xref: /llvm-project/mlir/test/Integration/Dialect/Vector/CPU/AMX/tilezero.mlir (revision 435114f9fe2139bec770e5a95799f4eab20639e7) 
1// RUN: mlir-opt %s -convert-vector-to-scf -lower-affine -convert-scf-to-cf -convert-vector-to-llvm="enable-amx" -finalize-memref-to-llvm -convert-func-to-llvm -reconcile-unrealized-casts | \
2// RUN: mlir-translate -mlir-to-llvmir | \
3// RUN: %lli --entry-function=entry --mattr="+amx-tile,+amx-int8,+amx-bf16" --dlopen=%mlir_c_runner_utils | \
4// RUN: FileCheck %s
5
6// Note: To run this test, your CPU must support AMX.
7
8func.func @tilezero(%arg0: memref<?x?xi32>, %i: index, %j: index) {
9  %1 = amx.tile_zero : vector<16x16xi32>
10  amx.tile_store %arg0[%i, %j], %1 : memref<?x?xi32>, vector<16x16xi32>
11  return
12}
13
14func.func @entry() -> i32 {
15  %i0 = arith.constant 0: i32
16  %i1 = arith.constant 1: i32
17  %c0 = arith.constant 0: index
18  %c1 = arith.constant 1: index
19  %c3 = arith.constant 3: index
20  %c19 = arith.constant 19: index
21
22  // Set up memory.
23  %a = memref.alloc(%c19, %c19) : memref<?x?xi32>
24  scf.for %i = %c0 to %c19 step %c1 {
25    scf.for %j = %c0 to %c19 step %c1 {
26      memref.store %i1, %a[%i, %j] : memref<?x?xi32>
27    }
28  }
29
30  // Call kernel.
31  call @tilezero(%a, %c1, %c1) : (memref<?x?xi32>, index, index) -> ()
32
33  // Print and verify that the tilezero is correctly strided within
34  // the enveloping 19x19 buffer.
35  //
36  // CHECK:      ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 )
37  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
38  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
39  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
40  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
41  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
42  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
43  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
44  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
45  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
46  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
47  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
48  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
49  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
50  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
51  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
52  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
53  // CHECK-NEXT: ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 )
54  // CHECK-NEXT: ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 )
55  //
56  scf.for %i = %c0 to %c19 step %c1 {
57    %av = vector.transfer_read %a[%i, %c0], %i0: memref<?x?xi32>, vector<19xi32>
58    vector.print %av : vector<19xi32>
59  }
60
61  // Call kernel with different indices.
62  call @tilezero(%a, %c0, %c3) : (memref<?x?xi32>, index, index) -> ()
63
64  // Print and verify that the tilezero is again correctly strided
65  // within the enveloping 19x19 buffer.
66  //
67  // CHECK-NEXT: ( 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
68  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
69  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
70  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
71  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
72  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
73  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
74  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
75  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
76  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
77  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
78  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
79  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
80  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
81  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
82  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 )
83  // CHECK-NEXT: ( 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1 )
84  // CHECK-NEXT: ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 )
85  // CHECK-NEXT: ( 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 )
86  //
87  scf.for %i = %c0 to %c19 step %c1 {
88    %av = vector.transfer_read %a[%i, %c0], %i0: memref<?x?xi32>, vector<19xi32>
89    vector.print %av : vector<19xi32>
90  }
91
92  // Release resources.
93  memref.dealloc %a : memref<?x?xi32>
94
95  return %i0 : i32
96}
97