RE: [PATCH v2] Hexagon (target/hexagon) implement mutability mask for GPRs

[Marco Liebel]

> -----Original Message-----
> From: Taylor Simpson <tsimpson@quicinc.com>
> Sent: Mittwoch, 21. Dezember 2022 21:06
> To: Marco Liebel <mliebel@qti.qualcomm.com>; Marco Liebel (QUIC)
> <quic_mliebel@quicinc.com>; qemu-devel@nongnu.org
> Cc: Brian Cain <bcain@quicinc.com>
> Subject: RE: [PATCH v2] Hexagon (target/hexagon) implement mutability
> mask for GPRs
> 
> 
> 
> > -----Original Message-----
> > From: Marco Liebel <mliebel@qti.qualcomm.com>
> > Sent: Wednesday, December 21, 2022 1:34 PM
> > To: Taylor Simpson <tsimpson@quicinc.com>; Marco Liebel (QUIC)
> > <quic_mliebel@quicinc.com>; qemu-devel@nongnu.org
> > Cc: Brian Cain <bcain@quicinc.com>
> > Subject: RE: [PATCH v2] Hexagon (target/hexagon) implement mutability
> > mask for GPRs
> >
> > > > +#define WRITE_REG_IN_PACKET(reg_name, output, input) \
> > > > +    asm volatile("{ " reg_name " = %1 }\n\t" \
> > >
> > > This is no different from the WRITE_REG above.  Instructions on a
> > > line with no curly braces are a single packet.
> > >
> >
> > Understood. The feedback on Brian's patch said to write tests that do
> > transfers in a packet. Should I write some? (Just not in the way I did
> > it above)
> 
> Put some more instructions in the packet with the assignment.  I recommend
> a read from the same register and verify you get the old value.
> 

Reading and writing a control register in a single packet isn't possible,
because CR instructions can only be executed in slot 3. I was thinking to
put a nop inside the packet, just so the packet gets generated. Or is there
something else that's useful, other than reading the previous value?