11:06 <_florent_> somlo: good news, we are now able to use your litex-hub/linux/litex-rebase branch with Linux-on-LiteX-VexRiscv, see: https://github.com/litex-hub/linux-on-litex-vexriscv/issues/176

11:57 <somlo> _florent_: nice!

13:41 <somlo> _florent_: trying to understand what appears to be a "quirk" of the antmicro litesdcard linux driver: https://github.com/litex-hub/linux/blob/litex-rebase/drivers/mmc/host/litex_mmc.c#L246

13:41 <somlo> here's my theory, please confirm or set me straight: by default, the microSD card assumes/expects bus width 1

13:42 <somlo> but the litesdcard driver doesn't actually support that default option, and only works with bus width 4, at all times

13:42 <somlo> (which might be a deviation from the "standard", but whatever :)

13:43 <somlo> so, instead of waiting for the linux mmc subsystem to do a bunch of data transfers using default width 1, *then* setting width to 4 at some convenient later point, the litesdcard linux driver MUST inject an ACMD6 to set bus width before the very first data transfer, or else things won't work

13:44 <somlo> if that's the case, I'll update the comments around that work to make it clear what's going on :)

13:44 <somlo> but if I'm wrong about my assumptions above, please kick me :)

16:36 <cr1901_modern> _florent_: Since lxterm doesn't work any more on Mac OS or Windows, is there any possibility we could revert to the old single-process version or provide the single-process version as a workaround for now? https://github.com/enjoy-digital/litex/issues/743#issuecomment-755902333

16:37 <cr1901_modern> If not, I'm in the process of adding SFL support to an alternate serial terminal impl

17:24 <st-gourichon-fid> Hi! I reinstalled tools and observed allocation failures on ICE40. Very few ICESTORM_RAM used, but extremely high usage of ICESTORM_LC, up to 1383%. Reverting yosys from "Yosys 0.9+3830 (git sha1 b72c2946, clang 10.0.0-4ubuntu1 -fPIC -Os)" to "Yosys 0.9+2406 (git sha1 aafaeb66, clang 10.0.0-4ubuntu1 -fPIC -Os)" fixed the issue.

17:24 <st-gourichon-fid> Open to any advice.

17:33 <zyp> cr1901_modern, I'm also curious what the reasoning for switching from threading to multiprocessing is, looks like just a random change tacked onto another feature: https://github.com/enjoy-digital/litex/commit/feeb2f7

17:37 <zyp> I don't see multiprocessing giving an obvious benefit to anything in this situation, but the disadvantage is obvious

17:38 <zyp> I should probably make that comment on the issue rather than on irc :)

17:56 <_florent_> cr1901_modern, zyp: thanks for the feedback, I'll look at that. In fact we should probably switch to asyncio

18:16 <zyp> true, asyncio would probably be a better approach, but more work :)

18:30 <_florent_> cr1901_modern: I just reverted LiteXTerm to use threading with https://github.com/enjoy-digital/litex/commit/460fada3acbc6626f15388e4259bec0058b4a2d6

19:20 <cr1901_modern> _florent_: Tyvm. A few lxterm changes incoming on my end

20:09 <cr1901_modern> Is the donut demo known to crash or otherwise produce some characters that aren't actually part of the donut?

20:22 <cr1901_modern> Hmm, I guess the donut demo's speed sensitive