anzz1

Rebased the branch to master for easier testing.

Even based on your tests alone I would think that merging this is a good idea. All the other platforms use the extensions already and MSVC is also *supposed* to...

@slaren And looking at the current [alternative](https://godbolt.org/z/WrTo5zaTb) paints a pretty clear picture :smile: And as seen in the before part, vex instructions like `vcvtsi2ss` are still used since AVX2 implies...

**Hold merging this until #546 is merged.**

Much appreciated ! :+1: Thanks to everyone taking part to this, with special thanks to @nicknitewolf and @KASR to take the time to do benchmarks. While a 5% speed increase...

For future considerations: In GitHub actions, each row of the `run:` commands are executed separately, irregardless what the previous input returned. The step determines "success" from the error code returned...

Separated build and test steps as separate tasks as pointed out by @Green-Sky . Obviously should've done that in the first place, idk how I didn't think of that. Also...

> > reverse prompt with extra space seems to not work for me atleast, llama.cpp goes on as if there was no reverse prompt in that case > > that...

> actually i think, it is because the space is part of the next token, so there is no tailing space to catch... True, the space after "user:" can be...

> The reverse prompt can span multiple tokens. However, there is no way for it to interrupt generation mid-token. (That's why I opted to use token vectors rather than strings...