reference, declaration → definition
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

; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 \
; RUN:   | FileCheck -check-prefix=CHECK -check-prefix=CHECK-SCALAR %s
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z14 \
; RUN:   | FileCheck -check-prefix=CHECK -check-prefix=CHECK-VECTOR %s

declare float @llvm.fma.f32(float %f1, float %f2, float %f3)

define float @f1(float %f1, float %f2, float %acc) {
; CHECK-LABEL: f1:
; CHECK-SCALAR: maebr %f4, %f0, %f2
; CHECK-SCALAR: ler %f0, %f4
; CHECK-VECTOR: wfmasb %f0, %f0, %f2, %f4
; CHECK: br %r14
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f2(float %f1, float *%ptr, float %acc) {
; CHECK-LABEL: f2:
; CHECK: maeb %f2, %f0, 0(%r2)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f3(float %f1, float *%base, float %acc) {
; CHECK-LABEL: f3:
; CHECK: maeb %f2, %f0, 4092(%r2)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %ptr = getelementptr float, float *%base, i64 1023
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f4(float %f1, float *%base, float %acc) {
; The important thing here is that we don't generate an out-of-range
; displacement.  Other sequences besides this one would be OK.
;
; CHECK-LABEL: f4:
; CHECK: aghi %r2, 4096
; CHECK: maeb %f2, %f0, 0(%r2)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %ptr = getelementptr float, float *%base, i64 1024
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f5(float %f1, float *%base, float %acc) {
; Here too the important thing is that we don't generate an out-of-range
; displacement.  Other sequences besides this one would be OK.
;
; CHECK-LABEL: f5:
; CHECK: aghi %r2, -4
; CHECK: maeb %f2, %f0, 0(%r2)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %ptr = getelementptr float, float *%base, i64 -1
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f6(float %f1, float *%base, i64 %index, float %acc) {
; CHECK-LABEL: f6:
; CHECK: sllg %r1, %r3, 2
; CHECK: maeb %f2, %f0, 0(%r1,%r2)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %ptr = getelementptr float, float *%base, i64 %index
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f7(float %f1, float *%base, i64 %index, float %acc) {
; CHECK-LABEL: f7:
; CHECK: sllg %r1, %r3, 2
; CHECK: maeb %f2, %f0, 4092({{%r1,%r2|%r2,%r1}})
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %index2 = add i64 %index, 1023
  %ptr = getelementptr float, float *%base, i64 %index2
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}

define float @f8(float %f1, float *%base, i64 %index, float %acc) {
; CHECK-LABEL: f8:
; CHECK: sllg %r1, %r3, 2
; CHECK: lay %r1, 4096({{%r1,%r2|%r2,%r1}})
; CHECK: maeb %f2, %f0, 0(%r1)
; CHECK-SCALAR: ler %f0, %f2
; CHECK-VECTOR: ldr %f0, %f2
; CHECK: br %r14
  %index2 = add i64 %index, 1024
  %ptr = getelementptr float, float *%base, i64 %index2
  %f2 = load float, float *%ptr
  %res = call float @llvm.fma.f32 (float %f1, float %f2, float %acc)
  ret float %res
}