Homec4science
Diffusion akantu (features/change-spiritv2-to-spiritx3) Browse

rAKA/examples/c++/scalability_test/plot.py
01d5a6530006features/change…

Last Change
Raw File

plot.py
View Options

#!/usr/bin/env python3
import os
import pandas as pd
import numpy as np
# needed to generate the plots on jed
import matplotlib
# matplotlib.use("TKAgg")
import matplotlib.pyplot as plt
# Same font as JOSS
plt.rcParams["font.sans-serif"] = "cmss10"
# Loading data
plots = {
"elastic gcc v5": {
"prefix": "timmings_",
"material": "elastic",
"compiler": "gcc",
"suffix": "_jed_v5.0.4"
},
"cohesive gcc v5": {
"prefix": "timmings_",
"material": "cohesive",
"compiler": "gcc",
"suffix": "_jed_v5.0.4"
},
"elastic gcc v4": {
"prefix": "timmings_",
"material": "elastic",
"compiler": "gcc",
"suffix": "_jed_v4.0.1"
},
"cohesive gcc v4": {
"prefix": "timmings_",
"material": "cohesive",
"compiler": "gcc",
"suffix": "_jed_v4.0.1"
},
}
fig, ax = plt.subplots(figsize=(4.5, 4))
# fig, ax = plt.subplots(1, 1)
plotting = "TTS"
handles = []
for plot_name, data in plots.items():
data["df"] = pd.read_csv(
os.path.join("results",
f"""{data["prefix"]}{data["material"]}_{data["compiler"]}{data["suffix"]}.csv"""),
sep=",",
skipinitialspace=True,
)
df = data["df"]
step = df["solve_step"] * df["solve_step nb_rep"]
if data["material"] == "cohesive":
step = step + df["check_cohesive_stress"] * df["check_cohesive_stress nb_rep"]
df["TTS"] = step
df["speedup"] = step[0] / step
df["mumps"] = df["static_solve"] * df["static_solve nb_rep"]
def plot_measure(ax, df, plotting, label, **kwargs):
"""Plot a given measure."""
grouped = df.groupby("psize") # compute stats grouped by number of procs
med = grouped.median()
min = grouped.min()
max = grouped.max()
min_psize = df["psize"][0]
print(list(med[plotting]))
(l,) = ax.plot(med.index, med[plotting], label=f"{label} (median)", **kwargs)
ax.fill_between(
med.index, min[plotting], max[plotting], color=l.get_color(), alpha=0.2
)
ax.plot(med.index, min_psize * med[plotting][min_psize] / med.index, ls="--", color=l.get_color())
# ax.boxplot(
# data[plotting]["grouped"], positions=psize, widths=[0.1 * s for s in psize]
# )
plot_measure(
ax,
plots["cohesive gcc v5"]["df"],
plotting,
"insertion",
marker="o",
)
plot_measure(
ax,
plots["elastic gcc v5"]["df"],
plotting,
"no insertion v5",
marker="o",
)
# Selecting appropriate tick values
psize = np.array(np.unique(plots[list(plots.keys())[0]]["df"]["psize"]))
labels = np.concatenate(
[[psize[0]], psize[1:][psize[1:] >= 2 * psize[:-1]], [psize[-1]]]
)
# for name, ax in axes.items():
ax.set_xscale("log", base=2)
ax.set_yscale("log")
ylabel = plotting if plotting != "TTS" else "Time to solution"
yunit = "s" if plotting != "speedup" else "-"
ax.set_xlabel("Nb Cores [-]")
ax.set_ylabel(f"""{ylabel} [{yunit}]""")
ax.set_xticks(ticks=labels, labels=map(str, labels))
# Constructing legend with min/max and ideal labels
handles, labels = ax.get_legend_handles_labels()
handles += [
matplotlib.lines.Line2D([], [], linestyle="--", color="k"),
matplotlib.patches.Patch(color="k", alpha=0.2),
]
labels += ["ideal", "min/max"]
ax.legend(handles=handles, labels=labels)
fig.tight_layout()
fig.savefig(f"{plotting}.svg", transparent=True, bbox_inches="tight", pad_inches=0.1)
plt.show()
c4science · Help