llvm dev - Aug 2012 - [LLVMdev] PROPOSAL: IR representation of detailed struct assignment information

On Aug 27, 2012, at 11:37 AM, gohman at apple.com wrote:
> On Aug 24, 2012, at 5:56 PM, Alex Rosenberg <alexr at leftfield.org>
wrote:
> 
> > If we can also describe the alignment padding inserted at the end of a
struct when it is placed in an array, then we can improve the current LoopIdiom
pass to build more memcpys. I would think that would be attached to the struct
definition.
> 
> Perhaps, though that would be unrelated in implementation and use from
> what I'm working on.
You're describing structure padding and doing so as metadata on memcpy
operations. I think it would be better to describe it in the struct definition
itself, which is where the end of struct alignment padding I mentioned would
also want to be described. Is padding really metadata? One might argue that the
end padding is, but between proper elements, it seems quite first-class to me.

+------------------------------------------------------------+
| Alexander M. Rosenberg        <mailto:alexr at leftfield.org> |
| Nobody cares what I say, so no disclaimer appears here.    |

On Aug 27, 2012, at 11:51 AM, Alex Rosenberg <alexr at leftfield.org>
wrote:
> On Aug 27, 2012, at 11:37 AM, gohman at apple.com wrote:
> 
>> On Aug 24, 2012, at 5:56 PM, Alex Rosenberg <alexr at
leftfield.org> wrote:
>> 
>>> If we can also describe the alignment padding inserted at the end
of a struct when it is placed in an array, then we can improve the current
LoopIdiom pass to build more memcpys. I would think that would be attached to
the struct definition.
>> 
>> Perhaps, though that would be unrelated in implementation and use from
>> what I'm working on.
> 
> You're describing structure padding and doing so as metadata on memcpy
operations. I think it would be better to describe it in the struct definition
itself, which is where the end of struct alignment padding I mentioned would
also want to be described. Is padding really metadata? One might argue that the
end padding is, but between proper elements, it seems quite first-class to me.
I like this approach.  I think it makes more sense on the struct definition.  

It would also be nice to make __attribute__((unused)) count as padding for this.
Vector math libraries in particular have an unused field on a vec3 to pad to a
vec4 and it would be nice to know thats just padding.

Pete
> 
> +------------------------------------------------------------+
> | Alexander M. Rosenberg        <mailto:alexr at leftfield.org> |
> | Nobody cares what I say, so no disclaimer appears here.    |
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev