First step to supporting LLVM’s opaque pointers done today in Burst - lots and lots of churn left though. Really hope the codegen benefits this is supposed to bring pan out!

@sheredom I'm not sure big benefits will manifest immediately, but I am generally excited to see some de-ossification of core IR, and some paying-down of technical debt. Too make design decisions in LLVM have been frozen for a long time.

@resistor the one place I hope might get better is that with all the bitcasts gone it might be easier to see the real operations behind the pointer. Too many times in the past I’d see the bitcast make the optimization fall apart!

@sheredom I certainly hope it will make existing optimizations less fragile!

Sign in to participate in the conversation
Gamedev Mastodon

Mastodon server focused on game development and related topics.