[Mesa-dev] [PATCH 1/2] gallium: add PIPE_COMPUTE_CAP_SUBGROUP_SIZE

[Giuseppe Bilotta]

On Sat, Jun 6, 2015 at 3:21 PM, Francisco Jerez <currojerez at riseup.net> wrote:
> Giuseppe Bilotta <giuseppe.bilotta at gmail.com> writes:
>
>> What I was trying to say is that while the cap itself is per-device,
>> the OpenCL property that relies on this cap isn't.
>> In this sense, I would expect the cap to report the actual _hardware_
>> property, and the higher level stack (OpenCL or whatever, if and when
>> it will be supported) to massage the value as appropriate (e.g. by
>> multiplying by 4x —the overcommit needed to keep the device pipes
>> full— and then dividing by the vector width of the kernel).
>>
> The problem is that it requires a lot of hardware-specific knowledge to
> find the right over-commit factor (instruction latencies, issue
> overhead, the fact that in some cases the pipeline is twice as wide),
> and whether and to what extent the kernel needs to be scalarized -- and
> the OpenCL state tracker is hardware-independent.

I see. In this sense, it does make sense that the hardware capability
itself tries to encompass as much information as possible, and thus
provide a value that the OpenCL state tracker can use without further
knowledge. I still suspect that this number by itself won't be
sufficient (so some other information _will_ be needed for
efficiency), but as long as it's the only one I do agree that using
(for ILO) 16 (or even 32) will be better than using the hardware SIMD
width of 4 (or 8).

As such, for me the patch in question can go in, with possibly some
extra details in the comment.

-- 
Giuseppe "Oblomov" Bilotta