Urutu
A Python based parallel programming library for GPUs
About
Urutu integrates CUDA and OpenCL into a single code. That is, write a single code in Python run it using CUDA or OpenCL or both. It is portable to different hardware and operating systems.
System Requirements
- Windows, Ubuntu (tested), Mac
- Python 2.7, GCC 4.6
- A GPU compatible with OpenCL or CUDA 5.0
- PyOpenCL or PyCUDA installed
Setup
$ git clone https://github.com/urutu/Urutu.git
$ cd Urutu
$ sudo python2 setup.py installFor installing with Python 3, install PyCUDA and PyOpenCL with Python 3. Then, clone this repository and use
$ git clone https://github.com/urutu/Urutu.git
$ 2to3 -w Urutu
$ cd Urutu
$ sudo python3 setup.py install=======
Sample Code
from Urutu import *
import numpy as np
@Urutu("CL")
def divmul(a, b, c, d):
__global is x, y
x = a[0:100]
y = b[0:100]
t, u, v, w = 10, 10.0, 'opencl', "open.cl"
c[tx] = x[tx] / y[tx]
d[tx] = x[tx] * y[tx]
return c, d
@Urutu("CU")
def addsub(a, b, e, f):
__shared is x , y
x = a[0:100]
y = b[0:100]
t, u, v, w = 11, 11.0, 'cuda', "cu.da"
e[tx] = x[tx] + y[tx]
f[tx] = x[tx] - y[tx]
return e, f
a = np.random.randint(10, size = 100)
b = np.random.randint(10, size = 100)
c = np.array(a, dtype = 'f')
d = np.empty_like(a)
e = np.empty_like(a)
f = np.empty_like(a)
print "The Array A is: \n", a
print "The Array B is: \n", b
print "Running on OpenCL.. \n", divmul([100,1,1], [1,1,1], a, b, c, d)
print "Running on CUDA.. \n", addsub([100,1,1], [1,1,1], a, b, e, f)Keywords made available
and elif if not
as else in or
def for is return
