01:30 <Sprite_tm> Hi all! I'm programming an ECP5 using the slave serial port. Which works great, after figuring out what commands to send. However, the slave port disappears after configuration. That stops me from being able to reconfigure the chip using that afterwards.

01:31 <Sprite_tm> The Lattice tools have an option to keep SSPI active, but it doesn't seem to be as simple as a configuration bit somewhere; activating that option changes a few things in the fusemap.

01:31 <Sprite_tm> So my question is: How would I go about getting the same effect in the open source tools?

01:31 <agg> not completely sure but you might want the --background option to ecppack

01:33 <Sprite_tm> That's worth a try, but I think that is a ctrl0 option only, it doesn't set the fusemap in the same way.

01:34 <agg> it does change the generated bitstream

01:35 <agg> ah, yea, it is a different thing though, and more to do with the MSPI mode I think

01:37 <agg> in your .lpf file put SSLAVE_SPI_PORT=ENABLE

01:37 <agg> uh, one S

01:37 <agg> SYSCONFIG SLAVE_SPI_PORT=ENABLE

01:39 <Sprite_tm> Oh hey, that actually seems to do something!

01:42 <Sprite_tm> Yep, that works. Yay, I can get rid of my PROGRAMN bodge wire!

01:43 <agg> :)

01:43 <agg> i have an ecp5 gpio wired to programn, heh

01:43 <agg> wish it had a WARMBOOT primitive like ice40

01:43 <Sprite_tm> Can I ask how you found that line btw? I found absolutely no references to it.

01:43 <agg> the lattice doc TN02039 (sysCONFIG for ECP5) said "Use the Spreadsheet View to ENABLE the SLAVE_SPI_PORT preference in your design"

01:44 <Sprite_tm> Yeah, did that on a previous project where I used a flash chip to configure the ECP5. It works, but that's it.

01:44 <agg> I grepped for SLAVE_SPI_PORT in nextpnr and saw it using it but not a clear example of how it's used, then I checked prjtrellis and it has an example .lpf file with it set to FALSE

01:44 <agg> or disable or whatever

01:44 <agg> so uh, not sure i'd describe it as documented per se

01:44 <Sprite_tm> Ah, excellent :) I have bene grepping prjtrellis, but I think I only used SSPI as a term.

01:45 <Sprite_tm> Hence me nearly reverse engineering the bitstream to see what's going on :)

01:45 <agg> the lpf for ecp5 is otherwise documented here ttps://github.com/YosysHQ/nextpnr/blob/master/docs/constraints.md#architecture-specific-io-constraints but no mention of this setting

01:45 <agg> ah, close then!

01:46 <agg> if you grep for it in nextpnr you can see exactly what it changes in the bitstream

01:47 <Sprite_tm> Yay, got everything else hooked up as well, me running 'make upload' now pushes a new image to my board over WiFi and triggers a reconfigure.

01:50 <Sprite_tm> Anyway, thanks for the help, I'd probably be chasing red herrings still otherwise :)

01:55 <agg> np, glad it's working!

04:01 <agg> as a followup to ecpdap (program ecp5s over jtag using cmsis-dap probes), i've just released spidap, which abuses jtag to access SPI flash from your cmsis-dap probe, handy for ice40 devices: https://crates.io/crates/spidap