.. DO NOT EDIT. .. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY. .. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE: .. "_as_gen/benchmarks/run_compute_histogram.py" .. LINE NUMBERS ARE GIVEN BELOW. .. only:: html .. note:: :class: sphx-glr-download-link-note :ref:`Go to the end ` to download the full example code. .. rst-class:: sphx-glr-example-title .. _sphx_glr__as_gen_benchmarks_run_compute_histogram.py: Compute histogram performance benchmarks ================================= This example benchmarks the compute histogram performance for the different algorithms available in Shamrock .. GENERATED FROM PYTHON SOURCE LINES 7-24 .. code-block:: Python :lineno-start: 9 import random import time import matplotlib.colors as colors import matplotlib.pyplot as plt import numpy as np import shamrock # If we use the shamrock executable to run this script instead of the python interpreter, # we should not initialize the system as the shamrock executable needs to handle specific MPI logic if not shamrock.sys.is_initialized(): shamrock.change_loglevel(1) shamrock.sys.init("0:0") .. GENERATED FROM PYTHON SOURCE LINES 27-31 .. code-block:: Python :lineno-start: 27 impl_control = shamrock.algs.compute_histogram_impl() print(impl_control.get_alg_name()) .. rst-class:: sphx-glr-script-out .. code-block:: none compute_histogram .. GENERATED FROM PYTHON SOURCE LINES 32-34 .. code-block:: Python :lineno-start: 32 impl_control.was_configured() .. rst-class:: sphx-glr-script-out .. code-block:: none False .. GENERATED FROM PYTHON SOURCE LINES 35-40 .. code-block:: Python :lineno-start: 35 default_config = impl_control.get_default_config() print(f"Current config: {impl_control.get_config()}") print(f"Default config: {default_config}") print(f"Available configs: {impl_control.get_avail_configs()}") .. rst-class:: sphx-glr-script-out .. code-block:: none Info: no autotuning registered for compute_histogram [Algs][rank=0] Info: no autotuning registered for compute_histogram [Algs][rank=0] Info: switching config for alg compute_histogram to cfg=naive_gpu [Algs][rank=0] Current config: naive_gpu Default config: naive_gpu Available configs: ['reference', 'naive_gpu', 'gpu_team_fetching', 'gpu_oversubscribe'] .. GENERATED FROM PYTHON SOURCE LINES 41-75 .. code-block:: Python :lineno-start: 41 bin_edges = np.linspace(0, 1, 2049) bin_edge_inf = bin_edges[:-1] bin_edge_sup = bin_edges[1:] rng = np.random.default_rng() positions = rng.random(int(1e6)) bin_edge_inf_f32 = bin_edge_inf.astype(np.float32) bin_edge_sup_f32 = bin_edge_sup.astype(np.float32) positions_f32 = positions.astype(np.float32) buf_bin_edge_inf = shamrock.backends.DeviceBuffer_f64() buf_bin_edge_sup = shamrock.backends.DeviceBuffer_f64() buf_positions = shamrock.backends.DeviceBuffer_f64() buf_bin_edge_inf.resize(len(bin_edge_inf)) buf_bin_edge_sup.resize(len(bin_edge_sup)) buf_positions.resize(len(positions)) buf_bin_edge_inf.copy_from_stdvec(bin_edge_inf) buf_bin_edge_sup.copy_from_stdvec(bin_edge_sup) buf_positions.copy_from_stdvec(positions) buf_bin_edge_inf_f32 = shamrock.backends.DeviceBuffer_f32() buf_bin_edge_sup_f32 = shamrock.backends.DeviceBuffer_f32() buf_positions_f32 = shamrock.backends.DeviceBuffer_f32() buf_bin_edge_inf_f32.resize(len(bin_edge_inf_f32)) buf_bin_edge_sup_f32.resize(len(bin_edge_sup_f32)) buf_positions_f32.resize(len(positions_f32)) buf_bin_edge_inf_f32.copy_from_stdvec(bin_edge_inf_f32) buf_bin_edge_sup_f32.copy_from_stdvec(bin_edge_sup_f32) buf_positions_f32.copy_from_stdvec(positions_f32) .. GENERATED FROM PYTHON SOURCE LINES 76-91 .. code-block:: Python :lineno-start: 76 results_f64 = {} results_f32 = {} avail_configs = impl_control.get_avail_configs() for config in avail_configs: impl_control.set_config(config) time_f64 = shamrock.algs.benchmark_compute_histogram_basic_f64( buf_bin_edge_inf, buf_bin_edge_sup, buf_positions ) time_f32 = shamrock.algs.benchmark_compute_histogram_basic_f32( buf_bin_edge_inf_f32, buf_bin_edge_sup_f32, buf_positions_f32 ) print(f"Config: {config}, Time f64: {time_f64 * 1000}ms, Time f32: {time_f32 * 1000}ms") results_f64[config] = time_f64 * 1000 results_f32[config] = time_f32 * 1000 .. rst-class:: sphx-glr-script-out .. code-block:: none Info: switching config for alg compute_histogram to cfg=reference [Algs][rank=0] Config: reference, Time f64: 2371.757527ms, Time f32: 2366.8604170000003ms Info: switching config for alg compute_histogram to cfg=naive_gpu [Algs][rank=0] Config: naive_gpu, Time f64: 894.8318885000001ms, Time f32: 893.2677315000001ms Info: switching config for alg compute_histogram to cfg=gpu_team_fetching [Algs][rank=0] Config: gpu_team_fetching, Time f64: 1212.12335ms, Time f32: 1252.430123ms Info: switching config for alg compute_histogram to cfg=gpu_oversubscribe [Algs][rank=0] Config: gpu_oversubscribe, Time f64: 2985.355484ms, Time f32: 2296.522044ms .. GENERATED FROM PYTHON SOURCE LINES 92-93 plot the histogram .. GENERATED FROM PYTHON SOURCE LINES 93-99 .. code-block:: Python :lineno-start: 93 result = shamrock.algs.compute_histogram_basic_f64( buf_bin_edge_inf, buf_bin_edge_sup, buf_positions ) plt.plot(result.copy_to_stdvec()) plt.show() .. image-sg:: /_as_gen/benchmarks/images/sphx_glr_run_compute_histogram_001.png :alt: run compute histogram :srcset: /_as_gen/benchmarks/images/sphx_glr_run_compute_histogram_001.png :class: sphx-glr-single-img .. GENERATED FROM PYTHON SOURCE LINES 100-101 plot the results .. GENERATED FROM PYTHON SOURCE LINES 101-127 .. code-block:: Python :lineno-start: 101 plt.figure(layout="constrained") configs = list(results_f64.keys()) vals_f64 = [results_f64[c] for c in configs] vals_f32 = [results_f32[c] for c in configs] x = np.arange(len(configs)) bar_w = 0.35 plt.bar(x - bar_w / 2, vals_f64, bar_w, label="f64") plt.bar(x + bar_w / 2, vals_f32, bar_w, label="f32") plt.xticks(x, configs, rotation=45, ha="right") for tick_label, cfg in zip(plt.gca().get_xticklabels(), configs): if cfg == default_config: tick_label.set_color("red") plt.ylabel("Time (ms)") plt.yscale("log") _ymin, _ymax = plt.gca().get_ylim() _ymin = 10 ** int(np.floor(np.log10(_ymin))) _ymax = 10 ** int(np.ceil(np.log10(_ymax))) plt.ylim(_ymin, _ymax * 1.1) plt.title("Compute histogram performance benchmarks") plt.legend() plt.grid(True, alpha=0.3) plt.show() .. image-sg:: /_as_gen/benchmarks/images/sphx_glr_run_compute_histogram_002.png :alt: Compute histogram performance benchmarks :srcset: /_as_gen/benchmarks/images/sphx_glr_run_compute_histogram_002.png :class: sphx-glr-single-img .. rst-class:: sphx-glr-timing **Total running time of the script:** (0 minutes 34.897 seconds) **Estimated memory usage:** 157 MB .. _sphx_glr_download__as_gen_benchmarks_run_compute_histogram.py: .. only:: html .. container:: sphx-glr-footer sphx-glr-footer-example .. container:: sphx-glr-download sphx-glr-download-jupyter :download:`Download Jupyter notebook: run_compute_histogram.ipynb ` .. container:: sphx-glr-download sphx-glr-download-python :download:`Download Python source code: run_compute_histogram.py ` .. container:: sphx-glr-download sphx-glr-download-zip :download:`Download zipped: run_compute_histogram.zip ` .. only:: html .. rst-class:: sphx-glr-signature `Gallery generated by Sphinx-Gallery `_