[PATCH] D14227: Add a new attribute: norecurse

[David Blaikie via llvm-commits]

On Mon, Nov 2, 2015 at 8:48 AM, James Molloy <james at jamesmolloy.co.uk>
wrote:

> Hi David,
>
> There was some discussion here that caused me to go down this route:
> http://reviews.llvm.org/D14148 . There are a couple of reasons why I
> think this is better as an attribute than an analysis, although I'm not
> strongly opposed to it being an analysis instead.
>
> * Analyses are easy to clobber, hard to preserve. The chances of having to
> recompute this information when needed is high. This is a weak argument in
> and of itself, though.
> * Norecurse does actually model language features. OpenCL mandates no
> recursion, so all of its BIFLs could be annotated with this so that
> optimizatons on non-recursive functions can take place even when BIFLs
> aren't available (function prototypes only).
> * In C++, "main" is non-recursive. This is not the case in C, and there is
> a longstanding gratuitous hack in GlobalOpt that completely breaks C in
> favour of C++ by string-matching "main". This attribute allows us to have
> Clang tell the backend this information in a language-safe manner.
>

Ah, right, I recall skimming that last bit coming up in review of the
related change or nearby recently - thanks for the reminder/context. (& the
OpenCL example too).


>
> Cheers,
>
> James
>
> On Mon, 2 Nov 2015 at 16:41 David Blaikie via llvm-commits <
> llvm-commits at lists.llvm.org> wrote:
>
>> On Mon, Nov 2, 2015 at 5:03 AM, James Molloy via llvm-commits <
>> llvm-commits at lists.llvm.org> wrote:
>>
>>> jmolloy created this revision.
>>> jmolloy added reviewers: manmanren, dexonsmith, joker.eph.
>>> jmolloy added a subscriber: llvm-commits.
>>> jmolloy set the repository for this revision to rL LLVM.
>>>
>>> This attribute allows the compiler to assume that the function never
>>> recurses into itself, either directly or indirectly (transitively). This
>>> can be used among other things to demote global variables to locals.
>>>
>>
>> Where would this attribute generally come from? Is it expected that this
>> would be exposed as a source language attribute? (any prior art there?)
>> Otherwise wouldn't this be an LLVM analysis, rather than a first-class IR
>> feature?
>>
>>
>>>
>>> The norecurse attribute indicates that the function does not call itself
>>> either directly or indirectly down any possible call path.
>>>
>>> Repository:
>>>   rL LLVM
>>>
>>> http://reviews.llvm.org/D14227
>>>
>>> Files:
>>>   docs/LangRef.rst
>>>   include/llvm/Bitcode/LLVMBitCodes.h
>>>   include/llvm/IR/Attributes.h
>>>   include/llvm/IR/Function.h
>>>   lib/AsmParser/LLLexer.cpp
>>>   lib/AsmParser/LLParser.cpp
>>>   lib/AsmParser/LLToken.h
>>>   lib/Bitcode/Reader/BitcodeReader.cpp
>>>   lib/Bitcode/Writer/BitcodeWriter.cpp
>>>   lib/IR/Attributes.cpp
>>>   lib/IR/Verifier.cpp
>>>   test/Bindings/llvm-c/invalid-bitcode.test
>>>   test/Bitcode/attributes.ll
>>>   test/Bitcode/compatibility.ll
>>>   test/Bitcode/invalid.ll
>>>   test/LTO/X86/invalid.ll
>>>
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>
>>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151102/fef924d2/attachment.html>