Re: [PATCH v4 13/16] target/riscv: compressed encodings for sspush and sspopchk

[Richard Henderson]

On 8/16/24 16:56, Deepak Gupta wrote:
> > > +  c_sspush        011 0  00001  00000 01 rs2=1 rs1=0 # c.sspush x1 carving out 
> > > of zcmops
> > > +  c_sspopchk      011 0  00101  00000 01 rs1=5 rd=0 # c.sspopchk x5 carving 
> > > out of zcmops
...
> > This indirection is pointless.  Have the decoder invoke the proper insn in the first 
> > place.  Identically with how we're treating 'addi', for instance.
>
> I was getting compilation error. How to reconcile with arugment sets between
> insn32.decode and insn16.decode. Earlier I was doing that and didn't need it.
> But after removing indirection in arguments and using single use format, type 
> for
> structs instruction arguments ends up conflicting and compiler complains.

Ah, for that you need to use named argument sets.  In this case,

sspush     .... &r2_s rs2=1 rs1=0
sspopchk   .... &r2   rs2=5 rd=0

in both insn32.decode and insn16.decode


r~