Announcement

Collapse
No announcement yet.

AMDGPU's xf86-video-amdgpu vs. Mode-Setting DDX Performance

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

  • s_j_newbury
    replied
    Originally posted by Sonadow View Post

    I tried something similar on my Skylake notebook and did not manage to get it running; i get no native res and no acceleration, and Fedora still wants to load i915. Did I do it wrongly or I missing out on something?
    modesetting isn't a DRM driver, it's a DDX, you can't replace i915 (a kernel module) with modesetting. You want to run modesetting Xorg DDX driver on top of the i915 kernel DRM module. Simply setting it in your xorg.conf driver section will do it. Otherwise the "intel" Xorg DDX driver is used by default.

    Leave a comment:


  • Luke
    replied
    Originally posted by smitty3268 View Post
    Yep, can confirm. Easylist, used by adblock, has a rule hiding

    Code:
    phoronix.com##div[style^="text-align: center;"]
    which contains all the charts. The 1st element on the page appears to be an ad, so i guess they got carried away with the rule. Perhaps it's automatic.
    Somehow this is no surprise. I remember about a year ago getting a complaint that a totally non-monetized, no ads allowed political website I am one of the admins of was being blocked in its entirety by Adblock as well as by the DC libraries and several wireless networks. A similar issue was reported by the activist-owned Riseup email server, which was several times falsely flagged by the Spamhaus project leading to blocked emails. In short, not all authors of blocklists can be trusted, and I've seen outright politicization before. Thus, all blocklists require manual editing both to remove undesired whitelists (e.g. ABP's "paying customers") and undesired blacklists. There's no really good solution to this unfortunately. Privacy Badger's behavior system, for instance, won't work on a "stateless" browser which is always run from a fresh start.

    Leave a comment:


  • liam
    replied
    Originally posted by eydee View Post

    The real question is, if this generic driver works just as well as vendor specific ones, why do we have them in the first place?
    Well, if it does work as well, then there is no reason.
    For embedded hardware with dedicated 2d acceleration sips then it makes much more sense to make use of those, if possible.

    Leave a comment:


  • Sonadow
    replied
    Originally posted by lumks View Post
    Code:
    [COLOR=#000000][FONT=monospace]cat /etc/X11/xorg.conf.d/20-radeon.conf [/FONT][/COLOR]
    
    [FONT=monospace]Section "Device"[/FONT]
    [FONT=monospace] Identifier "Device0"[/FONT]
    [FONT=monospace] Driver "modesetting"[/FONT]
    [FONT=monospace] Option "AccelMethod" "glamor"[/FONT]
    [FONT=monospace]EndSection[/FONT]
    to enable modesetting usage...
    I tried something similar on my Skylake notebook and did not manage to get it running; i get no native res and no acceleration, and Fedora still wants to load i915. Did I do it wrongly or I missing out on something?

    Leave a comment:


  • Michael
    replied
    Originally posted by darkbasic View Post


    Michael is it possible to fix it? I'm premium and I find it pretty annoying... I hope it isn't a "feature".
    darkbasic as mentioned in a few threads now, this isn't intended but isn't anything I can do to fix on my end. EasyList added a new phoronix.com rule a few days ago that is overly broad... Any div that's centered gets blocked. So I can't really do anything on my end, you need to either disable AB/EL/whatever it is or just wait for it to get fixed. Though in trying to find the EasyList bug report, it's not showing up so maybe it will already be fixed in your next EasyList update however that is done.

    Leave a comment:


  • darkbasic
    replied
    Originally posted by kurros View Post

    Someone in another post complained about this, I think it was their adblocker hiding the benchmarks.

    Michael is it possible to fix it? I'm premium and I find it pretty annoying... I hope it isn't a "feature".

    Leave a comment:


  • bridgman
    replied
    The generic driver probably does work as well as vendor specific ones for consumer use cases, but remember we're addressing the pro workstation market as well. Things like framelock/genlock will probably take a lot longer to become mainstream...

    Leave a comment:


  • ernstp
    replied
    During early development I had very different results with modesetting and amdgpu-ddx. Also getting nice and clean Vsync basically only worked with DRI3+amdgpu-ddx for Fury. Agd5f said there were a bunch of differences in the Vsync path.

    But the benchmarks are run without vsync of course...

    Leave a comment:


  • eydee
    replied
    Originally posted by liam View Post
    Why would we expect any changes in 3d performance when changing out the X driver?
    The real question is, if this generic driver works just as well as vendor specific ones, why do we have them in the first place?

    Leave a comment:


  • W.Irrkopf
    replied
    Originally posted by Michael View Post

    Just disable EasyList or whatever for Phoronix then. This isn't something I changed on my end but rather an EasyList fail as mentioned in a few other posts now.
    Thanks for the quick and detailed response! Now I regard this as a bug in Easylist and I will treat it as such.

    Leave a comment:

Working...
X