Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 1 | /* |
Sang-Hoon Park | 0094c02 | 2021-01-20 18:16:47 +0000 | [diff] [blame] | 2 | * Copyright (c) 2016-2021 Arm Limited. |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 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 "arm_compute/runtime/IScheduler.h" |
| 25 | |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 26 | #include "arm_compute/core/CPP/ICPPKernel.h" |
Anthony Barbier | 148b075 | 2018-09-11 14:19:39 +0100 | [diff] [blame] | 27 | #include "arm_compute/core/Error.h" |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 28 | #include "arm_compute/core/Window.h" |
Georgios Pinitas | 08302c1 | 2021-06-09 10:08:27 +0100 | [diff] [blame] | 29 | #include "src/common/cpuinfo/CpuInfo.h" |
Sang-Hoon Park | 68dd25f | 2020-10-19 16:00:11 +0100 | [diff] [blame] | 30 | #include "src/runtime/SchedulerUtils.h" |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 31 | |
| 32 | namespace arm_compute |
| 33 | { |
| 34 | IScheduler::IScheduler() |
| 35 | { |
Georgios Pinitas | 53d1227 | 2018-02-01 20:23:25 +0000 | [diff] [blame] | 36 | // Work out the best possible number of execution threads |
Georgios Pinitas | 08302c1 | 2021-06-09 10:08:27 +0100 | [diff] [blame] | 37 | _num_threads_hint = cpuinfo::num_threads_hint(); |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 38 | } |
| 39 | |
Pablo Tello | 7fad9b1 | 2018-03-14 17:55:27 +0000 | [diff] [blame] | 40 | CPUInfo &IScheduler::cpu_info() |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 41 | { |
Michalis Spyrou | 20fca52 | 2021-06-07 14:23:57 +0100 | [diff] [blame] | 42 | return CPUInfo::get(); |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 43 | } |
Georgios Pinitas | 53d1227 | 2018-02-01 20:23:25 +0000 | [diff] [blame] | 44 | |
Georgios Pinitas | 06e890b | 2020-07-09 18:38:34 +0100 | [diff] [blame] | 45 | void IScheduler::set_num_threads_with_affinity(unsigned int num_threads, BindFunc func) |
| 46 | { |
| 47 | ARM_COMPUTE_UNUSED(num_threads, func); |
| 48 | ARM_COMPUTE_ERROR("Feature for affinity setting is not implemented"); |
| 49 | } |
| 50 | |
Georgios Pinitas | 53d1227 | 2018-02-01 20:23:25 +0000 | [diff] [blame] | 51 | unsigned int IScheduler::num_threads_hint() const |
| 52 | { |
| 53 | return _num_threads_hint; |
| 54 | } |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 55 | |
Sang-Hoon Park | 0094c02 | 2021-01-20 18:16:47 +0000 | [diff] [blame] | 56 | void IScheduler::schedule_common(ICPPKernel *kernel, const Hints &hints, const Window &window, ITensorPack &tensors) |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 57 | { |
| 58 | ARM_COMPUTE_ERROR_ON_MSG(!kernel, "The child class didn't set the kernel"); |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 59 | #ifndef BARE_METAL |
Sang-Hoon Park | 0094c02 | 2021-01-20 18:16:47 +0000 | [diff] [blame] | 60 | const Window &max_window = window; |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 61 | if(hints.split_dimension() == IScheduler::split_dimensions_all) |
| 62 | { |
| 63 | /* |
| 64 | * if the split dim is size_t max then this signals we should parallelise over |
| 65 | * all dimensions |
| 66 | */ |
| 67 | const std::size_t m = max_window.num_iterations(Window::DimX); |
| 68 | const std::size_t n = max_window.num_iterations(Window::DimY); |
| 69 | |
| 70 | //in c++17 this can be swapped for auto [ m_threads, n_threads ] = split_2d(... |
| 71 | unsigned m_threads, n_threads; |
Sang-Hoon Park | 68dd25f | 2020-10-19 16:00:11 +0100 | [diff] [blame] | 72 | std::tie(m_threads, n_threads) = scheduler_utils::split_2d(this->num_threads(), m, n); |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 73 | |
| 74 | std::vector<IScheduler::Workload> workloads; |
| 75 | for(unsigned int ni = 0; ni != n_threads; ++ni) |
| 76 | { |
| 77 | for(unsigned int mi = 0; mi != m_threads; ++mi) |
| 78 | { |
| 79 | workloads.push_back( |
| 80 | [ni, mi, m_threads, n_threads, &max_window, &kernel](const ThreadInfo & info) |
| 81 | { |
| 82 | //narrow the window to our mi-ni workload |
| 83 | Window win = max_window.split_window(Window::DimX, mi, m_threads) |
| 84 | .split_window(Window::DimY, ni, n_threads); |
| 85 | |
| 86 | win.validate(); |
| 87 | |
| 88 | Window thread_locator; |
| 89 | thread_locator.set(Window::DimX, Window::Dimension(mi, m_threads)); |
| 90 | thread_locator.set(Window::DimY, Window::Dimension(ni, n_threads)); |
| 91 | |
| 92 | thread_locator.validate(); |
| 93 | |
| 94 | kernel->run_nd(win, info, thread_locator); |
| 95 | }); |
| 96 | } |
| 97 | } |
| 98 | run_workloads(workloads); |
| 99 | } |
| 100 | else |
| 101 | { |
| 102 | const unsigned int num_iterations = max_window.num_iterations(hints.split_dimension()); |
| 103 | const unsigned int num_threads = std::min(num_iterations, this->num_threads()); |
| 104 | |
| 105 | if(num_iterations == 0) |
| 106 | { |
| 107 | return; |
| 108 | } |
| 109 | |
| 110 | if(!kernel->is_parallelisable() || num_threads == 1) |
| 111 | { |
| 112 | ThreadInfo info; |
Michalis Spyrou | 20fca52 | 2021-06-07 14:23:57 +0100 | [diff] [blame] | 113 | info.cpu_info = &cpu_info(); |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 114 | if(tensors.empty()) |
| 115 | { |
| 116 | kernel->run(max_window, info); |
| 117 | } |
| 118 | else |
| 119 | { |
| 120 | kernel->run_op(tensors, max_window, info); |
| 121 | } |
| 122 | } |
| 123 | else |
| 124 | { |
| 125 | unsigned int num_windows = 0; |
| 126 | switch(hints.strategy()) |
| 127 | { |
| 128 | case StrategyHint::STATIC: |
| 129 | num_windows = num_threads; |
| 130 | break; |
| 131 | case StrategyHint::DYNAMIC: |
| 132 | { |
| 133 | const unsigned int granule_threshold = (hints.threshold() <= 0) ? num_threads : static_cast<unsigned int>(hints.threshold()); |
| 134 | // Make sure we don't use some windows which are too small as this might create some contention on the ThreadFeeder |
| 135 | num_windows = num_iterations > granule_threshold ? granule_threshold : num_iterations; |
| 136 | break; |
| 137 | } |
| 138 | default: |
| 139 | ARM_COMPUTE_ERROR("Unknown strategy"); |
| 140 | } |
| 141 | std::vector<IScheduler::Workload> workloads(num_windows); |
| 142 | for(unsigned int t = 0; t < num_windows; ++t) |
| 143 | { |
| 144 | //Capture 't' by copy, all the other variables by reference: |
| 145 | workloads[t] = [t, &hints, &max_window, &num_windows, &kernel, &tensors](const ThreadInfo & info) |
| 146 | { |
| 147 | Window win = max_window.split_window(hints.split_dimension(), t, num_windows); |
| 148 | win.validate(); |
| 149 | |
| 150 | if(tensors.empty()) |
| 151 | { |
| 152 | kernel->run(win, info); |
| 153 | } |
| 154 | else |
| 155 | { |
| 156 | kernel->run_op(tensors, win, info); |
| 157 | } |
| 158 | }; |
| 159 | } |
| 160 | run_workloads(workloads); |
| 161 | } |
| 162 | } |
Sang-Hoon Park | 0094c02 | 2021-01-20 18:16:47 +0000 | [diff] [blame] | 163 | #else /* !BARE_METAL */ |
| 164 | ARM_COMPUTE_UNUSED(kernel, hints, window, tensors); |
morgolock | 5111264 | 2020-08-20 14:51:39 +0100 | [diff] [blame] | 165 | #endif /* !BARE_METAL */ |
| 166 | } |
| 167 | |
Anthony Barbier | 148b075 | 2018-09-11 14:19:39 +0100 | [diff] [blame] | 168 | void IScheduler::run_tagged_workloads(std::vector<Workload> &workloads, const char *tag) |
| 169 | { |
| 170 | ARM_COMPUTE_UNUSED(tag); |
| 171 | run_workloads(workloads); |
| 172 | } |
| 173 | |
Moritz Pflanzer | beabe3b | 2017-08-31 14:56:32 +0100 | [diff] [blame] | 174 | } // namespace arm_compute |