I like GORC. This would be good for straightforward bitvector implementations of automata (Glushkov NFA, Thompson NFA). Frequently the limit to the speed of how fast this can go is fan-in on "OR" operations.
I've released version 0.91 of the RISC-V Bitmanip Spec a few days ago. https://github.com/riscv/riscv-bitmanip/blob/master/bitmanip-0.91.pdf … We added two new instructions, GORC and BFP, and changed the encodings of GREV and BMATXOR. The riscv-bitmanip branches of GNU tools and Spike already support the v0.91 bitmanip spec.
-
-
-
... though is the proposal to have 9 different extensions for this serious??!?
- Show replies
New conversation -
-
-
I think there is a ‘Basic’ and ‘Advanced’ grouping that will mostly fall out of what is already in most compiler IR or intrinsics, potentially including minor fixes for base ISA shortcomings (zero and sign extension of 32-bit comes to mind).
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
‘Advanced’ will be harder to decide. The question is how long to wait to add the ones we know we need. This is partially but not completely subjective. I’m happy to read and give feedback, but thinking a summary table with a column for existing IR and intrinsics would be useful.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
It’s brilliant work btw! Exemplary in fact, based on an earlier draft. A full review would mean implementing in a simulator as that is what I find makes one become aware of the edges. The scale of the draft leads to implementation time, hence the attraction to a simple subset.
-
Oh, I did write reference implementations for it all. https://twitter.com/oe1cxw/status/1164831905692278791 … I even have a PicoRV32-based reference system that can run the rv32 bitmanip compliance tests in a simulation.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.