python加速

之前一直用 conda版python, 发现可以直接装intel的numpy了.

https://software.intel.com/en-us/articles/installing-the-intel-distribution-for-python-and-intel-performance-libraries-with-pip-and

• 安装

root@zkk-VirtualBox:~# cat .pip/pip.conf
[global]
index-url = http://mirrors.aliyun.com/pypi/simple/
[install]
trusted-host = mirrors.aliyun.com
[download]
trusted-host = mirrors.aliyun.com
root@zkk-VirtualBox:~# pip3 install intel-numpy
Collecting intel-numpy
  Downloading http://mirrors.aliyun.com/pypi/packages/ef/b3/fb79b1f34dc83822ea4e57c9a889ee32a34087139c12c9f1c3473f060d4d/intel_numpy-1.15.1-cp36-cp36m-manylinux1_x86_64.whl (6.1MB)
    100% |████████████████████████████████| 6.1MB 12.3MB/s
Collecting tbb4py (from intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/af/88/06532a4fb130ce4d1573a0b13ae3c58d689fe950e9806c507ba44d13ab17/tbb4py-2019.0-cp36-cp36m-manylinux1_x86_64.whl (225kB)
    100% |████████████████████████████████| 235kB 6.9MB/s
Collecting mkl (from intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/9b/98/c892b77b755cb0c53491eabc88c49451a92e36fa5c5baf578e77b91ee31d/mkl-2019.0-py2.py3-none-manylinux1_x86_64.whl (261.0MB)
    100% |████████████████████████████████| 261.0MB 7.2MB/s
Collecting mkl-fft (from intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/f3/98/0783b1543ee73d47a258ebdffe002adcaabb18085653a6e83b42c6835c41/mkl_fft-1.0.6-cp36-cp36m-manylinux1_x86_64.whl (232kB)
    100% |████████████████████████████████| 235kB 3.0MB/s
Collecting mkl-random (from intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/c8/76/6255699dfa4e2265fadb91c27b991ad8113161835a76556396437fae1a7f/mkl_random-1.0.1.1-cp36-cp36m-manylinux1_x86_64.whl (396kB)
    100% |████████████████████████████████| 399kB 9.7MB/s
Collecting icc-rt (from intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/1e/29/0ee89498c2b3e96620add5f4730d4e55e685b292a47aa67f52975fed9f71/icc_rt-2019.0-py2.py3-none-manylinux1_x86_64.whl (11.2MB)
    100% |████████████████████████████████| 11.2MB 8.4MB/s
Collecting tbb==2019.* (from tbb4py->intel-numpy)
  Downloading http://mirrors.aliyun.com/pypi/packages/28/53/8bf93994fd985ea08c1ef64c7ec94a433611c6f6af5e3ddcc8c49d83fd5c/tbb-2019.0-py2.py3-none-manylinux1_x86_64.whl (895kB)
    100% |████████████████████████████████| 901kB 9.3MB/s
Requirement already satisfied: intel-openmp in /usr/local/lib/python3.6/dist-packages (from mkl->intel-numpy)
Installing collected packages: tbb, tbb4py, mkl, mkl-fft, mkl-random, icc-rt, intel-numpy
Successfully installed icc-rt-2019.0 intel-numpy-1.15.1 mkl-2019.0 mkl-fft-1.0.6 mkl-random-1.0.1.1 tbb-2019.0 tbb4py-2019.0

• import numpy 仍然有问题, 参考这个解决. https://askubuntu.com/questions/857528/python-3-cannot-import-third-party-packages

[zkk@zkk-VirtualBox  /usr/local/lib/python3.6/dist-packages]
$ cat site-packges.pth
../site-packages