[aarch64-port-dev ] jdk8: Get BUILTIN_SIM working again

[Edward Nevill]

On Tue, 2014-08-05 at 16:23 +0100, Andrew Dinn wrote:
> On 05/08/14 15:57, Edward Nevill wrote:
> > Oops,
> > 
> > The encode_iso_array patch broke the BUILTIN_SIM image because it
> > executes uqxtn instructions.
> > 
> > The following patch gets it going again.
> > 
> > Maybe we should create 'UseSIMD' and set it to false for BUILTIN_SIM
> > and use predicate(UseSIMD) in aarch64.ad?
> 
> Why not implement uqxtn in the simulator?

I don't think we should be implementing one off SIMD instructions.

If we are going to do it we should either implement SIMD (or decide on a subset of SIMD that we are going to implement).

There is already a significant quantity of SIMD used by the AES and CRC code (and not just the AES and CRC instructions, but other instructions in the AES and CRC code). This is already conditionalised on BUILTIN_SIM.

Regards,
Ed.