[PATCH] Add Cortex-A9 scheduling classes for vldm/vstm instructions that access more than 32 bytes

[Arnold Schwaighofer]

Except for:

   // For unknown LDMs, define the maximum number of writes, but only
   // make the first two consume resources.
-  SchedVar<A9LMUnknownPred, [A9WriteL1, A9WriteL1Hi,
-                             A9WriteL2, A9WriteL2Hi,
-                             A9WriteL3Hi, A9WriteL3Hi,
-                             A9WriteL4Hi, A9WriteL4Hi,
-                             A9WriteL5Hi, A9WriteL5Hi,
-                             A9WriteL6Hi, A9WriteL6Hi,
-                             A9WriteL7Hi, A9WriteL7Hi,
-                             A9WriteL8Hi, A9WriteL8Hi]>]> {
+  SchedVar<A9LMUnknownPred, A9WriteLMOpsList.Writes[0-15]>]> {

(This is not the same the former will require 2 resources of the LSUnit the later 16 resources of the LSUnit?)


the patch LGTM. When I wrote the Swift model, I noticed the same mistake but did not get around to fixing this.

Thanks,
Arnold

On Sep 2, 2013, at 10:56 AM, Renato Golin <renato.golin at linaro.org> wrote:

> Thanks Tim, no worries, I'm also not the best person to review that part.
> 
> Evan, can you have a look on this patch, or point us who's the best person to do so, please?
> 
> thanks,
> --renato
> 
> 
> On 2 September 2013 16:45, Tim Northover <tnorthover at apple.com> wrote:
> 
> On 2 Sep 2013, at 15:48, Renato Golin <renato.golin at linaro.org> wrote:
> > Patch looks good to me, do you mind having a look just to make sure there's nothing obviously wrong? I'm not an expert in the table-gen descriptions, so would be good to have a different set of eyes on this patch.
> 
> I'm afraid I decided I didn't know nearly enough about how the TableGen scheduling descriptions work to make a sensible comment.
> 
> Cheers.
> 
> Tim.
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits