AuroraMiddleware/bullet3
1
0
Fork 0
mirror of https://github.com/bulletphysics/bullet3 synced 2024-12-13 13:20:07 +00:00
Code Issues Releases Wiki Activity

Limit compilation parallelism

Browse Source 
On smaller platforms like the Raspberry Pi, which have a deficit of
physical memory, the default algorithm for determining compilation
parallelism is far too optimistic and usually results in crashing the
platform.

This modification limits the number of parallel jobs to the size of RAM
in Gb. On a mid-size PC (e.g. quad-core with 8+Gb of RAM), this will
make no difference. On a Pi (quad-core with 1Gb of RAM), this will limit
it to 1 job at a time (which is ideal).
This commit is contained in:
Dave Jones 2019-02-20 22:52:29 +01:00
parent fd161fa061
commit 7b5f81958a
1 changed files with 12 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
setup.py
View File

@ -22,6 +22,18 @@ def parallelCCompile(self, sources, output_dir=None, macros=None, include_dirs=N
cc_args = self._get_cc_args(pp_opts, debug, extra_preargs) cc_args = self._get_cc_args(pp_opts, debug, extra_preargs)
# parallel code # parallel code
N = 2*multiprocessing.cpu_count()# number of parallel compilations N = 2*multiprocessing.cpu_count()# number of parallel compilations
try:
# On Unix-like platforms attempt to obtain the total memory in the
# machine and limit the number of parallel jobs to the number of Gbs
# of RAM (to avoid killing smaller platforms like the Pi)
mem = os.sysconf('SC_PHYS_PAGES') * os.sysconf('SC_PAGE_SIZE') # bytes
except (AttributeError, ValueError):
# Couldn't query RAM; don't limit parallelism (it's probably a well
# equipped Windows / Mac OS X box)
pass
else:
mem = max(1, int(round(mem / 1024 ** 3))) # convert to Gb
N = min(mem, N)
def _single_compile(obj): def _single_compile(obj):
try: src, ext = build[obj] try: src, ext = build[obj]
except KeyError: return except KeyError: return