Eric Anholt (anholt) wrote,
Eric Anholt
anholt

X with glamor on vc4

Today I finally got X up on my vc4 driver using glamor.  As you can see, there are a bunch of visual issues, and what you can't see is that after a few frames of those gears the hardware locked up and didn't come back.  It's still major progress.
2014-08-21 16.16.37
The code can be found in my vc4 branch of mesa and linux-2.6, and the glamor branch of my xf86-video-modesetting.  I think the driver's at the point now that someone else could potentially participate.  I've intentionally left a bunch of easy problems -- things like supporting the SCS, DST, DPH, and XPD opcodes, for which we have piglit tests (in glean) and are just a matter of translating the math from TGSI's vec4 instruction set (documented in tgsi.rst) to the scalar QIR opcodes.
Subscribe
  • Post a new comment

    Error

    default userpic

    Your IP address will be recorded 

    When you submit the form an invisible reCAPTCHA check will be performed.
    You must follow the Privacy Policy and Google Terms of use.
  • 4 comments

Anonymous

August 22 2014, 13:46:35 UTC 4 years ago

I was just wondering how comprehensive you find that the Piglet tests are. I have glanced through the list of tests, and there are a lot of them, but do you find them to be good quality and have good coverage?

I hear a lot of people complaining about the vast array of OpenGL ES incompatibilities between vendors (especially in the cheapo Android market). Is Piglet compliance going to be enough, or will you need to do build more test cases internally?
Khronos's GLES suite (GTF) and piglit are fairly comparable in coverage, but piglit is designed with the intent of making driver debug easy. I tend to spend my time debugging the piglit failures, and trust that GTF failures will go away as piglit passrate approaches 100%. Eventually, you sit down and debug the last bugs in GTF, then you give Khronos their money and get your certification.

Hopefully, as the open source community gets support for more mobile chips, the Android incompatility problem goes away because everyone starts using the same GL stack (so that all the areas where behavior is undefined in the spec end up happening to work out the same).
Hi

I'm currently working on Android for the PI using the current userland sources. It's basically a case of implementing the missing Android HAL libraries. I am curious as to whether a vc4 flavoured libdrm is part of this project? ( please excuse my ignorance wrt glamors workings )

Thanks
trevd

anholt

August 28 2014, 00:01:39 UTC 4 years ago Edited:  August 28 2014, 00:25:47 UTC

Nope, not currently -- I don't have any other consumers yet, so I haven't generalized my buffer management code (not that there's much there yet)