[llvm] [VPlan] Fix crash with inloop fmuladd reductions with blend (PR #131154)

[Florian Hahn via llvm-commits]

================
@@ -1106,6 +1106,90 @@ for.end:
   ret float %muladd
 }
 
+define float @reduction_fmuladd_blend(ptr %a, ptr %b, i64 %n, i1 %c) {
+; CHECK-LABEL: @reduction_fmuladd_blend(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_VEC:%.*]] = and i64 [[N]], -4
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C:%.*]], i64 0
+; CHECK-NEXT:    [[TMP0:%.*]] = xor <4 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison, i1 poison, i1 poison>
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i1> [[TMP0]], <4 x i1> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = fmul <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; CHECK-NEXT:    [[TMP5:%.*]] = select <4 x i1> [[TMP1]], <4 x float> [[TMP4]], <4 x float> splat (float -0.000000e+00)
+; CHECK-NEXT:    [[TMP6:%.*]] = call float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[TMP5]])
+; CHECK-NEXT:    [[TMP7]] = fadd float [[TMP6]], [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
+; CHECK-NEXT:    [[SUM:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SUM_NEXT:%.*]], [[LATCH]] ]
+; CHECK-NEXT:    br i1 [[C]], label [[FOO:%.*]], label [[BAR:%.*]]
+; CHECK:       foo:
+; CHECK-NEXT:    br label [[LATCH]]
+; CHECK:       bar:
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; CHECK-NEXT:    [[TMP9:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[MULADD:%.*]] = tail call float @llvm.fmuladd.f32(float [[TMP10]], float [[TMP9]], float [[SUM]])
+; CHECK-NEXT:    br label [[LATCH]]
+; CHECK:       latch:
+; CHECK-NEXT:    [[SUM_NEXT]] = phi float [ [[SUM]], [[FOO]] ], [ [[MULADD]], [[BAR]] ]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]
+; CHECK:       for.end:
+; CHECK-NEXT:    [[SUM_NEXT_LCSSA:%.*]] = phi float [ [[SUM_NEXT]], [[LATCH]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret float [[SUM_NEXT_LCSSA]]
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %latch ]
+  %sum = phi float [ 0.000000e+00, %entry ], [ %sum.next, %latch ]
+  %arrayidx = getelementptr inbounds float, ptr %a, i64 %iv
+  %0 = load float, ptr %arrayidx, align 4
+  %arrayidx2 = getelementptr inbounds float, ptr %b, i64 %iv
+  %1 = load float, ptr %arrayidx2, align 4
+
----------------
fhahn wrote:

```suggestion
```

https://github.com/llvm/llvm-project/pull/131154