KloudKoder

To address your specific comment about broadening the scope, I certainly don't oppose that. I just don't personally know of any practical uses for RM tweaking apart from interval arithmetic....

TL;DR: Some algorithms are polynomially slower without interval arithmetic; others just simply don't work. Alright, I promised to list some practical examples. On the one hand, you're obviously not going...

I have a few progress updates on this. Basically I've been trying to contact the team leaders or authors of various interval applications across a variety of languages. Based on...

@Chris00 "That implies that you will have to collaborate with the developers of Rust so we can make use of the rounding primitives." It's possible to port Inari to WASM...

Anyone still not satisfied with the use cases on the table? @conrad-watt Can we get something off the ground here so that the Inari team can prototype their library, probably...

@tlively Sorry for the slow reply. If you have any link to a previous phase 1 proposal that you think is well written, please share it. I'll try to make...

@Chris00 @tlively (and all concerned). Working on this. I'm trying to get the required instructions down to as little complexity as possible. The 4 standard RMs are a must, but...

@sunfishcode You raised some important considerations. First of all, at least in AMD64, you can't just do comparisons on floats. There needs to be a signal propagation to the RFLAGS...

First of all, I guess you've really found a way to avoid having "Unify Zeroes". Kudos! Secondly, what about your f64.copysign(1.0, x)? According to the [list of instructions](https://webassembly.github.io/spec/core/appendix/index-instructions.html), wouldn't that...

@tlively @titzer Thanks. Definitely copysign isn't going to fit the bill, unfortunately. It's necessary to branch based on the state of signs prior to interval multiplication, for instance. Because this...