[aarch64-port-dev ] RFR: Add support for A53 multiply accumulate (take 2)

[Andrew Haley]

On 12/02/2014 03:19 PM, Edward Nevill wrote:
> Hi,
> 
> The following patches overrides madd, msub, maddw, msubw, smaddl, smsubl, umaddl and umsubl and adds a nop if Ra != zr (ie it is doing a genuine mul accumulate, not just a mul).
> 
> This can be bypassed if it is known that the nop is not necessary. I have done this in three cases, 32 bit and 64 bit mod functions and some array indexing in InterpreterMacroAssembler::profile_switch_case which are the only uses of these instructions outside of C2.
> 
> OK?

Yes, thanks.

Andrew.