[patch] Start removing the old JIT
chisophugis at gmail.com
Sat Jul 26 09:41:12 PDT 2014
On Thu, Jul 24, 2014 at 6:30 PM, Philip Reames <listmail at philipreames.com>
> I fully accept your internal tool argument, but given the documentation
> is more or less "see what google says", making flag migrations obvious
> seems like a helpful thing. :)
Actually LLI supposedly has its own man page:
I have no idea how up-to-date it is though.
(so do many other of our tools: http://llvm.org/docs/CommandGuide/ )
-- Sean Silva
> On 07/24/2014 03:19 PM, Lang Hames wrote:
> Hi Philip,
> I think adding a short-term warning is a good idea. When the old JIT is
> removed however, the plan is for this option to be removed too. LLI is an
> internal tool, and doesn't guarantee backwards compatibility with flags.
> - Lang.
> On Thu, Jul 24, 2014 at 2:33 PM, Philip Reames <listmail at philipreames.com>
>> On 07/23/2014 09:26 PM, Rafael Espíndola wrote:
>>> - cl::opt<bool> UseMCJIT(
>>> - "use-mcjit", cl::desc("Enable use of the MC-based JIT (if
>>> - cl::init(false));
>> I would suggest that we leave this option in and switch the default.
>> This keeps command line compatibility for all users.
>> We should add an explicit warning/error for !UseMCJIT though. If someone
>> is explicitly opting out of MCJIT, we should probably let them know that's
>> not an option any more.
>> This would even make sense as a separate change a few days before
>> anything else. It would help flush out things which unknowingly or
>> implicitly depend on using the old JIT.
>> Also, where is the documentation change? That should be part of the very
>> first patch. :)
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the llvm-commits