<span style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); ">Dear list,</span><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">so far I used Enthoughts Python Distribution which contains a compiled version of numpy linked against MKL. Now, I want to implement my own extensions to numpy, so I need my build numpy on my own. So, I installed Intel Parallel studio including MKL and the C / Fortran compilers.</div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">I linked against the same libraries as Enthought:</div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><div style="border-style: initial; border-color: initial; ">In [2]: np.show_config()</div><div style="border-style: initial; border-color: initial; ">lapack_opt_info:</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; libraries = ['mkl_lapack95_lp64', 'iomp5', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'pthread']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; library_dirs = ['/opt/intel/lib/intel64', '/opt/intel/mkl/lib/intel64']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; define_macros = [('SCIPY_MKL_H', None)]</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; include_dirs = ['/opt/intel/include/']</div><div style="border-style: initial; border-color: initial; ">blas_opt_info:</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; libraries = ['iomp5', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'pthread']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; library_dirs = ['/opt/intel/lib/intel64', '/opt/intel/mkl/lib/intel64']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; define_macros = [('SCIPY_MKL_H', None)]</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; include_dirs = ['/opt/intel/include/']</div><div style="border-style: initial; border-color: initial; ">lapack_mkl_info:</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; libraries = ['mkl_lapack95_lp64', 'iomp5', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'pthread']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; library_dirs = ['/opt/intel/lib/intel64', '/opt/intel/mkl/lib/intel64']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; define_macros = [('SCIPY_MKL_H', None)]</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; include_dirs = ['/opt/intel/include/']</div><div style="border-style: initial; border-color: initial; ">blas_mkl_info:</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; libraries = ['iomp5', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'pthread']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; library_dirs = ['/opt/intel/lib/intel64', '/opt/intel/mkl/lib/intel64']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; define_macros = [('SCIPY_MKL_H', None)]</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; include_dirs = ['/opt/intel/include/']</div><div style="border-style: initial; border-color: initial; ">mkl_info:</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; libraries = ['iomp5', 'mkl_def', 'mkl_intel_lp64', 'mkl_intel_thread', 'mkl_core', 'pthread']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; library_dirs = ['/opt/intel/lib/intel64', '/opt/intel/mkl/lib/intel64']</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; define_macros = [('SCIPY_MKL_H', None)]</div><div style="border-style: initial; border-color: initial; ">&nbsp; &nbsp; include_dirs = ['/opt/intel/include/']</div></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">and used the intel compilers to build numpy. Since I have an i7 processor I compiled including AVX instructions. Yet, EPD is double as fast as my own build executing the simple benchmark from:&nbsp;<a href="http://dpinte.wordpress.com/2010/01/15/numpy-performance-improvement-with-the-mkl/" target="_blank" style="color: rgb(17, 85, 204); border-style: initial; border-color: initial; ">http://dpinte.wordpress.<wbr>com/2010/01/15/numpy-<wbr>performance-improvement-with-<wbr>the-mkl/</a></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">I expected at least comparable performance.&nbsp;How is such a decrease possible? Did I miss a significant part to make numpy really fast?</div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; "><br></div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">Thanks</div><div style="font-family: arial, sans-serif; background-color: rgba(255, 255, 255, 0.917969); border-style: initial; border-color: initial; ">Christoph</div>