Mark Reed

We should perhaps use logging.exception or traceback. The error message above just prints the X0 filename - likely due to running out of disk space? Looking at the code its...

We're missing tests and don't handle locale. Some thoughts on test **Stdin** Redirection - Note that --files0-from needs to pull a nul delimited list of filenames ``` find . -name...

For counting characters we can locale.getencoding() in python then a naive approach would be len(bytes.decode('utf-8')) which would not be performant. Ultimately we'd want to be able to scan for unicode...

UTF-8 looks like this - you can count bits for the character size once you see the left most bit set. Languages like Chinese will be all unicode characters. I...

No, haven't looked at split yet. I'm guessing for 139 he ran out of space. We can update the split code to print a better error message.

Working on this

The complex dot product exists for neon, but we're converting to f32 and want to operate on the bf16 inputs. The complex vector is real, imag, real, imag,,, Original //ab_real...

10% faster. There are not bf16 versions of the neg and rev32 so we still have to jump through hoops. I confirmed that the new function's output matches the old...

Good catch I was playing around with another compiler on there. Ubuntu 24.04's clang 18.1 sees the same bug when building this code. Issue opened: https://github.com/llvm/llvm-project/issues/107810 I'll try to move...

Clang is choking on the flipping of the sign bit. I haven't come up with an alternative to these two. No amount of moving code around fixes the clang bug...