[llvm-commits] [PATCH] Execution domain support for VMOV and VLDR

[Jakob Stoklund Olesen]

On Aug 29, 2012, at 2:11 AM, Tim Northover <t.p.northover at gmail.com> wrote:

>> Good point and good idea. I'll get a new set of patches ready
>> tomorrow, making the following changes unless there are more
>> developments overnight:
> 
> Ok. I've attached three patches as planned: the first does all of the
> refactoring (including S -> D reg conversion) and brings existing
> instructions in line with what we've discussed. The second two each
> add a new instruction to the framework: VMOVS and VLDRS.
> 
> Ok to commit?

> <0001-Refactor-setExecutionDomain-to-be-clearer-about-what.patch>

Looks good!

> <0002-Add-support-for-moving-pure-S-register-to-NEON-pipel.patch>

This one is odd, I think. Is it really a good idea to replace a vmovs with two shuffles? What is a vmovs doing in NEON 
code anyway? Is isel emitting vmovs for some shuffles?

This seems like it would be better done in isel.

> <0003-Add-domain-setting-for-VLDR-instructions-on-ARM.patch>

You can get the alignment from the instruction's memory operands, if present.

/jakob