Merge pull request #75 from danielwe/patch-1

Add info on threading

Author: staticfloat

README.md

@@ -36,6 +36,10 @@ Some additional functions that are also available:
 * `fdiv(x,y)`: divide (`x ./ y` in Base)
 * `sincos(x)`: returns `(sin(x), cos(x))`
 
+## Setting the number of threads
+
+Accelerate is multithreaded by default. You can set an upper limit on the number of threads through the `VECLIB_MAXIMUM_THREADS` environment variable; for example, for single-threaded execution, start Julia as `VECLIB_MAXIMUM_THREADS=1 julia`. Accelerate does not support the `BLAS.set_num_threads(nthreads)` and `BLAS.get_num_threads()` API used by other BLAS backends (`set_num_threads` is a no-op and `get_num_threads` returns a hardcoded default).
+
 ## Example
 
 To use the Accelerate BLAS and LAPACK, simply load the library: