reference, declarationdefinition
definition → references, declarations, derived classes, virtual overrides
reference to multiple definitions → definitions
unreferenced
    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
   82
   83
   84
   85
   86
   87
   88
   89
   90
   91
   92
   93
   94
   95
   96
   97
   98
   99
  100
  101
  102
  103
  104
  105
  106
  107
  108
  109
  110
  111
  112
  113
  114
  115
  116
  117
  118
  119
  120
  121
  122
  123
  124
  125
  126
  127
  128
  129
  130
  131
  132
  133
  134
  135
  136
  137
  138
  139
  140
  141
  142
  143
  144
  145
  146
  147
  148
  149
  150
  151
  152
  153
  154
  155
  156
  157
  158
  159
  160
  161
  162
  163
  164
  165
  166
  167
  168
  169
  170
  171
  172
  173
  174
  175
  176
  177
  178
  179
  180
  181
  182
  183
  184
  185
  186
  187
  188
  189
  190
  191
  192
  193
  194
  195
  196
  197
  198
  199
  200
  201
  202
  203
  204
  205
  206
  207
  208
  209
  210
  211
  212
  213
  214
  215
  216
  217
  218
  219
; Test memcpy using MVC.
;
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s

declare void @llvm.memcpy.p0i8.p0i8.i32(i8 *nocapture, i8 *nocapture, i32, i1) nounwind
declare void @llvm.memcpy.p0i8.p0i8.i64(i8 *nocapture, i8 *nocapture, i64, i1) nounwind
declare void @foo(i8 *, i8 *)

; Test a no-op move, i32 version.
define void @f1(i8* %dest, i8* %src) {
; CHECK-LABEL: f1:
; CHECK-NOT: %r2
; CHECK-NOT: %r3
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 0, i1 false)
  ret void
}

; Test a no-op move, i64 version.
define void @f2(i8* %dest, i8* %src) {
; CHECK-LABEL: f2:
; CHECK-NOT: %r2
; CHECK-NOT: %r3
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 0, i1 false)
  ret void
}

; Test a 1-byte move, i32 version.
define void @f3(i8* %dest, i8* %src) {
; CHECK-LABEL: f3:
; CHECK: mvc 0(1,%r2), 0(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 1, i1 false)
  ret void
}

; Test a 1-byte move, i64 version.
define void @f4(i8* %dest, i8* %src) {
; CHECK-LABEL: f4:
; CHECK: mvc 0(1,%r2), 0(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1, i1 false)
  ret void
}

; Test the upper range of a single MVC, i32 version.
define void @f5(i8* %dest, i8* %src) {
; CHECK-LABEL: f5:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 256, i1 false)
  ret void
}

; Test the upper range of a single MVC, i64 version.
define void @f6(i8* %dest, i8* %src) {
; CHECK-LABEL: f6:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 256, i1 false)
  ret void
}

; Test the first case that needs two MVCs.
define void @f7(i8* %dest, i8* %src) {
; CHECK-LABEL: f7:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: mvc 256(1,%r2), 256(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 257, i1 false)
  ret void
}

; Test the last-but-one case that needs two MVCs.
define void @f8(i8* %dest, i8* %src) {
; CHECK-LABEL: f8:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: mvc 256(255,%r2), 256(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 511, i1 false)
  ret void
}

; Test the last case that needs two MVCs.
define void @f9(i8* %dest, i8* %src) {
; CHECK-LABEL: f9:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: mvc 256(256,%r2), 256(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 512, i1 false)
  ret void
}

; Test an arbitrary value that uses straight-line code.
define void @f10(i8* %dest, i8* %src) {
; CHECK-LABEL: f10:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: mvc 256(256,%r2), 256(%r3)
; CHECK: mvc 512(256,%r2), 512(%r3)
; CHECK: mvc 768(256,%r2), 768(%r3)
; CHECK: mvc 1024(255,%r2), 1024(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1279, i1 false)
  ret void
}

; ...and again in cases where not all parts are in range of MVC.
define void @f11(i8* %srcbase, i8* %destbase) {
; CHECK-LABEL: f11:
; CHECK: mvc 4000(256,%r2), 3500(%r3)
; CHECK: lay [[NEWDEST:%r[1-5]]], 4256(%r2)
; CHECK: mvc 0(256,[[NEWDEST]]), 3756(%r3)
; CHECK: mvc 256(256,[[NEWDEST]]), 4012(%r3)
; CHECK: lay [[NEWSRC:%r[1-5]]], 4268(%r3)
; CHECK: mvc 512(256,[[NEWDEST]]), 0([[NEWSRC]])
; CHECK: mvc 768(255,[[NEWDEST]]), 256([[NEWSRC]])
; CHECK: br %r14
  %dest = getelementptr i8, i8* %srcbase, i64 4000
  %src = getelementptr i8, i8* %destbase, i64 3500
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1279, i1 false)
  ret void
}

; ...and again with a destination frame base that goes out of range.
define void @f12() {
; CHECK-LABEL: f12:
; CHECK: brasl %r14, foo@PLT
; CHECK: mvc 4076(256,%r15), 2100(%r15)
; CHECK: lay [[NEWDEST:%r[1-5]]], 4332(%r15)
; CHECK: mvc 0(256,[[NEWDEST]]), 2356(%r15)
; CHECK: mvc 256(256,[[NEWDEST]]), 2612(%r15)
; CHECK: mvc 512(256,[[NEWDEST]]), 2868(%r15)
; CHECK: mvc 768(255,[[NEWDEST]]), 3124(%r15)
; CHECK: brasl %r14, foo@PLT
; CHECK: br %r14
  %arr = alloca [6000 x i8]
  %dest = getelementptr [6000 x i8], [6000 x i8] *%arr, i64 0, i64 3900
  %src = getelementptr [6000 x i8], [6000 x i8] *%arr, i64 0, i64 1924
  call void @foo(i8* %dest, i8* %src)
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1279, i1 false)
  call void @foo(i8* %dest, i8* %src)
  ret void
}

; ...and again with a source frame base that goes out of range.
define void @f13() {
; CHECK-LABEL: f13:
; CHECK: brasl %r14, foo@PLT
; CHECK: mvc 200(256,%r15), 3826(%r15)
; CHECK: mvc 456(256,%r15), 4082(%r15)
; CHECK: lay [[NEWSRC:%r[1-5]]], 4338(%r15)
; CHECK: mvc 712(256,%r15), 0([[NEWSRC]])
; CHECK: mvc 968(256,%r15), 256([[NEWSRC]])
; CHECK: mvc 1224(255,%r15), 512([[NEWSRC]])
; CHECK: brasl %r14, foo@PLT
; CHECK: br %r14
  %arr = alloca [6000 x i8]
  %dest = getelementptr [6000 x i8], [6000 x i8] *%arr, i64 0, i64 24
  %src = getelementptr [6000 x i8], [6000 x i8] *%arr, i64 0, i64 3650
  call void @foo(i8* %dest, i8* %src)
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1279, i1 false)
  call void @foo(i8* %dest, i8* %src)
  ret void
}

; Test the last case that is done using straight-line code.
define void @f14(i8* %dest, i8* %src) {
; CHECK-LABEL: f14:
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: mvc 256(256,%r2), 256(%r3)
; CHECK: mvc 512(256,%r2), 512(%r3)
; CHECK: mvc 768(256,%r2), 768(%r3)
; CHECK: mvc 1024(256,%r2), 1024(%r3)
; CHECK: mvc 1280(256,%r2), 1280(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1536, i1 false)
  ret void
}

; Test the first case that is done using a loop.
define void @f15(i8* %dest, i8* %src) {
; CHECK-LABEL: f15:
; CHECK: lghi [[COUNT:%r[0-5]]], 6
; CHECK: [[LABEL:\.L[^:]*]]:
; CHECK: pfd 2, 768(%r2)
; CHECK: mvc 0(256,%r2), 0(%r3)
; CHECK: la %r2, 256(%r2)
; CHECK: la %r3, 256(%r3)
; CHECK: brctg [[COUNT]], [[LABEL]]
; CHECK: mvc 0(1,%r2), 0(%r3)
; CHECK: br %r14
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1537, i1 false)
  ret void
}

; ...and again with frame bases, where the base must be loaded into a
; register before the loop.
define void @f16() {
; CHECK-LABEL: f16:
; CHECK: brasl %r14, foo@PLT
; CHECK-DAG: lghi [[COUNT:%r[0-5]]], 6
; CHECK-DAG: la [[BASE:%r[0-5]]], 160(%r15)
; CHECK: [[LABEL:\.L[^:]*]]:
; CHECK: pfd 2, 2368([[BASE]])
; CHECK: mvc 1600(256,[[BASE]]), 0([[BASE]])
; CHECK: la [[BASE]], 256([[BASE]])
; CHECK: brctg [[COUNT]], [[LABEL]]
; CHECK: mvc 1600(1,[[BASE]]), 0([[BASE]])
; CHECK: brasl %r14, foo@PLT
; CHECK: br %r14
  %arr = alloca [3200 x i8]
  %dest = getelementptr [3200 x i8], [3200 x i8] *%arr, i64 0, i64 1600
  %src = getelementptr [3200 x i8], [3200 x i8] *%arr, i64 0, i64 0
  call void @foo(i8* %dest, i8* %src)
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dest, i8* %src, i64 1537, i1 false)
  call void @foo(i8* %dest, i8* %src)
  ret void
}