1; RUN: opt < %s -instcombine -S -data-layout=e-n32 | FileCheck %s --check-prefix=ALL --check-prefix=LE
2; RUN: opt < %s -instcombine -S -data-layout=E-n32 | FileCheck %s --check-prefix=ALL --check-prefix=BE
3
4declare i32 @memcmp(i8*, i8*, i64)
5
6; The alignment of this constant does not matter. We constant fold the load.
7
8@charbuf = private unnamed_addr constant [4 x i8] [i8 0, i8 0, i8 0, i8 1], align 1
9
10define i1 @memcmp_4bytes_unaligned_constant_i8(i8* align 4 %x) {
11; LE-LABEL: @memcmp_4bytes_unaligned_constant_i8(
12; LE-NEXT:    [[TMP1:%.*]] = bitcast i8* %x to i32*
13; LE-NEXT:    [[LHSV:%.*]] = load i32, i32* [[TMP1]], align 4
14; LE-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[LHSV]], 16777216
15; LE-NEXT:    ret i1 [[TMP2]]
16;
17; BE-LABEL: @memcmp_4bytes_unaligned_constant_i8(
18; BE-NEXT:    [[TMP1:%.*]] = bitcast i8* %x to i32*
19; BE-NEXT:    [[LHSV:%.*]] = load i32, i32* [[TMP1]], align 4
20; BE-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[LHSV]], 1
21; BE-NEXT:    ret i1 [[TMP2]]
22;
23  %call = tail call i32 @memcmp(i8* %x, i8* getelementptr inbounds ([4 x i8], [4 x i8]* @charbuf, i64 0, i64 0), i64 4)
24  %cmpeq0 = icmp eq i32 %call, 0
25  ret i1 %cmpeq0
26}
27
28; We still don't care about alignment of the constant. We are not limited to constant folding only i8 arrays.
29; It doesn't matter if the constant operand is the first operand to the memcmp.
30
31@intbuf_unaligned = private unnamed_addr constant [4 x i16] [i16 1, i16 2, i16 3, i16 4], align 1
32
33define i1 @memcmp_4bytes_unaligned_constant_i16(i8* align 4 %x) {
34; LE-LABEL: @memcmp_4bytes_unaligned_constant_i16(
35; LE-NEXT:    [[TMP1:%.*]] = bitcast i8* %x to i32*
36; LE-NEXT:    [[RHSV:%.*]] = load i32, i32* [[TMP1]], align 4
37; LE-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[RHSV]], 131073
38; LE-NEXT:    ret i1 [[TMP2]]
39;
40; BE-LABEL: @memcmp_4bytes_unaligned_constant_i16(
41; BE-NEXT:    [[TMP1:%.*]] = bitcast i8* %x to i32*
42; BE-NEXT:    [[RHSV:%.*]] = load i32, i32* [[TMP1]], align 4
43; BE-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[RHSV]], 65538
44; BE-NEXT:    ret i1 [[TMP2]]
45;
46  %call = tail call i32 @memcmp(i8* bitcast (i16* getelementptr inbounds ([4 x i16], [4 x i16]* @intbuf_unaligned, i64 0, i64 0) to i8*), i8* %x, i64 4)
47  %cmpeq0 = icmp eq i32 %call, 0
48  ret i1 %cmpeq0
49}
50
51; TODO: Any memcmp where all arguments are constants should be constant folded. Currently, we only handle i8 array constants.
52
53@intbuf = private unnamed_addr constant [2 x i32] [i32 0, i32 1], align 4
54
55define i1 @memcmp_3bytes_aligned_constant_i32(i8* align 4 %x) {
56; ALL-LABEL: @memcmp_3bytes_aligned_constant_i32(
57; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* bitcast (i32* getelementptr inbounds ([2 x i32], [2 x i32]* @intbuf, i64 0, i64 1) to i8*), i8* bitcast ([2 x i32]* @intbuf to i8*), i64 3)
58; ALL-NEXT:    [[CMPEQ0:%.*]] = icmp eq i32 [[CALL]], 0
59; ALL-NEXT:    ret i1 [[CMPEQ0]]
60;
61  %call = tail call i32 @memcmp(i8* bitcast (i32* getelementptr inbounds ([2 x i32], [2 x i32]* @intbuf, i64 0, i64 1) to i8*), i8* bitcast (i32* getelementptr inbounds ([2 x i32], [2 x i32]* @intbuf, i64 0, i64 0) to i8*), i64 3)
62  %cmpeq0 = icmp eq i32 %call, 0
63  ret i1 %cmpeq0
64}
65
66; A sloppy implementation would infinite loop by recreating the unused instructions.
67
68define i1 @memcmp_4bytes_one_unaligned_i8(i8* align 4 %x, i8* align 1 %y) {
69; ALL-LABEL: @memcmp_4bytes_one_unaligned_i8(
70; ALL-NEXT:    [[CALL:%.*]] = tail call i32 @memcmp(i8* %x, i8* %y, i64 4)
71; ALL-NEXT:    [[CMPEQ0:%.*]] = icmp eq i32 [[CALL]], 0
72; ALL-NEXT:    ret i1 [[CMPEQ0]]
73;
74  %bc = bitcast i8* %x to i32*
75  %lhsv = load i32, i32* %bc
76  %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 4)
77  %cmpeq0 = icmp eq i32 %call, 0
78  ret i1 %cmpeq0
79}
80
81