&op != user) {
120-
SmallVector<AffineForOp, 8> userIVs;
121-
getAffineForIVs(*user, &userIVs);
122-
// Check that userIVs don't contain the for loop around the op.
123-
if (llvm::is_contained(userIVs, getForInductionVarOwner(indVar)))
124-
return false;
125-
}
126-
}
127108
}
128109
}
129-
130-
if (op.getNumOperands() == 0 && !isa<AffineYieldOp>(op)) {
131-
LLVM_DEBUG(llvm::dbgs() << "Non-constant op with 0 operands\n");
132-
return false;
133-
}
134110
}
135111

136112
// Check operands.
113+
ValueRange iterArgs = loop.getRegionIterArgs();
137114
for (unsigned int i = 0; i < op.getNumOperands(); ++i) {
138115
auto *operandSrc = op.getOperand(i).getDefiningOp();
139116

140-
LLVM_DEBUG(
141-
op.getOperand(i).print(llvm::dbgs() << "Iterating on operand\n"));
142-
143117
// If the loop IV is the operand, this op isn't loop invariant.
144-
if (indVar == op.getOperand(i)) {
145-
LLVM_DEBUG(llvm::dbgs() << "Loop IV is the operand\n");
118+
if (iv == op.getOperand(i))
146119
return false;
147-
}
148120

149121
// If the one of the iter_args is the operand, this op isn't loop invariant.
150-
if (llvm::is_contained(iterArgs, op.getOperand(i))) {
151-
LLVM_DEBUG(llvm::dbgs() << "One of the iter_args is the operand\n");
122+
if (llvm::is_contained(iterArgs, op.getOperand(i)))
152123
return false;
153-
}
154124

155125
if (operandSrc) {
156-
LLVM_DEBUG(llvm::dbgs() << *operandSrc << "Iterating on operand src\n");
157-
158126
// If the value was defined in the loop (outside of the if/else region),
159127
// and that operation itself wasn't meant to be hoisted, then mark this
160128
// operation loop dependent.
@@ -170,14 +138,13 @@ static bool isOpLoopInvariant(Operation &op, Value indVar, ValueRange iterArgs,
170138

171139
// Checks if all ops in a region (i.e. list of blocks) are loop invariant.
172140
static bool
173-
areAllOpsInTheBlockListInvariant(Region &blockList, Value indVar,
174-
ValueRange iterArgs,
141+
areAllOpsInTheBlockListInvariant(Region &blockList, AffineForOp loop,
175142
SmallPtrSetImpl<Operation *> &opsWithUsers,
176143
SmallPtrSetImpl<Operation *> &opsToHoist) {
177144

178145
for (auto &b : blockList) {
179146
for (auto &op : b) {
180-
if (!isOpLoopInvariant(op, indVar, iterArgs, opsWithUsers, opsToHoist))
147+
if (!isOpLoopInvariant(op, loop, opsWithUsers, opsToHoist))
181148
return false;
182149
}
183150
}
@@ -187,40 +154,36 @@ areAllOpsInTheBlockListInvariant(Region &blockList, Value indVar,
187154

188155
// Returns true if the affine.if op can be hoisted.
189156
static bool
190-
checkInvarianceOfNestedIfOps(AffineIfOp ifOp, Value indVar, ValueRange iterArgs,
157+
checkInvarianceOfNestedIfOps(AffineIfOp ifOp, AffineForOp loop,
191158
SmallPtrSetImpl<Operation *> &opsWithUsers,
192159
SmallPtrSetImpl<Operation *> &opsToHoist) {
193-
if (!areAllOpsInTheBlockListInvariant(ifOp.getThenRegion(), indVar, iterArgs,
160+
if (!areAllOpsInTheBlockListInvariant(ifOp.getThenRegion(), loop,
194161
opsWithUsers, opsToHoist))
195162
return false;
196163

197-
if (!areAllOpsInTheBlockListInvariant(ifOp.getElseRegion(), indVar, iterArgs,
164+
if (!areAllOpsInTheBlockListInvariant(ifOp.getElseRegion(), loop,
198165
opsWithUsers, opsToHoist))
199166
return false;
200167

201168
return true;
202169
}
203170

204171
void LoopInvariantCodeMotion::runOnAffineForOp(AffineForOp forOp) {
205-
auto *loopBody = forOp.getBody();
206-
auto indVar = forOp.getInductionVar();
207-
ValueRange iterArgs = forOp.getRegionIterArgs();
208-
209172
// This is the place where hoisted instructions would reside.
210173
OpBuilder b(forOp.getOperation());
211174

212175
SmallPtrSet<Operation *, 8> opsToHoist;
213176
SmallVector<Operation *, 8> opsToMove;
214177
SmallPtrSet<Operation *, 8> opsWithUsers;
215178

216-
for (auto &op : *loopBody) {
179+
for (Operation &op : *forOp.getBody()) {
217180
// Register op in the set of ops that have users. This set is used
218181
// to prevent hoisting ops that depend on these ops that are
219182
// not being hoisted.
220183
if (!op.use_empty())
221184
opsWithUsers.insert(&op);
222185
if (!isa<AffineYieldOp>(op)) {
223-
if (isOpLoopInvariant(op, indVar, iterArgs, opsWithUsers, opsToHoist)) {
186+
if (isOpLoopInvariant(op, forOp, opsWithUsers, opsToHoist)) {
224187
opsToMove.push_back(&op);
225188
}
226189
}
@@ -231,18 +194,13 @@ void LoopInvariantCodeMotion::runOnAffineForOp(AffineForOp forOp) {
231194
for (auto *op : opsToMove) {
232195
op->moveBefore(forOp);
233196
}
234-
235-
LLVM_DEBUG(forOp->print(llvm::dbgs() << "Modified loop\n"));
236197
}
237198

238199
void LoopInvariantCodeMotion::runOnOperation() {
239200
// Walk through all loops in a function in innermost-loop-first order. This
240201
// way, we first LICM from the inner loop, and place the ops in
241202
// the outer loop, which in turn can be further LICM'ed.
242-
getOperation().walk([&](AffineForOp op) {
243-
LLVM_DEBUG(op->print(llvm::dbgs() << "\nOriginal loop\n"));
244-
runOnAffineForOp(op);
245-
});
203+
getOperation().walk([&](AffineForOp op) { runOnAffineForOp(op); });
246204
}
247205

248206
std::unique_ptr<OperationPass<func::FuncOp>>

mlir/test/Dialect/Affine/affine-loop-invariant-code-motion.mlir

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -855,15 +855,16 @@ func.func @affine_prefetch_invariant() {
855855
affine.for %i0 = 0 to 10 {
856856
affine.for %i1 = 0 to 10 {
857857
%1 = affine.load %0[%i0, %i1] : memref<10x10xf32>
858+
// A prefetch shouldn't be hoisted.
858859
affine.prefetch %0[%i0, %i0], write, locality<0>, data : memref<10x10xf32>
859860
}
860861
}
861862

862863
// CHECK: memref.alloc() : memref<10x10xf32>
863864
// CHECK-NEXT: affine.for %{{.*}} = 0 to 10 {
864-
// CHECK-NEXT: affine.prefetch
865865
// CHECK-NEXT: affine.for %{{.*}} = 0 to 10 {
866-
// CHECK-NEXT: %{{.*}} = affine.load %{{.*}}[%{{.*}} : memref<10x10xf32>
866+
// CHECK-NEXT: affine.load %{{.*}}[%{{.*}} : memref<10x10xf32>
867+
// CHECK-NEXT: affine.prefetch
867868
// CHECK-NEXT: }
868869
// CHECK-NEXT: }
869870
return

0 commit comments

Comments
 (0)