loopy

loopy copied to clipboard

Accumulator name after `lp.realize_reduction` is intractable

1

Maybe the newly introduced accumulator variables should be tagged with some specifications or the naming must be standardized. In feinsum, I have to rely on internals of realize_redudction (cf https://github.com/kaushikcfd/feinsum/blob/87e3a43dff6ebbc3709a3afc9df174ab2e12c8fa/src/feinsum/tuning/impls/ifj_fe_fej_to_ei.py#L447-L448)...

kaushikcfd

Matching an instruction based on the substitution rule it invokes

1

There's no way to express such match criteria, maybe we need one such `MatchExpressionBase`?

kaushikcfd

The function signature for the transforms aren't finalized yet. I'm happy to make changes to these with help from the reviewers. Draft because: - [x] Pass CI - [x] Change...

kaushikcfd

Used in e.g. `split_reduction_outward`, #711. cc @kaushikcfd

inducer

add TaggedExpression

7

TODOs: - [x] add to MemAccess, Sync _Please squash_

matthiasdiener

[Update] generate_loop_schedule_v2

2

Followup of #350

matthiasdiener

Switch immutables.Map and pyrsistent.PMap to constantdict

2

This currently a proof-of-concept. **Edit:** I removed the previous performance results, they were likely caused by some kind of caching of kernels. TODOs: - [x] ~~add `mutate` support to constantdict...

matthiasdiener

support aborting on cache miss

1

What do you think @inducer? This would not only make debugging cache misses easier, it could also be used to automate determinism tests (by setting `LOOPY_ABORT_ON_CACHE_MISS` to something trueish, and...

matthiasdiener

Allow vectorization of ternaries, indices

1

``` knl = lp.make_kernel( " { [i] : 0

nkoskelo

- [x] This will fail until https://github.com/inducer/pymbolic/pull/151 is in. - [ ] Maybe this will wait until pymbolic 2024.1 is released, so that the Firedrake test can pass?

inducer