Python/OpenCL

numpy

• https://docs.scipy.org/doc/numpy/user/quickstart.html

Data Types

import numpy as np

# Default int type is int64
a = np.array([1, 2, 3])
print(type(a[0]).__name__)

# Default float type is float64
b = np.array([1.0, 2.0, 3.0])
print(type(b[0]).__name__)

# Force int32
c = np.array([1, 2, 3]).astype(np.int32)
print(type(c[0]).__name__)

# Force float32
d = np.array([1.0, 2.0, 3.0]).astype(np.float32)
print(type(d[0]).__name__)

pyopencl

To increase an array of float32.

import os
import numpy as np
import pyopencl as cl

# Kernel code.
CL_INC_F32 = '''
__kernel void inc_f32(__global const float *a_g, __global float *res_g)
{
  int gid = get_global_id(0);
  res_g[gid] = a_g[gid] + 1;
}
'''

# Context and Queue
os.environ['PYOPENCL_CTX'] = '0:1' # Apple > HD Graphics 4000 of Macbook Air 2012 mid
ctx = cl.create_some_context()
queue = cl.CommandQueue(ctx)

# Memory allocation.
mf = cl.mem_flags
in_np = np.array([1, 2, 3, 4, 5]).astype(np.float32)
in_cl = cl.Buffer(ctx, mf.READ_ONLY | mf.COPY_HOST_PTR, hostbuf=in_np)
out_cl = cl.Buffer(ctx, mf.WRITE_ONLY, in_np.nbytes)
out_np = np.empty_like(in_np)

# Build OpenCL kernel code.
prg = cl.Program(ctx, CL_INC_F32).build()

# Run kernel.
prg.inc_f32(queue, in_np.shape, None, in_cl, out_cl)

# Convert results into numpy format.
cl.enqueue_copy(queue, out_np, out_cl)

print(in_np)
print(out_np)