@@ -72,10 +72,16 @@ VERSION = 0.3.6.dev
72
72
73
73
# You can define the maximum number of threads. Basically it should be less
74
74
# than or equal to the number of CPU threads. If you don't specify one, it's
75
- # automatically detected by the the script .
75
+ # automatically detected by the the build system .
76
76
# If SMT (aka. HT) is enabled on the system, it may or may not be beneficial to
77
77
# restrict NUM_THREADS to the number of physical cores. By default, the automatic
78
78
# detection includes logical CPUs, thus allowing the use of SMT.
79
+ # Users may opt at runtime to use less than NUM_THREADS threads.
80
+ #
81
+ # Note for package maintainers: you can build OpenBLAS with a large NUM_THREADS
82
+ # value (eg. 32-256) if you expect your users to use that many threads. Due to the way
83
+ # some internal structures are allocated, using a large NUM_THREADS value has a RAM
84
+ # footprint penalty, even if users reduce the actual number of threads at runtime.
79
85
# NUM_THREADS = 24
80
86
81
87
# If you have enabled USE_OPENMP and your application would call
@@ -138,6 +144,7 @@ NO_WARMUP = 1
138
144
# to the same core(s) as OpenBLAS, possibly binding all threads to a single core.
139
145
# For this reason, affinity handling is disabled by default. Can be safely enabled if nothing
140
146
# else modifies affinity settings.
147
+ # Note: enabling affinity has been known to cause problems with NumPy and R
141
148
NO_AFFINITY = 1
142
149
143
150
# If you are compiling for Linux and you have more than 16 numa nodes or more than 256 cpus
0 commit comments