def test_cublasDcopy(self):
x = np.random.rand(5).astype(np.float64)
x_gpu = gpuarray.to_gpu(x)
y_gpu = gpuarray.zeros_like(x_gpu)
cublas.cublasDcopy(self.cublas_handle, x_gpu.size, x_gpu.gpudata, 1,
y_gpu.gpudata, 1)
assert np.allclose(y_gpu.get(), x_gpu.get())
def fun_b_k(self, b_k_gpu, k):
cublas.cublasDcopy(self.h, self.b_gpu.size, self.b_gpu.gpudata, 1,
b_k_gpu.gpudata, 1)
for i in range(len(self.Ax_gpu)):
if i != k:
cublas.cublasDaxpy(self.h, b_k_gpu.size, np.float64(-1),
self.Ax_gpu[i].gpudata, 1, b_k_gpu.gpudata,
1)
def _zaxpy(self, handle, z_gpu, alpha, x_gpu, y_gpu):
# copy y to z
cublas.cublasDcopy(handle, y_gpu.size, y_gpu.gpudata, 1, z_gpu.gpudata,
1)
cublas.cublasDaxpy(handle, x_gpu.size, alpha, x_gpu.gpudata, 1,
z_gpu.gpudata, 1)