HTO status

I just watched 2019 LLVM Developers’ Meeting: W. Moses “Cross-Translation Unit Optimization via Annotated Headers” - YouTube. It says there are “plans to upstream once cleaned up (and submitted for publication)”.

~3Y had passed. Anyone knows what is the status of this? I thought it showed great promise.

As you realized, it got hold up. We started the process back then but then we hit a roadblock or two.

@wsmoses should comment :slight_smile:

1 Like

Even just exposing more IR attributes to clang could add value.

I agree. Question is if one by one is worth it or if we do it like HTO and expose “all”. Back in the day the latter got pushback.

If we had more evidence for benefit, I think we could make a better point. Would you be interested in trying something like this to come up with some improvements based on additional attributes? I’d be happy to help. We also have some more stuff along these lines downstream.

I’m familiar with this work, think it’s brilliant and would love to try and apply it to real world examples. I think there’s a good chance I’d be able to show added value - I was able to show it with much weaker tools to guide attributes. (namely optimization remarks)
Probably not this week, but would push to get to it. Would appreciate your help too, but can we please take this offline? ofekshilon at gmail dot com

Pardon the delay I was out sick last week.

Yeah we kind of got pulled in other priorities and didn’t manage to finish up streaming.

That said we’re hoping to do so in the fall. That also said if you’d be interested in helping and or have good use cases please share/reach out.

Maybe it would be useful to set up a call for next week?

@wsmoses not sure if you’re addressing me or @jdoerfert about the call - but I think I have interesting use cases and would like to try and help. If you’re having that call please add me.

Hi. Can this still happen?