[libre-riscv-dev] shader compiler progress update

Luke Kenneth Casson Leighton lkcl at lkcl.net
Fri Dec 21 23:36:25 GMT 2018


On Fri, Dec 21, 2018 at 11:10 PM Jacob Lifshay <programmerjake at gmail.com> wrote:

> I haven't made any commits to Kazan in the last week or so because I've
> been working on implementing the SIMT to masked-vectors conversion and been
> running into problems.

 understood.  i quite like the vector-masking programming model. the
Aspex ASP (Array-String Process) was however an absolute b*****d to
program (4096 2-bit SIMD maskable ALUs).  so... it's a faaairly big
compsci-sized deal :)

> The benefit of the Op*Merge implementation is that the input SPIR-V is
> already structured, though the structure is very complex to parse.

 hmmm... i have an idea, which always tends to help me: how would you
feel about writing an update for the crowdsupply campaign, about this?
just outline where things stand, what the end-goal is, and so on.
often i find that the more times (in different places) i talk about
stuff, clarity and ideas come.

 what you reckon?

l.



More information about the libre-riscv-dev mailing list