def test_cublasDtrsmBatched(self):
l, m, n = 11, 7, 5
A = np.random.rand(l, m, m).astype(np.float64)
B = np.random.rand(l, m, n).astype(np.float64)
A = np.array(list(map(np.triu, A)))
X = np.array([np.linalg.solve(a, b) for a, b in zip(A, B)])
alpha = np.float64(1.0)
a_gpu = gpuarray.to_gpu(A)
b_gpu = gpuarray.to_gpu(B)
a_arr = bptrs(a_gpu)
b_arr = bptrs(b_gpu)
cublas.cublasDtrsmBatched(self.cublas_handle, 'r', 'l', 'n', 'n', n, m,
alpha, a_arr.gpudata, m, b_arr.gpudata, n, l)
assert np.allclose(X, b_gpu.get(), 5)
def test_cublasDtrsmBatched(self):
l, m, n = 11, 7, 5
A = np.random.rand(l, m, m).astype(np.float64)
B = np.random.rand(l, m, n).astype(np.float64)
A = np.array(list(map(np.triu, A)))
X = np.array([np.linalg.solve(a, b) for a, b in zip(A, B)])
alpha = np.float64(1.0)
a_gpu = gpuarray.to_gpu(A)
b_gpu = gpuarray.to_gpu(B)
a_arr = bptrs(a_gpu)
b_arr = bptrs(b_gpu)
cublas.cublasDtrsmBatched(self.cublas_handle, 'r', 'l', 'n', 'n',
n, m, alpha,
a_arr.gpudata, m,
b_arr.gpudata, n, l)
assert np.allclose(X, b_gpu.get(), 5)