dependencies, and also to treat loops more carefully.

* Dependencies are now tracked based on a list of active potential
  sources, and the active set is narrowed by filtering through
  readers/writers
* Stricter handling of dependencies within loops by not assuming
  that loops are nonempty
* Changes the barrier insertion algorithm to act in two stages: first,
  barriers are recursively inserted into loops, then they are inserted
  at the outermost level. This is mostly meant to simplify the code.
* See also #18 on gitlab

Fix barrier insertion above top of loop (Fixes #17 on gitlab)

See merge request !73

Faster type inference

See merge request !68

Fix atomicity check

See merge request !71

Fix dependency graph drawing for {Call,ReturnFrom}Kernel

See merge request !70

Improve pickleability: Avoid interning instruction ids that are None,

which causes a TypeError (closes #14).

See merge request !69

which causes a TypeError (closes #14).

be faster in the common case (all the types are the same).

topological order, chunked according to SCCs of the dependency graph.

See also #15.

find_idempotence(): Replace fixed point iteration with a check based

on SCCs (closes #16).

See merge request !66

on SCCs (closes #16).

Bump kernel version to invalidate caches with broken schedule results (inducer/loopy!62)

Use CSE caching for mappers that may get called before the CSE -> assignment translation

See merge request !64

This reverts merge request !62

Remove ambiguity warning from the scheduler

Implement https://github.com/inducer/loopy/issues/74

See merge request !62

it is needed and useful in tests.

Speed up temporary shape deduction by using a batched access range

mapper call (closes #13).

See merge request !61

Fix doctest: Execute the block that disables syntax highlighting.

See merge request !60

mapper call (closes #13).

Allow integer reductions

Enables reductions on integer variables, and adds a correctness test for said reductions

See merge request !59