Announcement

Collapse
No announcement yet.

NIR Experimental Backend For RadeonSI Updated

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    Originally posted by darkbasic View Post

    Switching to Gallium would take a couple of years too for Intel.
    Right, but building it on top of Vulkan would take a couple years, and then another couple years to switch the driver. Using gallium the hard part is already done, and you're just transitioning the driver over to an already working stack.

    But like i said, Intel already has a working driver stack they seem happy with, so I don't see them manufacturing work for themselves like this.

    Comment


    • #12
      Originally posted by duby229 View Post

      I think I remember initially Gallium was developed with an experimental Intel driver. It was really I think just a driver used to implement the concepts and infrastructure. AMD came on board after it was basically working, advanced it considerably and added their range of hardware support.

      So I guess I'm just wondering if Intel as a member of Kronos, did they have some conception of GLnext or Vulkan back in those days? In after thought, they had to have known something else was coming in just a few years.
      Intel was never directly involved with Gallium. There was an i915 driver that Google created for their Chromebooks that you might be thinking of.

      I believe gallium was largely created by VMWare and the original hardware driver that really started using it widespread was Marek's r300 port, but i guess i don't know the details of when that i915 driver started in comparison. Or maybe vmware did some early experiments with the intel driver, but i'm pretty sure that would have just been because it was the best working driver at the time.

      GLNext/Vulkan really didn't come around until Valve started pushing it, which is quite a bit more recent than gallium. There was some talk of really changing stuff up around GL3 as well, but I think that was quite a bit different than the low-level API we ended up with in Vulkan. Gallium was based mostly off of DX10 concepts.
      Last edited by smitty3268; 06 July 2017, 02:14 PM.

      Comment


      • #13
        Originally posted by smitty3268 View Post

        Intel was never directly involved with Gallium. There was an i915 driver that Google created for their Chromebooks that you might be thinking of.

        I believe gallium was largely created by VMWare and the original hardware driver that really started using it widespread was Marek's r300 port, but i guess i don't know the details of when that i915 driver started in comparison. Or maybe vmware did some early experiments with the intel driver, but i'm pretty sure that would have just been because it was the best working driver at the time.

        GLNext/Vulkan really didn't come around until Valve started pushing it, which is quite a bit more recent than gallium. There was some talk of really changing stuff up around GL3 as well, but I think that was quite a bit different than the low-level API we ended up with in Vulkan. Gallium was based mostly off of DX10 concepts.
        Ah ok, thanks. you answered my questions well.

        Comment


        • #14
          Originally posted by smitty3268 View Post

          Right, but building it on top of Vulkan would take a couple years, and then another couple years to switch the driver. Using gallium the hard part is already done, and you're just transitioning the driver over to an already working stack.

          But like i said, Intel already has a working driver stack they seem happy with, so I don't see them manufacturing work for themselves like this.
          What would be far more interesting is a faster version of modesetting running on top of Vulkan. Would Vulkan help in such regard compared to OpenGL?
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #15
            Originally posted by darkbasic View Post

            What would be far more interesting is a faster version of modesetting running on top of Vulkan. Would Vulkan help in such regard compared to OpenGL?
            I don't know the answer to that one, but I suspect it might. I think there are places in modesetting where they have to go to great lengths to work around GL slowness that would probably be easier in Vulkan. Then again, maybe that's already mostly taken care of by now.

            It sounded like a few people were interested in using the no_error support being added to mesa to see if it helped glamor, at least on lower-end hardware. Not sure if anyone finished that or not.

            Comment

            Working...
            X