[cfe-dev] Expression Templates for AST matchers

[Sean Silva]

Although I'm mostly an armchair spectator right now w.r.t. clang tooling
stuff, I think that use cases like clang-query and in general the dynamic
AST matchers suggests that expression templates aren't a very good solution
because they will require reimplementing things from scratch for the
"dynamic" case.

Regardless, I don't think that the primary bottleneck is not doing enough
at compile time (of the matchers). Rather we aren't effectively indexing
the underlying data set we are trying to query (right now, we're more like
grep and less like google). Indexing starts paying off once you have to do
multiple queries* and can amortize the indexing cost across all those
queries. I think most uses for clang-tidy have enough ast matching in them
for indexing to pay off. Note that indexing doesn't have to return results
immediately; it just needs to prune the search space enough to pay off (a
good example is using a trigram index to prune the search space for regexp
matching http://swtch.com/~rsc/regexp/regexp4.html).

* "multiple" here also scales numerically with the number of TU's. An index
can avoid visiting certain TU's at all in many cases.

-- Sean Silva

On Tue, Apr 7, 2015 at 2:01 PM, Richard <legalize at xmission.com> wrote:

> OCLint, a lint checker using clang tooling, says:
> <http://docs.oclint.org/en/dev/internals/rules.html>
>
>     "Using AST matchers has more restrictions than AST visitor, and it
>     takes much longer time in analysis, this could lower the
>     performance of the tool. So, we always have to consider the
>     trade-off, and choose wisely."
>
> It seems that AST matchers, while convenient, are not the best
> performing option for processing source files.  I've done some simplistic
> measurements on clang-tidy and found that the majority of the execution
> time is spent in the matchers.  I believe this is the result of the
> matchers building runtime structures to match nodes.
>
> Has anyone investigated using the technique of expression templates to
> get the convenience of an AST matching domain-specific language (DSL)
> with the runtime efficiency of an AST visitor?
>
> If we were to pursue such a direction, would it be possible to use an
> expression template library like those in Boost for the implementation?
> I don't see any existing use of Boost in the LLVM/clang code base, so
> I infer that there is some avoidance of Boost for whatever reason.  I
> haven't done any expression template coding beyond using existing
> expression template based libraries like Boost.Spirit.  I have no idea
> what it would take to create an expression template based AST matcher
> DSL for clang, but it seems like it would be worthwhile as it would
> speed up all the tools that want to use the AST matcher style
> interface.
>
> Does anyone have any thoughts about this?  Pehraps Manuel Klimek?
> --
> "The Direct3D Graphics Pipeline" free book <
> http://tinyurl.com/d3d-pipeline>
>      The Computer Graphics Museum <http://ComputerGraphicsMuseum.org>
>          The Terminals Wiki <http://terminals.classiccmp.org>
>   Legalize Adulthood! (my blog) <http://LegalizeAdulthood.wordpress.com>
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150407/2728b65c/attachment.html>