|
150
|
1 // RUN: mlir-opt %s -convert-linalg-to-parallel-loops -split-input-file | FileCheck %s --dump-input-on-failure
|
|
|
2
|
|
|
3 #map0 = affine_map<(d0, d1) -> (d0, d1)>
|
|
|
4 func @linalg_generic_sum(%lhs: memref<2x2xf32>,
|
|
|
5 %rhs: memref<2x2xf32>,
|
|
|
6 %sum: memref<2x2xf32>) {
|
|
|
7 linalg.generic {
|
|
|
8 args_in = 2 : i64,
|
|
|
9 args_out = 1 : i64,
|
|
|
10 indexing_maps = [#map0, #map0, #map0],
|
|
|
11 iterator_types = ["parallel", "parallel"]
|
|
|
12 } %lhs, %rhs, %sum {
|
|
|
13 ^bb0(%lhs_in: f32, %rhs_in: f32, %sum_out: f32): // no predecessors
|
|
|
14 %0 = addf %lhs_in, %rhs_in : f32
|
|
|
15 linalg.yield %0 : f32
|
|
|
16 }: memref<2x2xf32>, memref<2x2xf32>, memref<2x2xf32>
|
|
|
17 return
|
|
|
18 }
|
|
|
19 // CHECK-LABEL: @linalg_generic_sum
|
|
|
20 // CHECK: (%[[LHS:.*]]:{{.*}}, %[[RHS:.*]]:{{.*}}, %[[SUM:.*]]:{{.*}})
|
|
|
21 // CHECK-DAG: %[[C2:.*]] = constant 2
|
|
|
22 // CHECK-DAG: %[[C0:.*]] = constant 0
|
|
|
23 // CHECK-DAG: %[[C1:.*]] = constant 1
|
|
173
|
24 // CHECK: scf.parallel (%[[I:.*]], %[[J:.*]]) = {{.*}}
|
|
150
|
25 // CHECK: %[[LHS_ELEM:.*]] = load %[[LHS]][%[[I]], %[[J]]]
|
|
|
26 // CHECK: %[[RHS_ELEM:.*]] = load %[[RHS]][%[[I]], %[[J]]]
|
|
|
27 // CHECK: %[[SUM:.*]] = addf %[[LHS_ELEM]], %[[RHS_ELEM]] : f32
|
|
|
28 // CHECK: store %[[SUM]], %{{.*}}[%[[I]], %[[J]]]
|
|
173
|
29 // CHECK: scf.yield
|
|
150
|
30
|
|
|
31 // -----
|
|
|
32
|
|
|
33 #accesses = [
|
|
173
|
34 affine_map<(d0, d1, d2, d3) -> (d0, d1, d2, d3)>,
|
|
|
35 affine_map<(d0, d1, d2, d3) -> (d0, d1, d3)>
|
|
150
|
36 ]
|
|
|
37 #trait = {
|
|
|
38 args_in = 1,
|
|
|
39 args_out = 1,
|
|
173
|
40 iterator_types = ["parallel", "parallel", "reduction", "parallel"],
|
|
150
|
41 indexing_maps = #accesses
|
|
|
42 }
|
|
|
43
|
|
173
|
44 func @lower_outer_parallel(%A: memref<?x?x?x?xf32>, %B: memref<?x?x?xf32>) {
|
|
150
|
45 linalg.generic #trait %A, %B {
|
|
|
46 ^bb0(%a: f32, %b: f32):
|
|
|
47 linalg.yield %a: f32
|
|
173
|
48 } : memref<?x?x?x?xf32>, memref<?x?x?xf32>
|
|
150
|
49 return
|
|
|
50 }
|
|
173
|
51 // CHECK-LABEL: @lower_outer_parallel
|
|
|
52 // CHECK-DAG: %[[C0:.*]] = constant 0
|
|
|
53 // CHECK-DAG: %[[C1:.*]] = constant 1
|
|
|
54 // CHECK-DAG: %[[D0:.*]] = dim %{{.*}}, 0
|
|
|
55 // CHECK-DAG: %[[D1:.*]] = dim %{{.*}}, 1
|
|
|
56 // CHECK-DAG: %[[D2:.*]] = dim %{{.*}}, 2
|
|
|
57 // CHECK-DAG: %[[D3:.*]] = dim %{{.*}}, 3
|
|
|
58 // CHECK: scf.parallel (%[[IV0:.*]], %[[IV1:.*]]) = (%[[C0]], %[[C0]]) to (%[[D0]], %[[D1]]) step (%[[C1]], %[[C1]])
|
|
|
59 // CHECK: scf.for %[[IV2:.*]] = %[[C0]] to %[[D2]] step %[[C1]]
|
|
|
60 // CHECK: scf.for %[[IV3:.*]] = %[[C0]] to %[[D3]] step %[[C1]]
|
|
|
61 // CHECK: load %{{.*}}[%[[IV0]], %[[IV1]], %[[IV2]], %[[IV3]]]
|
|
|
62 // CHECK: store %{{.*}}, %{{.*}}[%[[IV0]], %[[IV1]], %[[IV3]]]
|
