#asahi-dev · 2021-11-10 18:49
squags_my intuition is that most of the graph -> model binary stuff wouldn't be particularly useful for like, OpenCL/Tensorflow/whatever else, just need to know opcodes and DMA logistics#asahi-dev · 2021-11-10 18:46
squags_noted#asahi-dev · 2021-11-10 18:46
squags_I'm just curious how granular the instruction set will end up being#asahi-dev · 2021-11-10 18:45
squags_I've seen that geohot look at stuff yeah, I'm taking a look at some of the ANE Mach-O files#asahi-dev · 2021-11-10 18:44
squags_yeah my vague plan atm is to get a minimal ANE example going in MacOS, dump traces, then try getting it to run in m1n1/similar so I can fuzz out specifics#asahi-dev · 2021-11-10 18:37
squags_makes sense#asahi-dev · 2021-11-10 18:14
squags_just to be clear on the RE policy, logs and looking at intermediate/generated files are fair game, but disassembling disqualifies working on any Linux-end drivers?#asahi-dev · 2021-11-10 18:13
squags_might take a stab at some ANE RE bc it at least looks interesting compared to neural accelerators I've worked with before