Conversation

New blog post: "Bit-Twiddling: Optimising AArch64 Logical Immediate Encoding (and Decoding)"

dougallj.wordpress.com

Bit-Twiddling: Optimising AArch64 Logical Immediate Encoding (and Decoding)

I came up with a (seemingly) new method to encode bitmask immediate values on ARM64. This really isn’t worth optimising – clarity and verifiability are more important – but itR…

Zach Wegner

@zwegner

Replying to

@dougallj

Nice! For the decoding lookup table equation, you could use 128b unsigned division with (2^128-1) / (2^2^(6-n) + 1) + 1, but that's a tad ridiculous...

8:20 PM · Oct 30, 2021Twitter Web App

Replying to

This is great! I added a note about it to the post. Using division for repeating patterns like this is a brilliant idea, and I definitely wouldn't have thought of it myself.

Replying to

Thanks! Can't remember if I'd seen it for this specific case, but knowing that pattern in divisions it wasn't too hard to derive. The +1 was on purpose though, or at least I thought so--I worked the formula out in Python, and tried it in C to make sure it would work...

Show replies