[Libre-soc-isa] [Bug 560] big-endian little-endian SV regfile layout idea

bugzilla-daemon at libre-soc.org bugzilla-daemon at libre-soc.org
Tue Jan 5 23:57:35 GMT 2021


https://bugs.libre-soc.org/show_bug.cgi?id=560

--- Comment #92 from Jacob Lifshay <programmerjake at gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #91)
> rright.  ok.  i took a look at the byteswap diagram: the PartitionedMul not
> doing arbitrary permutations throws a royal spanner in the works, and we
> can't spend 2 months fixing that.  so we go with the flow.
> 
> with arbitrary permutations out, the "aligned" byteswapping is ok.  the gate
> count of 4-in MUXes looks to be around 8 (possibly less) which times 64 is
> around 512 gates.
> 
> multiplied up to 4 operands (src123 and dest) and 50 FUs this is 100,000
> gates.

Are you calculating based on 4 muxes per FU or 4 muxes per pipeline? I've been
saying we should have the muxes be per-pipeline rather than per-FU since that
will reduce the gate count by a factor of 5 or so, getting the gate count to
the 10-20k per-cpu mentioned earlier

I'll make another illustration.

> this is just about borderline tolerable.
> 
> i am therefore de-marking this as invalid and instead putting it as
> deferred.  if we have time (or if it proves to be so heavy a penalty
> elsewhere not to have it), it goes in.

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


More information about the Libre-SOC-ISA mailing list