xref: /llvm-project/llvm/test/CodeGen/RISCV/unroll-loop-cse.ll (revision 2e57156fead3c27136aa2f66caad9e4b5c890ca1)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s
3; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s | FileCheck %s
4
5@x = global [6 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4, i32 5], align 4
6@check = global [6 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4, i32 5], align 4
7
8; This test case checks whether the base address of an array is repeatedly
9; rematerialised within a unrolled loop.
10define signext i32 @unroll_loop_cse() {
11; CHECK-LABEL: unroll_loop_cse:
12; CHECK:       # %bb.0:
13; CHECK-NEXT:    lui a1, %hi(x)
14; CHECK-NEXT:    lw a3, %lo(x)(a1)
15; CHECK-NEXT:    lui a2, %hi(check)
16; CHECK-NEXT:    lw a4, %lo(check)(a2)
17; CHECK-NEXT:    li a0, 1
18; CHECK-NEXT:    bne a3, a4, .LBB0_6
19; CHECK-NEXT:  # %bb.1:
20; CHECK-NEXT:    addi a1, a1, %lo(x)
21; CHECK-NEXT:    lw a1, 4(a1)
22; CHECK-NEXT:    addi a2, a2, %lo(check)
23; CHECK-NEXT:    lw a2, 4(a2)
24; CHECK-NEXT:    bne a1, a2, .LBB0_6
25; CHECK-NEXT:  # %bb.2:
26; CHECK-NEXT:    lui a1, %hi(x)
27; CHECK-NEXT:    addi a1, a1, %lo(x)
28; CHECK-NEXT:    lw a3, 8(a1)
29; CHECK-NEXT:    lui a2, %hi(check)
30; CHECK-NEXT:    addi a2, a2, %lo(check)
31; CHECK-NEXT:    lw a4, 8(a2)
32; CHECK-NEXT:    bne a3, a4, .LBB0_6
33; CHECK-NEXT:  # %bb.3:
34; CHECK-NEXT:    lw a1, 12(a1)
35; CHECK-NEXT:    lw a2, 12(a2)
36; CHECK-NEXT:    bne a1, a2, .LBB0_6
37; CHECK-NEXT:  # %bb.4:
38; CHECK-NEXT:    lui a1, %hi(x)
39; CHECK-NEXT:    addi a1, a1, %lo(x)
40; CHECK-NEXT:    lw a3, 16(a1)
41; CHECK-NEXT:    lui a2, %hi(check)
42; CHECK-NEXT:    addi a2, a2, %lo(check)
43; CHECK-NEXT:    lw a4, 16(a2)
44; CHECK-NEXT:    bne a3, a4, .LBB0_6
45; CHECK-NEXT:  # %bb.5:
46; CHECK-NEXT:    lw a0, 20(a1)
47; CHECK-NEXT:    lw a1, 20(a2)
48; CHECK-NEXT:    xor a0, a0, a1
49; CHECK-NEXT:    snez a0, a0
50; CHECK-NEXT:  .LBB0_6:
51; CHECK-NEXT:    ret
52  %1 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 0), align 4
53  %2 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 0), align 4
54  %3 = icmp eq i32 %1, %2
55  br i1 %3, label %4, label %25
56
574:
58  %5 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 1), align 4
59  %6 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 1), align 4
60  %7 = icmp eq i32 %5, %6
61  br i1 %7, label %8, label %25
62
638:
64  %9 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 2), align 4
65  %10 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 2), align 4
66  %11 = icmp eq i32 %9, %10
67  br i1 %11, label %12, label %25
68
6912:
70  %13 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 3), align 4
71  %14 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 3), align 4
72  %15 = icmp eq i32 %13, %14
73  br i1 %15, label %16, label %25
74
7516:
76  %17 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 4), align 4
77  %18 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 4), align 4
78  %19 = icmp eq i32 %17, %18
79  br i1 %19, label %20, label %25
80
8120:
82  %21 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 5), align 4
83  %22 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 5), align 4
84  %23 = icmp ne i32 %21, %22
85  %24 = zext i1 %23 to i32
86  br label %25
87
8825:
89  %26 = phi i32 [ 1, %0 ], [ 1, %4 ], [ 1, %8 ], [ 1, %12 ], [ 1, %16 ], [ %24, %20 ]
90  ret i32 %26
91}
92
93