VariantXYZ

Just to piggy-back on this issue, in an older version of this CFW (I think 1.3.3), my PocketGo was able to suspend and wake at any moment by pressing the...

Hi @Rikorose , I went ahead and tried out your branch out of curiosity on my a55 (on a rockchip sbc which is the only fp16 compatible CPU I have)...

Some additional information (maybe relevant to tract): The inner_loop generic in core/src/ops/cnn/conv/depth_wise.rs is doing a back and forth conversion like every single line. Some other ops may be like this...

> I think that this is now OK with what's currently pushed @iains , I actually ran into this today trying to build off this commit: ``` commit 98b6b0b6e18ab7d6efa3cabfa3356a8fc8ab9719 (HEAD...

> what is this --oldincludedir=$(xcrun --show-sdk-path)/usr/include intended to achieve? It seemed to be necessary in an old gcc config when I was cross-compiling so I've been including it by default,...

Understood. I just saw your comment and wanted to note it was still occurring. (In my case, I was just experimenting with the 12 branch, but since the 13 one...

> Do you already have some kind of list of the ops that should be ported to f16? DeepFIlterNet2 (a very convenient example) at least is primarily bottlenecked by this...

@kali , I tried looking into doing some of the inner loop in assembly but had a few questions: ``` for (ix, v) in iter { let k = *kptr.add(ix);...

Got it, so it looks like writing a quick dot product for fp16/fp32 in C is fine for now. My next question is hopefully a more straightforward one: Are there...

Will try the first option. To clarify though, offset_pairs would be a 2D size_t array right? And you'd need to pass input_center_offset, right?