Vlasov-Ampere, CPU node vs GPU node: CPU node

Filter regionsRegionsall 5Depthall levels

cpu-node.h52026-10-17 12:00 UTCsynthetic 2x64-core CPU node8 ranks5 regions0.384858 s profiled

← Vlasov-Ampere, CPU node vs GPU node · GPU node ›

Summary

Wall time385 mssetup to finalize

Ranks8

Regions5

In regions99%of the session

Top hotspotsolve93.8% of the session

Load imbalance+9%solve, slowest on rank 2

solve (in simulation) is the largest hotspot: 2.669 s over 80 calls, 93.8% of the session.
The three largest hotspots account for 99% of the session.

Hotspots

The leaves of the call tree with the most time: regions without nested regions, and the own time parents spend outside theirs. Shares are of the session. Click one to find it in the table.

Region statistics

region

% session

total [s]

trend

scope_profiler.session (8x)

100.00%

2.844559

└─ (own)

1.12%

0.032000

Per rank

rank	calls	total [s]	avg [s]	min [s]	max [s]
0	1	0.350111	0.350111	0.350111	0.350111
1	1	0.334196	0.334196	0.334196	0.334196
2	1	0.384858	0.384858	0.384858	0.384858
3	1	0.363857	0.363857	0.363857	0.363857
4	1	0.345113	0.345113	0.345113	0.345113
5	1	0.356605	0.356605	0.356605	0.356605
6	1	0.368921	0.368921	0.368921	0.368921
7	1	0.340899	0.340899	0.340899	0.340899

└─ simulation (8x)

98.88%

2.812559

│ └─ (own)

0.00%

0.000000

Per rank

rank	calls	total [s]	avg [s]	min [s]	max [s]
0	1	0.346111	0.346111	0.346111	0.346111
1	1	0.330196	0.330196	0.330196	0.330196
2	1	0.380858	0.380858	0.380858	0.380858
3	1	0.359857	0.359857	0.359857	0.359857
4	1	0.341113	0.341113	0.341113	0.341113
5	1	0.352605	0.352605	0.352605	0.352605
6	1	0.364921	0.364921	0.364921	0.364921
7	1	0.336899	0.336899	0.336899	0.336899

│ └─ setup (8x)

1.12%

0.032000

Per rank

rank	calls	total [s]	avg [s]	min [s]	max [s]
0	1	0.004	0.004	0.004	0.004
1	1	0.004	0.004	0.004	0.004
2	1	0.004	0.004	0.004	0.004
3	1	0.004	0.004	0.004	0.004
4	1	0.004	0.004	0.004	0.004
5	1	0.004	0.004	0.004	0.004
6	1	0.004	0.004	0.004	0.004
7	1	0.004	0.004	0.004	0.004

│ └─ solve (80x)

93.81%

2.668559

Per rank

rank	calls	total [s]	avg [s]	min [s]	max [s]
0	10	0.328111	0.0328111	0.0328111	0.0328111
1	10	0.312196	0.0312196	0.0312196	0.0312196
2	10	0.362858	0.0362858	0.0362858	0.0362858
3	10	0.341857	0.0341857	0.0341857	0.0341857
4	10	0.323113	0.0323113	0.0323113	0.0323113
5	10	0.334605	0.0334605	0.0334605	0.0334605
6	10	0.346921	0.0346921	0.0346921	0.0346921
7	10	0.318899	0.0318899	0.0318899	0.0318899

│ └─ exchange (80x)

3.94%

0.112000

Per rank

rank	calls	total [s]	avg [s]	min [s]	max [s]
0	10	0.014	0.0014	0.0014	0.0014
1	10	0.014	0.0014	0.0014	0.0014
2	10	0.014	0.0014	0.0014	0.0014
3	10	0.014	0.0014	0.0014	0.0014
4	10	0.014	0.0014	0.0014	0.0014
5	10	0.014	0.0014	0.0014	0.0014
6	10	0.014	0.0014	0.0014	0.0014
7	10	0.014	0.0014	0.0014	0.0014

Durations are in seconds. (own) rows show a region's time excluding its children. % session uses wall-clock coverage; overlapping recursive calls count once. Click a row for its call site and per-rank breakdown.

Load balance

Each region's own time per rank, excluding nested regions, so a parent does not repeat its children's imbalance. Excess is how much longer the slowest rank spends than the average rank: the most balancing that region could save. Regions are ordered by it.

region	mean	min	max	slowest rank	imbalance	excess
solve	334 ms	312 ms	363 ms	2	+9%	29.3 ms
setup	4 ms	4 ms	4 ms	0	+0%	0 s
exchange	14 ms	14 ms	14 ms	0	+0%	0 s

Rank comparison

Each cell compares one rank's time in a region with the region's mean over the ranks. Rank 2, in red in the header, spends the most time in regions overall. Hover a cell for its time.

faster than the meanslower

region	mean	0	1	2	3	4	5	6	7
solve	334 ms	−1.6%	−6.4%	+8.8%	+2.5%	−3.1%	+0.3%	+4.0%	−4.4%
exchange	14 ms	0%	0%	0%	0%	0%	0%	0%	0%
setup	4 ms	0%	0%	0%	0%	0%	0%	0%	0%
all regions	352 ms	−1.6%	−6.1%	+8.3%	+2.4%	−3.0%	+0.3%	+3.8%	−4.2%

Charts

Region durations

Each bar shows a region's total recorded duration, summed over the selected ranks, longest first.

stack children

Timeline: CPU node (rank 0)

Each bar is one recorded region call on rank 0. Its position and width show when the call started and how long it ran; colors identify regions.

Rank imbalance

Each line compares a region's total duration by rank; its dashed line marks the mean across ranks. Points far from the mean identify stragglers.

Rank heatmap

This heatmap uses exclusive timings. Exclusive duration is the time spent in a region itself, excluding time spent in nested child regions; this prevents the enclosing session region from dominating the heatmap.

Figures

Download run.py (715 B)

"""Landau damping of a Langmuir wave with a Vlasov-Ampere solver: the script both runs use."""

import sys

import yaml
from scope_profiler import ProfileManager


def main(parameter_file):
    with open(parameter_file) as handle:
        parameters = yaml.safe_load(handle)
    with ProfileManager.session(label=parameters["label"]):
        with ProfileManager.region("setup"):
            state = initialise(parameters)
        for _ in range(parameters["steps"]):
            with ProfileManager.region("solve"):
                state = advance(state, parameters["dt"])
            with ProfileManager.region("exchange"):
                exchange_halos(state)


if __name__ == "__main__":
    main(sys.argv[1])

Download cpu-node.yml (204 B)

# Landau damping of a Langmuir wave, k = 0.5, 1 percent perturbation
label: cpu node
device: cpu
mpi_ranks: 8
omp_threads: 2
steps: 10
dt: 0.05
grid:
  cells: [128, 128, 128]
  particles_per_cell: 128000

Metadata

commit	`0000000`
system	`synthetic`
chip_information	`synthetic 2x64-core CPU node`
mpi_size	`8`
num_nodes	`4`
omp_num_threads	`2`
platform	`synthetic`
scope_profiler_version	`sample`
timestamp	`2026-10-17T12:00:00Z`
total_cores	`16`