File: memcpy-to-load.ll

package info (click to toggle)
llvm-toolchain-17 1%3A17.0.6-22
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 1,799,624 kB
  • sloc: cpp: 6,428,607; ansic: 1,383,196; asm: 793,408; python: 223,504; objc: 75,364; f90: 60,502; lisp: 33,869; pascal: 15,282; sh: 9,684; perl: 7,453; ml: 4,937; awk: 3,523; makefile: 2,889; javascript: 2,149; xml: 888; fortran: 619; cs: 573
file content (81 lines) | stat: -rw-r--r-- 3,063 bytes parent folder | download | duplicates (2)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S                         | FileCheck %s
; RUN: opt < %s -passes=instcombine -S -data-layout=n32        | FileCheck %s
; RUN: opt < %s -passes=instcombine -S -data-layout=n32:64     | FileCheck %s
; RUN: opt < %s -passes=instcombine -S -data-layout=n32:64:128 | FileCheck %s

declare void @llvm.memcpy.p0.p0.i32(ptr nocapture, ptr nocapture, i32, i1) nounwind

; memcpy can be expanded inline with load/store (based on the datalayout?)

define void @copy_1_byte(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_1_byte(
; CHECK-NEXT:    [[TMP1:%.*]] = load i8, ptr [[S:%.*]], align 1
; CHECK-NEXT:    store i8 [[TMP1]], ptr [[D:%.*]], align 1
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 1, i1 false)
  ret void
}

define void @copy_2_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_2_bytes(
; CHECK-NEXT:    [[TMP1:%.*]] = load i16, ptr [[S:%.*]], align 1
; CHECK-NEXT:    store i16 [[TMP1]], ptr [[D:%.*]], align 1
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 2, i1 false)
  ret void
}

; We don't expand small non-power-of-2. Should we? Might be a target-dependent choice.

define void @copy_3_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_3_bytes(
; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr noundef nonnull align 1 dereferenceable(3) [[D:%.*]], ptr noundef nonnull align 1 dereferenceable(3) [[S:%.*]], i32 3, i1 false)
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 3, i1 false)
  ret void
}

define void @copy_4_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_4_bytes(
; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[S:%.*]], align 1
; CHECK-NEXT:    store i32 [[TMP1]], ptr [[D:%.*]], align 1
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 4, i1 false)
  ret void
}

; We don't expand small non-power-of-2. Should we? Might be a target-dependent choice.

define void @copy_5_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_5_bytes(
; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr noundef nonnull align 1 dereferenceable(5) [[D:%.*]], ptr noundef nonnull align 1 dereferenceable(5) [[S:%.*]], i32 5, i1 false)
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 5, i1 false)
  ret void
}

define void @copy_8_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_8_bytes(
; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[S:%.*]], align 1
; CHECK-NEXT:    store i64 [[TMP1]], ptr [[D:%.*]], align 1
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 8, i1 false)
  ret void
}

define void @copy_16_bytes(ptr %d, ptr %s) {
; CHECK-LABEL: @copy_16_bytes(
; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr noundef nonnull align 1 dereferenceable(16) [[D:%.*]], ptr noundef nonnull align 1 dereferenceable(16) [[S:%.*]], i32 16, i1 false)
; CHECK-NEXT:    ret void
;
  call void @llvm.memcpy.p0.p0.i32(ptr %d, ptr %s, i32 16, i1 false)
  ret void
}