; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=memcpyopt -S %s -verify-memoryssa | FileCheck %s

; Redundant memmove.
define i32 @redundant_memmove() {
; CHECK-LABEL: @redundant_memmove(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
  %val = load i32, ptr %array, align 16
  ret i32 %val
}

; Used memmove, buffer is reset to zero.
define i32 @used_memmove_1() {
; CHECK-LABEL: @used_memmove_1(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    store i32 1, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  store i32 1, ptr %array.idx
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
  %val = load i32, ptr %array.idx, align 4
  ret i32 %val
}

; Used memmove.
define i32 @used_memmove_2() {
; CHECK-LABEL: @used_memmove_2(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    store i32 1, ptr [[ARRAY]], align 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  store i32 1, ptr %array
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
  %val = load i32, ptr %array.idx, align 4
  ret i32 %val
}

; Used memmove, buffer clobbered by opaque.
define i32 @used_memmove_3() {
; CHECK-LABEL: @used_memmove_3(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [25 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 100, i1 false)
; CHECK-NEXT:    call void @opaque(ptr [[ARRAY]])
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 96, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [25 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 100, i1 false)
  call void @opaque(ptr %array)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 96, i1 false)
  %val = load i32, ptr %array, align 16
  ret i32 %val
}

; Redundant memmove, not within the same basic block.
define i32 @redundant_memmove_different_bbs() {
; CHECK-LABEL: @redundant_memmove_different_bbs(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    br label [[USE:%.*]]
; CHECK:       use:
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT:    ret i32 [[VAL]]
;
entry:
  %array = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  br label %use

use:                                              ; preds = %entry
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
  %val = load i32, ptr %array, align 16
  ret i32 %val
}

@g_var = global [26 x i32] zeroinitializer, align 16

; Redundant memmove on a global variable.
define ptr @redundant_memmove_memset_global_variable() {
; CHECK-LABEL: @redundant_memmove_memset_global_variable(
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 @g_var, i8 0, i64 104, i1 false)
; CHECK-NEXT:    ret ptr @g_var
;
  call void @llvm.memset.p0.i64(ptr align 16 @g_var, i8 0, i64 104, i1 false)
  call void @llvm.memmove.p0.p0.i64(ptr align 16 @g_var, ptr align 4 getelementptr inbounds nuw (i8, ptr @g_var, i64 4), i64 100, i1 false)
  ret ptr @g_var
}

; Memset only partial.
define i32 @partial_memset() {
; CHECK-LABEL: @partial_memset(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 92
; CHECK-NEXT:    store i32 1, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 26, i1 false)
; CHECK-NEXT:    [[ARRAY_IDX_2:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX_2]], i64 100, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY_IDX]], align 4
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  %array.idx = getelementptr inbounds i8, ptr %array, i64 92
  store i32 1, ptr %array.idx
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 26, i1 false)
  %array.idx.2 = getelementptr inbounds i8, ptr %array, i64 4
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx.2, i64 100, i1 false)
  %val = load i32, ptr %array.idx, align 4
  ret i32 %val
}

; Memset length not constant.
define i32 @memset_length_not_constant(i64 %size) {
; CHECK-LABEL: @memset_length_not_constant(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 [[SIZE:%.*]], i1 false)
; CHECK-NEXT:    [[ARRAY_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY]], i64 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY]], ptr align 4 [[ARRAY_IDX]], i64 100, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY]], align 16
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 %size, i1 false)
  %array.idx = getelementptr inbounds i8, ptr %array, i64 4
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array, ptr align 4 %array.idx, i64 100, i1 false)
  %val = load i32, ptr %array, align 16
  ret i32 %val
}

; Memmove buffer not memset'd, different buffers.
define i32 @memset_memmove_dest_buffers_not_alias() {
; CHECK-LABEL: @memset_memmove_dest_buffers_not_alias(
; CHECK-NEXT:    [[ARRAY:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    [[ARRAY2:%.*]] = alloca [26 x i32], align 16
; CHECK-NEXT:    call void @llvm.memset.p0.i64(ptr align 16 [[ARRAY]], i8 0, i64 104, i1 false)
; CHECK-NEXT:    [[ARRAY2_IDX:%.*]] = getelementptr inbounds i8, ptr [[ARRAY2]], i64 4
; CHECK-NEXT:    call void @llvm.memmove.p0.p0.i64(ptr align 16 [[ARRAY2]], ptr align 4 [[ARRAY2_IDX]], i64 100, i1 false)
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[ARRAY2]], align 16
; CHECK-NEXT:    ret i32 [[VAL]]
;
  %array = alloca [26 x i32], align 16
  %array2 = alloca [26 x i32], align 16
  call void @llvm.memset.p0.i64(ptr align 16 %array, i8 0, i64 104, i1 false)
  %array2.idx = getelementptr inbounds i8, ptr %array2, i64 4
  call void @llvm.memmove.p0.p0.i64(ptr align 16 %array2, ptr align 4 %array2.idx, i64 100, i1 false)
  %val = load i32, ptr %array2, align 16
  ret i32 %val
}

declare void @opaque(ptr)
declare void @llvm.memset.p0.i64(ptr nocapture, i8, i64, i1)
declare void @llvm.memmove.p0.p0.i64(ptr nocapture, ptr nocapture, i64, i1)