Nouveau - Oct 2025 - [PATCH 5/7] gpu: nova-core: add extra conversion functions and traits

On 10/28/25 7:44 AM, Alexandre Courbot wrote:
> On Tue Oct 28, 2025 at 3:46 AM JST, John Hubbard wrote:
>> On 10/26/25 9:44 AM, Miguel Ojeda wrote:
>>> On Sun, Oct 26, 2025 at 3:40?PM Alexandre Courbot <acourbot at
nvidia.com> wrote:
>> ...
>>
>>> Regarding the `.into_as()` name, it makes sense, but it can be a
bit
>>> surprising when reading out of context... The standalone functions
are
>>> super clear, in comparison. But I am not sure what could be better.
>>> `into_in_this_arch()` or similar could emphasize that this will
only
>>> work in certain architectures, i.e. it is "an `into()` for
this arch"
>>> rather than the general one.
>>> That would go well with the idea that you didn't implement it
for
>>> other obvious types, which I guess was to avoid developers using
this
>>> instead of `into()` by mistake, right?
>>>
>>
>> Exactly: the into-as, from-as naming suffers from *appearing* to be
>> familiar and readable, but actually, the naming gives no hint as to
>> what it is really doing--nor how it is subtly different from the
>> basic from/as/into standard conversions.
>>
>> Instead, we need to add something (almost anything) to the name, to
>> make it clearly different from the from/as/into.
>>
>> into_for_arch() goes in that direction, for example.
> 
> I'd like to get more input on that, for I am not sure how we can stay
> succint in the naming, while carrying the relevant information.
That's too many constraints: if you want an extremely short name
that carries information, *and* avoids (as requested here) confusion
with existing "as" methods, then...you can't.

But you are allowed to be less succinct here, because the more
specialized and rare a case is, the longer you can make the name.
And here, you are definitely allowed a few more characters.

> `into_arch` does not sound much more explanatory than `into_as` - the
> intent with the latter was to say "I would normally have done an `as`,
> but instead here is a method that attests that this operations is indeed
> lossless and safe".
> 
> The best naming scheme I could think of is to have the methods carry the
> source or destination types: e.g. `from_usize` or `into_usize` (like the
> standalone functions), but that would require defining as many traits,
> and increase the number of imports - if we go that way, we might just as
> well drop the traits completely and use the standalone functions.
Accurate names are really desirable; maybe we shouldn't completely
close the door to the above approach.

thanks,
John Hubbard
> 
> `into_native` also comes to mind, but like `arch`, it can mean many
> things depending on the context.
> 
> ... I think I still believe that `into_as` is the clearest name, once
> one has read the documentation for the trait - which one should be
> expected to do anyway. :)

On Wed Oct 29, 2025 at 2:18 AM JST, John Hubbard wrote:
> On 10/28/25 7:44 AM, Alexandre Courbot wrote:
>> On Tue Oct 28, 2025 at 3:46 AM JST, John Hubbard wrote:
>>> On 10/26/25 9:44 AM, Miguel Ojeda wrote:
>>>> On Sun, Oct 26, 2025 at 3:40?PM Alexandre Courbot <acourbot
at nvidia.com> wrote:
>>> ...
>>>
>>>> Regarding the `.into_as()` name, it makes sense, but it can be
a bit
>>>> surprising when reading out of context... The standalone
functions are
>>>> super clear, in comparison. But I am not sure what could be
better.
>>>> `into_in_this_arch()` or similar could emphasize that this will
only
>>>> work in certain architectures, i.e. it is "an `into()` for
this arch"
>>>> rather than the general one.
>>>> That would go well with the idea that you didn't implement
it for
>>>> other obvious types, which I guess was to avoid developers
using this
>>>> instead of `into()` by mistake, right?
>>>>
>>>
>>> Exactly: the into-as, from-as naming suffers from *appearing* to be
>>> familiar and readable, but actually, the naming gives no hint as to
>>> what it is really doing--nor how it is subtly different from the
>>> basic from/as/into standard conversions.
>>>
>>> Instead, we need to add something (almost anything) to the name, to
>>> make it clearly different from the from/as/into.
>>>
>>> into_for_arch() goes in that direction, for example.
>> 
>> I'd like to get more input on that, for I am not sure how we can
stay
>> succint in the naming, while carrying the relevant information.
>
> That's too many constraints: if you want an extremely short name
> that carries information, *and* avoids (as requested here) confusion
> with existing "as" methods, then...you can't.
>
> But you are allowed to be less succinct here, because the more
> specialized and rare a case is, the longer you can make the name.
> And here, you are definitely allowed a few more characters.
>
>
>> `into_arch` does not sound much more explanatory than `into_as` - the
>> intent with the latter was to say "I would normally have done an
`as`,
>> but instead here is a method that attests that this operations is
indeed
>> lossless and safe".
>> 
>> The best naming scheme I could think of is to have the methods carry
the
>> source or destination types: e.g. `from_usize` or `into_usize` (like
the
>> standalone functions), but that would require defining as many traits,
>> and increase the number of imports - if we go that way, we might just
as
>> well drop the traits completely and use the standalone functions.
>
> Accurate names are really desirable; maybe we shouldn't completely
> close the door to the above approach.
I think we have reached the stage where any responsible adult would
shove this whole discussion into a LLM and see what it comes up with.

And the candidate is... `FromSafeCast`/`IntoSafeCast`. Which I have to
say sounds like a good middle ground? :) The intent is definitely to
perform a safe cast here.