[Mesa-dev] GL4.5 or bust...

[Francisco Jerez]

Jason Ekstrand <jason at jlekstrand.net> writes:

> On May 16, 2016 1:53 AM, "Kenneth Graunke" <kenneth at whitecape.org> wrote:
>>
>> On Monday, May 16, 2016 6:32:11 AM PDT Dave Airlie wrote:
>> > So I said this on irc over the weekend and it seemed like we had some
>> > consensus on holding off 12.0 until we could announce 4.5 on some
>> > hardware. This assumes the FP64 stuff is going in at least.
>> >
>> > So I decided to roll out the proposal here, which is that we finish
>> > GL4.5 features off for at least Skylake I think.
>>
>> Personally, I think it makes the most sense to proceed with the
>> time-based release as planned.  We've already packed a ton of great
>> things into it...i965 will jump from 3.3 to 4.2, and the other drivers
>> have made significant strides as well.  I'd like to have an i965 GL 4.x
>> release sooner rather than later.
>
> The three things I would absolutely like to land in 12.0 are: attrib64 so
> we can get 4.2, Curro's spilling *compiler* fixes (we can land those later
> and back-port if needed), and a set of format cleanups in the Vulkan driver
> I plan to send today.
>
Yeah, I plan on sending that series out today.

> Beyond that, I agree that landing SIMD32 would be really nice and worth
> delaying by a week but no more.  Getting actual 4.5 it looks like will take
> longer to do properly.
>
Right.  If other people agree on delaying the branch point by a week it
should be possible to at least merge the subset of SIMD32 that we're
missing to enable desktop compute shaders.  Personally I'm not sure if
that's a great idea [the 12.0 release is likely to be bumpy enough
already ;)], but if people want I'm willing to do the work -- Otherwise
I'm fine with doing it after the branch point which gives me some more
time to look into the performance issues of SIMD32 compilation.

>> I might be inclined to push it back a week or so if that would help us
>> land SIMD32 (only for compute, and only when required).  Then i965/Gen8+
>> could enable ARB_compute_shader and claim 4.3.  Supposedly Curro has most
>> of that code written, and Jason's offered to review it as soon as
> possible.
>> A lot of applications seem to rely on GL 4.3 and compute shaders.
>>
>> I'm all for rushing to GL 4.5, but why not do it in master after the
>> branch point?  We can always pull in the next release if we want, rather
>> than pushing this one out.  Looking ahead...we've got a bunch going on:
>>
>> - Finishing GL 4.5
>> - Finishing ES 3.2
>> - Landing Tim's shader cache
>> - More performance work
>> - Porting features to more platforms (i.e. fp64 on Haswell/Ivybridge)
>>
>> I think we could knock out a bunch of that out in another 3 months, and
>> it'd make for a pretty great 12.1 release.  I also think that varying
>> packing is going to continue to bite us, and feel like we're going to
>> have to rewrite that in order to have a good quality 4.4
>> implementation...
>>
>> That's just my current thinking, anyway.  I'm not sure what's best.
>>
>> --Ken
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 212 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160516/1808491d/attachment.sig>