[llvm-dev] Rotates, once again

[Krzysztof Parzyszek via llvm-dev]

On 7/2/2018 11:27 AM, Sanjay Patel via llvm-dev wrote:
> 
> Let's settle on the intrinsic definition(s).
> 
> 1. There was a suggestion to generalize rotate to a "valign" or "double 
> shift" (that's what x86 means with its poorly worded "double precision 
> shift"). How does that work with vector types? The options are a full 
> vector-size shift or a per-element shift. If it's the full vector, do we 
> still want/need a specialized rotate intrinsic for per-element? If it's 
> per-element, do we still want/need the other form for a full vector?

The scalar rotate moves bits and such an operation doesn't make much 
sense for moving data across lanes in vectors. I voted for the valign 
variant originally, but I think that a per-element rotate would be the 
natural vector version of the scalar operation.

It could still be doing the "double shift", since it's more general, it 
just shouldn't be called valign. A per-byte cross-lane vector rotate (an 
actual vector valign) can be implemented using shuffle, so I don't think 
that an intrinsic for that is necessary.


-Krzysztof

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
hosted by The Linux Foundation