01:15 <davidc__> Has anyone looked at producing "fixed" CSR layouts? (Or adding a feature where you can pin certain CSR addresses?)

01:15 <davidc__> I have an application where it'd be really handy to fix a couple CSR locations (such that they are always at a particular memory address)

08:10 <keesj> when is there a next open hardware design type conference in the EU?

08:13 <daveshah> keesj: not much info yet, but there is OSDA next March

08:13 <daveshah> https://www.date-conference.com/workshop/w05

08:13 <tpb> Title: W05 2nd Workshop Open-Soure Design Automation (OSDA 2020) | DATE 2020 (at www.date-conference.com)

08:14 <daveshah> (this was last years program https://osda.gitlab.io/)

08:14 <tpb> Title: Workshop on Open Source Design Automation (OSDA) -- in conjunction with the Design, Automation and Test in Europe Conference (DATE) (at osda.gitlab.io)

10:58 <_florent_> scanakci: i added the UART_POLLING mode for cases similar to yours: do some first BIOS tests with un-validated interrupts

10:58 <_florent_> scanakci: you should be able to enable it in your design by doing self.add_constant("UART_POLLING", None)

10:59 <_florent_> scanakci: i could look if that's not working/compiling

11:01 <_florent_> davidc__: yes i make heavy use of fixed CSR layout on systems that are deployed and where we want some peripherals to always be at the same location (for example FLASH module if a PCIe board to be sure we can update with any version of the software)

11:05 <_florent_> davidc__: you just need to do something like this: https://hastebin.com/nuwojaciqe.rb

11:05 <_florent_> davidc__: you can mix fixed and dynamic CSR, it's just that the fixed one are reserved first then the dynamic one use the empty locations

16:38 <davidc__> _florent_: yeah - I do that, but that presumes the peripheral itself stays consistent internally?

16:41 <davidc__> (I'm not 100% clear on the rules for how CSRs get ordered in a peripheral; so I'm gonna dig into that today)

17:23 <davidc__> (Nevermind; just allocation order since sorted by DUID. Thats easy enough!)

17:29 <scanakci> _florent_: thank you. I tried it for Rocket and looks like it works.

20:59 <davidc__> _florent_: hey, any reason that CommUDP uses a separate 'rx_socket' that binds to a specific port? (rather than looking for responses on the socket it uses for tx?)