On 11/13/20 1:10 AM, Richard Sandiford via Gcc-patches wrote:
> Just after GCC 10 stage 1 closed (oops), I posted a patch to add a new
> combine pass. One of its main aims was to allow instructions to move
> around where necessary in order to make a combination possible.
> It also tried to parallelise instructions that use the same resource.
>
> That pass contained its own code for maintaining limited def-use chains.
> When I posted the patch, Segher asked why we wanted yet another piece
> of pass-specific code to do that. Although I had specific reasons
> (which I explained at the time) I've gradually come round to agreeing
> that that was a flaw.
>
> This series of patches is the result of a Covid-time project to add
> a more general, pass-agnostic framework. There are two parts:
> adding the framework itself, and using it to make fwprop.c faster.
>
> The framework part
> ------------------
>
> The framework provides an optional, on-the-side SSA view of existing
> RTL instructions. Each instruction gets a list of definitions and a
> list of uses, with each use having a single definition. Phi nodes
> handle cases in which there are multiple possible definitions of a
> register on entry to a basic block. There are also routines for
> updating instructions while keeping the SSA representation intact.
>
> The aim is only to provide a different view of existing RTL instructions.
> Unlike gimple, and unlike (IIRC) the old RTL SSA project from way back,
> the new framework isn't a “native” SSA representation. This means that
> all inputs to a phi node for a register R are also definitions of
> register R; no move operation is “hidden” in the phi node.
Hmm, I'm trying to parse what the last phrase means. Does it mean that
the "hidden copy" problem for out-of-ssa is avoided? And if so, how is
that maintained over time. Things like copy-prop will tend to introduce
those issues even if they didn't originally exist.
>
> Like gimple, the framework treats memory as a single unified resource.
>
> A more in-depth summary is contained in the doc patch, but some
> other random notes:
>
> * At the moment, the SSA information is local to one pass, but it might
> be good to maintain it between passes in future.
Right. I think we can look at the passes near fwprop as good targets
for extending the lifetime over which we have an SSA framework. I note
CSE is just before the first fwprop and CSE is a hell of a lot easier in
an SSA world :-) It's unfortunately that there's no DCE passes abutting
fwprop as DCE is really easy in an SSA world.
>
> * The SSA code groups blocks into extended basic blocks, with the
> EBBs rather than individual blocks having phi nodes.
So I haven't looked at the patch, but the usual place to put PHIs is at
the dominance frontier. But extra PHIs just increase time/memory and
shouldn't affect correctness.
>
> * The framework also provides live range information for registers
> within an extended basic block and allows instructions to move within
> their EBB. It might be useful to allow further movement in future;
> I just don't have a use case for it yet.
Yup. You could do something like Click's algorithm to schedule the
instructions in a block to maximize CSE opportunities on top of this.
>
> * One advantage of the new infrastructure is that it gives
> recog_for_combine-like behaviour: if recog wants to add clobbers
> of things like the flags register, the SSA code will make sure
> that the flags register is free.
I look more at the intersection between combine and SSA as an
opportunity to combine on extended blocks, simplify the "does dataflow
allow this combination" logic, drop the need to build/maintain LOG_LINKS
and more generally simplify note distribution.
> * I've tried to optimise the code for both memory footprint and
> compile time. The first part involves quite a bit of overloading
> of pointers and various other kinds of reuse, so most of the new data
> structures use private member variables and public accessor functions.
> I know that style isn't universally popular, but I think it's
> justified here. Things could easily go wrong if passes tried
> to operate directly on the underlying data structures.
ACK.
>
> * Debug instructions get SSA information too, on a best-effort basis.
> Providing complete information would be significantly more expensive.
>
> * I wasn't sure for new C++ code whether to stick to the old C /* … */
> comments, or whether to switch to //. In the end I went for //,
> on the basis that:
>
> - The ranger code already does this.
>
> - // is certainly more idiomatic in C++.
>
> - // is in the lisp tradition of per-line comments and it matches the
> ;; used in .md files. I feel sure that GCC would have been written
> using // from the outset if that had been possible.
I think we're allowing both and realistically /* */ vs // shouldn't be
something we spend a lot of time arguing about :-)
>
> The patches only do this for new files. The aim is to ensure that
> each file is at least self-consistent.
ACK.
Anyway, given this posted before end of stage1, it deserves
consideration of gcc-11. It's (by far) the largest set in my gcc-11 queue.
jeff
