xref: /llvm-project/flang/test/Lower/OpenMP/parallel-reduction-add-byref.f90 (revision cdb3ebf1e62df060767863e1e683409d6077ca6e)
1! RUN: bbc -emit-hlfir --force-byref-reduction -fopenmp -o - %s 2>&1 | FileCheck %s
2! RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --force-byref-reduction -o - %s 2>&1 | FileCheck %s
3
4!CHECK-LABEL: omp.declare_reduction
5!CHECK-SAME: @[[RED_F32_NAME:.*]] : !fir.ref<f32>
6!CHECK-SAME: alloc {
7!CHECK:  %[[REF:.*]] = fir.alloca f32
8!CHECK:  omp.yield(%[[REF]] : !fir.ref<f32>)
9!CHECK-LABEL: } init {
10!CHECK: ^bb0(%{{.*}}: !fir.ref<f32>, %[[ALLOC:.*]]: !fir.ref<f32>):
11!CHECK:  %[[C0_1:.*]] = arith.constant 0.000000e+00 : f32
12!CHECKL  fir.store [[%C0_1]] to %[[ALLOC]] : !fir.ref<f32>
13!CHECK:  omp.yield(%[[ALLOC]] : !fir.ref<f32>)
14!CHECK: } combiner {
15!CHECK: ^bb0(%[[ARG0:.*]]: !fir.ref<f32>, %[[ARG1:.*]]: !fir.ref<f32>):
16!CHECK:  %[[LD0:.*]] = fir.load %[[ARG0]] : !fir.ref<f32>
17!CHECK:  %[[LD1:.*]] = fir.load %[[ARG1]] : !fir.ref<f32>
18!CHECK:  %[[RES:.*]] = arith.addf %[[LD0]], %[[LD1]] {{.*}}: f32
19!CHECK:  fir.store %[[RES]] to %[[ARG0]] : !fir.ref<f32>
20!CHECK:  omp.yield(%[[ARG0]] : !fir.ref<f32>)
21!CHECK: }
22
23!CHECK-LABEL: omp.declare_reduction
24!CHECK-SAME: @[[RED_I32_NAME:.*]] : !fir.ref<i32>
25!CHECK-SAME: alloc {
26!CHECK:  %[[REF:.*]] = fir.alloca i32
27!CHECK:  omp.yield(%[[REF]] : !fir.ref<i32>)
28!CHECK-LABEL: } init {
29!CHECK: ^bb0(%{{.*}}: !fir.ref<i32>, %[[ALLOC:.*]]: !fir.ref<i32>):
30!CHECK:  %[[C0_1:.*]] = arith.constant 0 : i32
31!CHECK:  fir.store %[[C0_1]] to %[[ALLOC]] : !fir.ref<i32>
32!CHECK:  omp.yield(%[[ALLOC]] : !fir.ref<i32>)
33!CHECK: } combiner {
34!CHECK: ^bb0(%[[ARG0:.*]]: !fir.ref<i32>, %[[ARG1:.*]]: !fir.ref<i32>):
35!CHECK:  %[[LD0:.*]] = fir.load %[[ARG0]] : !fir.ref<i32>
36!CHECK:  %[[LD1:.*]] = fir.load %[[ARG1]] : !fir.ref<i32>
37!CHECK:  %[[RES:.*]] = arith.addi %[[LD0]], %[[LD1]] : i32
38!CHECK:  fir.store %[[RES]] to %[[ARG0]] : !fir.ref<i32>
39!CHECK:  omp.yield(%[[ARG0]] : !fir.ref<i32>)
40!CHECK: }
41
42!CHECK-LABEL: func.func @_QPsimple_int_add
43!CHECK:  %[[IREF:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFsimple_int_addEi"}
44!CHECK:  %[[I_DECL:.*]]:2 = hlfir.declare %[[IREF]] {uniq_name = "_QFsimple_int_addEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
45!CHECK:  %[[I_START:.*]] = arith.constant 0 : i32
46!CHECK:  hlfir.assign %[[I_START]] to %[[I_DECL]]#0 : i32, !fir.ref<i32>
47!CHECK:  omp.parallel reduction(byref @[[RED_I32_NAME]] %[[I_DECL]]#0 -> %[[PRV:.+]] : !fir.ref<i32>) {
48!CHECK:    %[[P_DECL:.+]]:2 = hlfir.declare %[[PRV]] {{.*}} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
49!CHECK:    %[[LPRV:.+]] = fir.load %[[P_DECL]]#0 : !fir.ref<i32>
50!CHECK:    %[[I_INCR:.*]] = arith.constant 1 : i32
51!CHECK:    %[[RES:.+]] = arith.addi %[[LPRV]], %[[I_INCR]] : i32
52!CHECK:    hlfir.assign %[[RES]] to %[[P_DECL]]#0 : i32, !fir.ref<i32>
53!CHECK:    omp.terminator
54!CHECK:  }
55!CHECK: return
56subroutine simple_int_add
57    integer :: i
58    i = 0
59
60    !$omp parallel reduction(+:i)
61    i = i + 1
62    !$omp end parallel
63
64    print *, i
65end subroutine
66
67!CHECK-LABEL: func.func @_QPsimple_real_add
68!CHECK:  %[[RREF:.*]] = fir.alloca f32 {bindc_name = "r", uniq_name = "_QFsimple_real_addEr"}
69!CHECK:  %[[R_DECL:.*]]:2 = hlfir.declare %[[RREF]] {uniq_name = "_QFsimple_real_addEr"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
70!CHECK:  %[[R_START:.*]] = arith.constant 0.000000e+00 : f32
71!CHECK:  hlfir.assign %[[R_START]] to %[[R_DECL]]#0 : f32, !fir.ref<f32>
72!CHECK:  omp.parallel reduction(byref @[[RED_F32_NAME]] %[[R_DECL]]#0 -> %[[PRV:.+]] : !fir.ref<f32>) {
73!CHECK:    %[[P_DECL:.+]]:2 = hlfir.declare %[[PRV]] {{.*}} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
74!CHECK:    %[[LPRV:.+]] = fir.load %[[P_DECL]]#0 : !fir.ref<f32>
75!CHECK:    %[[R_INCR:.*]] = arith.constant 1.500000e+00 : f32
76!CHECK:    %[[RES:.+]] = arith.addf %[[LPRV]], %[[R_INCR]] {{.*}} : f32
77!CHECK:    hlfir.assign %[[RES]] to %[[P_DECL]]#0 : f32, !fir.ref<f32>
78!CHECK:    omp.terminator
79!CHECK:  }
80!CHECK: return
81subroutine simple_real_add
82    real :: r
83    r = 0.0
84
85    !$omp parallel reduction(+:r)
86    r = r + 1.5
87    !$omp end parallel
88
89    print *, r
90end subroutine
91
92!CHECK-LABEL: func.func @_QPint_real_add
93!CHECK:  %[[IREF:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFint_real_addEi"}
94!CHECK:  %[[I_DECL:.*]]:2 = hlfir.declare %[[IREF]] {uniq_name = "_QFint_real_addEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
95!CHECK:  %[[RREF:.*]] = fir.alloca f32 {bindc_name = "r", uniq_name = "_QFint_real_addEr"}
96!CHECK:  %[[R_DECL:.*]]:2 = hlfir.declare %[[RREF]] {uniq_name = "_QFint_real_addEr"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
97!CHECK:  %[[R_START:.*]] = arith.constant 0.000000e+00 : f32
98!CHECK:  hlfir.assign %[[R_START]] to %[[R_DECL]]#0 : f32, !fir.ref<f32>
99!CHECK:  %[[I_START:.*]] = arith.constant 0 : i32
100!CHECK:  hlfir.assign %[[I_START]] to %[[I_DECL]]#0 : i32, !fir.ref<i32>
101!CHECK:  omp.parallel reduction(byref @[[RED_I32_NAME]] %[[I_DECL]]#0 -> %[[IPRV:.+]], byref @[[RED_F32_NAME]] %[[R_DECL]]#0 -> %[[RPRV:.+]] : !fir.ref<i32>, !fir.ref<f32>) {
102!CHECK:    %[[IP_DECL:.+]]:2 = hlfir.declare %[[IPRV]] {{.*}} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
103!CHECK:    %[[RP_DECL:.+]]:2 = hlfir.declare %[[RPRV]] {{.*}} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
104!CHECK:    %[[R_INCR:.*]] = arith.constant 1.500000e+00 : f32
105!CHECK:    %[[R_LPRV:.+]] = fir.load %[[RP_DECL]]#0 : !fir.ref<f32>
106!CHECK:    %[[RES1:.+]] = arith.addf %[[R_INCR]], %[[R_LPRV]] {{.*}} : f32
107!CHECK:    hlfir.assign %[[RES1]] to %[[RP_DECL]]#0 : f32, !fir.ref<f32>
108!CHECK:    %[[I_LPRV:.+]] = fir.load %[[IP_DECL]]#0 : !fir.ref<i32>
109!CHECK:    %[[I_INCR:.*]] = arith.constant 3 : i32
110!CHECK:    %[[RES0:.+]] = arith.addi %[[I_LPRV]], %[[I_INCR]] : i32
111!CHECK:    hlfir.assign %[[RES0]] to %[[IP_DECL]]#0 : i32, !fir.ref<i32>
112!CHECK:    omp.terminator
113!CHECK:  }
114!CHECK: return
115subroutine int_real_add
116    real :: r
117    integer :: i
118
119    r = 0.0
120    i = 0
121
122    !$omp parallel reduction(+:i,r)
123    r = 1.5 + r
124    i = i + 3
125    !$omp end parallel
126
127    print *, r
128    print *, i
129end subroutine
130