[Libre-soc-bugs] [Bug 324] create POWER DIV pipeline

bugzilla-daemon at libre-soc.org bugzilla-daemon at libre-soc.org
Fri Jun 19 21:33:14 BST 2020


--- Comment #31 from Luke Kenneth Casson Leighton <lkcl at lkcl.net> ---
(In reply to Jacob Lifshay from comment #29)

> That trunc_div function isn't intended to be Power ISA compatible 

ah then that will be the source of confusion: i was expecting it to
provide the functionality of... of... something-that-i-don't-have-a-name-for...

> (or
> POWER9, which is not even what we're implementing -- we're implementing
> Libre-SOC's first core -- not POWER9, they are different cpus -- please
> avoid calling the Power ISA "POWER9", they are different things. That's like
> if AMD called their hypothetical new processor core "Sandy Bridge" or
> "Pentium", which are widely associated with Intel).

urrrr... :)  ok then we need to know the exact name.  POWER ISA V3.0B is a
bit of a handful.  POWER 3.0B?  we've been asked by OpenPOWER Foundation to
be compatible with that.  short-hand i was referring to it as POWER9.

> trunc_div is intended to be the equivalent of python's // operator, except
> that it rounds differently -- the same way Rust, C, C++, Java, x86, and
> Power do.
> Similarly with trunc_rem.

ok right: then it's... (POWER 3.0B?) compatible, would that be ok to call it?

if so, it would be part of the (using the appropriate name, now?)
Power30BCompatibleSelectableInt class.

however, right now, we're going for a collection of *functions* that are
Power 3.0B compatible, and those functions *use* SelectableInt, whose behavior
is expected to be directly equal to python int class.

when those functions require state (which they don't right now), it becomes
not just appropriate but necessary to move them to a class.

also if we move _those_ functions to a class, we have to also add (move)
helpers.ne, eq, lt, gt, ge and then *also* add sub, add, mul, everything
and then do a bit of a rewrite on parser.py to put it into use.

and that's a lot of work which is quite a major distraction.

i'm quite happy for that work to be done *if* you can see a good reason as to
why not having it is a major blocker to the October 2020 tapeout?

You are receiving this mail because:
You are on the CC list for the bug.

More information about the libre-soc-bugs mailing list