1 /*
2 * Copyright (c) 2022-2023 Arm Limited.
3 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24 #include "GpuKernelComponentGraph.h"
25
26 #include "arm_compute/dynamic_fusion/sketch/MemoryDescriptor.h"
27
28 namespace arm_compute
29 {
30 namespace experimental
31 {
32 namespace dynamic_fusion
33 {
get_tensor_ids(const std::vector<const ITensorInfo * > tensors)34 std::vector<DependencyGraph::TensorId> GpuKernelComponentGraph::get_tensor_ids(const std::vector<const ITensorInfo *> tensors)
35 {
36 std::vector<DependencyGraph::TensorId> tensor_ids{};
37 std::transform(
38 std::begin(tensors), std::end(tensors),
39 std::back_inserter(tensor_ids),
40 [](const auto & t)
41 {
42 return t->id();
43 });
44 return tensor_ids;
45 }
46
GpuKernelComponentGraph(GpuComponentServices * services)47 GpuKernelComponentGraph::GpuKernelComponentGraph(GpuComponentServices *services)
48 : _services{ services }, _components{}, _tensors{}, _dependency_graph{}
49 {
50 }
51
fuse(const MemoryDescriptorMap & mem_map) const52 GpuKernelComponentStream GpuKernelComponentGraph::fuse(const MemoryDescriptorMap &mem_map) const
53 {
54 GpuKernelComponentStream stream{ _services, mem_map };
55 const auto op_seq = _dependency_graph.build_operators_sequence();
56
57 stream.new_component_group();
58 for(auto op : op_seq)
59 {
60 const auto component = _components.at(op.op).get();
61 const auto success = stream.add_component(component);
62 if(!success) // Assume first failure was because the root component is unfusable
63 {
64 stream.new_component_group();
65 const auto success = stream.add_component(component);
66 ARM_COMPUTE_ERROR_ON(!success);
67 ARM_COMPUTE_UNUSED(success);
68 }
69 }
70
71 return stream;
72 }
73 } // namespace dynamic_fusion
74 } // namespace experimental
75 } // namespace arm_compute
76