[Hybrid] can we support Array tensor?

1

Hi,

can we support this case? @were

import tvm
from tvm.hybrid import script

@script
def test(A, f):
    c = output_tensor(A.shape, A.dtype)
    j = f[0]
    for i in range(A.shape[0]):
        c[i] = A[i]+1
    return c

A = tvm.placeholder((5,), name='a', dtype='float16')
f = [A]
fp = tvm.convert(f)
C = test(A, fp)
print C.op.body
2

Yes and No.
Some limited case is supported.
You should garantee the index is always compilation time constant.

3

Ok with compilation time constant, could you support for this case? thanks