Skip to content

Commit 0cb9c87

Browse files
zhczhongzhczhong
committed
update
1 parent 48090c4 commit 0cb9c87

File tree

3 files changed

+15
-16
lines changed

3 files changed

+15
-16
lines changed

‎test/mlir/test/gc/Integration/op/binary.mlir‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
// RUN: gc-opt %s --gc-gpu-pipeline -split-input-file | FileCheck %s
22

33
// CHECK-LABEL: llvm
4-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
4+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
55
func.func @multiply(%arg0: memref<1024x1024xf16>, %arg1: memref<1024x1024xf16>, %arg2: memref<1024x1024xf16>, %arg3: memref<1024x1024xf16>) {
66
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf16>
77
%1 = bufferization.to_tensor %arg1 restrict : memref<1024x1024xf16>
@@ -16,7 +16,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
1616
// -----
1717

1818
// CHECK-LABEL: llvm
19-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
19+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
2020
func.func @add(%arg0: memref<1024x1024xf16>, %arg1: memref<1024x1024xf16>, %arg2: memref<1024x1024xf16>, %arg3: memref<1024x1024xf16>) {
2121
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf16>
2222
%1 = bufferization.to_tensor %arg1 restrict : memref<1024x1024xf16>
@@ -30,7 +30,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
3030

3131
// -----
3232
// CHECK-LABEL: llvm
33-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
33+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
3434
func.func @subtract(%arg0: memref<1024x1024xf16>, %arg1: memref<1024x1024xf16>, %arg2: memref<1024x1024xf16>, %arg3: memref<1024x1024xf16>) {
3535
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf16>
3636
%1 = bufferization.to_tensor %arg1 restrict : memref<1024x1024xf16>
@@ -44,7 +44,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
4444

4545
// -----
4646
// CHECK-LABEL: llvm
47-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
47+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
4848
func.func @divide(%arg0: memref<1024x1024xf16>, %arg1: memref<1024x1024xf16>, %arg2: memref<1024x1024xf16>, %arg3: memref<1024x1024xf16>) {
4949
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf16>
5050
%1 = bufferization.to_tensor %arg1 restrict : memref<1024x1024xf16>

‎test/mlir/test/gc/Integration/op/matmul.mlir‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
// RUN: gc-opt %s --gc-gpu-pipeline -split-input-file | FileCheck %s
22

33
// CHECK-LABEL: llvm
4-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
4+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
55
func.func @matmul_f16(%arg0: memref<4096x4096xf16>, %arg1: memref<4096x4096xf16>, %arg2: memref<4096x4096xf16>) {
66
%0 = bufferization.to_tensor %arg0 restrict : memref<4096x4096xf16>
77
%1 = bufferization.to_tensor %arg1 restrict : memref<4096x4096xf16>
@@ -16,7 +16,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
1616

1717
// -----
1818
// CHECK-LABEL: llvm
19-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
19+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
2020
func.func @corner_shape_matmul_f16(%arg0: memref<521x521xf16>, %arg1: memref<521x521xf16>, %arg2: memref<521x521xf16>) {
2121
%0 = bufferization.to_tensor %arg0 restrict : memref<521x521xf16>
2222
%1 = bufferization.to_tensor %arg1 restrict : memref<521x521xf16>
@@ -31,7 +31,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
3131

3232
// -----
3333
// CHECK-LABEL: llvm
34-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
34+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>}{
3535
func.func @dynamic_matmul_f16(%arg0: memref<?x?xf16>, %arg1: memref<1024x1024xf16>, %arg2: memref<?x1024xf16>) {
3636
%0 = bufferization.to_tensor %arg0 restrict : memref<?x?xf16>
3737
%c0 = arith.constant 0 : index
@@ -50,7 +50,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
5050

5151
// -----
5252
// CHECK-LABEL: llvm
53-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
53+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
5454
func.func @matmul_bf16(%arg0: memref<4096x4096xbf16>, %arg1: memref<4096x4096xbf16>, %arg2: memref<4096x4096xbf16>) {
5555
%0 = bufferization.to_tensor %arg0 restrict : memref<4096x4096xbf16>
5656
%1 = bufferization.to_tensor %arg1 restrict : memref<4096x4096xbf16>
@@ -65,7 +65,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
6565

6666
// -----
6767
// CHECK-LABEL: llvm
68-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
68+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
6969
func.func @matmul_f32(%arg0: memref<4096x4096xf32>, %arg1: memref<4096x4096xf32>, %arg2: memref<4096x4096xf32>) {
7070
%0 = bufferization.to_tensor %arg0 restrict : memref<4096x4096xf32>
7171
%1 = bufferization.to_tensor %arg1 restrict : memref<4096x4096xf32>
@@ -76,4 +76,4 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
7676
bufferization.materialize_in_destination %4 in restrict writable %arg2 : (tensor<4096x4096xf32>, memref<4096x4096xf32>) -> ()
7777
return
7878
}
79-
}
79+
}

‎test/mlir/test/gc/Integration/op/relu.mlir‎

Lines changed: 5 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,7 @@
33

44
// -----
55
// CHECK-LABEL: llvm
6-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
6+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
77
func.func @relu_f16(%arg0: memref<1024x1024xf16>, %arg1: memref<1024x1024xf16>) {
88
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf16>
99
%1 = tensor.empty() : tensor<1024x1024xf16>
@@ -17,7 +17,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
1717

1818
// -----
1919
// CHECK-LABEL: llvm
20-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
20+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
2121
func.func @dynamic_relu(%arg0: memref<?x?xf16>, %arg1: memref<?x?xf16>) {
2222
%0 = bufferization.to_tensor %arg0 restrict : memref<?x?xf16>
2323
%c0 = arith.constant 0 : index
@@ -35,7 +35,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
3535

3636
// -----
3737
// CHECK-LABEL: llvm
38-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
38+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
3939
func.func @relu_bf16(%arg0: memref<1024x1024xbf16>, %arg1: memref<1024x1024xbf16>) {
4040
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xbf16>
4141
%1 = tensor.empty() : tensor<1024x1024xbf16>
@@ -49,7 +49,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
4949

5050
// -----
5151
// CHECK-LABEL: llvm
52-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
52+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
5353
func.func @relu_f32(%arg0: memref<1024x1024xf32>, %arg1: memref<1024x1024xf32>) {
5454
%0 = bufferization.to_tensor %arg0 restrict : memref<1024x1024xf32>
5555
%1 = tensor.empty() : tensor<1024x1024xf32>
@@ -63,7 +63,7 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
6363

6464
// -----
6565
// CHECK-LABEL: llvm
66-
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"CPU" : #dlti.target_device_spec<#dlti.dl_entry<"tile_size", 32 : i32>>>} {
66+
module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"GPU" : #dlti.target_device_spec<#dlti.dl_entry<"num_exec_units", 448 : i32>, #dlti.dl_entry<"num_exec_units_per_slice", 32 : i32>, #dlti.dl_entry<"num_threads_per_eu", 8 : i32>, #dlti.dl_entry<"L1_cache_size_in_bytes", 67108864 : i32>, #dlti.dl_entry<"max_vector_op_width", 256 : i32>, #dlti.dl_entry<"max_work_group_size", 1024 : i32>>>} {
6767
func.func @relu_f32_corner_shape(%arg0: memref<1061x1061xf32>, %arg1: memref<1061x1061xf32>) {
6868
%0 = bufferization.to_tensor %arg0 restrict : memref<1061x1061xf32>
6969
%1 = tensor.empty() : tensor<1061x1061xf32>
@@ -74,4 +74,3 @@ module @fragment_name attributes {"#dlti.sys_spec" = #dlti.target_system_spec<"C
7474
return
7575
}
7676
}
77-

0 commit comments

Comments
 (0)