From mboxrd@z Thu Jan 1 00:00:00 1970 From: bugzilla-daemon@freedesktop.org Subject: [Bug 92775] [radeon][TTM] Contention when evicting large buffers between VRAM and GTT Date: Fri, 25 Dec 2015 07:41:41 +0000 Message-ID: References: Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="===============0945782495==" Return-path: Received: from culpepper.freedesktop.org (unknown [131.252.210.165]) by gabe.freedesktop.org (Postfix) with ESMTP id DDE066E221 for ; Thu, 24 Dec 2015 23:41:41 -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 --===============0945782495== Content-Type: multipart/alternative; boundary="1451029301.aaD5Df0.23300"; charset="UTF-8" --1451029301.aaD5Df0.23300 Date: Fri, 25 Dec 2015 07:41:41 +0000 MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable https://bugs.freedesktop.org/show_bug.cgi?id=3D92775 --- Comment #1 from Michel D=C3=A4nzer --- (In reply to Shawn Starr from comment #0) > One solution discussed is to split up the transfer into smaller chunks in > radeon_ttm. Specifically, here's how I think a fallback could be implemented in the ker= nel driver which can never fail because of fragmentation or resource starvation: During initialization, reserve some pinned GTT memory for bounce buffers. W= hen a BO can't be bound to GTT for eviction as in the case reported here, inste= ad do the eviction directly from VRAM to CPU domain in one or several passes o= f: 1. Copy part of the BO from VRAM to one of the reserved bounce buffers in G= TT using the GPU. 2. Copy that part of the BO from the bounce buffer to the BO's system RAM p= ages using the CPU. --=20 You are receiving this mail because: You are the assignee for the bug. --1451029301.aaD5Df0.23300 Date: Fri, 25 Dec 2015 07:41:41 +0000 MIME-Version: 1.0 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

Comment= # 1 on bug 92775<= /a> from Michel D=C3=A4nzer 
(In reply to Shawn Starr from comment #0)
> One solution discussed is to split up the transf=
er into smaller chunks in
> radeon_ttm.

Specifically, here's how I think a fallback could be implemented in the ker=
nel
driver which can never fail because of fragmentation or resource starvation:

During initialization, reserve some pinned GTT memory for bounce buffers. W=
hen
a BO can't be bound to GTT for eviction as in the case reported here, inste=
ad
do the eviction directly from VRAM to CPU domain in one or several passes o=
f:
1. Copy part of the BO from VRAM to one of the reserved bounce buffers in G=
TT
using the GPU.
2. Copy that part of the BO from the bounce buffer to the BO's system RAM p=
ages
using the CPU.

You are receiving this mail because: =20=20=20=20=20=20
  • You are the assignee for the bug.
--1451029301.aaD5Df0.23300-- --===============0945782495== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: base64 Content-Disposition: inline X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHA6Ly9saXN0 cy5mcmVlZGVza3RvcC5vcmcvbWFpbG1hbi9saXN0aW5mby9kcmktZGV2ZWwK --===============0945782495==--