[llvm-dev] [cfe-dev] RFC: Supported Optimizations attribute

Wed Dec 5 09:22:45 PST 2018

I can’t help but wonder if the back-and-forth here is largely about naming.

> Authors of good-faith optimizations need to design their representations so that transformations that know nothing about their optimizations but merely preserve semantics and well-formed IR structure will not break their representations

That sounds like a pretty reasonable burden to place on good-faith optimization authors, if I’m following correctly; “preserve semantics and well-formed IR structure” is something we already require of transformations to consider them correct, right?

I think that some of the points raised come from a more extreme version where “supported_optimizations” are allowed to invent wild new invariants.  I wonder if it would help to shift the naming from the optimizations to the annotations.  In particular, the annotations we’re concerned with have been embedded into the semantics of the program.  I.e., if I’m following correctly, in the case of the annotations for devirtualization, these launder/strip operations have been written into the value graph – so there’s not just a store of p anymore, there’s a store of (launder of p), something like that?  So I’m wondering if a name like “semantic_annotations” would help – this function’s semantics include some operations that are really no-op annotations (but any transform that blindly treats them as opaque operations will preserve them sufficiently for the intended consumer).  But if you merge such a function into one that hasn’t had the same class of annotations written into its semantics, you get a function that’s only partially annotated and we need to convey that to the ultimate consumer.

-Joseph

From: cfe-dev <cfe-dev-bounces at lists.llvm.org> On Behalf Of John McCall via cfe-dev
Sent: Tuesday, December 4, 2018 6:22 PM
To: Philip Reames <listmail at philipreames.com>
Cc: llvm-dev <llvm-dev at lists.llvm.org>; Sanjoy Das <sanjoy at playingwithpointers.com>; Clang Dev <cfe-dev at lists.llvm.org>; Richard Smith <RichardSmith at google.com>
Subject: Re: [cfe-dev] [llvm-dev] RFC: Supported Optimizations attribute

On 4 Dec 2018, at 17:50, Philip Reames wrote:

Skimming along, apologies if I'm repeating something which already got said.

If I understand this correctly, the basic problem we're trying to solve is to use a local hint (the invariant.group) to make a global assumption about other code which might exist elsewhere outside the function.  The attribute proposed can basically be phrased as describing a universe of functions within which our desired global property holds.  There's an ambiguity about what is allowed to be assumed about code outside that universe.

I think it's important to note that we have a precedent of something similar to this in TBAA.  TBAA information coming from different modules has the same base problem.  We solve it by using the "root" of the TBAA tree as a scope descriptor, and essentially making two TBAA nodes from distinct roots incomparable.

Can someone explain concisely why a similar scheme couldn't be used to solve this problem?

TBAA is conservative in two ways:
- It allows two accesses to alias if they have TBAA nodes with different roots.
- It allows two accesses to alias if only one of them has a TBAA node.

The second is what doesn't generalize: there are optimizations where you need to
rely on transition points being explicitly identified. Looking at a function
with no identified transition points, you don't know whether it actually doesn't
transition or whether it was compiled without the transitions being explicitly
marked. There's no way to extend the TBAA idea to make that work.

On 12/4/18 11:24 AM, John McCall via llvm-dev wrote:

Note that IPO is generally permitted to partially inline or outline code,
and so good-faith optimizations that e.g. require two instructions to be moved
in tandem or not at all must use tokens to establish that unbreakable
relationship.

I think the way your framing this is dangerous.  We absolutely can not allow any annotation of this form to *weaken* the semantics of the existing IR.  We can and should impose a criteria that any extension of this variety strictly add information to the IR which might not have been previously inferred.  We can then design rules for how to preserve our new information as long as possible, but framing this in terms of disallowed transformations is really a non-starter.

That's exactly what I was trying to convey here. Authors of good-faith
optimizations need to design their representations so that transformations
that know nothing about their optimizations but merely preserve semantics
and well-formed IR structure will not break their representations. The only
transforms that need to know about the existence of good-faith optimizations
are interprocedural optimizations; furthermore, those optimizations don't
need to know about any good-faith optimizations specifically, they just need
to understand how to correctly update the supported_optimizations list.
That is a very small burden on IPO that enables an interesting class of
language-specific optimizations.

John.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181205/cee712b3/attachment-0001.html>