While looking at intel-graphics-compiler release notes[1] I noticed the platform deprecation warning, which sent me to this package's release notes[2]. There it said that anything older than gen10 is in legacy mode. If one were to ctrl-f for "Quality expectations" one'd fund a table saying that gen <11 are not even labeled. Does that mean that they are unsupported? One thing is sure: the slot does not reflect the official support status. Could you please clarify what does the SLOT represent? Maybe it should be reworked? [1]: https://github.com/intel/intel-graphics-compiler/releases/tag/igc-1.0.17537.20 [2]: https://github.com/intel/compute-runtime/releases/tag/24.35.30872.22
(In reply to Filip Kobierski from comment #0) > There it said that anything older than gen10 is in legacy mode. Older than Gen12 would be correct. > If one were to ctrl-f for "Quality expectations" one'd fund a table saying > that gen <11 are not even labeled. > > Does that mean that they are unsupported? Sorry, but I dont' understand, what you want to say. Broadwell = Gen 8 Skylake = Gen 9 Kaby Lake = Gen 9.5 Coffee Lake = Gen 9.5 Apollo Lake = Gen 9 Gemini Lake = Gen 9 (Gen 10 basically never existed) Ice Lake = Gen 11 Elkhart Lake = Gen 11 All newer GPUs listed in [2] are Gen12+ > One thing is sure: the slot does not reflect the official support status. Please explain, why you see this? I would not agree here. The legacy slot is for the legacy platforms [1]. This was once a normal release and decleared as legacy for those older GPUs, while current versions only support new GPUs. > Could you please clarify what does the SLOT represent? I am not sure what to say. SLOT=0/legacy is needed for Gen8-Gen11, while SLOT=0 is the current release version for everything with Gen12 and newer. > Maybe it should be reworked? What would you suggest? [1] https://github.com/intel/compute-runtime/blob/master/LEGACY_PLATFORMS.md [2] https://github.com/intel/compute-runtime/releases/tag/25.13.33276.16
> > There[0.1] it said that anything older than gen10 is in legacy mode. > Older than Gen12 would be correct. Tiger Lake is 11th gen and it's marked as production. Ice Lake (10th gen) is marked as legacy. Please note that the link points to release 24.35.30872.22. I translate the codenames according to wikipedia. > All newer GPUs listed in [1.2] are Gen12+ Tiger lake is marked as production though. LEGACY_PLATFORMS.md[1.1] explained the status of gen<10 to me. Also I have learned that Comet Lake, so gen10 was never supported... Now that I look at it I think the SLOT's fine but is not explained anywhere. Also the fact that :legacy is stable makes it worse. > What would you suggest? Create some documentation for it. longdescription would be a good place to start. With :legacy being stable I'd also put a postinst message that would tell the users to upgrade to ~amd64 if they have a newer processor. I think it's critical to mention the generation numbers (and maybe codenames?) there.
(In reply to Filip Kobierski from comment #2) > Tiger Lake is 11th gen and it's marked as production. This is not correct. Tiger Lake's GPU is Gen12. > Ice Lake (10th gen) is marked as legacy. No, this is not correct. Ice Lake's GPU is Gen11. > I translate the codenames according to wikipedia. Seems then not correct? > > All newer GPUs listed in [1.2] are Gen12+ > Tiger lake is marked as production though. > > > LEGACY_PLATFORMS.md[1.1] explained the status of gen<10 to me. > Also I have learned that Comet Lake, so gen10 was never supported... > Also the fact that :legacy is stable makes it worse. Well, :legacy was before slotmove the only stable version and support for newer versions didn't existed. IGC2/IGR will be promoted soon to stable too. > > What would you suggest? > Create some documentation for it. longdescription would be a good place to > start. I can do that. > With :legacy being stable I'd also put a postinst message that would tell > the users to upgrade to ~amd64 if they have a newer processor. I think it's > critical to mention the generation numbers (and maybe codenames?) there. Ok, I can add that.
> This is not correct. <codename>'s GPU is Gen12. Where do you get this information? I have assumed that CPU & GPU generations are the same. I can see that it's not the case now: https://en.wikipedia.org/wiki/Tiger_Lake#GPU For CPUs, Ice Lake is gen10[1] & Tiger Lake is gen11[2]. [1]: https://www.intel.com/content/www/us/en/ark/products/codename/74979/products-formerly-ice-lake.html [2]: https://www.intel.com/content/www/us/en/ark/products/codename/88759/products-formerly-tiger-lake.html --- Right now, from what I can tell, the only mention for the right hardware is in virtual/opencl: - dev-libs/intel-compute-runtime - integrated Intel GPUs from Gen12 onwards. 64-bit only; - dev-libs/intel-compute-runtime:legacy - integrated Intel GPUs from Gen8 up to Gen11. 64-bit only; I think the wording is quite poor since the GPUs are integrated with the CPU and at least from my experience intel generation = CPU generation. In https://github.com/gentoo/gentoo/pull/41842 I'd like make it just "integrated Intel GPUs" so that people can consult this package for which slot to choose. Do you think it's a good idea?
