Invalid result for HO function with bound instancemethod argument #63

Activity

1. quasilyte commented on Oct 5, 2018

   

   quasilyte

   on Oct 5, 2018

   Last edited by quasilyte

   ContributorAuthor

   More actions

   Both HPAT and Numba fail to compile this code:

   import numpy as np
   from numba import jit
   
   @jit(nopython=True)
   def indirect_ok(fn):
       return fn()
   
   @jit(nopython=True)
   def works_as_expected():
       a = np.array([1])
       return indirect_ok(a.sum)
   
   print(works_as_expected())

   For HPAT:

   import numpy as np
   import hpat
   
   @hpat.jit
   def indirect_ok(fn):
       return fn()
   
   @hpat.jit
   def works_as_expected():
       a = np.array([1])
       return indirect_ok(a.sum)
   
   print(works_as_expected())
2. ehsantn commented on Oct 5, 2018

   

   ehsantn

   on Oct 5, 2018

   Contributor

   More actions

   HPAT has an inline pass that inlines other jit functions. With inlining in place, I don't think these example will have higher order function call.

   However, this pass was a quick hack for a use case written long time ago and definitely needs to be revisited. For example, it might not set the IR definitions data structures properly.
3. quasilyte commented on Oct 5, 2018

   

   quasilyte

   on Oct 5, 2018

   Last edited by quasilyte

   ContributorAuthor

   More actions

   Ah, I see.
   With inlining disabled, HPAT gives same typing error for bound instancemethod passed as a function argument. Thanks.
4. ehsantn commented on Oct 8, 2018

   

   ehsantn

   on Oct 8, 2018

   Contributor

   More actions

   I looked at these examples some more. Looks like inlining is fine. For the top example, the weird function has a type unification issue, since either a dataframe (from df.head) or 0 is assigned to a variable based on a conditional. This will be caught when we make the hiframes pass a typed pass (the current pass has no way of knowing).

   I think the a.sum example triggers a bug in Numba's parfor preprocessing stage where a.sum should be converted to np.sum(a) but looks like the call doesn't have the array argument.