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
; REQUIRES: asserts
; RUN: llc < %s -mtriple=arm64-linux-gnu -mcpu=cortex-a53 -pre-RA-sched=source -enable-misched -verify-misched -debug-only=machine-scheduler -disable-machine-dce -o - 2>&1 > /dev/null | FileCheck %s
; RUN: llc < %s -mtriple=arm64-linux-gnu -mcpu=cortex-a53 -pre-RA-sched=source -enable-misched -verify-misched -debug-only=machine-scheduler -disable-machine-dce -o - -misched-limit=2 2>&1 > /dev/null | FileCheck %s
;
; The Cortex-A53 machine model will cause the MADD instruction to be scheduled
; much higher than the ADD instructions in order to hide latency. When not
; specifying a subtarget, the MADD will remain near the end of the block.
;
; CHECK: ********** MI Scheduling **********
; CHECK: main
; CHECK: *** Final schedule for %bb.2 ***
; CHECK: MADDWrrr
; CHECK: ADDWri
; CHECK: ********** INTERVALS **********
@main.x = private unnamed_addr constant [8 x i32] [i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1], align 4
@main.y = private unnamed_addr constant [8 x i32] [i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2], align 4

; Function Attrs: nounwind
define i32 @main() #0 {
entry:
  %retval = alloca i32, align 4
  %x = alloca [8 x i32], align 4
  %y = alloca [8 x i32], align 4
  %i = alloca i32, align 4
  %xx = alloca i32, align 4
  %yy = alloca i32, align 4
  store i32 0, i32* %retval
  %0 = bitcast [8 x i32]* %x to i8*
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %0, i8* align 4 bitcast ([8 x i32]* @main.x to i8*), i64 32, i1 false)
  %1 = bitcast [8 x i32]* %y to i8*
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %1, i8* align 4 bitcast ([8 x i32]* @main.y to i8*), i64 32, i1 false)
  store i32 0, i32* %xx, align 4
  store i32 0, i32* %yy, align 4
  store i32 0, i32* %i, align 4
  br label %for.cond

for.cond:                                         ; preds = %for.inc, %entry
  %2 = load i32, i32* %i, align 4
  %cmp = icmp slt i32 %2, 8
  br i1 %cmp, label %for.body, label %for.end

for.body:                                         ; preds = %for.cond
  %3 = load i32, i32* %i, align 4
  %idxprom = sext i32 %3 to i64
  %arrayidx = getelementptr inbounds [8 x i32], [8 x i32]* %x, i32 0, i64 %idxprom
  %4 = load i32, i32* %arrayidx, align 4
  %add = add nsw i32 %4, 1
  store i32 %add, i32* %xx, align 4
  %5 = load i32, i32* %xx, align 4
  %add1 = add nsw i32 %5, 12
  store i32 %add1, i32* %xx, align 4
  %6 = load i32, i32* %xx, align 4
  %add2 = add nsw i32 %6, 23
  store i32 %add2, i32* %xx, align 4
  %7 = load i32, i32* %xx, align 4
  %add3 = add nsw i32 %7, 34
  store i32 %add3, i32* %xx, align 4
  %8 = load i32, i32* %i, align 4
  %idxprom4 = sext i32 %8 to i64
  %arrayidx5 = getelementptr inbounds [8 x i32], [8 x i32]* %y, i32 0, i64 %idxprom4
  %9 = load i32, i32* %arrayidx5, align 4
  %10 = load i32, i32* %yy, align 4
  %mul = mul nsw i32 %10, %9
  store i32 %mul, i32* %yy, align 4
  br label %for.inc

for.inc:                                          ; preds = %for.body
  %11 = load i32, i32* %i, align 4
  %inc = add nsw i32 %11, 1
  store i32 %inc, i32* %i, align 4
  br label %for.cond

for.end:                                          ; preds = %for.cond
  %12 = load i32, i32* %xx, align 4
  %13 = load i32, i32* %yy, align 4
  %add6 = add nsw i32 %12, %13
  ret i32 %add6
}


; The Cortex-A53 machine model will cause the FDIVvvv_42 to be raised to
; hide latency. Whereas normally there would only be a single FADDvvv_4s
; after it, this test checks to make sure there are more than one.
;
; CHECK: ********** MI Scheduling **********
; CHECK: neon4xfloat:%bb.0
; CHECK: *** Final schedule for %bb.0 ***
; CHECK: FDIVv4f32
; CHECK: FADDv4f32
; CHECK: FADDv4f32
; CHECK: ********** INTERVALS **********
define <4 x float> @neon4xfloat(<4 x float> %A, <4 x float> %B) {
        %tmp1 = fadd <4 x float> %A, %B;
        %tmp2 = fadd <4 x float> %A, %tmp1;
        %tmp3 = fadd <4 x float> %A, %tmp2;
        %tmp4 = fadd <4 x float> %A, %tmp3;
        %tmp5 = fadd <4 x float> %A, %tmp4;
        %tmp6 = fadd <4 x float> %A, %tmp5;
        %tmp7 = fadd <4 x float> %A, %tmp6;
        %tmp8 = fadd <4 x float> %A, %tmp7;
        %tmp9 = fdiv <4 x float> %A, %B;
        %tmp10 = fadd <4 x float> %tmp8, %tmp9;

        ret <4 x float> %tmp10
}

; Function Attrs: nounwind
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i1) #1

attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { nounwind }


; Regression Test for PR19761
;   [ARM64] Cortex-a53 schedule mode can't handle NEON post-increment load
;
; Nothing explicit to check other than llc not crashing.
define { <16 x i8>, <16 x i8> } @test_v16i8_post_imm_ld2(i8* %A, i8** %ptr) {
  %ld2 = tail call { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2.v16i8.p0i8(i8* %A)
  %tmp = getelementptr i8, i8* %A, i32 32
  store i8* %tmp, i8** %ptr
  ret { <16 x i8>, <16 x i8> } %ld2
}

declare { <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld2.v16i8.p0i8(i8*)

; Regression Test for PR20057.
;
; Cortex-A53 machine model stalls on A53UnitFPMDS contention. Instructions that
; are otherwise ready are jammed in the pending queue.
; CHECK: ********** MI Scheduling **********
; CHECK: testResourceConflict
; CHECK: *** Final schedule for %bb.0 ***
; CHECK: BRK
; CHECK: ********** INTERVALS **********
define void @testResourceConflict(float* %ptr) {
entry:
  %add1 = fadd float undef, undef
  %mul2 = fmul float undef, undef
  %add3 = fadd float %mul2, undef
  %mul4 = fmul float undef, %add3
  %add5 = fadd float %mul4, undef
  %sub6 = fsub float 0.000000e+00, undef
  %sub7 = fsub float %add5, undef
  %div8 = fdiv float 1.000000e+00, undef
  %mul9 = fmul float %div8, %sub7
  %mul14 = fmul float %sub6, %div8
  %mul10 = fsub float -0.000000e+00, %mul14
  %mul15 = fmul float undef, %div8
  %mul11 = fsub float -0.000000e+00, %mul15
  %mul12 = fmul float 0.000000e+00, %div8
  %mul13 = fmul float %add1, %mul9
  %mul21 = fmul float %add5, %mul11
  %add22 = fadd float %mul13, %mul21
  store float %add22, float* %ptr, align 4
  %mul28 = fmul float %add1, %mul10
  %mul33 = fmul float %add5, %mul12
  %add34 = fadd float %mul33, %mul28
  store float %add34, float* %ptr, align 4
  %mul240 = fmul float undef, %mul9
  %add246 = fadd float %mul240, undef
  store float %add246, float* %ptr, align 4
  %mul52 = fmul float undef, %mul10
  %mul57 = fmul float undef, %mul12
  %add58 = fadd float %mul57, %mul52
  store float %add58, float* %ptr, align 4
  %mul27 = fmul float 0.000000e+00, %mul9
  %mul81 = fmul float undef, %mul10
  %add82 = fadd float %mul27, %mul81
  store float %add82, float* %ptr, align 4
  call void @llvm.trap()
  unreachable
}

declare void @llvm.trap()

; Regression test for PR20057: "permanent hazard"'
; Resource contention on LDST.
; CHECK: ********** MI Scheduling **********
; CHECK: testLdStConflict
; CHECK: *** Final schedule for %bb.1 ***
; CHECK: LD4Fourv2d
; CHECK: STRQui
; CHECK: ********** INTERVALS **********
define void @testLdStConflict(<2 x i64> %v) {
entry:
  br label %loop

loop:
  %0 = call { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4.v2i64.p0i8(i8* null)
  %ptr = bitcast i8* undef to <2 x i64>*
  store <2 x i64> %v, <2 x i64>* %ptr, align 4
  %ptr1 = bitcast i8* undef to <2 x i64>*
  store <2 x i64> %v, <2 x i64>* %ptr1, align 4
  %ptr2 = bitcast i8* undef to <2 x i64>*
  store <2 x i64> %v, <2 x i64>* %ptr2, align 4
  %ptr3 = bitcast i8* undef to <2 x i64>*
  store <2 x i64> %v, <2 x i64>* %ptr3, align 4
  %ptr4 = bitcast i8* undef to <2 x i64>*
  store <2 x i64> %v, <2 x i64>* %ptr4, align 4
  br label %loop
}

declare { <2 x i64>, <2 x i64>, <2 x i64>, <2 x i64> } @llvm.aarch64.neon.ld4.v2i64.p0i8(i8*)