What is happening regards to light parallelism (parallel for, parallel do, Parallel_Reduce) ? I read some time ago that it had been shelved, but other things still came out of the workshop at GNAT… but not that ?
I suppose the chunk features could be hard to implement, but they could put aside and still provide the rest, no ?
I haven’t seen anything recent on it. I know someone (either Tucker Taft or maybe AdaCore itself) released a library for it. Not the same as built in language support, but something you could dig into maybe:
It’s not like I have a immediate need for it… I’m just disappointed. It’s the kind of thing I get all tingly at the thought of using it everywhere. It’s the same thing as the pleasure of reading and writing clean mathematical proofs.
I just wish they would concentrate on this rather than pattern matching, case constructs and string interpolation ( ).
Well, there’s the Truffle bootstrap I’m working on… which should have the ability to parallelize the constructs as Truffle has a CUDA language and I plan on leveraging/interop with it to provide parallel-execution.
(Actually, if using an AST/IR interpreter, getting parallelism should be relatively easy: just load multiple TASK
s executing the parallel portions of the AST/IR.)
Yeah, I’m not that pleased with those directions either.
TBH, I’d rather focus on being able to describe the type-system within the type-system, which would also give us better proving and interfacing constructs.