[Vectorizer] precommit test for miscompilation (#120731)

we generate GEPs that are out of bounds but mark them as "inbound"

Author: fmayer

llvm/test/Transforms/LoopVectorize/X86/drop-inbounds-flags-for-reverse-vector-pointer.ll

@@ -0,0 +1,105 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -passes=loop-vectorize -force-vector-width=4 -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"
+
+; FIXME:  GEP flags on GEPs for reverse vector pointer need to be dropped when folding the tail.
+
+define i1 @fn(ptr %nno) #0 {
+; CHECK-LABEL: define i1 @fn(
+; CHECK-SAME: ptr [[NNO:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 10, i64 9, i64 8, i64 7>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 10, [[INDEX]]
+; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[INDEX]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], splat (i64 10)
+; CHECK-NEXT:    [[TMP2:%.*]] = and <4 x i64> [[VEC_IND]], splat (i64 1)
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i64> [[TMP2]], zeroinitializer
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw i32, ptr [[NNO]], i64 [[TMP22]]
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP23]], i32 0
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 -3
+; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP6]], i32 4, <4 x i1> [[REVERSE]], <4 x i32> poison)
+; CHECK-NEXT:    [[REVERSE1:%.*]] = shufflevector <4 x i32> [[WIDE_MASKED_LOAD]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[TMP7:%.*]] = shl <4 x i32> [[REVERSE1]], splat (i32 1)
+; CHECK-NEXT:    [[TMP8:%.*]] = urem <4 x i32> [[TMP7]], splat (i32 10)
+; CHECK-NEXT:    [[TMP9:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
+; CHECK-NEXT:    [[TMP10:%.*]] = select <4 x i1> [[TMP1]], <4 x i1> [[TMP9]], <4 x i1> zeroinitializer
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> [[REVERSE1]], <4 x i32> [[TMP8]]
+; CHECK-NEXT:    [[TMP11]] = or <4 x i32> [[PREDPHI]], [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP12:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[TMP11]], <4 x i32> [[VEC_PHI]]
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 -4)
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 12
+; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[TMP12]])
+; CHECK-NEXT:    br i1 true, label [[FOR_END36:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -2, [[MIDDLE_BLOCK]] ], [ 10, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP14]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    br label [[FOR_BODY20:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC35:%.*]] ]
+; CHECK-NEXT:    [[SUM_01:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[SUM_1:%.*]], [[FOR_INC35]] ]
+; CHECK-NEXT:    [[REM4:%.*]] = and i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[CMP21:%.*]] = icmp eq i64 [[REM4]], 0
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds nuw i32, ptr [[NNO]], i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[TMP15:%.*]] = load i32, ptr [[GEP]], align 4
+; CHECK-NEXT:    br i1 [[CMP21]], label [[IF_THEN22:%.*]], label [[FOR_INC35]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[MUL:%.*]] = shl i32 [[TMP15]], 1
+; CHECK-NEXT:    [[REM27:%.*]] = urem i32 [[MUL]], 10
+; CHECK-NEXT:    br label [[FOR_INC35]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[REM27_PN:%.*]] = phi i32 [ [[REM27]], [[IF_THEN22]] ], [ [[TMP15]], [[FOR_BODY20]] ]
+; CHECK-NEXT:    [[SUM_1]] = or i32 [[REM27_PN]], [[SUM_01]]
+; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
+; CHECK-NEXT:    [[CMP19_NOT:%.*]] = icmp eq i64 [[INDVARS_IV]], 0
+; CHECK-NEXT:    br i1 [[CMP19_NOT]], label [[FOR_END36]], label [[FOR_BODY20]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[SUM_1_LCSSA:%.*]] = phi i32 [ [[SUM_1]], [[FOR_INC35]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[CMP41:%.*]] = icmp eq i32 [[SUM_1_LCSSA]], 0
+; CHECK-NEXT:    ret i1 [[CMP41]]
+;
+entry:
+  br label %loop.header
+
+loop.header:                                       ; preds = %entry, %loop.latch
+  %iv = phi i64 [ 10, %entry ], [ %iv.next, %loop.latch ]
+  %sum.01 = phi i32 [ 0, %entry ], [ %sum.1, %loop.latch ]
+  %rem4 = and i64 %iv, 1
+  %cmp21 = icmp eq i64 %rem4, 0
+  %gep = getelementptr inbounds nuw i32, ptr %nno, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  br i1 %cmp21, label %if.then, label %loop.latch
+
+if.then:                                        ; preds = %loop.header
+  %mul = shl i32 %0, 1
+  %rem27 = urem i32 %mul, 10
+  br label %loop.latch
+
+loop.latch:                                        ; preds = %loop.header, %if.then
+  %rem27.pn = phi i32 [ %rem27, %if.then ], [ %0, %loop.header ]
+  %sum.1 = or i32 %rem27.pn, %sum.01
+  %iv.next = add nsw i64 %iv, -1
+  %cmp19.not = icmp eq i64 %iv, 0
+  br i1 %cmp19.not, label %exit, label %loop.header
+
+exit:                                        ; preds = %loop.latch
+  %sum.1.lcssa = phi i32 [ %sum.1, %loop.latch ]
+  %cmp41 = icmp eq i32 %sum.1.lcssa, 0
+  ret i1 %cmp41
+}
+
+attributes #0 = { "target-features"="+avx" }