[SCEVExpander] Don't try to reuse SCEVUnknown values

[nikic]

The expansion of a SCEVUnknown is trivial (it's just the wrapped value). If we try to reuse an existing value it might be a more complex expression that simplifies to the SCEVUnknown.

This is inspired by #114879, because SCEVExpander replacing a constant with a phi node is just silly. (I don't consider this a fix for that issue though.)

[llvmbot]

@llvm/pr-subscribers-llvm-transforms

Author: Nikita Popov (nikic)

Changes

The expansion of a SCEVUnknown is trivial (it's just the wrapped value). If we try to reuse an existing value it might be a more complex expression that simplifies to the SCEVUnknown.

This is inspired by #114879, because SCEVExpander replacing a constant with a phi node is just silly. (I don't consider this a fix for that issue though.)

---

Full diff: https://github.com/llvm/llvm-project/pull/115141.diff

4 Files Affected:

• (modified) llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp (+2-2)
• (modified) llvm/test/Transforms/LoopStrengthReduce/Power/incomplete-phi.ll (+1-3)
• (modified) llvm/test/Transforms/LoopStrengthReduce/lsr-term-fold.ll (+2-2)
• (modified) llvm/test/Transforms/LoopVectorize/pr45259.ll (+5-3)

diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index 39da38e4918176..791d528823972d 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -1467,8 +1467,8 @@ Value *SCEVExpander::FindValueInExprValueMap(
   if (!CanonicalMode && SE.containsAddRecurrence(S))
     return nullptr;
 
-  // If S is a constant, it may be worse to reuse an existing Value.
-  if (isa<SCEVConstant>(S))
+  // If S is a constant or unknown, it may be worse to reuse an existing Value.
+  if (isa<SCEVConstant>(S) || isa<SCEVUnknown>(S))
     return nullptr;
 
   for (Value *V : SE.getSCEVValues(S)) {
diff --git a/llvm/test/Transforms/LoopStrengthReduce/Power/incomplete-phi.ll b/llvm/test/Transforms/LoopStrengthReduce/Power/incomplete-phi.ll
index c57761c1a01f14..53aac1d9cf7f8d 100644
--- a/llvm/test/Transforms/LoopStrengthReduce/Power/incomplete-phi.ll
+++ b/llvm/test/Transforms/LoopStrengthReduce/Power/incomplete-phi.ll
@@ -21,12 +21,10 @@ define void @foo(ptr %arg) {
 ; CHECK-LABEL: define void @foo(
 ; CHECK-SAME: ptr [[ARG:%.*]]) {
 ; CHECK-NEXT:  bb:
-; CHECK-NEXT:    [[I:%.*]] = getelementptr [0 x %0], ptr [[ARG]], i64 0, i64 -1
-; CHECK-NEXT:    [[I2:%.*]] = getelementptr i8, ptr [[I]], i64 4
 ; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[ARG]], i64 396
 ; CHECK-NEXT:    br label [[BB3:%.*]]
 ; CHECK:       bb3:
-; CHECK-NEXT:    [[LSR_IV7:%.*]] = phi ptr [ [[SCEVGEP8:%.*]], [[BB18:%.*]] ], [ [[I2]], [[BB:%.*]] ]
+; CHECK-NEXT:    [[LSR_IV7:%.*]] = phi ptr [ [[SCEVGEP8:%.*]], [[BB18:%.*]] ], [ [[ARG]], [[BB:%.*]] ]
 ; CHECK-NEXT:    [[LSR_IV5:%.*]] = phi i64 [ [[LSR_IV_NEXT6:%.*]], [[BB18]] ], [ 4, [[BB]] ]
 ; CHECK-NEXT:    [[LSR_IV1:%.*]] = phi ptr [ [[SCEVGEP2:%.*]], [[BB18]] ], [ [[SCEVGEP]], [[BB]] ]
 ; CHECK-NEXT:    br i1 true, label [[BB22_PREHEADER:%.*]], label [[BB9_PREHEADER:%.*]]
diff --git a/llvm/test/Transforms/LoopStrengthReduce/lsr-term-fold.ll b/llvm/test/Transforms/LoopStrengthReduce/lsr-term-fold.ll
index 6f34dc843ae1ee..3c53befa67e230 100644
--- a/llvm/test/Transforms/LoopStrengthReduce/lsr-term-fold.ll
+++ b/llvm/test/Transforms/LoopStrengthReduce/lsr-term-fold.ll
@@ -113,7 +113,7 @@ for.end:                                          ; preds = %for.body
 define void @ptr_of_ptr_addrec(ptr %ptrptr, i32 %length) {
 ; CHECK-LABEL: @ptr_of_ptr_addrec(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[START_PTRPTR:%.*]] = getelementptr ptr, ptr [[PTRPTR:%.*]]
+; CHECK-NEXT:    [[START_PTRPTR1:%.*]] = getelementptr inbounds ptr, ptr [[START_PTRPTR:%.*]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[LENGTH:%.*]], -1
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
@@ -121,7 +121,7 @@ define void @ptr_of_ptr_addrec(ptr %ptrptr, i32 %length) {
 ; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[START_PTRPTR]], i64 [[TMP3]]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
-; CHECK-NEXT:    [[IT_04:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY]] ], [ [[START_PTRPTR]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[IT_04:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY]] ], [ [[START_PTRPTR1]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[IT_04]], align 8
 ; CHECK-NEXT:    tail call void @foo(ptr [[TMP4]])
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr ptr, ptr [[IT_04]], i64 1
diff --git a/llvm/test/Transforms/LoopVectorize/pr45259.ll b/llvm/test/Transforms/LoopVectorize/pr45259.ll
index 008971697775e4..9a154709aab813 100644
--- a/llvm/test/Transforms/LoopVectorize/pr45259.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr45259.ll
@@ -14,12 +14,14 @@ define i8 @widget(ptr %arr, i8 %t9) {
 ; CHECK-NEXT:    br i1 [[C]], label [[FOR_PREHEADER:%.*]], label [[BB6]]
 ; CHECK:       for.preheader:
 ; CHECK-NEXT:    [[T1_0_LCSSA:%.*]] = phi ptr [ [[T1_0]], [[BB6]] ]
+; CHECK-NEXT:    [[T1_0_LCSSA4:%.*]] = phi ptr [ [[T1_0]], [[BB6]] ]
+; CHECK-NEXT:    [[T1_0_LCSSA1:%.*]] = phi ptr [ [[T1_0]], [[BB6]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[ARR1]] to i32
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub i32 0, [[TMP0]]
 ; CHECK-NEXT:    [[T1_0_LCSSA3:%.*]] = ptrtoint ptr [[T1_0_LCSSA]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[T1_0_LCSSA3]] to i32
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[TMP1]], [[TMP2]]
-; CHECK-NEXT:    [[T1_0_LCSSA2:%.*]] = ptrtoint ptr [[T1_0_LCSSA]] to i64
+; CHECK-NEXT:    [[T1_0_LCSSA2:%.*]] = ptrtoint ptr [[T1_0_LCSSA4]] to i64
 ; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 4
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
 ; CHECK:       vector.scevcheck:
@@ -65,8 +67,8 @@ define i8 @widget(ptr %arr, i8 %t9) {
 ; CHECK-NEXT:    [[T3_I:%.*]] = icmp slt i8 [[IV_NEXT]], [[T9]]
 ; CHECK-NEXT:    [[T3_I8:%.*]] = zext i1 [[T3_I]] to i8
 ; CHECK-NEXT:    store i8 [[T3_I8]], ptr [[PTR]], align 1
-; CHECK-NEXT:    [[EC:%.*]] = icmp eq ptr [[T1_0_LCSSA]], [[PTR]]
-; CHECK-NEXT:    br i1 [[EC]], label [[FOR_EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq ptr [[T1_0_LCSSA1]], [[PTR]]
+; CHECK-NEXT:    br i1 [[EC]], label [[FOR_EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       for.exit:
 ; CHECK-NEXT:    [[IV_NEXT_LCSSA:%.*]] = phi i8 [ [[IV_NEXT]], [[FOR_BODY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    ret i8 [[IV_NEXT_LCSSA]]

[nikic]

Downside is that we might generate more duplicate LCSSA phis, but this seems like an independent SCEVExpander problem.

[dtcxzyw]

LG. No changes on real-world code.