[LV] Honor uniform-after-vectorization in setVectorizedCallDecision.

The legacy cost model always computes the cost for uniforms as cost of
VF = 1, but VPWidenCallRecipes would be created, as
setVectorizedCallDecisions would not consider uniform calls.

Fix setVectorizedCallDecision to set to Scalarize, if the call is
uniform-after-vectorization.

This fixes a bug in VPlan construction uncovered by the VPlan-based
cost model.

Fixes #111040.

Author: fhahn

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -6196,11 +6196,12 @@ void LoopVectorizationCostModel::setVectorizedCallDecision(ElementCount VF) {
           getScalarizationOverhead(CI, VF, CostKind);
 
       ScalarCost = ScalarCallCost * VF.getKnownMinValue() + ScalarizationCost;
-      // Honor ForcedScalars decision.
+      // Honor ForcedScalars and UniformAfterVectorization decisions.
       // TODO: For calls, it might still be more profitable to widen. Use
       // VPlan-based cost model to compare different options.
-      if (VF.isVector() && ForcedScalar != ForcedScalars.end() &&
-          ForcedScalar->second.contains(CI)) {
+      if (VF.isVector() && ((ForcedScalar != ForcedScalars.end() &&
+                             ForcedScalar->second.contains(CI)) ||
+                            isUniformAfterVectorization(CI, VF))) {
         setCallWideningDecision(CI, VF, CM_Scalarize, nullptr,
                                 Intrinsic::not_intrinsic, std::nullopt,
                                 ScalarCost);

llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll

@@ -0,0 +1,120 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -p loop-vectorize -S %s | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; Test for https://github.com/llvm/llvm-project/issues/111040
+define void @smax_call_uniform(ptr %dst, i64 %x) {
+; CHECK-LABEL: define void @smax_call_uniform(
+; CHECK-SAME: ptr [[DST:%.*]], i64 [[X:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[C:%.*]] = icmp ult i8 -68, -69
+; CHECK-NEXT:    [[MUL:%.*]] = mul nuw nsw i64 [[X]], 0
+; CHECK-NEXT:    br i1 true, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_UREM_CONTINUE6:.*]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
+; CHECK-NEXT:    [[TMP1:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i1> [[TMP0]], i32 0
+; CHECK-NEXT:    br i1 [[TMP2]], label %[[PRED_UREM_IF:.*]], label %[[PRED_UREM_CONTINUE:.*]]
+; CHECK:       [[PRED_UREM_IF]]:
+; CHECK-NEXT:    [[TMP3:%.*]] = urem i64 [[MUL]], [[X]]
+; CHECK-NEXT:    br label %[[PRED_UREM_CONTINUE]]
+; CHECK:       [[PRED_UREM_CONTINUE]]:
+; CHECK-NEXT:    [[TMP4:%.*]] = phi i64 [ poison, %[[VECTOR_BODY]] ], [ [[TMP3]], %[[PRED_UREM_IF]] ]
+; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP0]], i32 1
+; CHECK-NEXT:    br i1 [[TMP5]], label %[[PRED_UREM_IF1:.*]], label %[[PRED_UREM_CONTINUE2:.*]]
+; CHECK:       [[PRED_UREM_IF1]]:
+; CHECK-NEXT:    [[TMP6:%.*]] = urem i64 [[MUL]], [[X]]
+; CHECK-NEXT:    br label %[[PRED_UREM_CONTINUE2]]
+; CHECK:       [[PRED_UREM_CONTINUE2]]:
+; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i1> [[TMP1]], i32 0
+; CHECK-NEXT:    br i1 [[TMP7]], label %[[PRED_UREM_IF3:.*]], label %[[PRED_UREM_CONTINUE4:.*]]
+; CHECK:       [[PRED_UREM_IF3]]:
+; CHECK-NEXT:    [[TMP8:%.*]] = urem i64 [[MUL]], [[X]]
+; CHECK-NEXT:    br label %[[PRED_UREM_CONTINUE4]]
+; CHECK:       [[PRED_UREM_CONTINUE4]]:
+; CHECK-NEXT:    [[TMP9:%.*]] = phi i64 [ poison, %[[PRED_UREM_CONTINUE2]] ], [ [[TMP8]], %[[PRED_UREM_IF3]] ]
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i1> [[TMP1]], i32 1
+; CHECK-NEXT:    br i1 [[TMP10]], label %[[PRED_UREM_IF5:.*]], label %[[PRED_UREM_CONTINUE6]]
+; CHECK:       [[PRED_UREM_IF5]]:
+; CHECK-NEXT:    [[TMP11:%.*]] = urem i64 [[MUL]], [[X]]
+; CHECK-NEXT:    br label %[[PRED_UREM_CONTINUE6]]
+; CHECK:       [[PRED_UREM_CONTINUE6]]:
+; CHECK-NEXT:    [[TMP12:%.*]] = tail call i64 @llvm.smax.i64(i64 [[TMP4]], i64 0)
+; CHECK-NEXT:    [[TMP13:%.*]] = tail call i64 @llvm.smax.i64(i64 [[TMP9]], i64 0)
+; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x i1> [[TMP0]], i32 0
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[TMP14]], i64 [[TMP12]], i64 1
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <2 x i1> [[TMP1]], i32 0
+; CHECK-NEXT:    [[PREDPHI7:%.*]] = select i1 [[TMP15]], i64 [[TMP13]], i64 1
+; CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[PREDPHI]], 1
+; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[PREDPHI7]], 1
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP16]]
+; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP17]]
+; CHECK-NEXT:    store i64 0, ptr [[TMP18]], align 8
+; CHECK-NEXT:    store i64 0, ptr [[TMP19]], align 8
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 0
+; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
+; CHECK:       [[LOOP_HEADER]]:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; CHECK-NEXT:    br i1 [[C]], label %[[LOOP_LATCH]], label %[[ELSE:.*]]
+; CHECK:       [[ELSE]]:
+; CHECK-NEXT:    [[REM:%.*]] = urem i64 [[MUL]], [[X]]
+; CHECK-NEXT:    [[SMAX:%.*]] = tail call i64 @llvm.smax.i64(i64 [[REM]], i64 0)
+; CHECK-NEXT:    br label %[[LOOP_LATCH]]
+; CHECK:       [[LOOP_LATCH]]:
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, %[[LOOP_HEADER]] ], [ [[SMAX]], %[[ELSE]] ]
+; CHECK-NEXT:    [[ADD:%.*]] = add i64 [[P]], 1
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i64, ptr [[DST]], i64 [[ADD]]
+; CHECK-NEXT:    store i64 0, ptr [[GEP]], align 8
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 0
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %c = icmp ult i8 -68, -69
+  %mul = mul nsw nuw i64 %x, 0
+  br label %loop.header
+
+loop.header:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 %c, label %loop.latch, label %else
+
+else:
+  %rem = urem i64 %mul, %x
+  %smax = tail call i64 @llvm.smax.i64(i64 %rem, i64 0)
+  br label %loop.latch
+
+loop.latch:
+  %p = phi i64 [ 1, %loop.header ], [ %smax, %else ]
+  %add = add i64 %p, 1
+  %gep = getelementptr i64, ptr %dst, i64 %add
+  store i64 0, ptr %gep, align 8
+  %iv.next = add i64 %iv, 1
+  %ec = icmp eq i64 %iv.next, 0
+  br i1 %ec, label %exit, label %loop.header
+
+exit:
+  ret void
+}
+
+declare i64 @llvm.smax.i64(i64, i64)
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+;.