02:05 <swetland> so it's doing that because your linker script defines the magic symbol __global_pointer$

02:06 <swetland> which causes the linker to generate references to objects in the .sdata/.sbss/etc sections relative to gp

02:07 <swetland> hm actually it may just be that the gp is close to that mmio address so that may be a bit of a red herring

02:08 <swetland> it looks like all the code is due to you using uint32_t* to access an unaligned address

02:08 <swetland> the compiler is generating individual byte loads and stores for each byte in the word

03:32 <Hammdist> yeah that was it .. I asked on #gcc earlier and got unblocked with regards to that

18:02 <jevinskie[m]> _florent_: not sure if you saw this RFC but I thought you might be interested or have some input if you’re still examining amaranth https://github.com/amaranth-lang/amaranth/issues/693

18:02 <jevinskie[m]> And this sub RFC https://github.com/amaranth-lang/amaranth/issues/699

20:14 <_florent_> jevinskie[m]: thanks, Record could indeed be improved/simplified, the RFC seems more natural than Migen's Record and closer to what I would also probably do. I'm still experimenting with an alternative HDL and will evaluate it by writing new core with it (but have limited time since lots of things to do on LiteX and not really related to the HDL).