On 10/19/2016 06:18 PM, Richard Henderson wrote: > On 10/18/2016 07:31 PM, Marek Vasut wrote: >>> Processing a little more data can be preferable to fewer branch >>> prediction failures. And the best way to avoid those is to not have the >>> branch at all. Especially when it's unlikely that the data will be >>> created in the first place. >> >> OK, but I need to make sure stores to ZERO register are ignored, right ? > > Yes. > > Just remember that since nios2 doesn't have condition codes, where side > effects of an instruction can be useful alone, the *only* likely case of > ZERO as a destination register is in the (designated) NOP psuedo-op.
Got it, thanks. -- Best regards, Marek Vasut