Fix Bug in RemoveDeadValues Pass #148437

ronigoldman22 · 2025-07-13T14:57:29Z

This patch fixes a bug in the RemoveDeadValues pass where unused function arguments were not removed from the function signature in an edge case where the function returns void.
A corresponding test was added to the MLIR LIT test suite to cover this case.

github-actions · 2025-07-13T14:57:49Z

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 · 2025-07-13T14:58:17Z

@llvm/pr-subscribers-mlir-core

Author: None (ronigoldman22)

Changes

This patch fixes a bug in the RemoveDeadValues pass where unused function arguments were not removed from the function signature in an edge case where the function returns void.
A corresponding test was added to the MLIR LIT test suite to cover this case.

Full diff: https://github.com/llvm/llvm-project/pull/148437.diff

2 Files Affected:

(modified) mlir/lib/Transforms/RemoveDeadValues.cpp (+2-2)
(modified) mlir/test/Transforms/remove-dead-values.mlir (+40)

diff --git a/mlir/lib/Transforms/RemoveDeadValues.cpp b/mlir/lib/Transforms/RemoveDeadValues.cpp
index 608bdcb948176..ddd5f2ba1a7b7 100644
--- a/mlir/lib/Transforms/RemoveDeadValues.cpp
+++ b/mlir/lib/Transforms/RemoveDeadValues.cpp
@@ -345,8 +345,6 @@ static void processFuncOp(FunctionOpInterface funcOp, Operation *module,
   // since it forwards only to non-live value(s) (%1#1).
   Operation *lastReturnOp = funcOp.back().getTerminator();
   size_t numReturns = lastReturnOp->getNumOperands();
-  if (numReturns == 0)
-    return;
   BitVector nonLiveRets(numReturns, true);
   for (SymbolTable::SymbolUse use : uses) {
     Operation *callOp = use.getUser();
@@ -368,6 +366,8 @@ static void processFuncOp(FunctionOpInterface funcOp, Operation *module,
   cl.functions.push_back({funcOp, nonLiveArgs, nonLiveRets});
 
   // Do (5) and (6).
+  if (numReturns == 0)
+    return;
   for (SymbolTable::SymbolUse use : uses) {
     Operation *callOp = use.getUser();
     assert(isa<CallOpInterface>(callOp) && "expected a call-like user");
diff --git a/mlir/test/Transforms/remove-dead-values.mlir b/mlir/test/Transforms/remove-dead-values.mlir
index 3af95db3c0e24..dcaeefc3f2a6f 100644
--- a/mlir/test/Transforms/remove-dead-values.mlir
+++ b/mlir/test/Transforms/remove-dead-values.mlir
@@ -548,3 +548,43 @@ func.func @test_atomic_yield(%I: memref<10xf32>, %idx : index) {
   func.return
 }
 
+// CHECK-LABEL: module @return_void_with_unused_argument {
+// CHECK-LABEL:  func.func private @noop() {
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+// CHECK-LABEL:  func.func private @fn_return_void_with_unused_argument(%arg0: memref<1x1xsi8, 1>, %arg1: memref<1x1xsi16, 1>) {
+// CHECK-NEXT:    %alloc = memref.alloc() {alignment = 8 : i64} : memref<1x73xsi8, 3>
+// CHECK-NEXT:    %alloc_0 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi16, 3>
+// CHECK-NEXT:    %alloc_1 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi8, 3>
+// CHECK-NEXT:    memref.copy %arg0, %alloc_1 : memref<1x1xsi8, 1> to memref<1x1xsi8, 3>
+// CHECK-NEXT:    memref.copy %arg1, %alloc_0 : memref<1x1xsi16, 1> to memref<1x1xsi16, 3>
+// CHECK-NEXT:    call @noop() : () -> ()
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+// CHECK-LABEL:  func.func @main(%arg0: memref<1x73xsi8, 1>, %arg1: memref<1x1xsi8, 1>, %arg2: memref<1x1xsi16, 1>) -> memref<1x73xsi8, 1> {
+// CHECK-NEXT:    %alloc = memref.alloc() : memref<1x73xsi8, 1>
+// CHECK-NEXT:    call @fn_return_void_with_unused_argument(%arg1, %arg2) : (memref<1x1xsi8, 1>, memref<1x1xsi16, 1>) -> ()
+// CHECK-NEXT:    return %alloc : memref<1x73xsi8, 1>
+// CHECK-NEXT:  }
+// CHECK-NEXT:}
+module @return_void_with_unused_argument {
+  func.func private @noop(%arg0: memref<1x1xsi8, 3>, %arg1: memref<1x1xsi16, 3>, %arg2: memref<1x73xsi8, 1>) {
+    return
+  }
+  func.func private @fn_return_void_with_unused_argument(%arg0: memref<1x73xsi8, 1> , %arg1: memref<1x1xsi8, 1> , %arg2: memref<1x1xsi16, 1> , %arg3: memref<1x73xsi8, 1>) {
+    %alloc = memref.alloc() {alignment = 8 : i64} : memref<1x73xsi8, 3>
+    %alloc_0 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi16, 3>
+    %alloc_1 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi8, 3>
+    memref.copy %arg1, %alloc_1 : memref<1x1xsi8, 1> to memref<1x1xsi8, 3>
+    memref.copy %arg2, %alloc_0 : memref<1x1xsi16, 1> to memref<1x1xsi16, 3>
+    call @noop(%alloc_1, %alloc_0, %arg0) : 
+        (memref<1x1xsi8, 3>, memref<1x1xsi16, 3>, memref<1x73xsi8, 1>) -> ()
+    return
+  }
+  func.func @main(%arg0: memref<1x73xsi8, 1> , %arg1: memref<1x1xsi8, 1> , %arg2: memref<1x1xsi16, 1>) -> (memref<1x73xsi8, 1> ) {
+    %alloc = memref.alloc() : memref<1x73xsi8, 1>
+    call @fn_return_void_with_unused_argument(%arg0, %arg1, %arg2, %alloc) : (memref<1x73xsi8, 1>, memref<1x1xsi8, 1>, memref<1x1xsi16, 1>, memref<1x73xsi8, 1>) -> ()
+    return %alloc : memref<1x73xsi8, 1>
+  }
+}
+

llvmbot · 2025-07-13T14:58:18Z

@llvm/pr-subscribers-mlir

Author: None (ronigoldman22)

Changes

This patch fixes a bug in the RemoveDeadValues pass where unused function arguments were not removed from the function signature in an edge case where the function returns void.
A corresponding test was added to the MLIR LIT test suite to cover this case.

Full diff: https://github.com/llvm/llvm-project/pull/148437.diff

2 Files Affected:

(modified) mlir/lib/Transforms/RemoveDeadValues.cpp (+2-2)
(modified) mlir/test/Transforms/remove-dead-values.mlir (+40)

diff --git a/mlir/lib/Transforms/RemoveDeadValues.cpp b/mlir/lib/Transforms/RemoveDeadValues.cpp
index 608bdcb948176..ddd5f2ba1a7b7 100644
--- a/mlir/lib/Transforms/RemoveDeadValues.cpp
+++ b/mlir/lib/Transforms/RemoveDeadValues.cpp
@@ -345,8 +345,6 @@ static void processFuncOp(FunctionOpInterface funcOp, Operation *module,
   // since it forwards only to non-live value(s) (%1#1).
   Operation *lastReturnOp = funcOp.back().getTerminator();
   size_t numReturns = lastReturnOp->getNumOperands();
-  if (numReturns == 0)
-    return;
   BitVector nonLiveRets(numReturns, true);
   for (SymbolTable::SymbolUse use : uses) {
     Operation *callOp = use.getUser();
@@ -368,6 +366,8 @@ static void processFuncOp(FunctionOpInterface funcOp, Operation *module,
   cl.functions.push_back({funcOp, nonLiveArgs, nonLiveRets});
 
   // Do (5) and (6).
+  if (numReturns == 0)
+    return;
   for (SymbolTable::SymbolUse use : uses) {
     Operation *callOp = use.getUser();
     assert(isa<CallOpInterface>(callOp) && "expected a call-like user");
diff --git a/mlir/test/Transforms/remove-dead-values.mlir b/mlir/test/Transforms/remove-dead-values.mlir
index 3af95db3c0e24..dcaeefc3f2a6f 100644
--- a/mlir/test/Transforms/remove-dead-values.mlir
+++ b/mlir/test/Transforms/remove-dead-values.mlir
@@ -548,3 +548,43 @@ func.func @test_atomic_yield(%I: memref<10xf32>, %idx : index) {
   func.return
 }
 
+// CHECK-LABEL: module @return_void_with_unused_argument {
+// CHECK-LABEL:  func.func private @noop() {
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+// CHECK-LABEL:  func.func private @fn_return_void_with_unused_argument(%arg0: memref<1x1xsi8, 1>, %arg1: memref<1x1xsi16, 1>) {
+// CHECK-NEXT:    %alloc = memref.alloc() {alignment = 8 : i64} : memref<1x73xsi8, 3>
+// CHECK-NEXT:    %alloc_0 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi16, 3>
+// CHECK-NEXT:    %alloc_1 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi8, 3>
+// CHECK-NEXT:    memref.copy %arg0, %alloc_1 : memref<1x1xsi8, 1> to memref<1x1xsi8, 3>
+// CHECK-NEXT:    memref.copy %arg1, %alloc_0 : memref<1x1xsi16, 1> to memref<1x1xsi16, 3>
+// CHECK-NEXT:    call @noop() : () -> ()
+// CHECK-NEXT:    return
+// CHECK-NEXT:  }
+// CHECK-LABEL:  func.func @main(%arg0: memref<1x73xsi8, 1>, %arg1: memref<1x1xsi8, 1>, %arg2: memref<1x1xsi16, 1>) -> memref<1x73xsi8, 1> {
+// CHECK-NEXT:    %alloc = memref.alloc() : memref<1x73xsi8, 1>
+// CHECK-NEXT:    call @fn_return_void_with_unused_argument(%arg1, %arg2) : (memref<1x1xsi8, 1>, memref<1x1xsi16, 1>) -> ()
+// CHECK-NEXT:    return %alloc : memref<1x73xsi8, 1>
+// CHECK-NEXT:  }
+// CHECK-NEXT:}
+module @return_void_with_unused_argument {
+  func.func private @noop(%arg0: memref<1x1xsi8, 3>, %arg1: memref<1x1xsi16, 3>, %arg2: memref<1x73xsi8, 1>) {
+    return
+  }
+  func.func private @fn_return_void_with_unused_argument(%arg0: memref<1x73xsi8, 1> , %arg1: memref<1x1xsi8, 1> , %arg2: memref<1x1xsi16, 1> , %arg3: memref<1x73xsi8, 1>) {
+    %alloc = memref.alloc() {alignment = 8 : i64} : memref<1x73xsi8, 3>
+    %alloc_0 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi16, 3>
+    %alloc_1 = memref.alloc() {alignment = 8 : i64} : memref<1x1xsi8, 3>
+    memref.copy %arg1, %alloc_1 : memref<1x1xsi8, 1> to memref<1x1xsi8, 3>
+    memref.copy %arg2, %alloc_0 : memref<1x1xsi16, 1> to memref<1x1xsi16, 3>
+    call @noop(%alloc_1, %alloc_0, %arg0) : 
+        (memref<1x1xsi8, 3>, memref<1x1xsi16, 3>, memref<1x73xsi8, 1>) -> ()
+    return
+  }
+  func.func @main(%arg0: memref<1x73xsi8, 1> , %arg1: memref<1x1xsi8, 1> , %arg2: memref<1x1xsi16, 1>) -> (memref<1x73xsi8, 1> ) {
+    %alloc = memref.alloc() : memref<1x73xsi8, 1>
+    call @fn_return_void_with_unused_argument(%arg0, %arg1, %arg2, %alloc) : (memref<1x73xsi8, 1>, memref<1x1xsi8, 1>, memref<1x1xsi16, 1>, memref<1x73xsi8, 1>) -> ()
+    return %alloc : memref<1x73xsi8, 1>
+  }
+}
+

ftynse

Thanks for your patch, please make the test minimal and only checking the output relevant to the behavior being changed. In particular, never pattern-match specific SSA value names. See https://mlir.llvm.org/getting_started/TestingGuide/#contributor-guidelines.

badumbatish · 2025-07-17T11:39:38Z

does MLIR encourage pre-committing tests?

joker-eph · 2025-07-17T13:10:50Z

Not particularly, what would you do differently here? Is this codified anywhere in LLVM doc?

badumbatish · 2025-07-17T13:43:27Z

Not particularly, what would you do differently here? Is this codified anywhere in LLVM doc?

I don't have any particular suggestions on it :) i was just browsing MLIR PRs to learn more about it, thanks for the comment :)

joker-eph · 2025-07-17T15:52:51Z

I've seen in some areas of LLVM someone committing a test for a bug which is marked XFAIL, and then the PR fixing it would remove the XFAIL. Is this what you had in mind?

What's difficult with this in MLIR is that we haven't been in the habit of having 1 test per file, but instead grouping them, so you can't just XFAIL one test cases in a larger file.

badumbatish · 2025-07-21T04:08:33Z

@joker-eph oh i've never even thought of that, but with -split-input-file (present in this test file before the commit), wouldn't mlir-opt be running each test in a different process from each other, thus allowing // XFAIL: * to work?

I tested this on my local llvm build and it seems to be the case

joker-eph · 2025-07-23T22:02:06Z

No, from a lit point of view there is a single process. The -split-input-file option is implemented with in-process handling inside mlir-opt. Your test probably does not do what you think it does.

badumbatish · 2025-07-24T23:47:35Z

hmm ok i see

github-actions · 2025-07-25T00:24:23Z

@ronigoldman22 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!

This patch fixes a bug in the RemoveDeadValues pass where unused function arguments were not removed from the function signature in an edge case where the function returns void. A corresponding test was added to the MLIR LIT test suite to cover this case.

fix bug in remove-dead-values pass and add test for the case

b1564d0

llvmbot added mlir:core MLIR Core Infrastructure mlir labels Jul 13, 2025

ftynse requested changes Jul 13, 2025

View reviewed changes

ronigoldman22 force-pushed the fix-remove-dead-values-bug branch from 8dc188b to 6ee6862 Compare July 14, 2025 13:18

joker-eph approved these changes Jul 14, 2025

View reviewed changes

fix lit test

e116811

ronigoldman22 force-pushed the fix-remove-dead-values-bug branch from 6ee6862 to e116811 Compare July 14, 2025 13:24

joker-eph merged commit 9f724d0 into llvm:main Jul 25, 2025
9 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix Bug in RemoveDeadValues Pass #148437

Fix Bug in RemoveDeadValues Pass #148437

Uh oh!

ronigoldman22 commented Jul 13, 2025

Uh oh!

github-actions bot commented Jul 13, 2025

Uh oh!

llvmbot commented Jul 13, 2025

Uh oh!

llvmbot commented Jul 13, 2025

Uh oh!

ftynse left a comment

Uh oh!

badumbatish commented Jul 17, 2025

Uh oh!

joker-eph commented Jul 17, 2025

Uh oh!

badumbatish commented Jul 17, 2025 •

edited

Loading

Uh oh!

joker-eph commented Jul 17, 2025

Uh oh!

badumbatish commented Jul 21, 2025

Uh oh!

joker-eph commented Jul 23, 2025

Uh oh!

badumbatish commented Jul 24, 2025

Uh oh!

Uh oh!

github-actions bot commented Jul 25, 2025

Uh oh!

Uh oh!

Fix Bug in RemoveDeadValues Pass #148437

Fix Bug in RemoveDeadValues Pass #148437

Uh oh!

Conversation

ronigoldman22 commented Jul 13, 2025

Uh oh!

github-actions bot commented Jul 13, 2025

Uh oh!

llvmbot commented Jul 13, 2025

Uh oh!

llvmbot commented Jul 13, 2025

Uh oh!

ftynse left a comment

Choose a reason for hiding this comment

Uh oh!

badumbatish commented Jul 17, 2025

Uh oh!

joker-eph commented Jul 17, 2025

Uh oh!

badumbatish commented Jul 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joker-eph commented Jul 17, 2025

Uh oh!

badumbatish commented Jul 21, 2025

Uh oh!

joker-eph commented Jul 23, 2025

Uh oh!

badumbatish commented Jul 24, 2025

Uh oh!

Uh oh!

github-actions bot commented Jul 25, 2025

Uh oh!

Uh oh!

badumbatish commented Jul 17, 2025 •

edited

Loading