svgalib.chips - Information for Chips and Technologies Users

Information for Chips and Technologies Users

David Bateman <dbateman@eng.uts.edu.au>

23nd May 1997

0. Introduction

1. "libvga.config""options"

2. Unsupported Chips and Technologies chipsets

3. Known bugs

This is the really only my first attempt to get a working fully featured driver for the Chips and Technologies chipset to work with svgalib(7). As such the only machine that I know it will work on is my own. If you use this software then at this point I'm still very interested in hearing from you by e-mail. Include full details of your chipset, amount of videoram and whether you have a VL-Bus or PCI bus machine.

This server was written using the svgalib(7) patch from Sergio and Angelo Masci as a starting point. This version of the code resembled the XFree server code that was used up to XFree 3.1.2. As such it was incapable of programming the clocks, using linear addressing, Hi-Color, True-Color modes or the hardware acceleration. All of these features have since been added to the code. In addition support for the 65525, 65535, 65546, 65548, 65550 and 65554 have been included. The 64200 and 64300 chips are unsupported, however these chips are very similar to the 6554x chips which are supported.

At this point this code is only confirmed to work correctly on a 65545 VL-Bus machine. However as much of the code was stolen from my experiences with writing code for XFree I hope not to have too many problems with other machines. However if you run this code on a 65545/48 PCI machine or a 65550/54 machine then I am particularly interested in hearing of any success or failure stories.

The first thing to note is that the option parser for svgalib(7) is not very robust. Hence if you make some typing mistakes, you can have some very strange effects. I've set out below the libvga.config(5) options that are of particular interest to Chips and Technologies users. Normally this configuration file can be found at /etc/vga/libvga.config.

HorizSync MIN MAX

Often LCD panels has very different specifications for the horizontal sync than CRT's do. Hence often you'll need this option, particularly if you are using the XFree like modelines described below. The two floating point numbers specified will set the minimum and maximum allowed horizontal sync in kHz.

VertRefresh MIN MAX

Similar to the above, but this sets the LCD or CRT's vertical refresh rate in Hz.

modeline 640x480 20.00 640 688 704 776 480 480 481 486

This option allows you to specify XFree like modelines to use in preference to the in built modelines. Often LCD panels will need very different pixel clocks and timings than CRT's. Hence this option allows you to specify these. Note that the LCD panel timings are related to the panel size and not the mode size. Therefore by default the BIOS setting already uploaded into the registers are used by default. See the "UseModeline" option below if you wish to override these.

chipset C&T 5 1024

These option allows the user to specify the chipset to use and the amount of installed memory in kBytes. Currently supported chipsets are

0 65520 1 65525 2 65530 3 65535 4 65540 5 65545 6 65546 7 65548 8 65550 9 65554

TextClockFreq 25.175

One major difference between this code and the previously available support for the Chips and Technologies chipsets is that it supports the use of programmable clocks. Because of the way that the Chips and Technologies chips program the VCO from the registers, there is no way to be sure to recover the previously programmed clock value. Hence the driver assumes that the console clock is 25.175MHz. This will be wrong for many machines. However I have supplied this option to use a different value that might be more suitable for your machine.

nolinear

This option disables the use of the linear framebuffer. This might be useful for machines that have broken linear framebuffers. In lar the linear framebuffer doesn't seem to work with the achines.

linear

Allow, but don't enforce the use of the linear framebuffer. As this is the default anyway, I don't see that this option is much use.

setuplinear 0xC0000000

For VL-Bus machines I expect that the linear framebuffer starting address will be setup correctly. However to get the starting address for PCI machines requires access to the MEMBASE register in the PCI address space. Code to do this doesn't currently exist with svgalib(7), and so I've taken the easy option of just testing a few known PCI starting addresses. For now these are just 0xFE000000, 0xFD000000, 0x41000000 and 0xC0000000. If you have a different starting address then the linear framebuffer will be unusable. You might like to report your starting address to me so that I can include it in the probing code, but till then this option can be used to set up the correct address. This option just forces the given address to be the only one probed. It doesn't force the linear framebuffer to be used.

LCDPanelSize 800 600

For some machines the LCD panel size is incorrectly probed from the registers. This option forces the LCD panel size to be as specified. If you have a black band down one side of your LCD display you might very well need this option. Also if you are using the option "fix_panel_size" in XFree then this option has a similar effect. This option can be used in conjunction with the option "UseModeline" to program all the panel timings using the modeline values. Two machines that are known to need this option are the HP Omnibook 5000CTS and the NEC Versa 4080 800x600 TFT machines.

UseModeline

The flat panel timings are related to the panel size and not the size of the mode specified. For this reason the default behaviour of the svgalib(7) is to use the panel timings already installed in the chip. The user can force the panel timings to be recalculated from the modeline with this option. However the panel size will still be probed. Two machines that are known to need this option are the HP Omnibook 5000CTS and the Prostar 8200. You are advised to check the README.chips that come with XFree for more details.

NoBitBlt

This option disables the use of H/W acceleration. As far as I know the only thing that currently uses the H/W acceleration is libvgagl, so this might not be a problem anyway. However if you see corruption of the graphics on the screen try this option and see if it goes away.

Use18BitBus

For 24bpp on TFT screens, the server assumes that a 24bit bus is being used. This can result in a reddish tint to 24bpp mode for machines that actually have a 18 bit bus. This option, selects an 18 bit TFT bus. Note that using this option with a 24 bit bus machine will similarly discolour the screen. For other depths this option has no effect.

Center ENABLE/DISABLE or Stretch ENABLE/DISABLE

The default behaviour of svgalib(7) is to leave the stretching and centring registers completely alone. However for some machines this might result in poorly placed modes, or modes that don't fill the whole screen. These two options can be used to centre and stretch the mode on the screen. Note that for instance a Center DISABLE might follow a Center ENABLE in the config file. Only the last option takes effect.

The 64200 and 64300 chips are unsupported. However by specifying the chipset in your libvga.config as either a

chipset C&T 3 2048

Use 65535 for a 64200 assuming 2M of video ram, or

chipset C&T 7 2048

Use 65548 for a 64300 assuming 2Mb of video ram

then svgalib can be made to give limited support to these chipsets. Note that the paged addressing mode of the 65548 chip and earlier can only address upto 1Mb of video ram. If the additional memory is needed then linear addressing must be used!! Note that support of the 64xxx chips has not been tested at all, and the above is just a suggestion that I believe will work.

One persistent and annoying bug is that the text mode stretching on LCD displays is not always restored correctly for 65550 and 65554 machines. This is to do with the manner in which the extended registers are restored and what is being done with the synchronous reset while the registers are restored. As I don't have a 65550 or 65554 machine of my own on which to test this code, I have been unable to fix this problem. In most circumstances an LCD-CRT switch will restore the LCD stretching to the desired state.

David.

/etc/vga/libvga.config

svgalib(7), libvga.config(5).

of the driver and this documentation is David Bateman <dbateman@eng.uts.edu.au>. However, it was slightly reformatted by Michael Weller <eowmob@exp-math.uni-essen.de>.