11:17 <karlp> zapb_: just daydreaming about having say, a VM of some sort that we have compiled for all targets, that is uploaded to run "independent" scripts, instead of having to compile things like spi helpers for each architecture.

11:17 <karlp> it doesn't really come up with all flash loaders, as those are tied to the flash target already.

11:18 <karlp> but like #8400 is support for a spi peripheral, but only when it's on one particular arch, even though it's definitely used in other places.

11:21 <Haohmaru> oh, fancy

11:22 <Haohmaru> this was about... where openocd, for many targets, puts a small "firmware" in the RAM of the target to making programming faster, right?

11:22 <karlp> zapb_: but just silly overthinking really, far too much work for the amount of benefit

11:23 <Haohmaru> and those small firmwares need to be written by $someone, then they gotta be compiled, it gets hairy

18:24 <karlp> it already does taht for most flash targets, to get useful speeds. see https://sourceforge.net/p/openocd/code/ci/master/tree/contrib/loaders/flash/stm32/ and the directory above...

18:54 <antto> yeah, i mean, and your idea is to write a small virtual machine instead of this, so that the actual "helper algos" for the flashing can be written just for that VM, right?

18:54 <antto> and possibly without needing a pile of different compilers

18:56 <cambrian_invader> isn't the main difference that the flash registers are different?

18:57 <antto> well, not just that, there would also be chip-specific details, "you have to unlock this sh*t before you can do that sh*t" ... etc..

18:58 <antto> but you'll move that fiddling over to the code that runs on the VM

18:59 <antto> while the "real code" which you'd have to compile for every chip is gonna be just this VM, if i understand correctly

18:59 <antto> it's a fancy idea

19:00 <antto> reminds me of quake

19:09 <cambrian_invader> but like, these helpers are maybe 50 instructions long

19:24 <antto> *shrug*