From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 110781] Radeon: heavy r300 performance drop regression between 11.x and 19.x Date: Tue, 04 Jun 2019 08:24:41 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============1662339158==" Return-path: Received: from culpepper.freedesktop.org (culpepper.freedesktop.org [131.252.210.165]) by gabe.freedesktop.org (Postfix) with ESMTP id 4CD55897F5 for ; Tue, 4 Jun 2019 08:24:41 +0000 (UTC) In-Reply-To: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dri-devel-bounces@lists.freedesktop.org Sender: "dri-devel" To: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org --===============1662339158== Content-Type: multipart/alternative; boundary="15596366812.bA8C29a7F.28748" Content-Transfer-Encoding: 7bit --15596366812.bA8C29a7F.28748 Date: Tue, 4 Jun 2019 08:24:41 +0000 MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://bugs.freedesktop.org/ Auto-Submitted: auto-generated https://bugs.freedesktop.org/show_bug.cgi?id=3D110781 Richard Thier changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #144427|0 |1 is obsolete| | --- Comment #26 from Richard Thier --- Created attachment 144439 --> https://bugs.freedesktop.org/attachment.cgi?id=3D144439&action=3Dedit Hacky fix working also for 19.x versions I can confirm the problem with the "possible fix" is because radeon_get_heap_index does not handle both GTT and VRAM domains at the same time, but r300 still asks for that. Added a hacky fix that works on 19.x versions but forces VRAM domain. This highlights problematic areas for the new versions. I think this latter issu= e is because of bitrot happening with r300 as now we re-enabled a code path that= was never happening after the slowdown in 17.x as analysed before. PS.: Of course the "possible fix" still works when applied directly to the = 17.x commit that made things slow and that is the better direction, just not to = the latest code because of this issue. --=20 You are receiving this mail because: You are the assignee for the bug.= --15596366812.bA8C29a7F.28748 Date: Tue, 4 Jun 2019 08:24:41 +0000 MIME-Version: 1.0 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Bugzilla-URL: http://bugs.freedesktop.org/ Auto-Submitted: auto-generated Richard Thier changed bug 11078= 1
What Removed Added
Attachment #144427 is obsolete   1

Comme= nt # 26 on bug 11078= 1 from = Richard Thier 
Created attachment 144439 [details] [review]
Hacky fix working also for 19.x versions

I can confirm the problem with the "possible fix" is because
radeon_get_heap_index does not handle both GTT and VRAM domains at the same
time, but r300 still asks for that.

Added a hacky fix that works on 19.x versions but forces VRAM domain. This
highlights problematic areas for the new versions. I think this latter issu=
e is
because of bitrot happening with r300 as now we re-enabled a code path that=
 was
never happening after the slowdown in 17.x as analysed before.

PS.: Of course the "possible fix" still works when applied direct=
ly to the 17.x
commit that made things slow and that is the better direction, just not to =
the
latest code because of this issue.

You are receiving this mail because:
  • You are the assignee for the bug.
= --15596366812.bA8C29a7F.28748-- --===============1662339158== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHBzOi8vbGlz dHMuZnJlZWRlc2t0b3Aub3JnL21haWxtYW4vbGlzdGluZm8vZHJpLWRldmVs --===============1662339158==--