[WIP] Efficient storage of sum data types #522
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Riscky
commented
Jun 30, 2022
| @@ -259,6 +263,14 @@ packVec16 a b c d e f g h i j k l m n o p = runST $ do | |||
| ByteArray ba# <- unsafeFreezeByteArray mba | |||
| return $! Vec ba# | |||
|
|
|||
| replicateVecN :: forall a n . (KnownNat n, Prim a) => a -> Vec n a | |||
There was a problem hiding this comment.
This doesn't really belong in this PR, but the function seems useful enough to keep around
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR contains the Accelerate side of my thesis work.
It integrates the POSable library, which generically transforms Haskell 98 data types in the recursive tagged union layout.
This memory layout, specifically geared to sum types, is quite memory efficient.
The work is not finished (in fact, it doesn't even compile yet).
Before merging this, we should decide what to do with pure product types (tuples).
Do we represent those with a unified tag or with the tags inside the product?
I think the latter might be preferable, because this keeps
zipWitha no-op.An (incomplete) list of things that should be done before that this can be merged:
Matchableclass (based on the current implementation for polymorphicEither)Bool,MaybeandEither)Backends also have to be updated with the new AST constructors.
No interesting code has to be generated for the union constructors (just type casts).
It is important however that we keep padding equal between execution environments (CPU/GPU).
The updated
Caseconstructors have to be implemented as ranged case statements instead of comparisons.There is also a whole slew of possible improvements that are discussed in my thesis reports, which are probably out of scope for this PR.
How has this been tested?
Not yet, as this does not compile yet.
Types of changes
Checklist
Note
The commit history of this PR is quite terrible and should definitely be cleaned up.
I don't think this should be just a single commit, but I'm also not sure how I'd split this up in nicely defined commits.