flang/test/Lower/OpenMP/wsloop-reduction-add.f90

   1 ! RUN: bbc -emit-fir -fopenmp %s -o - | FileCheck %s
   2 ! RUN: %flang_fc1 -emit-fir -fopenmp %s -o - | FileCheck %s
   3
   4 !CHECK-LABEL: omp.reduction.declare
   5 !CHECK-SAME: @[[RED_F64_NAME:.*]] : f64 init {
   6 !CHECK: ^bb0(%{{.*}}: f64):
   7 !CHECK:  %[[C0_1:.*]] = arith.constant 0.000000e+00 : f64
   8 !CHECK:  omp.yield(%[[C0_1]] : f64)
   9 !CHECK: } combiner {
  10 !CHECK: ^bb0(%[[ARG0:.*]]: f64, %[[ARG1:.*]]: f64):
  11 !CHECK:  %[[RES:.*]] = arith.addf %[[ARG0]], %[[ARG1]] {{.*}}: f64
  12 !CHECK:  omp.yield(%[[RES]] : f64)
  13 !CHECK: }
  14
  15 !CHECK-LABEL: omp.reduction.declare
  16 !CHECK-SAME: @[[RED_I64_NAME:.*]] : i64 init {
  17 !CHECK: ^bb0(%{{.*}}: i64):
  18 !CHECK:  %[[C0_1:.*]] = arith.constant 0 : i64
  19 !CHECK:  omp.yield(%[[C0_1]] : i64)
  20 !CHECK: } combiner {
  21 !CHECK: ^bb0(%[[ARG0:.*]]: i64, %[[ARG1:.*]]: i64):
  22 !CHECK:  %[[RES:.*]] = arith.addi %[[ARG0]], %[[ARG1]] : i64
  23 !CHECK:  omp.yield(%[[RES]] : i64)
  24 !CHECK: }
  25
  26 !CHECK-LABEL: omp.reduction.declare
  27 !CHECK-SAME: @[[RED_F32_NAME:.*]] : f32 init {
  28 !CHECK: ^bb0(%{{.*}}: f32):
  29 !CHECK:  %[[C0_1:.*]] = arith.constant 0.000000e+00 : f32
  30 !CHECK:  omp.yield(%[[C0_1]] : f32)
  31 !CHECK: } combiner {
  32 !CHECK: ^bb0(%[[ARG0:.*]]: f32, %[[ARG1:.*]]: f32):
  33 !CHECK:  %[[RES:.*]] = arith.addf %[[ARG0]], %[[ARG1]] {{.*}}: f32
  34 !CHECK:  omp.yield(%[[RES]] : f32)
  35 !CHECK: }
  36
  37 !CHECK-LABEL: omp.reduction.declare
  38 !CHECK-SAME: @[[RED_I32_NAME:.*]] : i32 init {
  39 !CHECK: ^bb0(%{{.*}}: i32):
  40 !CHECK:  %[[C0_1:.*]] = arith.constant 0 : i32
  41 !CHECK:  omp.yield(%[[C0_1]] : i32)
  42 !CHECK: } combiner {
  43 !CHECK: ^bb0(%[[ARG0:.*]]: i32, %[[ARG1:.*]]: i32):
  44 !CHECK:  %[[RES:.*]] = arith.addi %[[ARG0]], %[[ARG1]] : i32
  45 !CHECK:  omp.yield(%[[RES]] : i32)
  46 !CHECK: }
  47
  48 !CHECK-LABEL: func.func @_QPsimple_int_reduction
  49 !CHECK:  %[[XREF:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFsimple_int_reductionEx"}
  50 !CHECK:  %[[C0_2:.*]] = arith.constant 0 : i32
  51 !CHECK:  fir.store %[[C0_2]] to %[[XREF]] : !fir.ref<i32>
  52 !CHECK:  omp.parallel
  53 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
  54 !CHECK:    %[[C1_1:.*]] = arith.constant 1 : i32
  55 !CHECK:    %[[C100:.*]] = arith.constant 100 : i32
  56 !CHECK:    %[[C1_2:.*]] = arith.constant 1 : i32
  57 !CHECK:    omp.wsloop   reduction(@[[RED_I32_NAME]] -> %[[XREF]] : !fir.ref<i32>) for  (%[[IVAL:.*]]) : i32 = (%[[C1_1]]) to (%[[C100]]) inclusive step (%[[C1_2]])
  58 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
  59 !CHECK:      %[[I_PVT_VAL:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
  60 !CHECK:      omp.reduction %[[I_PVT_VAL]], %[[XREF]] : i32, !fir.ref<i32>
  61 !CHECK:      omp.yield
  62 !CHECK:    omp.terminator
  63 !CHECK:  return
  64 subroutine simple_int_reduction
  65   integer :: x
  66   x = 0
  67   !$omp parallel
  68   !$omp do reduction(+:x)
  69   do i=1, 100
  70     x = x + i
  71   end do
  72   !$omp end do
  73   !$omp end parallel
  74 end subroutine
  75
  76 !CHECK-LABEL: func.func @_QPsimple_real_reduction
  77 !CHECK:  %[[XREF:.*]] = fir.alloca f32 {bindc_name = "x", uniq_name = "_QFsimple_real_reductionEx"}
  78 !CHECK:  %[[C0_2:.*]] = arith.constant 0.000000e+00 : f32
  79 !CHECK:  fir.store %[[C0_2]] to %[[XREF]] : !fir.ref<f32>
  80 !CHECK:  omp.parallel
  81 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
  82 !CHECK:    %[[C1_1:.*]] = arith.constant 1 : i32
  83 !CHECK:    %[[C100:.*]] = arith.constant 100 : i32
  84 !CHECK:    %[[C1_2:.*]] = arith.constant 1 : i32
  85 !CHECK:    omp.wsloop   reduction(@[[RED_F32_NAME]] -> %[[XREF]] : !fir.ref<f32>) for  (%[[IVAL:.*]]) : i32 = (%[[C1_1]]) to (%[[C100]]) inclusive step (%[[C1_2]])
  86 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
  87 !CHECK:      %[[I_PVT_VAL_i32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
  88 !CHECK:      %[[I_PVT_VAL_f32:.*]] = fir.convert %[[I_PVT_VAL_i32]] : (i32) -> f32
  89 !CHECK:      omp.reduction %[[I_PVT_VAL_f32]], %[[XREF]] : f32, !fir.ref<f32>
  90 !CHECK:      omp.yield
  91 !CHECK:    omp.terminator
  92 !CHECK:  return
  93 subroutine simple_real_reduction
  94   real :: x
  95   x = 0.0
  96   !$omp parallel
  97   !$omp do reduction(+:x)
  98   do i=1, 100
  99     x = x + i
 100   end do
 101   !$omp end do
 102   !$omp end parallel
 103 end subroutine
 104
 105 !CHECK-LABEL: func.func @_QPsimple_int_reduction_switch_order
 106 !CHECK:  %[[XREF:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFsimple_int_reduction_switch_orderEx"}
 107 !CHECK:  %[[C0_2:.*]] = arith.constant 0 : i32
 108 !CHECK:  fir.store %[[C0_2]] to %[[XREF]] : !fir.ref<i32>
 109 !CHECK:  omp.parallel
 110 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
 111 !CHECK:    %[[C1_1:.*]] = arith.constant 1 : i32
 112 !CHECK:    %[[C100:.*]] = arith.constant 100 : i32
 113 !CHECK:    %[[C1_2:.*]] = arith.constant 1 : i32
 114 !CHECK:    omp.wsloop   reduction(@[[RED_I32_NAME]] -> %[[XREF]] : !fir.ref<i32>) for  (%[[IVAL:.*]]) : i32 = (%[[C1_1]]) to (%[[C100]]) inclusive step (%[[C1_2]])
 115 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
 116 !CHECK:      %[[I_PVT_VAL:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 117 !CHECK:      omp.reduction %[[I_PVT_VAL]], %[[XREF]] : i32, !fir.ref<i32>
 118 !CHECK:      omp.yield
 119 !CHECK:    omp.terminator
 120 !CHECK:  return
 121 subroutine simple_int_reduction_switch_order
 122   integer :: x
 123   x = 0
 124   !$omp parallel
 125   !$omp do reduction(+:x)
 126   do i=1, 100
 127     x = i + x
 128   end do
 129   !$omp end do
 130   !$omp end parallel
 131 end subroutine
 132
 133 !CHECK-LABEL: func.func @_QPsimple_real_reduction_switch_order
 134 !CHECK:  %[[XREF:.*]] = fir.alloca f32 {bindc_name = "x", uniq_name = "_QFsimple_real_reduction_switch_orderEx"}
 135 !CHECK:  %[[C0_2:.*]] = arith.constant 0.000000e+00 : f32
 136 !CHECK:  fir.store %[[C0_2]] to %[[XREF]] : !fir.ref<f32>
 137 !CHECK:  omp.parallel
 138 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
 139 !CHECK:    %[[C1_1:.*]] = arith.constant 1 : i32
 140 !CHECK:    %[[C100:.*]] = arith.constant 100 : i32
 141 !CHECK:    %[[C1_2:.*]] = arith.constant 1 : i32
 142 !CHECK:    omp.wsloop   reduction(@[[RED_F32_NAME]] -> %[[XREF]] : !fir.ref<f32>) for  (%[[IVAL:.*]]) : i32 = (%[[C1_1]]) to (%[[C100]]) inclusive step (%[[C1_2]])
 143 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
 144 !CHECK:      %[[I_PVT_VAL_i32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 145 !CHECK:      %[[I_PVT_VAL_f32:.*]] = fir.convert %[[I_PVT_VAL_i32]] : (i32) -> f32
 146 !CHECK:      omp.reduction %[[I_PVT_VAL_f32]], %[[XREF]] : f32, !fir.ref<f32>
 147 !CHECK:      omp.yield
 148 !CHECK:    omp.terminator
 149 !CHECK:  return
 150 subroutine simple_real_reduction_switch_order
 151   real :: x
 152   x = 0.0
 153   !$omp parallel
 154   !$omp do reduction(+:x)
 155   do i=1, 100
 156     x = i + x
 157   end do
 158   !$omp end do
 159   !$omp end parallel
 160 end subroutine
 161
 162 !CHECK-LABEL: func.func @_QPmultiple_int_reductions_same_type
 163 !CHECK:  %[[XREF:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFmultiple_int_reductions_same_typeEx"}
 164 !CHECK:  %[[YREF:.*]] = fir.alloca i32 {bindc_name = "y", uniq_name = "_QFmultiple_int_reductions_same_typeEy"}
 165 !CHECK:  %[[ZREF:.*]] = fir.alloca i32 {bindc_name = "z", uniq_name = "_QFmultiple_int_reductions_same_typeEz"}
 166 !CHECK:  omp.parallel
 167 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
 168 !CHECK:    omp.wsloop   reduction(@[[RED_I32_NAME]] -> %[[XREF]] : !fir.ref<i32>, @[[RED_I32_NAME]] -> %[[YREF]] : !fir.ref<i32>, @[[RED_I32_NAME]] -> %[[ZREF]] : !fir.ref<i32>) for  (%[[IVAL]]) : i32
 169 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
 170 !CHECK:      %[[I_PVT_VAL1:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 171 !CHECK:      omp.reduction %[[I_PVT_VAL1]], %[[XREF]] : i32, !fir.ref<i32>
 172 !CHECK:      %[[I_PVT_VAL2:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 173 !CHECK:      omp.reduction %[[I_PVT_VAL2]], %[[YREF]] : i32, !fir.ref<i32>
 174 !CHECK:      %[[I_PVT_VAL3:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 175 !CHECK:      omp.reduction %[[I_PVT_VAL3]], %[[ZREF]] : i32, !fir.ref<i32>
 176 !CHECK:      omp.yield
 177 !CHECK:    omp.terminator
 178 !CHECK:  return
 179 subroutine multiple_int_reductions_same_type
 180   integer :: x,y,z
 181   x = 0
 182   y = 0
 183   z = 0
 184   !$omp parallel
 185   !$omp do reduction(+:x,y,z)
 186   do i=1, 100
 187     x = x + i
 188     y = y + i
 189     z = z + i
 190   end do
 191   !$omp end do
 192   !$omp end parallel
 193 end subroutine
 194
 195 !CHECK-LABEL: func.func @_QPmultiple_real_reductions_same_type
 196 !CHECK:  %[[XREF:.*]] = fir.alloca f32 {bindc_name = "x", uniq_name = "_QFmultiple_real_reductions_same_typeEx"}
 197 !CHECK:  %[[YREF:.*]] = fir.alloca f32 {bindc_name = "y", uniq_name = "_QFmultiple_real_reductions_same_typeEy"}
 198 !CHECK:  %[[ZREF:.*]] = fir.alloca f32 {bindc_name = "z", uniq_name = "_QFmultiple_real_reductions_same_typeEz"}
 199 !CHECK:  omp.parallel
 200 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
 201 !CHECK:    omp.wsloop   reduction(@[[RED_F32_NAME]] -> %[[XREF]] : !fir.ref<f32>, @[[RED_F32_NAME]] -> %[[YREF]] : !fir.ref<f32>, @[[RED_F32_NAME]] -> %[[ZREF]] : !fir.ref<f32>) for  (%[[IVAL]]) : i32
 202 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
 203 !CHECK:      %[[I_PVT_VAL1_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 204 !CHECK:      %[[I_PVT_VAL1_F32:.*]] = fir.convert %[[I_PVT_VAL1_I32]] : (i32) -> f32
 205 !CHECK:      omp.reduction %[[I_PVT_VAL1_F32]], %[[XREF]] : f32, !fir.ref<f32>
 206 !CHECK:      %[[I_PVT_VAL2_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 207 !CHECK:      %[[I_PVT_VAL2_F32:.*]] = fir.convert %[[I_PVT_VAL2_I32]] : (i32) -> f32
 208 !CHECK:      omp.reduction %[[I_PVT_VAL2_F32]], %[[YREF]] : f32, !fir.ref<f32>
 209 !CHECK:      %[[I_PVT_VAL3_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 210 !CHECK:      %[[I_PVT_VAL3_F32:.*]] = fir.convert %[[I_PVT_VAL3_I32]] : (i32) -> f32
 211 !CHECK:      omp.reduction %[[I_PVT_VAL3_F32]], %[[ZREF]] : f32, !fir.ref<f32>
 212 !CHECK:      omp.yield
 213 !CHECK:    omp.terminator
 214 !CHECK:  return
 215 subroutine multiple_real_reductions_same_type
 216   real :: x,y,z
 217   x = 0.0
 218   y = 0.0
 219   z = 0.0
 220   !$omp parallel
 221   !$omp do reduction(+:x,y,z)
 222   do i=1, 100
 223     x = x + i
 224     y = y + i
 225     z = z + i
 226   end do
 227   !$omp end do
 228   !$omp end parallel
 229 end subroutine
 230
 231 !CHECK-LABEL: func.func @_QPmultiple_reductions_different_type
 232 !CHECK:  %[[WREF:.*]] = fir.alloca f64 {bindc_name = "w", uniq_name = "_QFmultiple_reductions_different_typeEw"}
 233 !CHECK:  %[[XREF:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFmultiple_reductions_different_typeEx"}
 234 !CHECK:  %[[YREF:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFmultiple_reductions_different_typeEy"}
 235 !CHECK:  %[[ZREF:.*]] = fir.alloca f32 {bindc_name = "z", uniq_name = "_QFmultiple_reductions_different_typeEz"}
 236 !CHECK:  omp.parallel
 237 !CHECK:    %[[I_PVT_REF:.*]] = fir.alloca i32 {adapt.valuebyref, pinned}
 238 !CHECK:    omp.wsloop   reduction(@[[RED_I32_NAME]] -> %[[XREF]]  : !fir.ref<i32>, @[[RED_I64_NAME]] -> %[[YREF]] : !fir.ref<i64>, @[[RED_F32_NAME]] -> %[[ZREF]]  : !fir.ref<f32>, @[[RED_F64_NAME]] -> %[[WREF]]  : !fir.ref<f64>) for  (%[[IVAL:.*]]) : i32
 239 !CHECK:      fir.store %[[IVAL]] to %[[I_PVT_REF]] : !fir.ref<i32>
 240 !CHECK:      %[[I_PVT_VAL1_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 241 !CHECK:      omp.reduction %[[I_PVT_VAL1_I32]], %[[XREF]] : i32, !fir.ref<i32>
 242 !CHECK:      %[[I_PVT_VAL2_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 243 !CHECK:      %[[I_PVT_VAL2_I64:.*]] = fir.convert %[[I_PVT_VAL2_I32]] : (i32) -> i64
 244 !CHECK:      omp.reduction %[[I_PVT_VAL2_I64]], %[[YREF]] : i64, !fir.ref<i64>
 245 !CHECK:      %[[I_PVT_VAL3_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 246 !CHECK:      %[[I_PVT_VAL3_F32:.*]] = fir.convert %[[I_PVT_VAL3_I32]] : (i32) -> f32
 247 !CHECK:      omp.reduction %[[I_PVT_VAL3_F32]], %[[ZREF]] : f32, !fir.ref<f32>
 248 !CHECK:      %[[I_PVT_VAL4_I32:.*]] = fir.load %[[I_PVT_REF]] : !fir.ref<i32>
 249 !CHECK:      %[[I_PVT_VAL4_F64:.*]] = fir.convert %[[I_PVT_VAL4_I32]] : (i32) -> f64
 250 !CHECK:      omp.reduction %[[I_PVT_VAL4_F64]], %[[WREF]] : f64, !fir.ref<f64>
 251 !CHECK:      omp.yield
 252 !CHECK:    omp.terminator
 253 !CHECK:  return
 254 subroutine multiple_reductions_different_type
 255   integer :: x
 256   integer(kind=8) :: y
 257   real :: z
 258   real(kind=8) :: w
 259   x = 0
 260   y = 0
 261   z = 0.0
 262   w = 0.0
 263   !$omp parallel
 264   !$omp do reduction(+:x,y,z,w)
 265   do i=1, 100
 266     x = x + i
 267     y = y + i
 268     z = z + i
 269     w = w + i
 270   end do
 271   !$omp end do
 272   !$omp end parallel
 273 end subroutine