Information for Cirrus Chipset Users Harm Hanemaayer (H.Hanemaayer@inter.nl.net), Randy Hendry (randy@sgi.com) (64xx), Corin Anderson (corina@the4cs.com) 5 November 1998 1. Supported chipsets There are two different SVGA drivers for Cirrus chipsets, one called ``cir- rus'' and one called ``cl64xx''. The ``cirrus'' driver is used in the 256-color SVGA server (with acceleration) and the mono server (without accel- eration). The SVGA server supports 16, 24, and 32 bits-per-pixel truecolor modes on some configurations. The ``cl64xx'' driver is used in the 256-color SVGA, 16-color and mono servers. Note that except where stated otherwise, this document is referring to the ``cirrus'' driver. The following chipsets by Cirrus Logic are supported: CL-GD5420 ISA SVGA chipset, 1Mbyte; maximum dot clock is 45 MHz (256 color server). Acceleration with extended write modes (used for scrolling and solid filling in this driver). This chipset can not support 1024x768 non-interlaced in 256 colors. CL-GD5422 Enhanced version of the 5420 (32-bit internal memory interface). Maximum dot clock is 80 MHz. CL-GD6205/6215/6225/6235 Laptop chipsets more or less compatible with the 5420. The only dot clock supported is 25 MHz (more on an external display). Some problems have been reported with these chipsets (especially on external displays). Take note of the "noaccel" option. CL-GD6420/6440 These chipsets are not compatible with the 542x series, but are supported by the ``cl64xx'' driver. It is used in recent laptops, and bears some similarity to old Cirrus chipsets (5410/AVGA2). The driver may also work for other 64xx chips. The configuration identifiers for this driver are "cl6420" and "cl6440". This driver is discussed in detail in section The cl64xx Driver (sec- tion 6., page 1). CL-GD5424 Basically VLB version of the 5422, but resembles the 5426 in some respects. CL-GD5426 Supports both ISA bus and VLB, and up to 2Mbyte of memory. Has BitBLT engine for improved acceleration (BitBlt, image transfer, text). Dot clock limit is 85 MHz. CL-GD5428 Enhanced version of the 5426. CL-GD5429 Enhanced version of the 5428; officially supports higher MCLK and has memory-mapped I/O. CL-GD5430 Similar to 5429, but with 543x core (32-bit host interface). Does not have 64-bit memory mode. CL-GD5434 `Alpine' family chip with 64-bit internal memory interface. The chip can only support 64-bit mode if equipped with 2 Mbytes of memory; cards with only 1 Mbyte are severely limited. Supports dot clocks up to 110 MHz (later chips support 135 MHz). CL-GD5436 Highly optimized 5434. CL-GD5440 Similar to the CL-GD5430, and detected as such. CL-GD5446 Another member of the Alpine family of 2D accelerators; similar to the CL-GD5436. CL-GD5480 Newer Alpine family chip that support synchronous graphics RAM (SGRAM). CL-GD5462, CL-GD5464 and CL-GD5465 The Laguna VisualMedia family of 2D Accelerators. These chips use Rambus RDRAM memory. The '62 is a 64-bit 2D accelerator, including a BitBlit engine, video windows (not currently used by the server), and 64x64 HW cursor. Mono modes have not been tested. The CL-GD5464 is the next chip in the Laguna family, and the CL-GD5465 is the latest member, both have been tested. CL-GD7541/7542/7543/7548 Laptop chipsets more or less compatible with the 5428/3x. While has it been tested on some configurations, not all configuration may work correctly. CL-GD7555 Limited untested support, without auto-detection, has been pro- vided for this chip which is a 64-bit extension of the 754x fam- ily. Use a Chipset "clgd7555" line. Here's a list of maximum dot clocks for each supported depth: mono 8 bpp (256c) 16 bpp 24 bpp 32 bpp CL-GD62x5 45 MHz 45 MHz CL-GD5420 80 MHz 45 MHz (1) CL-GD542x/512K 80 MHz 45 MHz CL-GD5422/24 80 MHz 80 MHz 40 MHz 27 MHz CL-GD5426/28 85 MHz 85 MHz 45 MHz (2) 28 MHz CL-GD5429 85 MHz 85 MHz 50 MHz 28 MHz CL-GD5430 85 MHz 85 MHz 45 MHz (2) 28 MHz CL-GD5434/1Mb 85 MHz 85 MHz 42 MHz 28 MHz CL-GD5434/2Mb 85 MHz 110/135 MHz 85 MHz 28 MHz 45/50 MHz (2) CL-GD5436/1Mb 85 MHz 110 MHz (3) 60 MHz (3) 40 MHz (3) CL-GD5436/2Mb 85 MHz 135 MHz 85 MHz 85 MHz (3) 60 MHz (3) CL-GD5446/1Mb 85 MHz 110 MHz (3) 60 MHz (3) 40 MHz (3) CL-GD5446/2Mb 85 MHz 135 MHz 85 MHz 85 MHz (3) 60 MHz (3) CL-GD5462 170 MHz 170 Mhz 170 MHz 170 MHz 135 MHz CL-GD5464/65 170 MHz 230 Mhz 170 MHz 170 MHz 135 MHz CL-GD5480 85 MHz 200 MHz 100 MHz 100 MHz 50 MHz CL-GD754x 80 MHz 80 MHz 40 MHz (4) (5) (1) with 512K memory. (2) 50 MHz with high MCLK setting. (3) Depends on memory clock. (4) This may be too low for some chips. (5) This depth may actually work if it is enabled and tested. Rough virtual/physical screen resolution limits for different amounts of video memory: mono 8 bpp 16 bpp 24 bpp 32 bpp 256K 800x600 640x400 512K 1152x900 800x600 640x400 1024K 1600x1200 1152x900 800x600 680x510 2048K 2304x1728 1600x1200 1152x900 960x720 800x600 4096K 2304x1728 2272x1704 1600x1200 1360x1020 1152x900 For 546x chips, the above table isn't quite accurate. While the virtual width may be any size, the screen pitch will be rounded up to the nearest value in the table below. Thus, each line on the screen will take more video memory than just what is displayed. To maximize video memory, then, choose the virtual desktop width from the table of pixel widths below: 8bpp: 640, 1024, 1280, 1664, 2048, 2560, 3328, 4096, 5120, 6656 16bpp: 320, 512, 640, 832, 1024, 1280, 1664, 2048, 2560, 3328 24bpp: 640, 1024, 1280, 1664, 2048, 2560, 3328, 4096, 5120, 6656 32bpp: 160, 256, 320, 416, 512, 640, 832, 1024, 1280, 1664 For other Cirrus chips, it's advisable to have a virtual width that is a mul- tiple of 32 if acceleration is used. The horizontal monitor timings must be below 2048. To run XF86_SVGA at a higher color depth, pass options to the X server as follows: startx -- -bpp 16 5-6-5 RGB ('64K color', XGA) startx -- -bpp 16 -weight 555 5-5-5 RGB ('Hicolor') (not on 5462) startx -- -bpp 24 8-8-8 RGB truecolor startx -- -bpp 32 8-8-8 XRGB truecolor (543X/46/6X only) 2. Basic configuration It is recommended that you generate an XF86Config file using the `XF86Setup' or `xf86config' program, which should produce a working high-resolution 8bpp configuration. You may want to include mode timings in the Monitor section that better fit your monitor (e.g 1152x900 modes). The driver options are described in detail in the next section; here the basic options are hinted at. For all chipsets, a Clockchip "cirrus" line in the Device section can be use- ful. This allows the use of any dot clocks, instead of one out of the fixed set of dot clocks supported by the driver. This is required if you want a 12.6 MHz dot clock for low-resolution modes. However, when this option used, clock frequencies be unstable leading to strange effects, so only use it if absolutely required. For any chip with a BitBLT engine, the new XAA (XFree86 Acceleration Archi- tecture) is used. This code is new and still in a beta stage. If graphics redrawing goes wrong, try the "noaccel" option; if it is using memory-mapped I/O, "no_mmio" might be sufficient. In order to be able to run at a depth of 16bpp, 24bpp, or 32bpp, and to improve performance at 8bpp, linear addressing must be enabled. This is gen- erally In order to be able to run at a depth of 16bpp, 24bpp or 32bpp, and to improve performance at 8bpp, linear addressing must be enabled. Linear addressing is the default mode of operation on any PCI-bus configuration; use "nolinear" to disable it. For other bus types, it is generally possible on 543x local bus cards, and if you have less than 16Mb of system memory, on local bus 542x cards and ISA 543x cards. You must specify the "linear" option and possibly a Membase address. See the following sections for a detailed description. Memory-mapped I/O is the default mode of operation for any Alpine family chip. For the 5429, the "mmio" option may be used to enable it, but it has not been tested. Finally, if you have 546X chip, it will be on either a PCI or AGP bus. As such, there is no problem about memory mapped I/O or linear frame buffer address spaces running into system memory. The PCI spaces are mapped way up near the 4GB point. Because the mmio and linear frame buffer don't conflict at all on the system, the "linear", Membase, and "mmio" options are ignored (memory mapped I/O and linear addressing are always used). 3. XF86Config options Don't use the `Clocks' command. The clocks are fixed (i.e. not probed), and there should be no variation between cards (other than the maximum supported clock for each chipset). The following options are of particular interest to the Cirrus driver. Each of them must be specified in the `svga' driver section of the XF86Config file, within the Screen subsections of the depths to which they are applica- ble (you can enable options for all depths by specifying them in the Device section). Option "noaccel" This option will disable the use of any accelerated functions. This is likely to help with problems related to bugs in accelera- tion functions, and perhaps high dot clocks and DRAM timing, at the cost of performance (which will still be reasonable on a local bus). Option "fast_dram" "med_dram" "slow_dram" (5424/6/8/9, 543x, 5446, 546x)" These options set the internal memory clock (MCLK, or BCLK for the 546x) register to another value. The default value programmed by the BIOS is usu- ally OK, don't mess with these options unless absolutely required. The "fast_dram" option will cause the driver to set the internal memory clock (MCLK) register of the video card to a higher value (recent chips use an even higher value by default). Normally, this register is not touched but it appears that the standard CL- GD542x BIOS initializes it to a value that is somewhat on the low side (limited by the chip specification), which has a negative influence on performance of high dot clock modes. This is espe- cially true if extended RAS timing is being used (this is indi- cated in the server probe). The actual speed of DRAM is not a critical factor in the determining whether this option is appropriate; one CL-GD5426-based card with 80ns DRAM using Extended RAS timing, which came with a DOS driver utility to set the MCLK to this value (0x22), seems to run stable at higher MCLK. There are also (mainly brand name) cards whose customized BIOS does initialize to a higher non-standard value. The "slow_dram" option will set the MCLK to the value used by the standard CL-GD542x BIOS (0x1c). Symptoms of a MCLK that is too high can be vertical bands of flickering pixels on the screen, erroneous pixels appearing in text, and loosing pixels in the textmode font after running X (note that very similar effects can be caused by an MCLK setting that is too low). Upon start-up, the driver will report the value of the MCLK reg- ister (check this first), and also any changes that are made. Typical MCLK values: 0x1c (50 MHz) This is usually the BIOS default. It is forced by the "slow_dram" option. 0x1f (55 MHz) Value used by the "med_dram" option. Highest value that 542x based cards seem to be able to handle with linear addressing enabled. 0x22 (60 MHz) Value that most (Extended RAS) 542x cards seem to be able to handle, used by the "fast_dram" option. The official maximum of the 542x chips is 50 MHz. The official spec. for the 5434 is also 50 MHz (0x1c) and that for the 5429 and 5430 is probably 60 MHz (0x22). Current revisions of the 5434 (E and greater) support 60 MHz MCLK in graphics modes, and the driver will program this automatically. If it causes problems, use the "slow_dram" option. The driver takes the MCLK into account for clock limits that are determined by DRAM bandwidth. For the 546x chips, the BCLK is the Rambus access clock. Typical values live in the range of 258 MHz to 300 MHz. If you have troubles, such as a black checkerboard pattern on the screen, try using the "med_dram" or "slow_dram" options. In all cases, if you are not having any problems (performance or stability at high dot clocks), it is best not to use any of the DRAM options. Option "no_bitblt" This option, when used with a 5426/28/29/3x/46/6x/754x, will have the effect of disabling the use of the BitBLT engine (which the 5424 does not have), while retaining some acceleration. This will be useful for problems related to functions that use the BitBLT engine. Performance is significantly decreased. Option "no_imageblt" This option is now obsolete. The "xaa_no_color_exp" option has a somewhat similar effect. chipset "clgd54xx" Force detection of the given chipset. Useful if you have a sup- ported chipset that is not properly detected, or if you have an unsupported chip that might be compatible with a supported one. videoram 1024 (or another value) This option will override the detected amount of video memory, and pretend the given amount of memory is present on the card. This is useful on cards with 2Mbyte of memory whose DRAM configu- ration is not compatible with the way the driver enables the upper megabyte of memory, or if the memory detection goes wrong. It must be specified in the Device section. Option "fifo_conservative" (5424/6/8/9/3x/46/6x/754x) This option will set the CRT FIFO threshold to a conservative value for high dot clocks (>= 65 MHz), reducing performance but hopefully alleviating problems with what can be described as flashing `streaks', `jitter' or horizontally repeated display areas on the screen (especially when a BitBLT operation is in progress, e.g. scrolling). Option "fifo_aggressive" (5424/6/8/9/3x/46/6x/754x) This option will set the CRT FIFO threshold to an aggressive value; it will be the same as that used for lower dot clocks. Theoretically it improves performance at high dot clocks, but it does not help in the vast majority of cases. In some cases with 546x chips, however, this option can help reduce horizontal streaks or otherwise fix abnormal display problems (display shifted to the left, etc.). Option "no_2mb_banksel" (542x) This option will cause the driver not to set the `DRAM bank select' bit to enable the upper megabyte of memory on a 2Mbyte card. This should be helpful with cards equipped with 512Kx8 DRAMs, as opposed to 256Kx4/16 DRAMs, when using a virtual screen configuration that uses more than 1Mbyte of memory. Option "probe_clocks" This option will force probing of dot clocks on the card. This should not be necessary, since the clocks are fixed and the same for all Cirrus chipsets. Clockchip "cirrus" This enables programmable clocks. It must be specified in the Device section. With this option, the clocks the modes use will be automatically selected. Do not specify any Clocks line. This option makes a 12.5 MHz clock possible for a 320x200 Doublescan mode. Note that some frequencies may be unstable (resulting in a `wavy' screen). Only tried and tested frequencies (like the default clocks) are guaranteed to be stable. Option "linear" (542x/6/8/9/3x/754x on VL-bus) This enables linear addressing, which is the mapping of the entire framebuffer to a high address beyond system memory, so that SVGA bank switching is not necessary. It enhances perfor- mance at 256 colors, and is currently required for 16bpp, 24bpp, and 32bpp. See section 4 for details. Option "nolinear" (542x/6/8/9/3x/754x on PCI bus) Linear addressing is the default mode of operation on any PCI-bus chip. For these configurations, this option disables linear addressing. Membase 0x00e00000 (or a different address) (542x/6/8/9/3x/46/754x) This sets the physical memory base address of the linear frame- buffer. It must be specified in the Device section. It is required for non-PCI linear addressing configurations. Option "favour_bitblt" (5426 only) This option is now obsolete. Option "mmio" (5429, 7548) This enables the use of memory-mapped I/O to talk to the BitBLT engine on the 543x/5429, which is a bit faster. This is option has no effect when not using the BitBLT engine (e.g. when using "no_bitblt"). Option "no_mmio" (543x/4x) This disables the use of memory-mapped I/O to talk to the BitBLT engine on any chip for which it is the default mode of operation. Option "sw_cursor" (542x/3x/46/6x) This disables use of the hardware cursor provided by the chip. Try this if the cursor seems to have problems. In particular, use this when using dot clocks greater than 85 MHz on the 5434/6 since those chips don't fully support the hardware cursor at those clocks. Option "clgd6225_lcd" Provides a work-around for problems on the LCD screen of some 62x5 laptop chipsets with maximum white colors. Option "no_pixmap_cache" When XAA is used (on any BitBLT chip), this option disables the use of a pixmap cache in XAA. It could help with certain drawing bugs. Option "xaa_no_color_exp" When XAA is used, this option disables the use of hardware color expansion features by XAA. Again, this might help with certain drawing bugs. Option "no_stretch" (754x) Disable automatic stretching (horizontal and vertical expansion) of 640x480 on a 800x600 LCD. Option "pci_retry" (546x) Enables a performance feature for PCI based cards. When this feature is enabled, the driver code will attempt to transmit data on the PCI bus as fast as possible. For the most part, this option is safe, but may cause trouble with other PCI devices such as PCI network cards, sound cards, SCSI controllers, etc. When this option is not selected, a safer approach (polling the VGA's command queue) is taken. 4. Mode issues The accelerated 256-color driver uses 16K bytes of scratch space in video memory, and the hardware cursor also uses 1K (2K on the '6X). Consequently, a 1024x1024 virtual resolution should not be used with a 1Mbyte card. On older chips, the use of a higher dot clock frequencies has a negative effect on the performance of graphics operations, especially BitBlt, when little video memory bandwidth is left for drawing (the amount is displayed during start-up for 542x/3x/46/6x chips). For the 542x/3x chips, with default MCLK setting (0x1c) and a 32-bit memory interface, performance with a 65 MHz dot clock can be half of that with a dot clock of 25 MHz. So if you are short on memory bandwidth and experience blitting slowness, try using the lowest dot clock that is acceptable; for example, on a 14" or 15" screen 800x600 with high refresh (50 MHz dot clock) is not so bad, with a large vir- tual screen. 5434-based cards with 2Mbyte of memory do much better at high dot clocks; the DRAM bandwidth is basically double that of the 542x series. The 543x chips also make more efficient use of the available DRAM bandwidth. The same goes for the 544x. 5. Linear addressing and 16bpp/24bpp/32bpp modes Currently the framebuffer code 16-bit, 24-bit, and 32-bit pixels in the SVGA server requires linear addressing. Option "linear" can be specified in a depth-specific screen section to enable linear addressing; a MemBase setting (in the device section) is probably also required (although they are both automatically selected with PCI cards, like 5446, 546x, and some 543x based cards). There are a number of different card configurations. If you have a 542x/543x on the ISA bus, and you have 16Mb or more of system memory, linear addressing is impossible. 16bpp is out, sorry. If you have less than 14Mb of memory, you may be able to map the framebuffer at 14Mb, using `MemBase 0x00e00000'. That's five zeros after the `e'. Unfortunately many ISA cards don't support linear addressing. If you have a 5424/26/28/29 on VESA local bus, the situation is more compli- cated. There are two different types of cards w.r.t. linear addressing: o Cards that can only map in the lower 16Mb, like cards on the ISA bus. This is the case with most cards. The same restrictions apply (i.e. you must have less than 16Mb of memory). o Cards that connect address line A26 and always map at 64Mb + 14Mb or 64Mb. In this case specify `MemBase 0x04e00000' or `MemBase 0x04000000'. This assumes you have a VLB motherboard implementation that implements A26. Alternatively the card may map to 0x2000000, and recent cards like the 5429 usually map to 0x03e00000 (62Mb). You will probably have to rely on trial and error. If you have less than 16Mb memory, the `wrong' membase setting will result in no graphics being dis- played, but you can probably exit with ctrl-alt-backspace. If you have >= 16Mb memory, the first type of card (and even the second type with a stupid VLB motherboard) will result in a crash (probably a spontaneous hard reboot). It may be possible to find out the type by visual inspection. If the card has a pin at A26, it is likely to map beyond 64Mb. To do this, take the card out. At the VESA local bus pins (this is the smaller strip of connector pins at the non-slot side of the card), consider the right side (this is the side of the board where all the chips are mounted). There are 45 pins here. They are numbered 1 to 45, from the inside (i.e. the one nearest to the card end is 45). Counting from the inside, the 17th pin is probably not present, then there are pins at 18-20. The 21st is A30, the 22nd is A28 and the 23rd is A26. So, if we have no pins at at 21-23, the card doesn't map beyond 64Mb. If there's only a gap of two pins at 21 and 22 (or they are both present) and there's a pin at 23, the card does probably map beyond 64Mb. If there's a little logic near that pin on the card, it's more likely. With a 543x on the local bus things are simpler (the Cirrus Logic windows drivers use it), but it is not quite without problems. With a card on the PCI bus, there is a PCI configuration register that holds the framebuffer base address, which is read automatically by the driver if a PCI card is detected. The `scanpci' program can read out the PCI configura- tion and show the base address. On the VESA local bus, most 543x cards have a default mapping address of 64Mb, with jumper options for 2048Mb and 32Mb. This is probably described in the documentation that came with the card, or look in the MS-Windows sys- tem.ini file (something with linearaddr = ). These dif- ferent settings were added by Cirrus Logic after finding that many VLB moth- erboard implementations don't implement different address pins. The driver assumes a default of 64Mb if MemBase isn't specified. A few examples for Mem- Base: MemBase 0x02000000 32Mb MemBase 0x04000000 64Mb MemBase 0x80000000 2048Mb Finally, for 546X cards, you are in luck: there are no "issues" to worry about. The '6X will always use linear addressing and memory-mapped I/O, and will use the memory addresses up near 4GB. Yay for PCI! The 16bpp and 32bpp modes are now fully accelerated, thanks to XAA. On more recent chips like the 5436/46 and the 546X, 24bpp is also fully accelerated. So although there are now up to 4 times as many bits to display, the X server shouldn't feel overly sluggish. Note also that the 24bpp and 32bpp modes are only supported on a limited set of cards, and with at least 2Mb of memory. In the XF86Config "Screen" section, a "Display" subsection must be defined for each depth that you want to run, with separate Modes and virtual screen size. Example (2Mb of video memory): Section "screen" SubSection "Display" Depth 8 Virtual 1280 1024 ViewPort 0 0 Modes "640x480" "800x600" "1024x768" Option "linear" EndSubSection SubSection "Display" Depth 16 Virtual 1024 992 ViewPort 0 0 Modes "640x480" "800x600" "1024x768" Option "linear" EndSubSection SubSection "Display" Depth 32 Virtual 832 600 ViewPort 0 0 Modes "640x480" "800x600" Option "linear" EndSubSection EndSection 6. The ``cl64xx'' Driver The cl64xx driver supports the cl-gd6440 found in many laptops. For example, Nan Tan Computer's NP9200, NP3600, etc., which are OEM-ed by Sager, ProStar, etc. and Texas Instruments TI4000 series are supported. The driver works in LCD-only, CRT-only, and the chip's SimulScan mode which allows one to use both the LCD and external CRT displays simultaneously. The LCD and Simulscan modes' resolution is 640x480 while, for CRT-only, the stan- dard VESA modes of 640x480, 600x800, and 1024x768 have been tested. Inter- laced 1024x768 mode has never been debugged and does not work on the machines tested. The chip has a documented maximum operating limit for its dot clock that is related to its core voltage. Specifically, for 5.0V the maximum dot clock is 65MHz and for 3.3V the maximum dot clock is 40MHz. The driver checks the core voltage and limits the maximum dot clock to the corresponding value. This translates to a maximum resolution of about 1024x768 at a 60Hz refresh rate. The internal frequency generator can be programmed higher than these limits and is done so during server startup when the clocks are probed which momentarily exceeding the chip's operating limit. Once a set of valid clocks is obtained, I would recommend using Clocks lines in XF86Config. Doing so will also decrease startup time significantly. The clocks may be obtained by running the X server -probeonly (see the XFree86 man page for more informa- tion about -probeonly). The data book indicates that only a configuration of one megabyte of video memory is supported by the chip. This size has been directly set in the driver. If one finds a need, one should be able to override the default size in XF86Config. Also, with 1MB of video memory, one should be able to have a virtual screen size of e.g. 1024x1024 and this is possible in CRT-only screen mode. However, whenever the LCD is in use (LCD and SimulScan), the chip uses a portion of upper video ram for its own internal acceleration purposes. Thus, the maximum video memory available for virtual resolution in LCD modes is about 0.75MB e.g. 1024x768. If you set the virtual resolution above this, you will see what might be described as a compressed aliased band when the accelerated area is displayed. Currently, the driver does not support switching of screen modes among LCD, CRT, and SimulScan, and, at least on the NP9200, the mode must be chosen at OS boot time (e.g. Linux's LILO) while the BIOS is still active. It should be possible to add screen mode type selection as a ModeLine flag option in XF86Config to allow for dynamic screen mode selection from within the X server. Finally, the driver does not currently support any of the powerdown saving features of the chip nor does it shut off the LCD's backlight on screen blank. I hope to implement all these features in future releases. Some notes regarding the CL-GD6420: The amount of video memory may not always be detected correctly. The driver source code includes two methods, one defined out. Better specify the amount of video memory with a VideoRam line in the Device section. Use the standard 640x480 60 Hz standard mode timing with 25.175 MHz dot clock for CRT or SIMulscan mode; for LCD-only operation, use the same mode timing but with a dot clock of 16.257 MHz. Standard 56 Hz 800x600 is also supported on the CRT. The primary contact for the cl6440 problems with ``cl64xx'' driver is Randy Hendry . 7. Trouble shooting with the ``cirrus'' driver First of all, make sure that the default modes selected from your XF86Config is supported by your monitor, i.e. make sure the horizontal sync limit is correct. It is best to start with standard 640x480x256 with a 25.175 MHz clock (by specifying a single horizontal sync of 31.5) to make sure the driver works on your configuration. The default mode used will always be the first mode listed in the modes line, with the highest dot clock listed for that resolution in the timing section. Note that some VESA standard mode timings may give problems on some monitors (try increasing the horizontal sync pulse, i.e. the difference between the middle two horizontal timing values, or try multiples of 16 or 32 for all of the horizontal timing parameters). There is a video signal, but the screen doesn't sync. You are using a mode that your monitor cannot handle. If it is a non-standard mode, maybe you need to tweak the timings a bit. If it is a standard mode and frequency that your monitor should be able to handle, try to find different timings for a similar mode and frequency combination. Horizontal jitter at high dot clocks. This problem shows especially when drawing operations such as scrolling are in progress. If you're using a 542x/3x/46/6x/754x, try the "fifo_conservative" option. Failing that, you can try the "fast_dram" option, or use a lower dot clock. If that is not sufficient, the "noaccel" option or "no_bitblt" will probably help. When using a 546x, option "fifo_aggressive" can also be tried. `Wavy' screen. Horizontal waving or jittering of the whole screen, continuously (independent from drawing operations). You are probably using a dot clock that is too high; it is also possible that there is interference with a close MCLK. Try a lower dot clock. You can also try to tweak the mode timings; try increasing the second horizontal value somewhat. Here's a 65 MHz dot clock 1024x768 mode (about 60 Hz) that might help: "1024x768" 65 1024 1116 1228 1328 768 783 789 818 If you are using programmable clocks with Clockchip "cirrus", try disabling it and using the default set of clocks. Crash or hang after start-up (probably with a black screen). Try the "noaccel" option. If that works, try Option "no_bitblt" for somewhat better performance. Check that the BIOS settings are OK; in particular, disable caching of 0xa0000-0xaffff. Disabling hidden DRAM refresh may also help. Crash, hang, or trash on the screen after a graphics operation. This may be related to a bug in one of the accelerated functions, or a problem with the BitBLT engine. Try the "noaccel" option, or the "no_bitblt" option. Also check the BIOS settings. `Blitter timeout' messages from the server. Same as for the above entry. Screen is `wrapped' vertically. (542x/3x/46) This indicates a DRAM configuration problem. If your card has two megabytes of memory, try the "no_2mb_banksel" option, or use videoram "1024" if you only use 1 Mbyte for the virtual screen. Corrupted text in terminal window. This has been reported on non-standard video implementations. Use the "no_bitblt" option. Streaks or hangs with laptop chipset This can happen if the dot clock is high enough to leave very little bandwidth for drawing (e.g. 40 MHz on a 512K card), and (5422-style) acceleration is used. Occasional erroneous pixels in text, pixel dust when moving window-frame Probably related to MCLK setting that is too high (can happen with linear addressing even though banked mode runs OK). Chipset is not detected. Try forcing the chipset to a type that is most similar to what you have. Incorrect little lines (mostly white) appear occasionally This may be related to a problem with system-to-video-memory Bit- BLT operations. Try the "no_imageblt" option if it annoys you. Textmode is not properly restored This has been reported on some configurations. In XFree86 3.1 the SVGA server probe would corrupt a register on the 543x, requiring a Chipset line. Normally you should be able to restore the textmode font using a utility that sets it (setfont, runx, restorefont on Linux). Erratic system behaviour at very high dot clocks It is possible that high dot clocks on the video card interfere with other components in the system (e.g. disk I/O), because of a bad card and/or motherboard design. It has been observed on some PCI 5428-based cards (which are very rare, since the 5428 chip doesn't support PCI). No mouse cursor, or cursor appears twice on screen With high dot clocks, the graphics card's hardware cursor doesn't operate correctly. Try option "sw_cursor" or use a lower screen refresh. Random/garbage pixels on far right or bottom of screen (546x) This problem is usually associated with using a virtual screen size larger than the screen display size. The garbage pixels are unused portions of the frame buffer that result from padding each scanline to an integral number of memory tiles. To eliminate the extra pixels, use a screen display mode whose pixel width is evenly divisible by 128 / bits per pixel. Screen is wrapped horizontally on right side (546x) Same as above entry. The screen is initially displayed correctly, but then turns all white. (546x)" This problem usually happens at high bit depths and while the screen is changing rapidly (catting a long file or dragging a large window around). The RamBus memory is being overdriven. Use Option "med_dram", or, if the problem persists, Option "slow_dram". For other screen drawing related problems, try the "noaccel" option (if "no_bitblt" doesn't help). If are having driver-related problems that are not addressed by this docu- ment, or if you have found bugs in accelerated functions, you can try con- tacting the XFree86 team (the current driver maintainer, Corin Anderson, can be reached at corina@the4cs.com), or post in the Usenet newsgroup "comp.win- dows.x.i386unix". In fact, reports (success or failure) are very welcome, especially on config- urations that have not been tested. You can do this via the BetaReport form (mail to report@XFree86.org). You may want to keep an eye on forthcoming beta releases at www.xfree86.org. 8. Tested Configurations Version 3.3.3 has had the following configurations tested: CL-GD5446 with 2MB memory on PCI bus CL-GD5464 with 2MB memory on PCI bus CL-GD5465 with 4MB memory on PCI bus CL-GD5480 with 4MB memory on PCI bus CL-GD5465 with 4MB memory on AGP bus For version 3.3, the following configurations have received a certain amount of testing: CL-GD5446 with 2MB memory on PCI bus Support for dot clocks > 85 MHz has been fixed. At 16bpp, it has been reported that some stippled edges of window frames may be corrupted or show the wrong colors. The option "xaa_no_pixmap_cache" eliminates the problem. CL-GD5464 with 4MB memory on PCI bus CL-GD7543 on PCI bus This is a list of configurations that has received testing with one or more of the changes introduced in version XFree86 3.2A. The amount of testing is very small for some of the configurations, and the summaries may be incom- plete. If you can contribute, please do so. For the latest information check the latest version of this document on www.xfree86.org. CL-GD5426 on VL-bus This configuration was only tested with an early version of the XAA code. CL-GD5434 with 2MB memory on VL-bus MMIO operation is supported. This configuration was only tested with an early version of the XAA code. CL-GD5436 with 2MB memory on PCI-bus Works OK. Non-MMIO operation might have problems. CL-GD5446 with 2MB memory on PCI bus Works OK in MMIO mode. 32bpp probably doesn't work. The support for dot clocks > 85 MHz at 8bpp may or may not work. CL-GD5462 with 2MB memory on PCI bus CL-GD5462 with 4MB memory on PCI bus CL-GD5464 with 4MB memory on PCI bus Works OK at 8bpp, 16bpp, 24bpp and 32bpp. CL-GD5464 works OK at 16bpp, -weight 555. CL-GD7543 on PCI bus Works for 8bpp, 16bpp on TFT display (TI TravelMate 5000). Although the previous version, 3.2, was reported to broken, on some configurations it worked, while others were reported not to work correctly. On 800x600 displays, the recommended dot clock is 40 MHz for TFT and 33.7 MHz for a DSTN panel, with corresponding horizontal syncs of 33.7 kHz for TFT and 38.6 kHz for DSTN. How- ever, reports indicate that the VESA standard 40 MHz 800x600 tim- ing may cause problems. The solution is decrease the fourth hori- zontal timing number or use a dot clock of 36 MHz. Some configurations for which no up-to-date testing data is available: CL-GD5429 on VL-bus BitBLT operation should be fixed in 3.2. MMIO does not work, but not tested with with 3.2 or 3.2A. CL-GD5430 on PCI-bus Works OK. 24bpp was broken, but should be fixed in later versions (3.2A). CL-GD5430, and CL-GD5436 and CL-GD5446 with 1MB memory It would be nice to know whether these chips needs the same treatment at 16bpp as the CL-GD5434 with 1MB memory does. CL-GD5434 with 1MB memory on PCI bus 8bpp, 16pp and 24bpp work OK. 16bpp no longer has "static" prob- lems. MMIO operation is supported. CL-GD5436 and CL-GD5446 with 1MB memory In particular the FIFO settings for this configuration are uncer- tain. CL-GD7541 CL-GD7548 Should be compatible with 7543, but untested. Reports indicate that it worked with 3.2, and there's no reason why it shouldn't work with 3.2A. 9. Driver Changes Changes since XFree86 3.3.2: o Fix transparent screen-to-screen copies on 546x. o The built-in screen saver now correctly blanks the screen on 546x chips. o Driver prevents the use of the HW cursor on the 546x when the screen height is greater than 1023 scanlines (fix to double pointer problem). o CPU-to-screen BitBLT transfers disabled on the 5465. This fix should prevent 5465 AGP lockups. o Fixed mode display problem with 5480 at high resolutions. Changes since XFree86 3.2A: o A bug that caused a server crash with memory-mapped I/O operation on some chips has been fixed. o Correct handling of dot clocks > 85 MHz on the 5436 and 5446. o Preliminary support for the CL-GD7555 (no detection yet). o Support has been added for the CL-GD5480 and CL-GD5465. o 32bpp mode has been fixed on some Alpine family chips. o Support for dot clocks up to 230 MHz has been added for Laguna family chips. Changes since XFree86 3.2: o Enhanced acceleration using XAA on all chips with a BitBLT engine. o Enhanced acceleration using XAA for the Laguna series (546x). o 24bpp mode on 5430 is fixed. o Improved support for 754x, including support for LCD stretching/center- ing. Generated from XFree86: xc/programs/Xserver/hw/xfree86/doc/sgml/cirrus.sgml,v 3.23.2.6 1998/11/07 13:37:51 dawes Exp $ $XConsortium: cirrus.sgml /main/12 1996/10/28 05:43:32 kaleb $