libera/#petalisp - IRC Chatlog
Search
8:43:32
heisig
I spent most of my time recently on finishing https://github.com/marcoheisig/sb-simd.
8:44:43
heisig
If everything continues that way, we'll have a world-class SIMD library as an SBCL contrib, soon.
8:46:46
heisig
He has motivated me to add a mechanism for automatic batching to Petalisp. After this change, it will almost always be possible to have Petalisp functions work on whole batches.
8:47:37
heisig
So a function working on (m x n) arrays producing (m) arrays will silently accept (m x n x k) arrays and turn them into (m x k) arrays.
8:48:10
heisig
And I have an almost complete patch for using a more efficient representation for lazy arrays.
8:49:14
heisig
It essentially boils down to using structs instead of classes, which should give faster allocation, faster traversal times, and better type safety.
8:50:41
heisig
The good news is that we handed in a grant proposal, so there is a good chance I will continue developing Petalisp after handing in my thesis.