[Libre-soc-dev] mfocr and mtocrf v3.0B spec ambiguity

[Luke Kenneth Casson Leighton]

On Fri, Aug 28, 2020 at 9:43 AM Luke Kenneth Casson Leighton
<lkcl at lkcl.net> wrote:
> On Friday, August 28, 2020, Paul Mackerras <paulus at ozlabs.org> wrote:
>> Well, only if compilers or assembly language programmers use mfocrf
>> with more than one bit set in FXM, which would be a perverse thing to
>> do...
>
> :)
>
> they aren't likely to, i would agree, if they don't know it's possible.  still, clarifying the spec here to take into account bit 20 is.. sensible.

ah.

i did a first cut implementation in nmigen HDL, then was puzzled as to
why the unit test for the simulation failed.  of course, the simulator
uses the v3.0B pseudocode which, of course, i hadn't yet updated.

duh.

# Move To One Condition Register Field

XFX-Form

* mtocrf FXM,RS

Pseudo-code:

    count <- 0
    do i = 0 to 7
      if FXM[i] = 1 then
        n <- i
        count <- count + 1
    if count = 1 then
        CR[4*n+32:4*n+35] <- (RS)[4*n+32:4*n+35]
    else CR <- undefined

wait... ah :)  i don't mean bit 20, do i - i mean (IBM MSB0 numbering)
bit 11.  apologies for the coincidence that in microwatt source code,
insn[20] is POWER9 XFX-Form bit 11.  when looking at the XFX-Form i
thought, "oink, bit 20 is reserved".  sorry about that.

# 1.6.9 XFX-FORM
   |0     |6        |11|12             |20|21    |31 |
   | OPCD |  RT     |1 |  FXM          |/ |   XO | / |
   | OPCD |  RS     |0 |  FXM          |/ |   XO | / |
   | OPCD |  RS     |1 |  FXM          |/ |   XO | / |

i will need to alter the pseudocode so that it no longer has
"undefined" behaviour, simply selecting the highest priority (one-hot)
and always placing a value into CR. there should be no need to have a
new field (modify section 1.6.9).

likewise the mfocrf pseudocode would need altering to ignore "if count
= 1" and always select a one-hot.

am i right in thinking that the for-loop does not need altering?  that
it can just be this:

    do i = 0 to 7
      if FXM[i] = 1 then
        n <- i
    CR[4*n+32:4*n+35] <- (RS)[4*n+32:4*n+35]

or does it need to be:

    do i = 7 to 0       # inverted order
      if FXM[i] = 1 then
        n <- i
    CR[4*n+32:4*n+35] <- (RS)[4*n+32:4*n+35]

l.