[mlir][SCF] Multiply lower bound in loop range folding #111875

superlopuh · 2024-10-10T17:09:56Z

github-actions · 2024-10-10T17:10:15Z

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

llvmbot · 2024-10-10T17:10:48Z

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-scf

Author: Sasha Lopoukhine (superlopuh)

Changes

Fixes #83482

Full diff: https://github.com/llvm/llvm-project/pull/111875.diff

2 Files Affected:

(modified) mlir/lib/Dialect/SCF/Transforms/LoopRangeFolding.cpp (+2)
(modified) mlir/test/Dialect/SCF/loop-range.mlir (+60-20)

diff --git a/mlir/lib/Dialect/SCF/Transforms/LoopRangeFolding.cpp b/mlir/lib/Dialect/SCF/Transforms/LoopRangeFolding.cpp
index 9ca441f6431835..0c2a71eb25fe5d 100644
--- a/mlir/lib/Dialect/SCF/Transforms/LoopRangeFolding.cpp
+++ b/mlir/lib/Dialect/SCF/Transforms/LoopRangeFolding.cpp
@@ -72,9 +72,11 @@ void ForLoopRangeFolding::runOnOperation() {
         op.setUpperBound(ubFold->getResult(0));
 
       } else if (isa<arith::MulIOp>(user)) {
+        Operation *lbFold = b.clone(*user, lbMap);
         Operation *ubFold = b.clone(*user, ubMap);
         Operation *stepFold = b.clone(*user, stepMap);
 
+        op.setLowerBound(lbFold->getResult(0));
         op.setUpperBound(ubFold->getResult(0));
         op.setStep(stepFold->getResult(0));
       }
diff --git a/mlir/test/Dialect/SCF/loop-range.mlir b/mlir/test/Dialect/SCF/loop-range.mlir
index cd3b4861fc18d1..b0a421256af39a 100644
--- a/mlir/test/Dialect/SCF/loop-range.mlir
+++ b/mlir/test/Dialect/SCF/loop-range.mlir
@@ -21,12 +21,13 @@ func.func @fold_one_loop(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
 // CHECK:       %[[C4:.*]] = arith.constant 4 : index
 // CHECK:       %[[I0:.*]] = arith.addi %[[ARG2]], %[[C0]] : index
 // CHECK:       %[[I1:.*]] = arith.addi %[[ARG2]], %[[ARG1]] : index
-// CHECK:       %[[I2:.*]] = arith.muli %[[I1]], %[[C4]] : index
-// CHECK:       %[[I3:.*]] = arith.muli %[[C1]], %[[C4]] : index
-// CHECK:       scf.for %[[I:.*]] = %[[I0]] to %[[I2]] step %[[I3]] {
-// CHECK:         %[[I4:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
-// CHECK:         %[[I5:.*]] = arith.muli %[[I4]], %[[I4]] : i32
-// CHECK:         memref.store %[[I5]], %[[ARG0]]{{\[}}%[[I]]
+// CHECK:       %[[I2:.*]] = arith.muli %[[I0]], %[[C4]] : index
+// CHECK:       %[[I3:.*]] = arith.muli %[[I1]], %[[C4]] : index
+// CHECK:       %[[I4:.*]] = arith.muli %[[C1]], %[[C4]] : index
+// CHECK:       scf.for %[[I:.*]] = %[[I2]] to %[[I3]] step %[[I4]] {
+// CHECK:         %[[I5:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
+// CHECK:         %[[I6:.*]] = arith.muli %[[I5]], %[[I5]] : i32
+// CHECK:         memref.store %[[I6]], %[[ARG0]]{{\[}}%[[I]]
 
 func.func @fold_one_loop2(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
   %c0 = arith.constant 0 : index
@@ -54,12 +55,48 @@ func.func @fold_one_loop2(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
 // CHECK:       scf.for %[[J:.*]] = %[[C0]] to %[[C10]] step %[[C1]] {
 // CHECK:         %[[I0:.*]] = arith.addi %[[ARG2]], %[[C0]] : index
 // CHECK:         %[[I1:.*]] = arith.addi %[[ARG2]], %[[ARG1]] : index
-// CHECK:         %[[I2:.*]] = arith.muli %[[I1]], %[[C4]] : index
-// CHECK:         %[[I3:.*]] = arith.muli %[[C1]], %[[C4]] : index
-// CHECK:         scf.for %[[I:.*]] = %[[I0]] to %[[I2]] step %[[I3]] {
-// CHECK:           %[[I4:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
-// CHECK:           %[[I5:.*]] = arith.muli %[[I4]], %[[I4]] : i32
-// CHECK:           memref.store %[[I5]], %[[ARG0]]{{\[}}%[[I]]
+// CHECK:         %[[I2:.*]] = arith.muli %[[I0]], %[[C4]] : index
+// CHECK:         %[[I3:.*]] = arith.muli %[[I1]], %[[C4]] : index
+// CHECK:         %[[I4:.*]] = arith.muli %[[C1]], %[[C4]] : index
+// CHECK:         scf.for %[[I:.*]] = %[[I2]] to %[[I3]] step %[[I4]] {
+// CHECK:           %[[I5:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
+// CHECK:           %[[I6:.*]] = arith.muli %[[I5]], %[[I5]] : i32
+// CHECK:           memref.store %[[I6]], %[[ARG0]]{{\[}}%[[I]]
+
+func.func @fold_one_loop3(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
+  %c0 = arith.constant 0 : index
+  %c1 = arith.constant 1 : index
+  %c4 = arith.constant 4 : index
+  %c10 = arith.constant 10 : index
+  scf.for %j = %c0 to %c10 step %c1 {
+    scf.for %i = %c1 to %arg1 step %c1 {
+      %0 = arith.addi %arg2, %i : index
+      %1 = arith.muli %0, %c4 : index
+      %2 = memref.load %arg0[%1] : memref<?xi32>
+      %3 = arith.muli %2, %2 : i32
+      memref.store %3, %arg0[%1] : memref<?xi32>
+    }
+  }
+  return
+}
+
+// CHECK-LABEL: func @fold_one_loop3
+// CHECK-SAME:   (%[[ARG0:.*]]: {{.*}}, %[[ARG1:.*]]: {{.*}}, %[[ARG2:.*]]: {{.*}}
+// CHECK:       %[[C0:.*]] = arith.constant 0 : index
+// CHECK:       %[[C1:.*]] = arith.constant 1 : index
+// CHECK:       %[[C4:.*]] = arith.constant 4 : index
+// CHECK:       %[[C10:.*]] = arith.constant 10 : index
+// CHECK:       scf.for %[[J:.*]] = %[[C0]] to %[[C10]] step %[[C1]] {
+// CHECK:         %[[I0:.*]] = arith.addi %[[ARG2]], %[[C1]] : index
+// CHECK:         %[[I1:.*]] = arith.addi %[[ARG2]], %[[ARG1]] : index
+// CHECK:         %[[I2:.*]] = arith.muli %[[I0]], %[[C4]] : index
+// CHECK:         %[[I3:.*]] = arith.muli %[[I1]], %[[C4]] : index
+// CHECK:         %[[I4:.*]] = arith.muli %[[C1]], %[[C4]] : index
+// CHECK:         scf.for %[[I:.*]] = %[[I2]] to %[[I3]] step %[[I4]] {
+// CHECK:           %[[I5:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
+// CHECK:           %[[I6:.*]] = arith.muli %[[I5]], %[[I5]] : i32
+// CHECK:           memref.store %[[I6]], %[[ARG0]]{{\[}}%[[I]]
+
 
 func.func @fold_two_loops(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
   %c0 = arith.constant 0 : index
@@ -86,14 +123,17 @@ func.func @fold_two_loops(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {
 // CHECK:       %[[C10:.*]] = arith.constant 10 : index
 // CHECK:       %[[I0:.*]] = arith.addi %[[ARG2]], %[[C0]] : index
 // CHECK:       %[[I1:.*]] = arith.addi %[[ARG2]], %[[C10]] : index
-// CHECK:       scf.for %[[J:.*]] = %[[I0]] to %[[I1]] step %[[C1]] {
-// CHECK:         %[[I1:.*]] = arith.addi %[[ARG2]], %[[ARG1]] : index
-// CHECK:         %[[I2:.*]] = arith.muli %[[I1]], %[[C4]] : index
-// CHECK:         %[[I3:.*]] = arith.muli %[[C1]], %[[C4]] : index
-// CHECK:         scf.for %[[I:.*]] = %[[J]] to %[[I2]] step %[[I3]] {
-// CHECK:           %[[I4:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
-// CHECK:           %[[I5:.*]] = arith.muli %[[I4]], %[[I4]] : i32
-// CHECK:           memref.store %[[I5]], %[[ARG0]]{{\[}}%[[I]]
+// CHECK:       %[[I2:.*]] = arith.muli %[[I0]], %[[C4]] : index
+// CHECK:       %[[I3:.*]] = arith.muli %[[I1]], %[[C4]] : index
+// CHECK:       %[[I4:.*]] = arith.muli %[[C1]], %[[C4]] : index
+// CHECK:       scf.for %[[J:.*]] = %[[I2]] to %[[I3]] step %[[I4]] {
+// CHECK:         %[[I5:.*]] = arith.addi %[[ARG2]], %[[ARG1]] : index
+// CHECK:         %[[I6:.*]] = arith.muli %[[I5]], %[[C4]] : index
+// CHECK:         %[[I7:.*]] = arith.muli %[[C1]], %[[C4]] : index
+// CHECK:         scf.for %[[I:.*]] = %[[J]] to %[[I6]] step %[[I7]] {
+// CHECK:           %[[I8:.*]] = memref.load %[[ARG0]]{{\[}}%[[I]]
+// CHECK:           %[[I9:.*]] = arith.muli %[[I8]], %[[I8]] : i32
+// CHECK:           memref.store %[[I9]], %[[ARG0]]{{\[}}%[[I]]
 
 // If an instruction's operands are not defined outside the loop, we cannot
 // perform the optimization, as is the case with the arith.muli below. (If

zero9178

Makes a lot of sense to me and seems like an interesting oversight. Only have one nit in the tests.

zero9178 · 2024-10-11T08:18:28Z

mlir/test/Dialect/SCF/loop-range.mlir

+// CHECK:           %[[I6:.*]] = arith.muli %[[I5]], %[[I5]] : i32
+// CHECK:           memref.store %[[I6]], %[[ARG0]]{{\[}}%[[I]]
+
+func.func @fold_one_loop3(%arg0: memref<?xi32>, %arg1: index, %arg2: index) {


How is this test different from @fold_one_loop2? To me it looks like the test changes you made for the other test cases seem to cover the lower bound changes. I'd just remove the test otherwise

github-actions · 2024-10-14T18:15:33Z

@superlopuh Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!

Fixes llvm#83482

multiply lower bound in loop range folding

9a9169c

llvmbot added mlir mlir:scf labels Oct 10, 2024

zero9178 reviewed Oct 11, 2024

View reviewed changes

zero9178 changed the title ~~multiply lower bound in loop range folding~~ [mlir][SCF] Multiply lower bound in loop range folding Oct 11, 2024

remove unnecessary test case

230e32c

zero9178 approved these changes Oct 14, 2024

View reviewed changes

zero9178 merged commit 36a4055 into llvm:main Oct 14, 2024
8 checks passed

superlopuh deleted the mlir-scf-fix branch October 14, 2024 22:00

DanielCChen pushed a commit to DanielCChen/llvm-project that referenced this pull request Oct 16, 2024

[mlir][SCF] Multiply lower bound in loop range folding (llvm#111875)

372205a

Fixes llvm#83482

bricknerb pushed a commit to bricknerb/llvm-project that referenced this pull request Oct 17, 2024

[mlir][SCF] Multiply lower bound in loop range folding (llvm#111875)

d3cae09

Fixes llvm#83482

EricWF pushed a commit to efcs/llvm-project that referenced this pull request Oct 22, 2024

[mlir][SCF] Multiply lower bound in loop range folding (llvm#111875)

691b133

Fixes llvm#83482

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[mlir][SCF] Multiply lower bound in loop range folding #111875

[mlir][SCF] Multiply lower bound in loop range folding #111875

superlopuh commented Oct 10, 2024

github-actions bot commented Oct 10, 2024

llvmbot commented Oct 10, 2024 •

edited

Loading

zero9178 left a comment

zero9178 Oct 11, 2024

github-actions bot commented Oct 14, 2024

[mlir][SCF] Multiply lower bound in loop range folding #111875

[mlir][SCF] Multiply lower bound in loop range folding #111875

Conversation

superlopuh commented Oct 10, 2024

github-actions bot commented Oct 10, 2024

llvmbot commented Oct 10, 2024 • edited Loading

zero9178 left a comment

Choose a reason for hiding this comment

zero9178 Oct 11, 2024

Choose a reason for hiding this comment

github-actions bot commented Oct 14, 2024

llvmbot commented Oct 10, 2024 •

edited

Loading