def test_gufunc(self):
gufunc = GUVectorize(matmulcore, '(m,n),(n,p)->(m,p)',
target=self.target)
gufunc.add((float32[:, :], float32[:, :], float32[:, :]))
gufunc = gufunc.build_ufunc()
self.check_matmul_gufunc(gufunc)
def test_gufunc(self):
gufunc = GUVectorize(matmulcore, '(m,n),(n,p)->(m,p)',
target=self.target)
gufunc.add((float32[:, :], float32[:, :], float32[:, :]))
gufunc = gufunc.build_ufunc()
self.check_matmul_gufunc(gufunc)
def test_ufunc_like(self):
# Test problem that the stride of "scalar" gufunc argument not properly
# handled when the actual argument is an array,
# causing the same value (first value) being repeated.
gufunc = GUVectorize(axpy, '(), (), () -> ()', target=self.target)
gufunc.add('(intp, intp, intp, intp[:])')
gufunc = gufunc.build_ufunc()
x = np.arange(10, dtype=np.intp)
out = gufunc(x, x, x)
np.testing.assert_equal(out, x * x + x)
def test_ufunc_like(self):
# Test problem that the stride of "scalar" gufunc argument not properly
# handled when the actual argument is an array,
# causing the same value (first value) being repeated.
gufunc = GUVectorize(axpy, '(), (), () -> ()', target=self.target)
gufunc.add('(intp, intp, intp, intp[:])')
gufunc = gufunc.build_ufunc()
x = np.arange(10, dtype=np.intp)
out = gufunc(x, x, x)
np.testing.assert_equal(out, x * x + x)
def test_gufunc(self):
gufunc = GUVectorize(matmulcore, '(m,n),(n,p)->(m,p)', target='cpu')
gufunc.add(argtypes=[float32[:, :], float32[:, :], float32[:, :]])
gufunc = gufunc.build_ufunc()
matrix_ct = 1001 # an odd number to test thread/block division in CUDA
A = np.arange(matrix_ct * 2 * 4, dtype=np.float32).reshape(matrix_ct, 2, 4)
B = np.arange(matrix_ct * 4 * 5, dtype=np.float32).reshape(matrix_ct, 4, 5)
C = gufunc(A, B)
Gold = ut.matrix_multiply(A, B)
self.assertTrue(np.allclose(C, Gold))
def wrap(func):
guvec = GUVectorize(func, sig, nopython=True)
for t in coretypes:
guvec.add(t)
g = guvec.build_ufunc()
def h(*args, **kwargs):
out = g(*args, **kwargs)
view = xnd.from_buffer(out)
ret = xnd.empty(view.type)
np.copyto(np.array(ret, copy=False), out)
return ret
return h
def test_gufunc(self):
gufunc = GUVectorize(matmulcore, '(m,n),(n,p)->(m,p)',
target=self.target)
gufunc.add((float32[:, :], float32[:, :], float32[:, :]))
gufunc = gufunc.build_ufunc()
matrix_ct = 1001
A = np.arange(matrix_ct * 2 * 4, dtype=np.float32).reshape(matrix_ct, 2, 4)
B = np.arange(matrix_ct * 4 * 5, dtype=np.float32).reshape(matrix_ct, 4, 5)
C = gufunc(A, B)
Gold = ut.matrix_multiply(A, B)
self.assertTrue(np.allclose(C, Gold))
def test_gufunc(self):
gufunc = GUVectorize(matmulcore, '(m,n),(n,p)->(m,p)', target='cpu')
gufunc.add(argtypes=[float32[:, :], float32[:, :], float32[:, :]])
gufunc = gufunc.build_ufunc()
matrix_ct = 1001 # an odd number to test thread/block division in CUDA
A = np.arange(matrix_ct * 2 * 4,
dtype=np.float32).reshape(matrix_ct, 2, 4)
B = np.arange(matrix_ct * 4 * 5,
dtype=np.float32).reshape(matrix_ct, 4, 5)
C = gufunc(A, B)
Gold = ut.matrix_multiply(A, B)
self.assertTrue(np.allclose(C, Gold))
def test_gufunc(self, target='cpu'):
gufunc = GUVectorize(matmulcore,
'(m,n),(n,p)->(m,p)',
target=self.target)
gufunc.add((float32[:, :], float32[:, :], float32[:, :]))
gufunc = gufunc.build_ufunc()
matrix_ct = 1001
A = np.arange(matrix_ct * 2 * 4,
dtype=np.float32).reshape(matrix_ct, 2, 4)
B = np.arange(matrix_ct * 4 * 5,
dtype=np.float32).reshape(matrix_ct, 4, 5)
C = gufunc(A, B)
Gold = ut.matrix_multiply(A, B)
self.assertTrue(np.allclose(C, Gold))
