[10766] in Hotline Meeting
VR299
daemon@ATHENA.MIT.EDU (jchase@Athena.MIT.EDU)
Mon Sep 21 12:16:22 1992
From: jchase@Athena.MIT.EDU
To: hotline@Athena.MIT.EDU
Date: Mon, 21 Sep 92 12:16:07 EDT
------- Forwarded Message
Received: from ATHENA.MIT.EDU by po6.MIT.EDU (5.61/4.7) id AA18235; Mon, 21 Sep 92 12:13:23 EDT
Received: from [18.71.0.12] by Athena.MIT.EDU with SMTP
id AA11755; Mon, 21 Sep 92 09:44:16 EDT
From: epeisach@Athena.MIT.EDU
Received: by paris (5.57/4.7) id AA05890; Mon, 21 Sep 92 09:25:00 -0400
Date: Mon, 21 Sep 92 09:25:00 -0400
Message-Id: <9209211325.AA05890@paris>
To: op@Athena.MIT.EDU
Subject: Re: sgint overflows
The messages relate to VS3100 with color display cards. This error is from the
X server if certain operations timeout. (The kernel may also log it as well
in chaging colormaps and cursors).
What causes this problem:
1) To understand how the SG/QD graphics cards work, one must understand that they
are not memory mapped I/O devicies like all other X displays that we have.
This means to put an image on the screen, you set it up in some off
screen memory and then you tell the other processor (yes there is another
processor for the display) to transfer the bits down what I consider
a slow stream (like a straw)
2) There is about 2K of memory on the SG board for storage. (say of fonts)
3) The X server sometimes needs to transfer data out of the frame buffer
when it wants to install new pictures etc that are overlapped.
4) The X server tries to use the extra 2K whenever possible to store extra
info so it doesn't have the straw affect.
5) People who put large background images on their display will find
poor performance because the data will not fit in the offscreen region.
Hence bad sucking problem.
Because of (5) I often see that error message and an apparent hang. It's just
moving alot of data.
The technical manual for the hardware I believe I gave to Craig (or Chris V.)
I found it on a garbage pile about 2 years ago and grabbed. It is our only copy.
Ezra
------- End of Forwarded Message
