add Lattice Nexus vendor platform (using nextpnr-nexus and prjoxide)

[slagernate]

sorry, github newb. Won't do that again. Tested again with blinking all LEDs. Output:

python3 -m amaranth_boards.lifcl_evn
init..
IDCODE: 0x110f1043 (LIFCL-40)
NX Status Register: 0x0000150100400100
reset..
NX Status Register: 0x0000150100400e50
programming..
NX Status Register: 0x0000150100400100
Bye.

Notes: --Removed mention of ECP5 in lattice_nexus.py --Removed OSC

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 89.44%. Comparing base (b36e7e0) to head (51f5769). Report is 9 commits behind head on main.

:exclamation: Current head 51f5769 differs from pull request most recent head 8ccc083. Consider uploading reports for the commit 8ccc083 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #759      +/-   ##
==========================================
- Coverage   89.94%   89.44%   -0.50%     
==========================================
  Files          44       43       -1     
  Lines       10647     9066    -1581     
  Branches     2595     2150     -445     
==========================================
- Hits         9576     8109    -1467     
+ Misses        882      781     -101     
+ Partials      189      176      -13     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[richardeoin]

FWIW I've been using amaranth with radiant and a nexus LFD2NX-40 part, and intend to contribute a platform definition when I'm happy with it (likely in a few weeks). I'll check back to see what the status of this is then.

[whitequark]

I'm currently waiting on a LIFCL-EVN board so that I can do regression testing for this platform.

[whitequark]

Good news: I've received the devboard I can test this on! I'll merge it soon.

[slagernate]

@richardeoin how's your platform file looking?

[richardeoin]

@slagernate Here's the platform file I'm using with LFD2NX-40 + Radiant. I tweaked my formatting to make it somewhat more diff-able against yours.

Some differences

• Adds a two-stage FF sync into i_GSR_N of GSR (is this necessary?)
• Adds an OSCA if self.default_clk == "OSCA"
• Specifies p_GSR="DISABLED" on IDDRX1 and IDDRX2 instances 😱 Probably don't do this
• Less tortured templating syntax for generating the .pdc constraints

I initially tested without triggering a GSR after startup and observed some very unexpected behaviour from IDDR ODDR and BB primitives. After that I assume that triggering a GSR after startup is mandatory for correct operation on these parts. Skip my workaround of putting p_GSR="DISABLED" on IDDRX instances.

I've only tested with radiant and didn't use radiant's programmer tool at all.

[slagernate]

See latest suggested commit where I've pulled in your platform file @richardeoin . This works for the Radiant and Oxide / NextPnr tool chain with the LIFCL-40-EVN (Crosslink-NX Eval Board).

[whitequark]

The current plan is to go through the backlog of PRs after https://github.com/amaranth-lang/rfcs/pull/2 gets merged.

[richardeoin]

There have been subsequent improvements to amaranth/vendor and this PR needs at a minimum to be updated to include those.

@slagernate are you willing to maintain this PR (no is an ok answer!! :) )? It would be great to see it merged, and if you can't I'd happy to do the work (and have my own LFD2NX-40 test hardware).

[slagernate]

@richardeoin more than willing to update the PR, just need to find some spare time (perhaps in the next week). Have been using Amaranth + Radiant quite a bit (the PCLKDIV primitive in Oxide is missing :| ). But also totally happy if you'd like to take the reins!

[whitequark]

This PR has languished for a bit too long--I'll add it to the 0.5 milestone to make sure it's merged before then.

[slagernate]

Should be compatible with RFC#18 now.

[slagernate]

Updated to use tcl_quote instead of tcl_escape.

[whitequark]

Superseded by #1342, since recent internal refactoring made this PR unmergeable. However, #1342 is based on this work--thank you!