[Libre-soc-bugs] [Bug 942] next things to work on -- bigint rsa mul algorithm for bigint presentation

bugzilla-daemon at libre-soc.org bugzilla-daemon at libre-soc.org
Tue Oct 4 20:03:47 BST 2022


--- Comment #3 from Jacob Lifshay <programmerjake at gmail.com> ---
(In reply to Jacob Lifshay from comment #2)
> (In reply to Luke Kenneth Casson Leighton from comment #1)
> > (In reply to Jacob Lifshay from comment #0)
> > 
> > > It would be run on a simplistic model cpu with 1 256-bit simd add unit and 1
> > > 256-bit wide-mul unit (the model would be adjustable) and scalar units, I
> > > would create a simulator for that, the simulator would cover ALU pipeline
> > > timing and dependencies but not much else.
> > 
> > everything but this is a great idea.  there's no point spending time
> > on doing hardware-cycle-accurate pipelining when cavatools is specifically
> > designed to do exactly that, and already has the infrastructure in place.
> > any time spent on developing any hardware-cycle-accurate simulations is
> > therefore both time and money completely wasted.
> the point is to have pretty animations of how data moves in the cpu so
> people understand how the wide-mul merging works (which requires something
> to construct those animations), not to have specific timing for exact
> performance comparisons.
> the sw to construct those animations is the simplistic SIMD ALU simulator --
> imho spending a week writing that is waay better than creating animations by
> hand that only cover 1 specific run of 1 specific algorithm and is likely to
> have mistakes because it was done by hand.

it is not a waste of time or money because it has a completely different
purpose and deadline than cavatools -- it generates animations rather than
tries to quickly simulate a whole program, it is simplistic and ignores cache
latency, cache hits, branch mispredictions, scalar op latency, etc. cavatools
is much more detailed and complete and won't be ready any time soon and won't
produce blender animations so is unsuitable for a demo for the bigint

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

More information about the libre-soc-bugs mailing list