CODEWITHSUNDEEP
pythonmatplotlibrhel
AndyB
AndyB

Reputation: 11

matplotlib subplots runs very slow only on some workstations

I'm using matplotlib in scripts on two different RHEL linux workstations (both RHEL 7.9). On one workstation it takes less than a second (0.18s) to call plt.subplots, while the other can take more than 45s!! Below is a simple timing script I used to check the timing. I've tried different versions of matplotlib (via miniconda), but it doesn't seem to make a difference. The workstation that loads plt.subplots slowly should be faster based on specs (see below). Any ideas of other things to check?

Thanks in advance for any help! Andy

import matplotlib.pyplot as plt
import time

fig_time_s = time.time()

fig, ax  = plt.subplots(figsize=(9.0, 4.0), dpi=200)

fig_time_e = time.time()
fig_time  = fig_time_e - fig_time_s
print('FIG TIME: {:7.2f}s'.format(fig_time))

version of matplotlib: |# Name |Version |Build |Channel |---|---|---|---| |matplotlib |3.2.2 | 1 |conda-forge |matplotlib-base |3.2.2 |py37h30547a4_0 |conda-forge

These are the specs of the workstations:

Property faster slower
Architecture: x86_64 x86_64
CPU op-mode(s): 32-bit, 64-bit 32-bit, 64-bit
Byte Order: Little Endian Little Endian
CPU(s): 6 48
On-line CPU(s) list: 0-5 0-47
Thread(s) per core: 1 2
Core(s) per socket: 6 24
Socket(s): 1 1
NUMA node(s): 1 1
Vendor ID: GenuineIntel GenuineIntel
CPU family: 6 6
Model: 63 85
Model name: Intel(R) Xeon(R) CPU E5-2609 v3 @ 1.90GHz Intel(R) Xeon(R) Platinum 8168 CPU @ 2.70GHz
Stepping: 2 4
CPU MHz: 1899.884 1199.871
CPU max MHz: 1900.0000 3700.0000
CPU min MHz: 1200.0000 1200.0000
BogoMIPS: 3791.33 5400.00
Virtualization: VT-x VT-x
L1d cache: 32K 32K
L1i cache: 32K 32K
L2 cache: 256K 1024K
L3 cache: 15360K 33792K
NUMA node0 CPU(s): 0-5 0-47
NUMA node0 CPU(s): 0-47

Upvotes: 0

Views: 246

Answers (1)

AndyB
AndyB

Reputation: 11

Update: our Sys Admin found the issue. There was some sort of timeout happening when using X11 forwarding (ssh -Y). Removing the -Y flag, or changing to -x solved the issue. We found that adding

import matplotlib
matplotlib.use('Agg')

to the script will also solve the slowdown issue.

Upvotes: 1

Related Questions