From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 93147] [regression bisected] Stuttering in games caused by commit 4dfd6486 "drm: Use vblank timestamps to guesstimate how many vblanks were missed" Date: Thu, 03 Dec 2015 15:02:23 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============0253595739==" Return-path: Received: from culpepper.freedesktop.org (unknown [131.252.210.165]) by gabe.freedesktop.org (Postfix) with ESMTP id 502837A15B for ; Thu, 3 Dec 2015 07:02:23 -0800 (PST) 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 --===============0253595739== Content-Type: multipart/alternative; boundary="1449154943.a2E2771.8946"; charset="UTF-8" --1449154943.a2E2771.8946 Date: Thu, 3 Dec 2015 15:02:23 +0000 MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" https://bugs.freedesktop.org/show_bug.cgi?id=93147 --- Comment #10 from Mario Kleiner --- Created attachment 120301 --> https://bugs.freedesktop.org/attachment.cgi?id=120301&action=edit First proposed patch to fix this on radeon-kms Ok, the attached patch works on my test systems with my timing tests and hardware timing measurement equipment and i think it hopefully avoids the races Ville and me could come up with so far. One thing which might make sense for improved efficiency is to replace the udelay(5) in the radeon_flip_work_func() with a usleep_range() that avoids polling and sleeps the right minimum amount of time. I'll try to try that today. For the line buffer sizes for < DCE4 i just guessed values which are hopefully big enough to cover earlier asics. I think too big is not a big problem for correctness, but potentially for performance under higher graphics load. -- You are receiving this mail because: You are the assignee for the bug. --1449154943.a2E2771.8946 Date: Thu, 3 Dec 2015 15:02:23 +0000 MIME-Version: 1.0 Content-Type: text/html; charset="UTF-8"

Comment # 10 on bug 93147 from 
Created attachment 120301 [details] [review]
First proposed patch to fix this on radeon-kms

Ok, the attached patch works on my test systems with my timing tests and
hardware timing measurement equipment and i think it hopefully avoids the races
Ville and me could come up with so far.

One thing which might make sense for improved efficiency is to replace the
udelay(5) in the radeon_flip_work_func() with a usleep_range() that avoids
polling and sleeps the right minimum amount of time. I'll try to try that
today.

For the line buffer sizes for < DCE4 i just guessed values which are hopefully
big enough to cover earlier asics. I think too big is not a big problem for
correctness, but potentially for performance under higher graphics load.

You are receiving this mail because:
  • You are the assignee for the bug.
--1449154943.a2E2771.8946-- --===============0253595739== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHA6Ly9saXN0 cy5mcmVlZGVza3RvcC5vcmcvbWFpbG1hbi9saXN0aW5mby9kcmktZGV2ZWwK --===============0253595739==--