import chroma.gpu.tools as tools
class workQueue(object):
def __init__(self, context ):
# we get important information about work queues here
self.context = context
if api.is_gpu_api_opencl():
self.device = context.get_info( cl.context_info.DEVICES )[0]
self.shared_mem_size = self.device.get_info( cl.device_info.LOCAL_MEM_SIZE )
self.work_group_size = self.device.get_info( cl.device_info.MAX_WORK_GROUP_SIZE )
self.work_item_sizes = self.device.get_info( cl.device_info.MAX_WORK_ITEM_SIZES )
self.work_item_dims = self.device.get_info( cl.device_info.MAX_WORK_ITEM_DIMENSIONS )
self.max_compute_units = self.device.get_info( cl.device_info.MAX_COMPUTE_UNITS )
else:
self.device = context.get_device()
self.shared_mem_size = self.device.max_shared_memory_per_block
self.work_group_size = self.device.max_threads_per_block
self.work_item_sizes = self.device.max_block_dim_x
self.work_item_dimes = 3
self.max_compute_units = self.device.multiprocessor_count
def print_dev_info(self):
print self.device, self.shared_mem_size, self.work_group_size, self.work_group_size, self.max_compute_units
if __name__ == "__main__":
# Testing.
os.environ['PYOPENCL_CTX']='0:1'
context = tools.get_context()
w = workQueue( context )
import os, sys
os.environ["PYOPENCL_CTX"] ='1'
import numpy as np
import pyopencl as cl
import pyopencl.array as clarray
import chroma.gpu.tools as tools
float3 = clarray.vec.float3
print "float3 type: ",float3
ctx = tools.get_context()
queue = cl.CommandQueue(ctx)
dev = ctx.get_info( cl.context_info.DEVICES )[0]
print 'device %s' % dev.get_info( cl.device_info.NAME )
mod = tools.get_module( 'linalg_test.cl', ctx, include_source_directory=False )
size = {'block': (256,), 'grid': (1,)}
a_np = np.zeros((size['block'][0],3), dtype=np.float32)
b_np = np.zeros((size['block'][0],3), dtype=np.float32)
c_np = np.float32(np.random.random_sample())
mf = cl.mem_flags
a_vec_np = np.zeros(size['block'][0], dtype=float3)
b_vec_np = np.zeros(size['block'][0], dtype=float3)
d_vec_np = np.zeros(size['block'][0], dtype=float3)
#c_vec_np = np.float32(np.random.random_sample())
#float3add = mod.get_function('float3add')
#float3addequal = mod.get_function('float3addequal')
#float3sub = mod.get_function('float3sub')