[PATCH] [RuntimeDyld] Make ObjectBuffer/ObjectFile -> ObjectImage conversion a static method.

Lang Hames lhames at gmail.com
Wed Mar 5 12:06:00 PST 2014

Yep - the Obj local should go away. I'll make sure I do that if/when
this gets committed.

createObjectImageFromFile (and createObjectImage) should return
unique_ptrs. I intend to change them in future work. (That's fairly
orthogonal to the issues addressed in this patch, and I didn't want to
complicate review).

I kept loadObject's return value just to minimize the scope of this
patch, but I agree that it should probably go away, since it's just
returning its argument again.

- Lang.

On Wed, Mar 5, 2014 at 11:45 AM, David Blaikie <dblaikie at gmail.com> wrote:
> I'm not sure I know enough about the actual code here to have high
> level semantic suggestions, but one nit:
>   std::unique_ptr<ObjectImage> Obj(std::move(InputObject));
> seems unnecessary - you could just use the parameter directly (rename
> it to "Obj" for example)
> At some point I guess createObjectImageFromFile will return a
> unique_ptr? (so you don't have to call reset with the result in
> RuntimeDyld::loadObject(ObjectFile*))
> Curiosity:
> What does Dld->loadOObject(std::move(InputImage)) do? It's odd that it
> takes an ObjectImage and returns one - does it return the same object?
> Why not just have it take the thing by reference rather than taking
> and returning ownership?
> On Wed, Mar 5, 2014 at 11:20 AM, Lang Hames <lhames at gmail.com> wrote:
>> Hi All,
>> I would like to ensure that the target architecture for a
>> RuntimeDyldImpl instance is known at construction time. Once this is
>> possible, target specific logic (relocations/stub management) can be
>> pushed down into the target specific subclasses, rather than living in
>> large switch statements. This will improve both performance and
>> readability.
>> To support this work, this patch changes the way ObjectImages are
>> constructed from ObjectFiles and ObjectBuffers: it replaces virtual
>> methods on RuntimeDyldImpl with static methods on the relevant
>> subclasses (RuntimeDyldELF, RuntimeDyldMachO). All the information
>> needed to call the correct subclass method is already available at the
>> call sites, so we don't lose any flexibility by switching to static
>> methods here.
>> Cheers,
>> Lang.
>> _______________________________________________
>> 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