[Libre-soc-dev] change of tapeout date, and roadmap for 180nm ADIC

[Cole Poirier]

On Wed, Sep 23, 2020 at 11:04 AM Luke Kenneth Casson Leighton
<lkcl at lkcl.net> wrote:
>
> i had a conference call today with staf, dmitri and jean-paul to
> discuss the practical details of the 180nm ASIC.  staf will be putting
> the agenda into the bugtracker later.  it was 3 hours, seriously long.
>
> staf has had a lot more to do on the cell library than expected so we
> are moving to 2 dec 2020.
>
> jean-paul needs at least one month to do nothing but checking, so this
> is 2 nov latest.
>
> we have 2 weeks in which to solve thing like the litex issues and get
> "a first version" (no new features as of today).  this will give jean
> paul a stable base to work from.
>
> this will include checking of longest path and to ensure setup time
> and hold time on all FFs are ok.
>
> in the intervening time the benefits of being able to boot linux i.e.
> an MMU are such a leap that it is compelling to try.  also Staf would
> like to try an SRAM block and dcache and icache is a way to do that.
>
> so for a second iteration we have 5 weeks to try to get MMU in, test
> JTAG, and staf can design a fixed size custom SRAM block.
>
> if we cannot get the 2nd iteration working then we fall back to the 1st.
>
> staf says that with no clock timing constraint checking in coriolis2
> it is unwise to try multiple clocks.  therefore no PLL and only
> digital dividing of an incoming signal, 50mhz is perfectly reasonable.
>
> we also thought it may be worthwhile to do a 2nd 180nm or 130nm ASIC
> middle of next year.  this one could have the PLL, and also there will
> be interns working for Jean Paul on timing and buffers which will
> improve speed and layout.
>
> notes to go on bugtracker later.

!!!

Well, that is quite the change! Working on icache currently so I guess
nothing has changed for me :)

I understand your dislike of branches Luke, however, I do feel like
Jacob that in this case a branch is the appropriate given the critical
nature of the dec-fallback code.

Cole