llvm/test/Transforms/LoopVectorize/find-last-iv-sinkable-expr-epilogue.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6
; RUN: opt -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -epilogue-vectorization-force-VF=4 -S %s | FileCheck %s

; Tests for FindLastIV reductions where the selected value is an expression
; of the induction variable with epilogue vectorization enabled.

define i64 @findlast_mul_expr_epilogue(ptr %a, i64 %n) {
; CHECK-LABEL: define i64 @findlast_mul_expr_epilogue(
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP12:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX6]]
; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD9]], splat (i64 42)
; CHECK-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i64> [[VEC_IND7]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX6]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT11]] = add nuw nsw <4 x i64> [[VEC_IND7]], splat (i64 4)
; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vector.reduce.smax.v4i64(<4 x i64> [[TMP12]])
; CHECK-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 3
; CHECK-NEXT:    [[TMP16:%.*]] = icmp ne i64 [[TMP14]], -9223372036854775808
; CHECK-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i64 [[TMP15]], i64 -1
; CHECK-NEXT:    [[CMP_N12:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N12]], label %[[EXIT1:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP17]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[LOAD:%.*]] = load i64, ptr [[GEP]], align 8
; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i64 [[LOAD]], 42
; CHECK-NEXT:    [[EXPR:%.*]] = mul i64 [[IV]], 3
; CHECK-NEXT:    [[SEL]] = select i1 [[CMP]], i64 [[EXPR]], i64 [[RDX]]
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EXIT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EXIT]], label %[[EXIT1]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[EXIT1]]:
; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i64 [ [[SEL]], %[[LOOP]] ], [ [[TMP17]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[SEL_LCSSA]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i64 [ -1, %entry ], [ %sel, %loop ]
  %gep = getelementptr inbounds i64, ptr %a, i64 %iv
  %load = load i64, ptr %gep, align 8
  %cmp = icmp eq i64 %load, 42
  %expr = mul i64 %iv, 3
  %sel = select i1 %cmp, i64 %expr, i64 %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, %n
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i64 %sel
}

define i64 @findlast_shl_expr_epilogue(ptr %a, i64 %n) {
; CHECK-LABEL: define i64 @findlast_shl_expr_epilogue(
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP12:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX6]]
; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD9]], splat (i64 42)
; CHECK-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i64> [[VEC_IND7]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX6]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT11]] = add nuw nsw <4 x i64> [[VEC_IND7]], splat (i64 4)
; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vector.reduce.smax.v4i64(<4 x i64> [[TMP12]])
; CHECK-NEXT:    [[TMP15:%.*]] = shl i64 [[TMP14]], 2
; CHECK-NEXT:    [[TMP16:%.*]] = icmp ne i64 [[TMP14]], -9223372036854775808
; CHECK-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i64 [[TMP15]], i64 -1
; CHECK-NEXT:    [[CMP_N12:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N12]], label %[[EXIT1:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP17]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[LOAD:%.*]] = load i64, ptr [[GEP]], align 8
; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i64 [[LOAD]], 42
; CHECK-NEXT:    [[EXPR:%.*]] = shl i64 [[IV]], 2
; CHECK-NEXT:    [[SEL]] = select i1 [[CMP]], i64 [[EXPR]], i64 [[RDX]]
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EXIT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EXIT]], label %[[EXIT1]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[EXIT1]]:
; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i64 [ [[SEL]], %[[LOOP]] ], [ [[TMP17]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[SEL_LCSSA]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i64 [ -1, %entry ], [ %sel, %loop ]
  %gep = getelementptr inbounds i64, ptr %a, i64 %iv
  %load = load i64, ptr %gep, align 8
  %cmp = icmp eq i64 %load, 42
  %expr = shl i64 %iv, 2
  %sel = select i1 %cmp, i64 %expr, i64 %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, %n
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i64 %sel
}

define i64 @findlast_or_expr_epilogue(ptr %a, i64 %n) {
; CHECK-LABEL: define i64 @findlast_or_expr_epilogue(
; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT10:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP12:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX6]]
; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD9]], splat (i64 42)
; CHECK-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i64> [[VEC_IND7]], <4 x i64> [[VEC_PHI]]
; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX6]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT11]] = add nuw nsw <4 x i64> [[VEC_IND7]], splat (i64 4)
; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vector.reduce.smax.v4i64(<4 x i64> [[TMP12]])
; CHECK-NEXT:    [[TMP15:%.*]] = or i64 [[TMP14]], 1
; CHECK-NEXT:    [[TMP16:%.*]] = icmp ne i64 [[TMP14]], -9223372036854775808
; CHECK-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i64 [[TMP15]], i64 -1
; CHECK-NEXT:    [[CMP_N12:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N12]], label %[[EXIT1:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP17]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[LOOP:.*]]
; CHECK:       [[LOOP]]:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    [[LOAD:%.*]] = load i64, ptr [[GEP]], align 8
; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i64 [[LOAD]], 42
; CHECK-NEXT:    [[EXPR:%.*]] = or i64 [[IV]], 1
; CHECK-NEXT:    [[SEL]] = select i1 [[CMP]], i64 [[EXPR]], i64 [[RDX]]
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EXIT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EXIT]], label %[[EXIT1]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK:       [[EXIT1]]:
; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i64 [ [[SEL]], %[[LOOP]] ], [ [[TMP17]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT:    ret i64 [[SEL_LCSSA]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %rdx = phi i64 [ -1, %entry ], [ %sel, %loop ]
  %gep = getelementptr inbounds i64, ptr %a, i64 %iv
  %load = load i64, ptr %gep, align 8
  %cmp = icmp eq i64 %load, 42
  %expr = or i64 %iv, 1
  %sel = select i1 %cmp, i64 %expr, i64 %rdx
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp eq i64 %iv.next, %n
  br i1 %exitcond, label %exit, label %loop

exit:
  ret i64 %sel
}