1 // RUN: tco %s | FileCheck %s
3 // CHECK-LINE: define void @f1
4 // CHECK: (ptr nocapture %[[A:[^,]*]], {{.*}}, float %[[F:.*]])
5 func.func @f1(%a : !fir.ref<!fir.array<?x?xf32>>, %n : index, %m : index, %o : index, %p : index, %f : f32) {
6 %c1 = arith.constant 1 : index
7 %s = fir.shape_shift %o, %n, %p, %m : (index, index, index, index) -> !fir.shapeshift<2>
8 %vIn = fir.array_load %a(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
10 %r = fir.do_loop %j = %p to %m step %c1 iter_args(%v1 = %vIn) -> !fir.array<?x?xf32> {
12 %r = fir.do_loop %i = %o to %n step %c1 iter_args(%v = %v1) -> !fir.array<?x?xf32> {
13 // CHECK: %[[AOFF:.*]] = getelementptr float, ptr %[[A]], i64
14 // CHECK: store float %[[F]], ptr %[[AOFF]]
15 %r = fir.array_update %v, %f, %i, %j : (!fir.array<?x?xf32>, f32, index, index) -> !fir.array<?x?xf32>
16 fir.result %r : !fir.array<?x?xf32>
18 fir.result %r : !fir.array<?x?xf32>
20 fir.array_merge_store %vIn, %r to %a : !fir.array<?x?xf32>, !fir.array<?x?xf32>, !fir.ref<!fir.array<?x?xf32>>
25 // CHECK-LINE: define void @f2
26 // CHECK: (ptr nocapture %[[A:[^,]*]], {{.*}}, float %[[F:.*]])
27 func.func @f2(%a : !fir.ref<!fir.array<?x?xf32>>, %b : !fir.ref<!fir.array<?x?xf32>>, %n : index, %m : index, %o : index, %p : index, %f : f32) {
28 %c1 = arith.constant 1 : index
29 %s = fir.shape_shift %o, %n, %p, %m : (index, index, index, index) -> !fir.shapeshift<2>
30 %vIn = fir.array_load %a(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
31 %wIn = fir.array_load %b(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
33 %r = fir.do_loop %j = %p to %m step %c1 iter_args(%v1 = %vIn) -> !fir.array<?x?xf32> {
35 %r = fir.do_loop %i = %o to %n step %c1 iter_args(%v = %v1) -> !fir.array<?x?xf32> {
36 %x = fir.array_fetch %wIn, %i, %j : (!fir.array<?x?xf32>, index, index) -> f32
37 %y = arith.addf %x, %f : f32
38 // CHECK: %[[AOFF:.*]] = getelementptr float, ptr %[[A]], i64
39 %r = fir.array_update %v, %y, %i, %j : (!fir.array<?x?xf32>, f32, index, index) -> !fir.array<?x?xf32>
40 fir.result %r : !fir.array<?x?xf32>
42 fir.result %r : !fir.array<?x?xf32>
44 fir.array_merge_store %vIn, %r to %a : !fir.array<?x?xf32>, !fir.array<?x?xf32>, !fir.ref<!fir.array<?x?xf32>>
49 // CHECK-LINE: define void @f3
50 // CHECK: (ptr nocapture %[[A:[^,]*]], {{.*}}, float %[[F:.*]])
51 func.func @f3(%a : !fir.ref<!fir.array<?x?xf32>>, %b : !fir.ref<!fir.array<?x?xf32>>, %n : index, %m : index, %o : index, %p : index, %f : f32) {
52 %c1 = arith.constant 1 : index
53 %s = fir.shape_shift %o, %n, %p, %m : (index, index, index, index) -> !fir.shapeshift<2>
54 %vIn = fir.array_load %a(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
55 %wIn = fir.array_load %b(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
57 %r = fir.do_loop %j = %p to %m step %c1 iter_args(%v1 = %vIn) -> !fir.array<?x?xf32> {
59 %r = fir.do_loop %i = %o to %n step %c1 iter_args(%v = %v1) -> !fir.array<?x?xf32> {
60 %x = fir.array_fetch %wIn, %i, %j : (!fir.array<?x?xf32>, index, index) -> f32
61 %y = arith.addf %x, %f : f32
62 // CHECK: %[[AOFF:.*]] = getelementptr float, ptr %[[A]], i64
63 %i2 = arith.addi %i, %c1 : index
64 %r = fir.array_update %v, %y, %i2, %j : (!fir.array<?x?xf32>, f32, index, index) -> !fir.array<?x?xf32>
65 fir.result %r : !fir.array<?x?xf32>
67 fir.result %r : !fir.array<?x?xf32>
69 fir.array_merge_store %vIn, %r to %a : !fir.array<?x?xf32>, !fir.array<?x?xf32>, !fir.ref<!fir.array<?x?xf32>>
74 // CHECK-LINE: define void @f4
75 // CHECK: (ptr nocapture %[[A:[^,]*]], {{.*}}, float %[[F:.*]])
76 func.func @f4(%a : !fir.ref<!fir.array<?x?xf32>>, %b : !fir.ref<!fir.array<?x?xf32>>, %n : index, %m : index, %o : index, %p : index, %f : f32) {
77 %c1 = arith.constant 1 : index
78 %s = fir.shape_shift %o, %n, %p, %m : (index, index, index, index) -> !fir.shapeshift<2>
79 %vIn = fir.array_load %a(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
80 %wIn = fir.array_load %b(%s) : (!fir.ref<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.array<?x?xf32>
82 %r = fir.do_loop %j = %p to %m step %c1 iter_args(%v1 = %vIn) -> !fir.array<?x?xf32> {
84 %r = fir.do_loop %i = %o to %n step %c1 iter_args(%v = %v1) -> !fir.array<?x?xf32> {
85 %x2 = fir.array_fetch %vIn, %i, %j : (!fir.array<?x?xf32>, index, index) -> f32
86 %x = fir.array_fetch %wIn, %i, %j : (!fir.array<?x?xf32>, index, index) -> f32
87 %y = arith.addf %x, %f : f32
88 %y2 = arith.addf %y, %x2 : f32
89 // CHECK: %[[AOFF:.*]] = getelementptr float, ptr %[[A]], i64
90 %i2 = arith.addi %i, %c1 : index
91 %r = fir.array_update %v, %y2, %i2, %j : (!fir.array<?x?xf32>, f32, index, index) -> !fir.array<?x?xf32>
92 fir.result %r : !fir.array<?x?xf32>
94 fir.result %r : !fir.array<?x?xf32>
96 fir.array_merge_store %vIn, %r to %a : !fir.array<?x?xf32>, !fir.array<?x?xf32>, !fir.ref<!fir.array<?x?xf32>>
101 // Array expression assignment with potentially non contiguous arrays (e.g.
102 // `a = b + f`, with and v assumed shapes.
103 // Tests that the stride from the descriptor is used.
104 // CHECK-LINE: define void @f5
105 // CHECK: (ptr %[[A:.*]], ptr %[[B:.*]], float %[[F:.*]])
106 func.func @f5(%arg0: !fir.box<!fir.array<?xf32>>, %arg1: !fir.box<!fir.array<?xf32>>, %arg2: f32) {
107 %c0 = arith.constant 0 : index
108 %c1 = arith.constant 1 : index
109 %0:3 = fir.box_dims %arg0, %c0 : (!fir.box<!fir.array<?xf32>>, index) -> (index, index, index)
110 %1 = arith.subi %0#1, %c1 : index
111 %2 = fir.array_load %arg0 : (!fir.box<!fir.array<?xf32>>) -> !fir.array<?xf32>
112 %3 = fir.array_load %arg1 : (!fir.box<!fir.array<?xf32>>) -> !fir.array<?xf32>
114 %4 = fir.do_loop %arg3 = %c0 to %1 step %c1 iter_args(%arg4 = %2) -> (!fir.array<?xf32>) {
115 // CHECK: %[[B_STRIDE_GEP:.*]] = getelementptr {{.*}}, ptr %[[B]], i32 0, i32 7, i32 0, i32 2
116 // CHECK: %[[B_STRIDE:.*]] = load i64, ptr %[[B_STRIDE_GEP]]
117 // CHECK: %[[B_DIM_OFFSET:.*]] = mul nsw i64 %{{.*}}, %[[B_STRIDE]]
118 // CHECK: %[[B_OFFSET:.*]] = add nsw i64 %[[B_DIM_OFFSET]], 0
119 // CHECK: %[[B_BASE_GEP:.*]] = getelementptr {{.*}}, ptr %{{.*}}, i32 0, i32 0
120 // CHECK: %[[B_BASE:.*]] = load ptr, ptr %[[B_BASE_GEP]]
121 // CHECK: %[[B_VOID_ADDR:.*]] = getelementptr i8, ptr %[[B_BASE]], i64 %[[B_OFFSET]]
122 // CHECK: %[[B_VAL:.*]] = load float, ptr %[[B_VOID_ADDR]]
123 // CHECK: fadd float %[[B_VAL]], %[[F]]
124 %5 = fir.array_fetch %3, %arg3 : (!fir.array<?xf32>, index) -> f32
125 %6 = arith.addf %5, %arg2 : f32
126 %7 = fir.array_update %arg4, %6, %arg3 : (!fir.array<?xf32>, f32, index) -> !fir.array<?xf32>
127 fir.result %7 : !fir.array<?xf32>
129 fir.array_merge_store %2, %4 to %arg0 : !fir.array<?xf32>, !fir.array<?xf32>, !fir.box<!fir.array<?xf32>>
134 // Overlapping array expression assignment with a potentially non
135 // contiguous array (e.g. `a(2:10:1) = a(1:9:1) + f`, with a assumed shape).
136 // Test that a temp is created.
137 // CHECK-LINE: define void @f6
138 // CHECK: (ptr %[[A:[^,]*]], float %[[F:.*]])
139 func.func @f6(%arg0: !fir.box<!fir.array<?xf32>>, %arg1: f32) {
140 %c0 = arith.constant 0 : index
141 %c1 = arith.constant 1 : index
142 %c2 = arith.constant 2 : index
143 %c9 = arith.constant 9 : index
144 %c10 = arith.constant 10 : index
146 // CHECK: %[[EXT_GEP:.*]] = getelementptr {{.*}} %[[A]], i32 0, i32 7, i64 0, i32 1
147 // CHECK: %[[EXTENT:.*]] = load i64, ptr %[[EXT_GEP]]
148 // CHECK: %[[SIZE:.*]] = mul i64 ptrtoint (ptr getelementptr (float, ptr null, i32 1) to i64), %[[EXTENT]]
149 // CHECK: %[[MALLOC:.*]] = call ptr @malloc(i64 %[[SIZE]])
150 %1 = fir.slice %c2, %c10, %c1 : (index, index, index) -> !fir.slice<1>
151 %2 = fir.array_load %arg0 [%1] : (!fir.box<!fir.array<?xf32>>, !fir.slice<1>) -> !fir.array<?xf32>
152 %3 = fir.slice %c1, %c9, %c1 : (index, index, index) -> !fir.slice<1>
153 %4 = fir.array_load %arg0 [%3] : (!fir.box<!fir.array<?xf32>>, !fir.slice<1>) -> !fir.array<?xf32>
154 %5 = fir.do_loop %arg2 = %c0 to %c9 step %c1 iter_args(%arg3 = %2) -> (!fir.array<?xf32>) {
155 %6 = fir.array_fetch %4, %arg2 : (!fir.array<?xf32>, index) -> f32
156 %7 = arith.addf %6, %arg1 : f32
157 %8 = fir.array_update %arg3, %7, %arg2 : (!fir.array<?xf32>, f32, index) -> !fir.array<?xf32>
158 fir.result %8 : !fir.array<?xf32>
160 fir.array_merge_store %2, %5 to %arg0[%1] : !fir.array<?xf32>, !fir.array<?xf32>, !fir.box<!fir.array<?xf32>>, !fir.slice<1>
165 // Non contiguous array with lower bounds (x = y(100), with y(4:))
166 // Test array_coor offset computation.
167 // CHECK-LABEL: define void @f7(
168 // CHECK: ptr nocapture %[[X:[^,]*]], ptr %[[Y:.*]])
169 func.func @f7(%arg0: !fir.ref<f32>, %arg1: !fir.box<!fir.array<?xf32>>) {
170 %c4 = arith.constant 4 : index
171 %c100 = arith.constant 100 : index
172 %0 = fir.shift %c4 : (index) -> !fir.shift<1>
173 // CHECK: %[[STRIDE_GEP:.*]] = getelementptr {{.*}}, ptr %[[Y]], i32 0, i32 7, i32 0, i32 2
174 // CHECK: %[[STRIDE:.*]] = load i64, ptr %[[STRIDE_GEP]]
175 // CHECK: mul nsw i64 96, %[[STRIDE]]
176 %1 = fir.array_coor %arg1(%0) %c100 : (!fir.box<!fir.array<?xf32>>, !fir.shift<1>, index) -> !fir.ref<f32>
177 %2 = fir.load %1 : !fir.ref<f32>
178 fir.store %2 to %arg0 : !fir.ref<f32>
182 // Test A(:, :)%x reference codegen with A constant shape.
183 // CHECK-LABEL: define void @f8(
184 // CHECK-SAME: ptr nocapture %[[A:.*]], i32 %[[I:.*]])
185 func.func @f8(%a : !fir.ref<!fir.array<2x2x!fir.type<t{i:i32}>>>, %i : i32) {
186 %c0 = arith.constant 0 : index
187 %c1 = arith.constant 1 : index
188 %c2 = arith.constant 2 : index
189 %1 = fir.field_index i, !fir.type<t{i:i32}>
190 %2 = fir.shape %c2, %c2 : (index, index) -> !fir.shape<2>
191 %3 = fir.slice %c1, %c2, %c1, %c1, %c2, %c1 path %1 : (index, index, index, index, index, index, !fir.field) -> !fir.slice<2>
192 // CHECK: %[[GEP:.*]] = getelementptr %t, ptr %[[A]], i64 0, i32 0
193 %4 = fir.array_coor %a(%2) [%3] %c1, %c1 : (!fir.ref<!fir.array<2x2x!fir.type<t{i:i32}>>>, !fir.shape<2>, !fir.slice<2>, index, index) -> !fir.ref<i32>
194 // CHECK: store i32 %[[I]], ptr %[[GEP]], align 4
195 fir.store %i to %4 : !fir.ref<i32>
199 // Test casts in in array_coor offset computation when type parameters are not i64
200 // CHECK-LABEL: define ptr @f9(
201 // CHECK-SAME: i32 %[[I:.*]], i64 %{{.*}}, i64 %{{.*}}, ptr nocapture %[[C:.*]])
202 func.func @f9(%i: i32, %e : i64, %j: i64, %c: !fir.ref<!fir.array<?x?x!fir.char<1,?>>>) -> !fir.ref<!fir.char<1,?>> {
203 %s = fir.shape %e, %e : (i64, i64) -> !fir.shape<2>
204 // CHECK: %[[CAST:.*]] = sext i32 %[[I]] to i64
205 // CHECK: %[[OFFSET:.*]] = mul nsw i64 %{{.*}}, %[[CAST]]
206 // CHECK: getelementptr i8, ptr %[[C]], i64 %[[OFFSET]]
207 %a = fir.array_coor %c(%s) %j, %j typeparams %i : (!fir.ref<!fir.array<?x?x!fir.char<1,?>>>, !fir.shape<2>, i64, i64, i32) -> !fir.ref<!fir.char<1,?>>
208 return %a : !fir.ref<!fir.char<1,?>>