Subject: Re: [llvm-dev] [RFC][PIR] Parallel LLVM IR -- Stage 0 -- IR extension
Are there actually people interested in this?
I’m definitely interested too. I will have some high-level comments after next week.
We will continue working anyway but it might not make sense to put it
on reviews and announce it on the ML if nobody cares.
I expect this to takes many months, unless it is on the critical path of some core devs, because it seems to me to be a huge feature to consider and quite a time sink (which is why even people that care haven’t been able to invest enough time into this).
There are enough commercial and open-source projects working actively on this that I think the issue is reaching critical mass. A few of us — Hal, Jeff (Vetter), Johannes, TB Schardl, Xinmin, I, and a few others — have been discussing design ideas, building on lessons from some of these ongoing projects:
* Intel’s new OpenMP compiler
* ARES from Los Alamos and Oak Ridge
* PIR from Saarland.
* TAPIR from MIT
* HPVM from Illinois
Our goal is to put together a design for how to add representations for parallel programs that can leverage LLVM capabilities and features without requiring excessive changes to the existing infrastructure. We’ll submit design docs, patches for review, and one or more prototype compilers built on the design.
We’re having a conference call this week or early next week, and a face-to-face meeting at the end of next week to work on the details.
My impression is that this is major enough that it is unlikely to land in LLVM 5.0,
That’s probably true.
and even before the next US dev meeting.
I’m hoping we’ll have info submitted for review well before that.
Ideally we’d have enough discussions and review before then that we could have a BoF there to make sure we can move forward with the right representation. I hope the EuroLLVM meeting will be the opportunity for you to get some good feedback on this!
+1 on both.
// Vikram S. Adve
// Professor, Department of Computer Science
// University of Illinois at Urbana-Champaign