No announcement yet.

Coreboot's Flashrom Working On Radeon GPU Flashing Support

  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by agd5f View Post
    Many of these command tables are also board specific as the programming sequences are dependent on the board design. So no real need to change any of that either.
    Well, there was that time when ATI forgot about the DDWG required hotplug detect pin in their connector specification of atombios. So HPD event validity and finding where it originated was also not part of standard testing routines at ATI. So none of the R500 boards had been tested for valid hot plug detection, and in a few measly percent, the board makers had strayed enough from the reference board that it mattered. This was fixed in R600, i do not remember whether this was fixed for even the firegl cards.

    Similarly, R5xx, and most of R6xx used the wrong bit to watch for vsync after a modeswitch was triggered. We at RadeonHD had used this bit as well, as i decided to trust the BIOS more than the documentation, and it seemed to work. It only worked in 99.5% of the cases, and when some bugs appeared, the error was quickly correct, in the pure C code. The forked driver of course went and copied our change, as it did with most things. No attempt was made to fix the firmware, and it was only fixed for the displayport enabled rx6xx iirc.

    And that's just the ones i remember from soon 15ys ago. There's plenty of reason to change and/or fix this stuff.

    My prime reason to turn this into C code is control, always was and always will be. Which is why modesetting happened, and we got rid of drivers depending on int10.

    My main argument in case of RadeonHD was: do we want to be bug compatible with the fglrx driver, or do we want to work around hw bugs directly? How anyone could argue against that is beyond me, but it seems like you and your friends succeeded.
    Last edited by libv; 07 January 2022, 03:00 PM.