Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

grass.benchmark: shuffle runs for benchmarking by number of cores #2152

Merged
merged 1 commit into from
Feb 8, 2022
Merged

grass.benchmark: shuffle runs for benchmarking by number of cores #2152

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 20 additions & 16 deletions python/grass/benchmark/runners.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@
"""Basic functions for benchmarking modules"""

import shutil
import random
from types import SimpleNamespace

import grass.script as gs
Expand Down Expand Up @@ -68,7 +69,7 @@ def benchmark_single(module, label, repeat=5):
)


def benchmark_nprocs(module, label, max_nprocs, repeat=5):
def benchmark_nprocs(module, label, max_nprocs, repeat=5, shuffle=True):
"""Benchmark module using values of nprocs up to *max_nprocs*.

*module* is an instance of PyGRASS Module class or any object which
Expand All @@ -82,6 +83,8 @@ def benchmark_nprocs(module, label, max_nprocs, repeat=5):
to generate a continuous range of integer values from 1 up to *max_nprocs*.
*repeat* sets how many times the each run is repeated.
So, the module will run ``max_nprocs * repeat`` times.
Runs are executed in random order, set *shuffle* to false if they
need to be executed in order based on number of threads.

*label* is a text to add to the result (for user-facing display).
Optional *nprocs* is passed to the module if present.
Expand All @@ -102,27 +105,28 @@ def benchmark_nprocs(module, label, max_nprocs, repeat=5):
avg_times = []
all_times = []
nprocs_list = list(range(1, max_nprocs + 1))
for nprocs in nprocs_list:
print("\u2500" * term_size.columns)
print(f"Benchmark with {nprocs} thread(s)...\n")
time_sum = 0
measured_times = []
for _ in range(repeat):
module.update(nprocs=nprocs)
module.run()
print(f"{module.time}s")
time_sum += module.time
measured_times.append(module.time)

avg = time_sum / repeat
nprocs_list_shuffled = sorted(nprocs_list * repeat)
if shuffle:
random.shuffle(nprocs_list_shuffled)
times = {}
print("\u2500" * term_size.columns)
for nprocs in nprocs_list_shuffled:
module.update(nprocs=nprocs)
module.run()
print(f"Run with {nprocs} thread(s) took {module.time}s\n")
if nprocs in times:
times[nprocs] += [module.time]
else:
times[nprocs] = [module.time]
for nprocs in sorted(times):
avg = sum(times[nprocs]) / repeat
avg_times.append(avg)
all_times.append(measured_times)
all_times.append(times[nprocs])
if nprocs == 1:
serial_avg = avg
if avg < min_avg:
min_avg = avg
min_time = nprocs
print(f"\nResult - {avg}s")

print("\u2500" * term_size.columns)
if serial_avg is not None:
Expand Down