07:51 <_florent_> mithro: thanks, good idea, i'll add this (that's already in Linux-on-LiteX-VexRiscv README but could emphasize things here yes)

07:56 <_florent_> loxodes: i could have a closer look tomorrow if you want

08:26 <loxodes> _florent_: thanks for the offer, first let me see if I can come up with a minimal project that replicates the issue

08:36 <loxodes> I see there is now a wishbone dma writer core, I'll try that

20:52 <loxodes> _florent_: If you think this is worth investigating, I've replicated the problem on a fresh project which uses the wishbone dma writer core here: https://github.com/loxodes/dma_burst

20:52 <tpb> Title: GitHub - loxodes/dma_burst: Project to illustrate issue with writing to memory on LiteX (at github.com)

21:22 <zyp> loxodes, that looks like a fun issue to debug

21:27 <zyp> loxodes, there seems to be a pattern to it, notice how it only succeeds for 0b000xxx, 0b001xxx, 0b011xxx and 0b111xxx

21:28 <zyp> I bet if you extend it from 64 to 256, it'll also fail for 64-119 and 128-247

22:30 <mithro> _florent_: I would like to figure out a way that you can "automagically" also use any FuseSoC module in LiteX

22:45 <loxodes> zyp: good point, I'll try and dig into patterns on where it fails. extened to 256, it tends to fail on either 16-23, 32-55, 96-255 or 16-31, 40-55, 96-255