[Bug 110781] Radeon: heavy r300 performance drop regression between 11.x and 19.x

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=110781

--- Comment #25 from Richard Thier <u9vata at gmail.com> ---
In the assert problem this is the domain and flag meanings in case of only my
hack is in place:

domain(6) == RADEON_DOMAIN_GTT(2) | RADEON_DOMAIN_VRAM(4)
flags(1) == RADEON_FLAG_GTT_WC (1 << 0)

So I get -1 even if I "outcomment" this check in radeon_get_heap_index(..):

    /* Resources with interprocess sharing don't use any winsys allocators. 
    if (!(flags & RADEON_FLAG_NO_INTERPROCESS_SHARING))
        return -1;*/

There is a switch statement in this function where they switch based on the
domain parameter:

    switch (domain) {
    case RADEON_DOMAIN_VRAM:
      ...
    case RADEON_DOMAIN_GTT:
      ...
    default:
        break;
    }
    return -1;

I suspect we return -1 here because we go into the "default" case. I think that
is right because domain is: 6 == 2 | 4 == GTT+VRAM at the same time.

Maybe once it was meaningful to have both GTT (is that GART?) and VRAM at the
same time being or-ed together but not anymore?

I will try to add a hack to make the domain always just VRAM to see what
happens... I know that is bad, but just for the analysis.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/dri-devel/attachments/20190604/2883c80b/attachment.html>