When on AMD/Intel platforms, copies for unaligned arrays are disabled.

This is because AMD/Intel processors can deal very efficiently with
unaligned data, so avoiding the copy is generally a good thing (up to
2x speed-ups wrt doing a copy can easily be seen).

However, this optimization is not active when Numexpr is compiled with
Intel VML (Vector Math Library).  The rational is that (surprisingly
enough), VML has not optimized the access to unaligned arrays, and hence
the performance of its functions can drop a lot (as much as 10x in some
cases!).

Finally, the ``cpuinfo.py`` module has been added to the code in order
to detect AMD/Intel CPUs easily.  Thanks to Pearu Peterson for providing
this nice piece of code.

LICENSES/cpuinfo.txt

@@ -0,0 +1,31 @@
+Copyright statement for `cpuinfo` module.
+
+Copyright 2002 Pearu Peterson all rights reserved,
+Pearu Peterson <pearu@cens.ioc.ee>
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright
+      notice, this list of conditions and the following disclaimer.
+
+    * Redistributions in binary form must reproduce the above
+      copyright notice, this list of conditions and the following
+      disclaimer in the documentation and/or other materials provided
+      with the distribution.
+
+    * Neither the name of Pearu Peterson nor the names of its
+      contributors may be used to endorse or promote products derived
+      from this software without specific prior written permission.
+
+HIS SOFTWARE IS PROVIDED BY PEARU PETERSON ''AS IS'' AND ANY EXPRESS
+OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL PEARU PETERSON BE LIABLE FOR ANY DIRECT,
+INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
+IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGE.

bench/boolean_timing.py

@@ -114,8 +114,8 @@ def compare(expression=False):
 
 if __name__ == '__main__':
 
-    print 'Python version:    %s' % sys.version
-    print "NumPy version:     %s" % numpy.__version__
+    import numexpr
+    numexpr.print_versions()
 
     if len(sys.argv) > 1:
         expression = sys.argv[1]

bench/timing.py

@@ -126,6 +126,9 @@ def compare(check_only=False):
     return average
 
 if __name__ == '__main__':
+    import numexpr
+    numexpr.print_versions()
+
     averages = []
     for i in range(iterations):
         averages.append(compare())

bench/unaligned-bench.py

@@ -7,7 +7,10 @@
 import numexpr as ne
 
 niter = 10
-shape = (1000, 10000)
+#shape = (1000*10000)   # unidimensional test
+shape = (1000, 10000)   # multidimensional test
+
+ne.print_versions()
 
 Z_fast = np.zeros(shape, dtype=[('x',np.float64),('y',np.int64)])
 Z_slow = np.zeros(shape, dtype=[('x',np.float64),('y',np.bool)])

bench/vml_timing.py

@@ -114,10 +114,8 @@ def compare(expression=False):
     print
 
 if __name__ == '__main__':
-
-    print 'Python version:    %s' % sys.version
-    print "NumPy version:     %s" % numpy.__version__
-    print "use vml            %s" % numexpr.use_vml
+    import numexpr
+    numexpr.print_versions()
 
     numexpr.set_vml_accuracy_mode('low')
     numexpr.set_vml_num_threads(2)

numexpr/__init__.py

@@ -18,6 +18,12 @@
 else:
     use_vml = False
 
+from cpuinfo import cpu
+if cpu.is_AMD() or cpu.is_Intel():
+    is_cpu_amd_intel = True
+else:
+    is_cpu_amd_intel = False
+
 import os.path
 from numexpr.expressions import E
 from numexpr.necompiler import NumExpr, disassemble, evaluate