<div dir="ltr">Yea. I think AMD chains the node they're expanding into, but they don't chain it into an _existing_ chain. e.g. adding A->B to the DAG is ok. But adding A->B and next C->D with B->C is the problem. I appreciate the input</div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Thu, Jul 16, 2020 at 2:04 PM Matt Arsenault <<a href="mailto:arsenm2@gmail.com">arsenm2@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><br>
<br>
> On Jul 16, 2020, at 17:00, Hendrik Greving <<a href="mailto:hgreving@google.com" target="_blank">hgreving@google.com</a>> wrote:<br>
> <br>
> > No, non-sideeffecting operations can be legalized as compiler-rt calls<br>
> <br>
> Right, but not as "regular" nodes with side-effects? I guess you could search and analyze the DAG manually but that seems hacky. Maybe something that one day LLVM could support natively.<br>
<br>
<br>
You can’t add arbitrary chains or glue to the regular nodes, but you can define a custom node you select the same way with your chain/glue. You don’t need to preprocess the IR and can do in the custom lowering. This is what AMDGPU does for FDIV (see AMDGPUISD::FMA_W_CHAIN). GlobalISel avoids these complications by not having nodes or chains, and just instructions with side effects, so in that sense this is a solved problem.<br>
<br>
-Matt</blockquote></div>