Tweaked the way we compute the number of threads to use when building.

setup.py

@@ -137,7 +137,7 @@ class CMakeBuild(build_ext):
         else:
             cmake_args += ['-DCMAKE_BUILD_TYPE=' + cfg]
             # Do a parallel build
-            build_args += ['--', '-j'+str(num_available_cpu_cores(4))]
+            build_args += ['--', '-j'+str(num_available_cpu_cores(2))]
 
         build_folder = os.path.abspath(self.build_temp)
 
@@ -162,9 +162,8 @@ def num_available_cpu_cores(ram_per_build_process_in_gb):
     try:
         mem_bytes = os.sysconf('SC_PAGE_SIZE') * os.sysconf('SC_PHYS_PAGES')  
         mem_gib = mem_bytes/(1024.**3)
-        # Assume hyperthreading so divide by 2
+        num_cores = multiprocessing.cpu_count() 
-        num_cores = int(ceil(multiprocessing.cpu_count()/2.0)) 
+        # make sure we have enough ram for each build process.
-        # Require 4gb of ram per build thread.
         mem_cores = int(floor(mem_gib/float(ram_per_build_process_in_gb)+0.5));
         # We are limited either by RAM or CPU cores.  So pick the limiting amount
         # and return that.