[llvm-dev] Diff to add ARMv6L to Target parser

Bob Wilson via llvm-dev llvm-dev at lists.llvm.org
Fri Jan 8 10:16:13 PST 2016


> On Jan 7, 2016, at 2:21 PM, Eric Christopher <echristo at gmail.com> wrote:
> 
> 
> 
> On Thu, Jan 7, 2016 at 2:17 PM William Dillon <william at housedillon.com <mailto:william at housedillon.com>> wrote:
> Yikes!!
> 
> It looks like things have changed considerably!  I’ll need to start this from scratch.  A few questions, though:
> 
> I have a goal for this to be in the Swift 2.2 release, is that feasible?

Maybe. We’re using an older version of llvm/clang for Swift 2.2 to give us a stable platform to work with, but if you get a change into trunk first, and if it is relatively low-risk, we could consider back-porting it to the release branch for Swift 2.2. We do want to be careful to avoid destabilizing the branch, though, and the criteria for accepting changes will only get more strict over time.

> If so, will the current LLVM head end up in the branch for 2.2 when that time comes?

No. We already created the llvm/clang release branches to be used with Swift 2.2 to give us a longer time to stabilize them. We do not plan to rebranch again from trunk for Swift 2.2.

> 
> Given that the coordination costs to attempt any kind of change in swift that requires a change in LLVM are so high, I’m tempted to keep the logic of stripping the ‘l’s from armv7l and armv6l inside swift itself.  It really seems like the wrong approach, but sometimes the wrong answer is the best answer, depending on the circumstances.
> 
> 
> For all of these questions you should talk to Bob Wilson who is the llvm release manager for the swift project.
> 
> -eric
>  
> - Will
> 
>> On Jan 7, 2016, at 10:05 AM, Eric Christopher <echristo at gmail.com <mailto:echristo at gmail.com>> wrote:
>> 
>> The swift repository is old and many thousand revisions behind llvm. Please don't use it as a base when submitting to the llvm project.
>> 
>> 
>> On Thu, Jan 7, 2016, 10:02 AM William Dillon <william at housedillon.com <mailto:william at housedillon.com>> wrote:
>> Oops, I neglected to reply-all….
>> 
>> The current stable branch at github still has it:
>> 
>> https://github.com/apple/swift-llvm/blob/stable/include/llvm/Support/ARMTargetParser.def#L106 <https://github.com/apple/swift-llvm/blob/stable/include/llvm/Support/ARMTargetParser.def#L106>
>> 
>> Should I get the head of the non-swift repository and generate a new diff?
>> 
>> Also, I suspect that it’s not a good idea to have armv6l map to armv6k, because that seems like quite an assumption to make.  Clearly, armv6 sub architectures that aren’t v6k will still be v6l in linux. (provided they’re little-endian).
>> I’ve already made that change, and it would be included in any revised diff that I send out.
>> 
>> Thanks,
>> - Will
>> 
>>> On Jan 6, 2016, at 10:02 AM, Artyom Skrobov <Artyom.Skrobov at arm.com <mailto:Artyom.Skrobov at arm.com>> wrote:
>>> 
>>> William, what revision of LLVM is your patch based on?
>>> 
>>> The trunk hasn't had ARM_ARCH("armv6hl") since r252903 (Nov 12th)
>>> 
>>> 
>>> -----Original Message-----
>>> From: William Dillon [mailto:william at housedillon.com <mailto:william at housedillon.com>] 
>>> Sent: 06 January 2016 17:55
>>> To: Renato Golin
>>> Cc: Tim Northover; LLVM Dev; nd; Artyom Skrobov; Daniel Sanders; Eric Christopher
>>> Subject: Re: [llvm-dev] Diff to add ARMv6L to Target parser
>>> 
>>> Taking the suggestions of the group under consideration, I’ve generated a new diff.  The thing to note is that armv6l is now treated identically to armv6hl.  I’ve also added a unit test.
>>> This seems to me to be the least invasive method, and holds to existing conventions as closely as possible.
>>> 
>>> Thoughts?
>>> 
>> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160108/c163711c/attachment-0001.html>


More information about the llvm-dev mailing list