/* Execute the first example */
uint8 mask1 = (uint8)(1, 2, 0, 1, 3, 1, 2, 3);
float4 input = (float4)(0.25f, 0.5f, 0.75f, 1.0f);
*s1 = shuffle(input, mask1);
/* Execute the second example */
uchar16 mask2 = (uchar16)(6, 10, 5, 2, 8, 0, 9, 14,
7, 5, 12, 3, 11, 15, 1, 13);
char8 input1 = (char8)('l', 'o', 'f', 'c', 'a', 'u', 's', 'f');
char8 input2 = (char8)('f', 'e', 'h', 't', 'n', 'n', '2', 'i');
*s2 = shuffle2(input1, input2, mask2);
}
'''
# Get device and context, create command queue and program
dev = utility.get_default_device()
context = cl.Context(devices=[dev], properties=None, dev_type=None, cache_dir=None)
queue = cl.CommandQueue(context, dev, properties=None)
# Build program in the specified context using the kernel source code
prog = cl.Program(context, kernel_src)
try:
prog.build(options=['-Werror'], devices=[dev], cache_dir=None)
except:
print('Build log:')
print(prog.get_build_info(dev, cl.program_build_info.LOG))
raise
# Data and buffers
s1 = cl.array.vec.zeros_float8()
s2 = np.empty(shape=(16,), dtype=np.character) # zeros_char16 would also work
a = 0;
b = 0;
/* Increment without atomic add */
a++;
/* Increment with atomic add */
atomic_inc(&b);
x[0] = a;
x[1] = b;
}
'''
# Get device and context, create command queue and program
dev = utility.get_default_device()
context = cl.Context(devices=[dev],
properties=None,
dev_type=None,
cache_dir=None)
queue = cl.CommandQueue(context, dev, properties=None)
# Build program in the specified context using the kernel source code
prog = cl.Program(context, kernel_src)
try:
prog.build(options=['-Werror'], devices=[dev], cache_dir=None)
except:
print('Build log:')
print(prog.get_build_info(dev, cl.program_build_info.LOG))
raise
while(LOCK(mutex));
*sum += 1;
UNLOCK(mutex);
int waiting = 1;
while(waiting) {
while(LOCK(mutex));
if(*sum == get_global_size(0)) {
waiting = 0;
}
UNLOCK(mutex);
}
}
'''
# Get device and context, create command queue and program
dev = utility.get_default_device(use_gpu=False)
context = cl.Context(devices=[dev], properties=None, dev_type=None, cache_dir=None)
queue = cl.CommandQueue(context, dev, properties=None)
# Check for cl_khr_global_int32_base_atomics availability
if 'cl_khr_global_int32_base_atomics' not in dev.extensions.strip().split(' '):
raise RuntimeError('Selected device does not support int32 atomic operations.')
# Build program in the specified context using the kernel source code
prog = cl.Program(context, kernel_src)
try:
prog.build(options=['-Werror'], devices=[dev], cache_dir=None)
except:
print('Build log:')
print(prog.get_build_info(dev, cl.program_build_info.LOG))
raise