Re: [PATCH v2] RISC-V: Introduce -mrvv-allow-misalign.

[Palmer Dabbelt <palmer@dabbelt.com>]

On Fri, 24 May 2024 16:31:48 PDT (-0700), jeffreyalaw@gmail.com wrote:
>
>
> On 5/24/24 11:14 AM, Palmer Dabbelt wrote:
>> On Fri, 24 May 2024 09:19:09 PDT (-0700), Robin Dapp wrote:
>>>> We should have something in doc/invoke too, this one is going to be
>>>> tricky for users.  We'll also have to define how this interacts with
>>>> the existing -mstrict-align.
>>>
>>> Addressed the rest in the attached v2 which also fixes tests.
>>> I'm really not sure about -mstrict-align.  I would have hoped that using
>>> -mstrict-align we'd never run into any movmisalign situation but that
>>> might be wishful thinking.  Do we need to specify an
>>> interaction, though?  For now the new options disables movmisalign so
>>> if we hit that despite -mstrict-align we'd still not vectorize it.
>>
>> I think we just need to write it down.  I think there's two ways to
>> encode this: either we treat scalar and vector as independent, or we
>> couple them.  If we treat them independently then we end up with four
>> cases, it's not clear if they're all interesting.  IIUC with this patch
>> we'd be able to encode
> Given the ISA documents them as independent, I think we should follow
> suit and allow them to vary independently.

I'm only reading Zicclsm as saying both scalar and vector misaligned 
accesses are supported, but nothing about the performance.

>> * -mstrict-align: Both scalar and vector misaligned accesses are
>>   unsupported (-mrvv-allow-misalign doesn't matter).  I'm not sure if
>>   there's hardware there, but given we have systems that don't support
>>   scalar misaligned accesses it seems reasonable to assume they'll also
>>   not support vector misaligned accesses.
>> * -mno-strict-align -mno-rvv-allow-misalign: Scalar misaligned are
>>   supported, vector misaligned aren't supported.  This matches our best
>>   theory of how the k230 and k1 behave, so it also seems reasonable to
>>   support.
>> * -mno-strict-align -mrvv-allow-misalign: Both scalar and vector
>>   misaligned accesses are supported.  This seems reasonable to support
>>   as it's how I'd hope big cores end up being designed, though again
>>   there's no hardware.
> I'd almost lean towards -m[no-]scalar-strict-align and
> -m[no-]vector-strict-align and deprecate -mstrict-align (aliasing it to
> the scalar alignment option).  But I'll go with consensus here.

Seems reasonable to me.  Just having a regular naming scheme for the 
scalar/vector makes it clear what we're doing, and it's not like having 
the extra name for -mscalar-strict-align really costs anything.

>> The fourth case is kind of wacky: scalar misaligned is unsupported,
>> vector misaligned is supported.  I'm not really sure why we'd end up
>> with a system like that, but HW vendors do wacky things so it's kind of
>> hard to predict.
> I've worked on one of these :-)  The thinking from the designers was
> unaligned scalar access just wasn't that important, particularly with
> mem* and str* using the vector rather than scalar ops.

OK then ;)