@@ -35,6 +35,21 @@ Building NumPy requires the following software installed:
35
35
Python__ http://www.python.org
36
36
nose__ http://somethingaboutorange.com/mrl/projects/nose/
37
37
38
+ Basic Installation
39
+ ==================
40
+
41
+ To install numpy run:
42
+
43
+ python setup.py build -j 4 install --prefix $HOME/.local
44
+
45
+ This will compile numpy on 4 CPUs and install it into the specified prefix.
46
+ To perform an inplace build that can be run from the source folder run:
47
+
48
+ python setup.py build_ext --inplace -j 4
49
+
50
+ The number of build jobs can also be specified via the environment variable
51
+ NPY_NUM_BUILD_JOBS.
52
+
38
53
Fortran ABI mismatch
39
54
====================
40
55
@@ -65,40 +80,34 @@ this means that g77 has been used. If libgfortran.so is a dependency,
65
80
gfortran has been used. If both are dependencies, this means both have been
66
81
used, which is almost always a very bad idea.
67
82
68
- Building with ATLAS support
69
- ===========================
70
-
71
- Ubuntu 8.10 (Intrepid)
72
- ----------------------
73
-
74
- You can install the necessary packages for optimized ATLAS with this
75
- command:
76
-
77
- sudo apt-get install libatlas-base-dev
78
-
79
- If you have a recent CPU with SIMD support (SSE, SSE2, etc...), you should
80
- also install the corresponding package for optimal performance. For
81
- example, for SSE2:
83
+ Building with optimized BLAS support
84
+ ====================================
82
85
83
- sudo apt-get install libatlas3gf-sse2
86
+ Ubuntu/Debian
87
+ -------------
84
88
85
- *NOTE*: if you build your own atlas, Intrepid changed its default fortran
86
- compiler to gfortran. So you should rebuild everything from scratch,
87
- including lapack, to use it on Intrepid.
89
+ In order to build with optimized a BLAS providing development package must be installed.
90
+ Options are for example:
88
91
89
- Ubuntu 8.04 and lower
90
- ---------------------
92
+ - libblas-dev
93
+ reference BLAS not very optimized
94
+ - libatlas-base-dev
95
+ generic tuned ATLAS, it is recommended to tune it to the available hardware,
96
+ see /usr/share/doc/libatlas3-base/README.Debian for instructions
97
+ - libopenblas-base
98
+ fast and runtime detected so no tuning required but as of version 2.11 still
99
+ suffers from correctness issues on some CPUs, test your applications
100
+ thoughly.
91
101
92
- You can install the necessary packages for optimized ATLAS with this
93
- command :
102
+ The actual implementation can be exchanged also after installation via the
103
+ alternatives mechanism :
94
104
95
- sudo apt-get install atlas3-base-dev
105
+ update-alternatives --config libblas.so.3
106
+ update-alternatives --config liblapack.so.3
96
107
97
- If you have a recent CPU with SIMD support (SSE, SSE2, etc...), you should
98
- also install the corresponding package for optimal performance. For
99
- example, for SSE2:
108
+ Or by preloading a specific BLAS library with
109
+ LD_PRELOAD=/usr/lib/atlas-base/atlas/libblas.so.3 python ...
100
110
101
- sudo apt-get install atlas3-sse2
102
111
103
112
Windows 64 bits notes
104
113
=====================
0 commit comments