Re: [RFC PATCH 0/3] kernel: add support for 256-bit IO access

[Linus Torvalds]

On Thu, Mar 22, 2018 at 5:48 AM, David Laight <David.Laight@xxxxxxxxxx> wrote:
>
> So if we needed to do PIO reads using the AVX2 (or better AVX-512)
> registers would make a significant difference.
> Fortunately we can 'dma' most of the data we need to transfer.

I think this is the really fundamental issue.

A device that expects PIO to do some kind of high-performance
transaction is a *broken* device.

It really is that simple. We don't bend over for misdesigned hardware
crap unless it is really common.

> I've traced writes before, they are a lot faster and are limited
> by things in the fpga fabric (they appear back to back).

The write combine buffer really should be much more effective than any
AVX or similar can ever be.

               Linus