[SCEV] Don't add predicates already implied by UnionPredicate. (#93397)

Update SCEVUnionPredicate::add to only add predicates from another union
predicate, if they aren't alread implied by the union predicate we add
them to.

Note that there exists logic elsewhere to avoid adding predicates if
they are already implied, but this logic misses cases when only some
predicates of a union predicate are implied by the current set of
predicates.

PR: #93397

Author: fhahn

llvm/lib/Analysis/ScalarEvolution.cpp

@@ -14751,7 +14751,9 @@ void SCEVUnionPredicate::add(const SCEVPredicate *N) {
     return;
   }
 
-  Preds.push_back(N);
+  // Only add predicate if it is not already implied by this union predicate.
+  if (!implies(N))
+    Preds.push_back(N);
 }
 
 PredicatedScalarEvolution::PredicatedScalarEvolution(ScalarEvolution &SE,

llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll

@@ -152,7 +152,6 @@ for.end:                                          ; preds = %for.cond
 
 @h = global i64 0
 
-; TODO: Currently we generate SCEV check code for the same predicate twice.
 define void @implied_wrap_predicate(ptr %A, ptr %B, ptr %C) {
 ; CHECK-LABEL: define void @implied_wrap_predicate
 ; CHECK-SAME: (ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]]) {
@@ -184,17 +183,11 @@ define void @implied_wrap_predicate(ptr %A, ptr %B, ptr %C) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp ult i16 [[TMP15]], 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp ugt i64 [[TMP8]], 65535
 ; CHECK-NEXT:    [[TMP18:%.*]] = or i1 [[TMP16]], [[TMP17]]
-; CHECK-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP8]] to i16
-; CHECK-NEXT:    [[TMP20:%.*]] = add i16 1, [[TMP19]]
-; CHECK-NEXT:    [[TMP21:%.*]] = icmp ult i16 [[TMP20]], 1
-; CHECK-NEXT:    [[TMP22:%.*]] = icmp ugt i64 [[TMP8]], 65535
-; CHECK-NEXT:    [[TMP23:%.*]] = or i1 [[TMP21]], [[TMP22]]
-; CHECK-NEXT:    [[TMP24:%.*]] = or i1 [[TMP13]], [[TMP18]]
-; CHECK-NEXT:    [[TMP25:%.*]] = or i1 [[TMP24]], [[TMP23]]
-; CHECK-NEXT:    br i1 [[TMP25]], label [[SCALAR_PH]], label [[VECTOR_MEMCHECK:%.*]]
+; CHECK-NEXT:    [[TMP19:%.*]] = or i1 [[TMP13]], [[TMP18]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[SCALAR_PH]], label [[VECTOR_MEMCHECK:%.*]]
 ; CHECK:       vector.memcheck:
-; CHECK-NEXT:    [[TMP26:%.*]] = sub i64 [[C2]], [[A3]]
-; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP26]], 32
+; CHECK-NEXT:    [[TMP20:%.*]] = sub i64 [[C2]], [[A3]]
+; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP20]], 32
 ; CHECK-NEXT:    br i1 [[DIFF_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP4]], 4
@@ -206,16 +199,16 @@ define void @implied_wrap_predicate(ptr %A, ptr %B, ptr %C) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
-; CHECK-NEXT:    [[TMP27:%.*]] = add i64 [[OFFSET_IDX]], 0
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i64, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP29]], align 4
-; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i64, ptr [[C]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i64, ptr [[TMP30]], i32 0
-; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP31]], align 4
+; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP21]]
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr i64, ptr [[TMP22]], i32 0
+; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP23]], align 4
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i64, ptr [[C]], i64 [[TMP21]]
+; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i64, ptr [[TMP24]], i32 0
+; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP25]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]