Samsung/opencl4py

Name: opencl4py

Owner: Samsung

Description: Python cffi OpenCL bindings and helper classes

Created: 2014-03-25 06:47:54.0

Updated: 2017-05-31 15:46:19.0

Pushed: 2016-01-03 20:31:18.0

Homepage: null

Size: 167

Language: Python

GitHub Committers

User	Most Recent Commit	# Commits

Other Committers

User	Email	Most Recent Commit	# Commits

README

opencl4py

Python cffi OpenCL bindings and helper classes.

Tested with Python 2.7, Python 3.3, Python 3.4 and PyPy on Linux, MacOSX and Windows.

To use clBLAS, libclBLAS.so (clBLAS.dll) should be present.

Not all OpenCL api is currently covered.

To install the module run:

install .

or just copy src/opencl4py to any place where python interpreter will be able to find it.

To run the tests, execute:

for Python 2.7:

ONPATH=src nosetests -w tests

for Python 3.3, 3.4:

ONPATH=src nosetests3 -w tests

for PyPy:

ONPATH=src pypy -m nose -w tests

Example usage:

rt opencl4py as cl
rt logging
rt numpy


_name__ == "__main__":
logging.basicConfig(level=logging.DEBUG)
platforms = cl.Platforms()
logging.info("OpenCL devices available:\n\n%s\n",
             platforms.dump_devices())
ctx = platforms.create_some_context()
queue = ctx.create_queue(ctx.devices[0])
prg = ctx.create_program(
    """
    __kernel void test(__global const float *a, __global const float *b,
                       __global float *c, const float k) {
      size_t i = get_global_id(0);
      c[i] = (a[i] + b[i]) * k;
    }
    """)
krn = prg.get_kernel("test")
a = numpy.arange(1000000, dtype=numpy.float32)
b = numpy.arange(1000000, dtype=numpy.float32)
c = numpy.empty(1000000, dtype=numpy.float32)
k = numpy.array([0.5], dtype=numpy.float32)
a_buf = ctx.create_buffer(cl.CL_MEM_READ_ONLY | cl.CL_MEM_COPY_HOST_PTR,
                          a)
b_buf = ctx.create_buffer(cl.CL_MEM_READ_ONLY | cl.CL_MEM_COPY_HOST_PTR,
                          b)
c_buf = ctx.create_buffer(cl.CL_MEM_WRITE_ONLY | cl.CL_MEM_ALLOC_HOST_PTR,
                          size=c.nbytes)
krn.set_args(a_buf, b_buf, c_buf, k[0:1])
queue.execute_kernel(krn, [a.size], None)
queue.read_buffer(c_buf, c)
max_diff = numpy.fabs(c - (a + b) * k[0]).max()
logging.info("max_diff = %.6f", max_diff)

Released under Simplified BSD License. Copyright (c) 2014, Samsung Electronics Co.,Ltd.