[llvm-commits] [llvm] r77060 - in /llvm/trunk: include/llvm/Target/ lib/ExecutionEngine/JIT/ lib/Support/ lib/Target/ARM/TargetInfo/ lib/Target/Alpha/TargetInfo/ lib/Target/CBackend/TargetInfo/ lib/Target/CellSPU/TargetInfo/ lib/Target/CppBackend

[Daniel Dunbar]

On Sat, Jul 25, 2009 at 10:49 AM, Chris Lattner<clattner at apple.com> wrote:
> On Jul 25, 2009, at 3:10 AM, Daniel Dunbar wrote:
>> Author: ddunbar
>> Date: Sat Jul 25 05:09:50 2009
>> New Revision: 77060
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=77060&view=rev
>> Log:
>> Simplify JIT target selection.
>> - Instead of requiring targets to define a JIT quality match
>> function, we just
>>   have them specify if they support a JIT.
>>
>> - Target selection for the JIT just gets the host triple and looks
>> for the best
>>   target which matches the triple and has a JIT.
>
> Will JIT construction eventually go through the registry mechanism?
> If so, you can just see if the JIT ctor method is non-null for the
> target instead of needing an extra bool.  This also lets you know that
> the target has a jit *and it is linked in*.

What is "JIT construction", the addCodeEmitter hook?

But yes, I agree this is a silly solution, the target shouldn't have
to explain that it supports the JIT. I don't plan to refactor any
parts of the target machine to decouple things more, but it would make
sense to do.

 - Daniel