[PATCH] IR: Make metadata typeless in assembly

Duncan P. N. Exon Smith dexonsmith at apple.com
Wed Dec 10 22:40:02 PST 2014


> On 2014 Dec 10, at 20:35, Robinson, Paul <Paul_Robinson at playstation.sony.com> wrote:
> 
> Note: in case it's confusing, some of the testcase changes in the code
> patch are killing newlines so the upgrade script can do it's magic.
> 
> That part really feels like a purely mechanical change that oughta be separate.
> --paulr

Thanks Paul, good point.  Committed that in r224002.

Attaching the rebased patch.  Let me know if you have any comments on LangRef!

-------------- next part --------------
A non-text attachment was scrubbed...
Name: metadata-value-split-assembly-2.patch
Type: application/octet-stream
Size: 25503 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141210/6668c127/attachment.obj>
-------------- next part --------------

>  
> From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of Duncan P. N. Exon Smith
> Sent: Wednesday, December 10, 2014 6:17 PM
> To: llvm-commits
> Subject: [PATCH] IR: Make metadata typeless in assembly
>  
> Now that `Metadata` is typeless, reflect that in the assembly.
> 
> There's really only one reasonably thing to do here: strip the
> `metadata` type (except in arguments to `call` instructions, where it
> still has meaning), and add the symmetric `value` keyword for all
> metadata operands that bridge to `Value`.
> 
> I'm confident in the code changes (although obviously I'll relish any
> criticism offered), but I'd love a review of `docs/LangRef.rst`.  I
> wasn't really sure how much exposition was appropriate, so I kept it
> pretty simple.
> 
> Also, I wrote an upgrade script that handle all of the tests in llvm
> after applying the main patch.  It even handles much of cfe (doing
> mostly the right thing with `CHECK` lines).  I'll be attaching it the PR
> after I make my commit, but feel free to spot-check its enormous result,
> which I've attached as llvm-tests.patch.  (I'll squash that in with the
> code changes when I commit.)
> 
> Note: in case it's confusing, some of the testcase changes in the code
> patch are killing newlines so the upgrade script can do it's magic.
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits



More information about the llvm-commits mailing list