""" Dusty disc performance results ============================== Solver rate of the dusty disc perf test (``dustydisc_perftest.py``) on an H200, as a function of the number of dust species. """ import matplotlib as mpl import matplotlib.pyplot as plt import numpy as np mpl.rcParams.update( { "font.family": "serif", "mathtext.fontset": "cm", "font.size": 14, "axes.labelsize": 16, "axes.titlesize": 16, "xtick.labelsize": 13, "ytick.labelsize": 13, "legend.fontsize": 13, "axes.facecolor": "white", "axes.linewidth": 1.0, "xtick.direction": "in", "ytick.direction": "in", "xtick.top": True, "ytick.right": True, "xtick.major.size": 8, "ytick.major.size": 8, "xtick.minor.visible": True, "ytick.minor.visible": True, "legend.frameon": True, "legend.fancybox": False, "legend.edgecolor": "black", } ) RESULTS_H200_BASELINE = [ {"kernel": "M4", "ndust": 0, "rate": np.float64(20743174.163988266)}, {"kernel": "M6", "ndust": 0, "rate": np.float64(13799390.088761112)}, {"kernel": "M6", "ndust": 1, "rate": np.float64(12633702.042326497)}, {"kernel": "M6", "ndust": 2, "rate": np.float64(12087424.370965026)}, {"kernel": "M6", "ndust": 3, "rate": np.float64(11881595.212850157)}, {"kernel": "M6", "ndust": 4, "rate": np.float64(11300411.830739705)}, {"kernel": "M6", "ndust": 5, "rate": np.float64(11222855.348508552)}, {"kernel": "M6", "ndust": 6, "rate": np.float64(10755208.603137823)}, {"kernel": "M6", "ndust": 7, "rate": np.float64(10688006.42580047)}, {"kernel": "M6", "ndust": 8, "rate": np.float64(10142698.857492624)}, {"kernel": "M6", "ndust": 9, "rate": np.float64(10214823.874985265)}, {"kernel": "M6", "ndust": 10, "rate": np.float64(9761393.55028728)}, {"kernel": "M6", "ndust": 11, "rate": np.float64(9792483.14187795)}, {"kernel": "M6", "ndust": 12, "rate": np.float64(9219263.745632991)}, {"kernel": "M6", "ndust": 13, "rate": np.float64(9388890.664223792)}, {"kernel": "M6", "ndust": 14, "rate": np.float64(8948305.414125301)}, {"kernel": "M6", "ndust": 15, "rate": np.float64(9006443.219386235)}, {"kernel": "M6", "ndust": 16, "rate": np.float64(8591518.90067488)}, {"kernel": "M6", "ndust": 17, "rate": np.float64(8663145.798459316)}, {"kernel": "M6", "ndust": 18, "rate": np.float64(8252145.146574976)}, {"kernel": "M6", "ndust": 19, "rate": np.float64(8367198.978522308)}, {"kernel": "M6", "ndust": 20, "rate": np.float64(7823688.987683165)}, {"kernel": "M6", "ndust": 21, "rate": np.float64(8078987.197870458)}, {"kernel": "M6", "ndust": 22, "rate": np.float64(7692061.528587866)}, {"kernel": "M6", "ndust": 23, "rate": np.float64(7820297.588177133)}, {"kernel": "M6", "ndust": 24, "rate": np.float64(7334507.940110926)}, {"kernel": "M6", "ndust": 25, "rate": np.float64(7578401.6513876775)}, {"kernel": "M6", "ndust": 26, "rate": np.float64(7200301.321665853)}, {"kernel": "M6", "ndust": 27, "rate": np.float64(7341230.143859156)}, {"kernel": "M6", "ndust": 28, "rate": np.float64(6858490.479193717)}, {"kernel": "M6", "ndust": 29, "rate": np.float64(7151851.505432058)}, {"kernel": "M6", "ndust": 30, "rate": np.float64(6780589.608352743)}, {"kernel": "M6", "ndust": 31, "rate": np.float64(6957826.151235339)}, {"kernel": "M6", "ndust": 32, "rate": np.float64(6871322.974704008)}, {"kernel": "M6", "ndust": 33, "rate": np.float64(6766758.272378567)}, {"kernel": "M6", "ndust": 34, "rate": np.float64(6416124.696613544)}, {"kernel": "M6", "ndust": 35, "rate": np.float64(6595107.764492851)}, {"kernel": "M6", "ndust": 36, "rate": np.float64(6124688.97834106)}, {"kernel": "M6", "ndust": 37, "rate": np.float64(6428013.769450867)}, {"kernel": "M6", "ndust": 38, "rate": np.float64(6106585.969761499)}, {"kernel": "M6", "ndust": 39, "rate": np.float64(6261067.583511858)}, {"kernel": "M6", "ndust": 40, "rate": np.float64(5846567.91142668)}, {"kernel": "M6", "ndust": 41, "rate": np.float64(6130846.680557474)}, {"kernel": "M6", "ndust": 42, "rate": np.float64(5781410.068906898)}, {"kernel": "M6", "ndust": 43, "rate": np.float64(5978457.374960809)}, {"kernel": "M6", "ndust": 44, "rate": np.float64(5510755.835376551)}, {"kernel": "M6", "ndust": 45, "rate": np.float64(5847274.351241548)}, {"kernel": "M6", "ndust": 46, "rate": np.float64(5533625.961877414)}, {"kernel": "M6", "ndust": 47, "rate": np.float64(5712080.273564)}, {"kernel": "M6", "ndust": 48, "rate": np.float64(5405527.541108767)}, {"kernel": "M6", "ndust": 49, "rate": np.float64(5590447.5575348595)}, {"kernel": "M6", "ndust": 50, "rate": np.float64(5307054.447021384)}, {"kernel": "M6", "ndust": 51, "rate": np.float64(5474498.7673062915)}, {"kernel": "M6", "ndust": 52, "rate": np.float64(5065365.2100912845)}, {"kernel": "M6", "ndust": 53, "rate": np.float64(5353810.653996148)}, {"kernel": "M6", "ndust": 54, "rate": np.float64(5066047.936317459)}, {"kernel": "M6", "ndust": 55, "rate": np.float64(5242961.026497445)}, {"kernel": "M6", "ndust": 56, "rate": np.float64(4883427.61396701)}, {"kernel": "M6", "ndust": 57, "rate": np.float64(5132179.980544963)}, {"kernel": "M6", "ndust": 58, "rate": np.float64(4853087.799315191)}, {"kernel": "M6", "ndust": 59, "rate": np.float64(5020866.021284079)}, {"kernel": "M6", "ndust": 60, "rate": np.float64(4618314.988822764)}, {"kernel": "M6", "ndust": 61, "rate": np.float64(4921603.58455902)}, {"kernel": "M6", "ndust": 62, "rate": np.float64(4648092.049223436)}, ] def _lookup_rate(results, kernel, ndust): for entry in results: if entry["kernel"] == kernel and entry["ndust"] == ndust: return float(entry["rate"]) return None def plot_perf_results(results, title="Dusty disc performance"): visible_ndust = {1, 5, 10, 20, 30, 40, 50, 60, 70} labels = ["gas only M4", "gas only M6"] rates_off = [ _lookup_rate(results, "M4", 0), _lookup_rate(results, "M6", 0), ] ndust_values = sorted({entry["ndust"] for entry in results if entry["ndust"] > 0}) for ndust in ndust_values: labels.append(str(ndust)) rates_off.append(_lookup_rate(results, "M6", ndust)) fig, ax = plt.subplots(figsize=(6.4, 6)) gas_count = 2 gas_spread = 2.0 ndust_gap = 4.0 x = np.arange(len(labels), dtype=float) x[1:] += gas_spread x[gas_count:] += ndust_gap width = 2 gas_x = x[:gas_count] gas_y = rates_off[:gas_count] dust_x = x[gas_count:] dust_y = rates_off[gas_count:] ax.bar(gas_x, gas_y, width, label="coala off", color="C0") ax.plot(dust_x, dust_y, color="C0", marker="o", markersize=3) separator_x = (x[gas_count - 1] + x[gas_count]) / 2 ax.axvline(separator_x, color="0.4", linewidth=1.2, linestyle="-", zorder=0) # ax.set_yscale("log") # ax.set_ylim(1e6,3e7) ax.set_ylabel("rate (particles/s)") ax.set_xlabel("dust species", labelpad=-16) ax.set_title(title) baseline_rate = _lookup_rate(results, "M6", 0) if baseline_rate: ax.tick_params(axis="y", right=False, which="both") secax = ax.secondary_yaxis( "right", functions=(lambda r: r / baseline_rate, lambda s: s * baseline_rate) ) secax.set_ylabel("speedup relative to M6 (gas only)") for frac in (1.00, 0.75, 0.5): y = frac * baseline_rate ax.axhline(y, color="red", linewidth=1.0, linestyle="--", zorder=0) ax.text( x[-1], y, f"{frac:.0%}", color="red", fontsize=11, va="bottom", ha="right", ) ax.set_xticks(x) visible_indices = {0, 1} for ndust in visible_ndust: visible_indices.add(1 + ndust) tick_labels = [labels[i] if i in visible_indices else "" for i in range(len(labels))] ax.set_xticklabels(tick_labels, rotation=90, fontsize=12) for i, tick in enumerate(ax.xaxis.get_major_ticks()): if i not in visible_indices: tick.tick1line.set_visible(False) tick.tick2line.set_visible(False) # ax.legend() ax.grid(axis="y", which="both", linestyle=":", alpha=0.5) fig.tight_layout() return fig, ax # %% # Plot the results fig, ax = plot_perf_results(RESULTS_H200_BASELINE, title="Dusty disc performance (H200)") plt.show()