[llvm-dev] RFC: Resolving TBAA issues (original) (raw)

Hal Finkel via llvm-dev llvm-dev at lists.llvm.org
Sat Aug 19 12:33:12 PDT 2017

On 08/19/2017 02:05 PM, Daniel Berlin wrote:

FWIW: to be completely concrete; from my perspective, the main thing we'd get out of switching is something that is more complete already.

past that "I don't want language-specific kinds of nodes in the grammar, and I believe that it's not necessary under some reasonable variant of this scheme. Maybe I'm wrong." If your position is that you'd be fine with a variant of this scheme, than i don't really think we disagree at all :)

That's exactly my position. Good :-)

Here's what I find most intriguing about this: In the way we'd discussed extending the current scheme to handle unions, we extend the current way that we traverse the graph such that, if there are multiple fields in one of the types with the same offset (i.e. a union), when we need to walk up the graph through all fields. While I still believe this is unlikely to be problematic in practice, we're now exploring many paths, and the asymptotic complexity doesn't thrill me.

If I'm thinking about this correctly, this type of encoding makes dealing with unions much more efficient. The frontend knows the access path, and can encode that directly. We don't need to explore many paths through a graph to find it (i.e. find a potential set of paths through the graph that indicate potential aliasing). Instead we can just examine the two paths directly encoded. If they both have the same union type at the same offset, then they can alias. If they have the same union type but incompatible offsets, they can't. No path-finding required. Also, if I'm thinking about this correctly, doing it this way is also more accurate (because it can distinguish between structurally-identical union members, and the proposed extension we'd discussed previously cannot (i.e., it would give a conservative answer)).

Thoughts?

Thanks again, Hal

On Sat, Aug 19, 2017 at 12:00 PM, Hal Finkel <hfinkel at anl.gov_ _<mailto:hfinkel at anl.gov>> wrote: On 08/17/2017 11:25 PM, Daniel Berlin wrote:

<just want to focus on these parts for a second. *All* of these_ _representations are really access path representations, just_ _encoded slightly different ways, and, as a result, with various_ _parts of the rules in slightly different places> Imagine that we took the enhancement we previously discussed, but instead of implementing it directly, we just directly encoded for every access the path from the access type to the root. I think it would look very much like this proposal.

Something like it, yes. Note that this representation also has special vtable and union groups, for example, and assumes very c-like rules in several places around the sequencing of access types. Also note that in the language of access paths, C allows overlapping that does not exist in, for example, Java Is this a statement generally favoring or disfavoring having the frontend encode the path explicitly? (This is true in both the points-to and type-based domains). Ada has discriminated unions (and you could not use the "union" in this proposal to represent them) Yes, but discriminated unions also have a dynamic element to them. As a result, I suspect we'd need a scheme that captures that (or else we'd need to model them like a struct with a field and something like a C union). etc.

We generally explain our current TBAA rules by saying that they're generic but motivated by C/C++ rules. We do say that but that's not really what our implementation does in any way. Really? I thought it was motivated by C/C++ rules. When you say that it's "not really what our implementation does", is this because it drops a lot of potential information in the translation to our generic form? What we do is completely and totally unrelated to types as they exist in the original language. It is a completely and totally generic implementation with no rules related to the original language. The original language, for example, has dynamic typing rules, object lifetime rules, etc. We have none of these. The types do not, in fact,even always have the same relationship we give them. We have a tree and some nodes, and we happen to name them after types. Like this proposal, they also represent computed access paths, in a much simpler language. The nodes represent alias set members (either one or many) The edges represent either "access at offset" (where the default is 0 if unweighted). We have a simple rule that that processes the access paths related to ancestry. I certainly understand all of this ;) You can view it as either a grammar parsing or a graph reachability problem depending on what works for you (in fact, it's really a Dyck-CFL reachability problem on bidirected trees) The reachability rule is usually given as: If node Target is reachable from node Source, offset another through either it's children or the upwards edges (or was it downwards, i always screw up which direction we go in), they may-alias You could also implement it as a real grammar parsing problem (IE you quite literally could generate strings from tag + offset, and parse them against the grammar, and see if the terminal nodes contain your target). They are equivalent. This representation would be the same in that regard. Our current lowering for clang happens to kind of look like c/c++ structures converted to a tree. However, this is actually just inefficient, space wise, and done because it's simple to lower it this way. I agree (although this wasn't happenstance; the representation was designed with an expected lowering scheme in mind, at least for C/C++). Because the accesses are completely unrelated to the original types, and require no language rules to interpret, you could also just partition the things that alias by the language rules in the frontend, and then output a tree that represents the possible unique paths. IE figure out all the answers, then re-encode it as precisely as possible. This trades time for space. This representation is super generic. That is, the language being used here is super simple, and the reachability rule is super simple. I could have the frontend generate these trees based on anything i like. I could, in fact, encode steensgaard points-to results as this tree without any trouble. The access path language described in this proposal is more complex and complete, and directly closer to access paths you find in C/C++. It has bitfields, vtables, and unions. It is more complex, but I think this is partly in the description. AFAIKT, only "union" has special properties under the scheme. Bit fields, vtables, etc. are all just particular types with no particular special rules. I don't like special rules at all for any named entities, but as there's only one such entity right now ("union"), this is an encoding choice we could bikeshed. The reachability rules are more complex. Is it possible to express other languages in that set of access path rules? Sure. For example, you can, as above, generate the set of answers, and then re-encode it into access paths. Right now, the work it takes in the front end is minimal, and has a fairly efficient space encoding. if i want to say two things alias, i just gotta be able to reach one from the other. If i want to say two things do not alias, i just gotta be able to not reach one from the other only using certain types of edges In our current language, all that takes in a frontend is "find longest no-aliasing part of tree. Go to parent, add new child". In the proposed language, the lowering is more complex. Is it doable? Sure, of course, not gonna claim otherwise. But the more "features" of the access path you add and expect the middle end to handle, instead of the front-end expanding them, and the more those feature's reachability rules are related to a specific language the more language-specific it gets. That's the only tradeoff we are making here, representationally. How much does the frontend have to understand about how the middle end walks these things, in order to generate whatever it wants as precisely as possible We can make whichever we want express everything we want in N^3 time :) The real question is "do we try to add on to what we have in ways that work for multiple languages, and are expressed neutrally in a simple reachability language" or "do we add language-specific kinds of nodes to the grammar, and have reachability rules that are fairly language specific". I don't want language-specific kinds of nodes in the grammar, and I believe that it's not necessary under some reasonable variant of this scheme. Maybe I'm wrong. IE do you add, say, discriminatedunion nodes to our current representation for ada, or "vtableaccess" nodes to our current representation for C++ vtable accesses Or do you instead generate a metadata that has a unidirectional edge reachability (IE up only), or whatever it takes to do vtables generically. Both are completely and totally viable paths, and it's all about which way you want to go. But they definitely have a difference in terms of language-specificness. To be clear, if the system is not generic, I'm far less interested. Thanks again, Hal -- Hal Finkel Lead, Compiler Technology and Programming Languages Leadership Computing Facility Argonne National Laboratory

-- Hal Finkel Lead, Compiler Technology and Programming Languages Leadership Computing Facility Argonne National Laboratory

-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170819/4ce20b3b/attachment.html>

More information about the llvm-dev mailing list