[llvm-dev] Deleting function IR after codegen

[Pete Cooper via llvm-dev]

> On Mar 7, 2016, at 5:05 PM, Justin Bogner <mail at justinbogner.com> wrote:
> 
> Pete Cooper via llvm-dev <llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>> writes:
>> Hi all
>> 
>> After codegen for a given function, the IR should no longer be needed.
>> In the AsmPrinter we convert from MI->MCInstr, and then we never go
>> back and look at the IR during the MC layer.
>> 
>> I’ve prototyped a simple pass which can be (optionally) scheduled to
>> do just this.  It is added at the end of addPassesToEmitFile.  It is
>> optional so that clang can continue to leak the IR with
>> --disable-free, but i would propose that all other tools, and
>> especially LTO, would enable it.  The savings are 20% of peak memory
>> in LTO of clang itself.
> 
> Did you happen to measure whether it's even worth keeping for
> --disable-free? I suppose the cost of deleting it might be enough to
> worry about for very small compilations, in which case we'd probably
> want to keep the current behavior.
I didn’t, no.  We could though.

I’ve CCed Vedant as I know he looked at the cost of freeing the IR fairly recently.  I think he said it was about 3-4% compile time, but he can correct me if i have that wrong.
> 
>> I could attach a patch, but first i’d really like to know if anyone is
>> fundamentally opposed to this.
> 
> I'm definitely interested in seeing the patch. ISTM that anywhere the MC
> layer needs to access IR would be a bit of a layering violation already,
> so we'd want to clean that up anyway.
Personally I think its a layering violation.  At least I think it is one for the AsmPrinter FunctionPass to be walking the use lists of Globals.  I don’t think FunctionPasses at any level should be able to do that.

Cheers,
Pete
> 
>> I should note, a couple of issues have come up in the prototype.
>> - llvm::getDISubprogram was walking the function body to find the
>>  subprogram.  This is trivial to fix as functions now have !dbg on
>>  them.
>> - The AsmPrinter is calling canBeOmittedFromSymbolTable on
>>  GlobalValue’s which then walks all their uses.  I think this should be
>>  done earlier in codegen as an analysis whose results are available to
>>  the AsmPrinter.
>> - BB’s whose addresses are taken, i.e. jump tables, can’t be deleted.
>>  Those functions will just keep their IR around so no changes there.
>> 
>> With the above issues fixed, I can run make check and pass all the tests.
>> 
>> Comments very welcome.
>> 
>> Cheers,
>> Pete
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org <mailto:llvm-dev at lists.llvm.org>
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev <http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160307/344e2642/attachment.html>