void py_curand_get_direction_vectors(
curandDirectionVectorSet_t set, py::object dst, int count)
{
int n = 0;
py_buffer_wrapper buf_wrapper;
buf_wrapper.get(dst.ptr(), PyBUF_ANY_CONTIGUOUS | PyBUF_WRITABLE);
void *buf = buf_wrapper.m_buf.buf;
PYCUDA_BUFFER_SIZE_T len = buf_wrapper.m_buf.len;
if (CURAND_DIRECTION_VECTORS_32_JOEKUO6 == set
#if CUDAPP_CUDA_VERSION >= 4000
|| CURAND_SCRAMBLED_DIRECTION_VECTORS_32_JOEKUO6 == set
#endif
) {
curandDirectionVectors32_t *vectors;
CURAND_CALL_GUARDED(curandGetDirectionVectors32, (&vectors, set));
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(curandDirectionVectors32_t);
memcpy((unsigned int *)buf+n*20000*sizeof(curandDirectionVectors32_t)/sizeof(unsigned int), vectors, size);
count -= size/sizeof(curandDirectionVectors32_t);
n++;
}
}
#if CUDAPP_CUDA_VERSION >= 4000
if (CURAND_DIRECTION_VECTORS_64_JOEKUO6 == set
|| CURAND_SCRAMBLED_DIRECTION_VECTORS_64_JOEKUO6 == set) {
curandDirectionVectors64_t *vectors;
CURAND_CALL_GUARDED(curandGetDirectionVectors64, (&vectors, set));
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(curandDirectionVectors64_t);
memcpy((unsigned long long *)buf+n*20000*sizeof(curandDirectionVectors64_t)/sizeof(unsigned long long), vectors, size);
count -= size/sizeof(curandDirectionVectors64_t);
n++;
}
}
#endif
}
void py_curand_get_direction_vectors(
curandDirectionVectorSet_t set, py::object dst, int count)
{
void *buf;
PYCUDA_BUFFER_SIZE_T len;
int n = 0;
if (PyObject_AsWriteBuffer(dst.ptr(), &buf, &len))
throw py::error_already_set();
if (CURAND_DIRECTION_VECTORS_32_JOEKUO6 == set
#if CUDAPP_CUDA_VERSION >= 4000
|| CURAND_SCRAMBLED_DIRECTION_VECTORS_32_JOEKUO6 == set
#endif
) {
curandDirectionVectors32_t *vectors;
CURAND_CALL_GUARDED(curandGetDirectionVectors32, (&vectors, set));
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(curandDirectionVectors32_t);
memcpy((unsigned int *)buf+n*20000*sizeof(curandDirectionVectors32_t)/sizeof(unsigned int), vectors, size);
count -= size/sizeof(curandDirectionVectors32_t);
n++;
}
}
#if CUDAPP_CUDA_VERSION >= 4000
if (CURAND_DIRECTION_VECTORS_64_JOEKUO6 == set
|| CURAND_SCRAMBLED_DIRECTION_VECTORS_64_JOEKUO6 == set) {
curandDirectionVectors64_t *vectors;
CURAND_CALL_GUARDED(curandGetDirectionVectors64, (&vectors, set));
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(curandDirectionVectors64_t);
memcpy((unsigned long long *)buf+n*20000*sizeof(curandDirectionVectors64_t)/sizeof(unsigned long long), vectors, size);
count -= size/sizeof(curandDirectionVectors64_t);
n++;
}
}
#endif
}
void py_curand_get_scramble_constants32(py::object dst, int count)
{
void *buf;
PYCUDA_BUFFER_SIZE_T len;
int n = 0;
if (PyObject_AsWriteBuffer(dst.ptr(), &buf, &len))
throw py::error_already_set();
unsigned int *vectors;
CURAND_CALL_GUARDED(curandGetScrambleConstants32, (&vectors));
// Documentation does not mention number of dimensions
// Assuming the same as in getDirectionVectors*
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(unsigned int);
memcpy((unsigned int *)buf+n*20000, vectors, size);
count -= size/sizeof(unsigned int);
n++;
}
}
void py_curand_get_scramble_constants32(py::object dst, int count)
{
int n = 0;
py_buffer_wrapper buf_wrapper;
buf_wrapper.get(dst.ptr(), PyBUF_ANY_CONTIGUOUS | PyBUF_WRITABLE);
void *buf = buf_wrapper.m_buf.buf;
PYCUDA_BUFFER_SIZE_T len = buf_wrapper.m_buf.len;
unsigned int *vectors;
CURAND_CALL_GUARDED(curandGetScrambleConstants32, (&vectors));
// Documentation does not mention number of dimensions
// Assuming the same as in getDirectionVectors*
while (count > 0) {
int size = ((count > 20000) ? 20000 : count)*sizeof(unsigned int);
memcpy((unsigned int *)buf+n*20000, vectors, size);
count -= size/sizeof(unsigned int);
n++;
}
}
