Re: [Xpert]pre-allocated colormap?
KP> Can DPS handle a non-uniform set of colors? No. The protocol requires a colour cube (not necessarily cubic) and a gray ramp (possibly being the diagonal of the colour cube); in addition, the relevant colourmap indices need to be in arithmetic progression. The current implementation additionally requires that the colour cube be cubic. KP> If you made DPS use Render, these questions would all be moot :-) I'll most definitely move DPS to be above RENDER when I next have some time to work on it (something that hasn't happened in over six months, though). But I'll still use Ghostscript's code for dithering. Juliusz ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 16 o'clock on Sep 23, Juliusz Chroboczek wrote: > The 9 + 4 x 4 x 4 scheme gives excellent results (to my eyes) while > leaving a lot of colour cells free. I was suggesting that you should > try implementing a similar colourmap for RENDER, and see what the > results look like. I think a 4x4x4 cube has too many dark entries; I'd rather weight the values so that it uses more light entries instead. Can DPS handle a non-uniform set of colors? Render is happy with essentially any set, but I suspect DPS will want to use relatively simplistic dithering... > Just like you, I consider 8 bit colour displays to be obsolete, and I > am not interested in putting much effort into making DPS work well on > those. (8 and even 4 bit grayscale displays are quite nice, on the > other hand.) If you made DPS use Render, these questions would all be moot :-) [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
>> On an eight-bit visual, XDPS will allocate a 9-level gray ramp, and a >> 4x4x4 colour cube, for a total of 73 colourmap entries. (Colour >> allocation happens on the client side in XDPS.) >> Can RENDER deal with such a configuration? KP> RENDER must have the colors allocated within the server; there's no KP> provision for the extension to use a client-provided colormap. I understand that. I was asking whether RENDER can deal with a gray ramp of a different size than the diagonal of the colour cube. The 9 + 4 x 4 x 4 scheme gives excellent results (to my eyes) while leaving a lot of colour cells free. I was suggesting that you should try implementing a similar colourmap for RENDER, and see what the results look like. > if you happened to create a colormap compatible with DPS for Render Just like you, I consider 8 bit colour displays to be obsolete, and I am not interested in putting much effort into making DPS work well on those. (8 and even 4 bit grayscale displays are quite nice, on the other hand.) Juliusz ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 19 o'clock on Sep 22, Juliusz Chroboczek wrote: > On an eight-bit visual, XDPS will allocate a 9-level gray ramp, and a > 4x4x4 colour cube, for a total of 73 colourmap entries. (Colour > allocation happens on the client side in XDPS.) > Can RENDER deal with such a configuration? RENDER must have the colors allocated within the server; there's no provision for the extension to use a client-provided colormap. That being the case, the current render code is happy to use any colormap whatsoever, so if you happened to create a colormap compatible with DPS for Render, it will use it and DPS can share. -keith ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On an eight-bit visual, XDPS will allocate a 9-level gray ramp, and a 4x4x4 colour cube, for a total of 73 colourmap entries. (Colour allocation happens on the client side in XDPS.) Can RENDER deal with such a configuration? Juliusz ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sun, 16 Sep 2001 [EMAIL PROTECTED] wrote: > 3) do we have any idea how many times things fail if static color is > made the default? I've certainly occasionally run into such servers, > but not recently, and therefore don't have a feel as to how often things break. The legacy PseudoColor apps will break. Most of the time they don't do any palette tricks, they just allocate cells instead of colors. If they just called XAllocColor they'd have worked with all visuals. If the default is StaticColor they probably won't work at any depth anymore. I suspect there are another class of applications which try to allocate cells only if they see a depth 8 visual. Those would have previously worked in all depths, but may now break in depth 8. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
1) since monitors vary, I think some to match colors into an existing preallocated cube is OK. Presumably we can figure out how much slop is tolerable. I don't think rws would be too upset (nor am I), if we can come up with a defensible description of what we're doing. It is unpredictable action that seems really wrong to me. 2) the size of the cube allocated should probably be settable when the server starts. We've lived with similar situations with Netscape in the past, and I think we'll survive in the future this way. 3) do we have any idea how many times things fail if static color is made the default? I've certainly occasionally run into such servers, but not recently, and therefore don't have a feel as to how often things break. 4) presumably, we could make it optional how much slop to tolerate; this interacts with the amount of space left unallocated in some fashion. Some amount of unallocated space is needed for antique apps that use writable entries to run at all without flashing. - Jim -- Jim Gettys Cambridge Research Laboratory Compaq Computer Corporation [EMAIL PROTECTED] ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Sidik, I know the image display hackery done for astronomy well: I used to write this code myself (at SAO, in the OIR group, for some of the early CCD systems) before I started hacking window systems. So I can comment on this on both sides of the issue: as window system designer and former astronomy hacker. A large part of why pseudocolor exists in X was to support such applications on the 1 mip class machines of the late 1980's, where displays were often over 2 megabyte/second busses (with a tailwind). But that is now over a decade in the past. We knew then (or at least I did) that pseudocolor was evil and would eventually go away, but accepted it as certain apps, such as those used by astronomers and for other imaging applications, were not possible without it on that class hardware. Thankfully, that hardware is now hardware of the (long) past Any computer system with PCI class display hardware is fast enough to emulate colormap tricks in completely in software that should be adequate for interactive, real time contrast enhancement/display trickery completely on the client side, particularly using the shared memory extension. I also know that most astronomers have access to most of their source code, and it is, I believe, the right thing in your case to rewrite the app, since hardware even capable of pseudocolor at all is becoming hard to come by. This doesn't handle the case of what to do when source is not available. I'll comment on the general discussions separately. - Jim -- Jim Gettys Cambridge Research Laboratory Compaq Computer Corporation [EMAIL PROTECTED] ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 16 o'clock on Sep 15, Mark Vojkovich wrote: >Like by temporarily lowering the sigBits? Changing current allocation > behavior makes me uneasy. Doesn't the specification say something > about this? AllocColor: "This request allocates a read-only colormap entry corresponding to the closest RGB values provided by the hardware. It also returns the pixel and the RGB values actually used. Multiple clients requesting the same effective RGB values can be assigned the same read-only entry, allowing entries to be shared." This has always been interpreted to say that PseudoColor allocations return the best possible color and fail if no exactly matching cell is available. And, I'm sure rws would concur that this was the intent. However, we might consider stretching this to allow matching of existing entries that were not precisely the same as the requested value as a way to avoid application failure or flashing. This is what StaticColor visuals already do, so it's not as if we'd be breaking new ground here. The server could allocate a set of fixed colors that would be used to match all read-only requests; read-write requests would use the remainder of the colormap. That seems like it would give applications the benefits of PseudoColor visuals while forcing other applications to share the available read-only cells better than they do today. It would be easy to do; just make read-only allocations from clients match existing read-only cells in the colormap, then let the Render extension build the read-only cells and you're all set. > Besides the quality will drop noticeable. Yes, applications allocating read-only colors will see a significant reduction in fidelity. Those same applications usually have an option to create their own private colormap, in which case they'll have complete access to the available hardware capabilities. > How about a config file option for the size of Render's color cube? That's my plan; just specify the maximum number of colors available to Render in the config file and make it suffer. When presented with 0 colors, it should disable itself. [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sat, 15 Sep 2001, Keith Packard wrote: > What do others think about making read-only PseudoColor allocations a bit > less strict and permitting them to match nearby entries with not exactly > the same RGB values? That would reduce color fidelity while also avoiding > application failure in many cases. > Like by temporarily lowering the sigBits? Changing current allocation behavior makes me uneasy. Doesn't the specification say something about this? I would have thought that to be illegal. Besides the quality will drop noticeable. I used to run in 8 bit most of the time on my 2 Meg cards. There was a big difference between 6 and 8 sigBits in xv and Netscape. How about a config file option for the size of Render's color cube? Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 15 o'clock on Sep 15, Mark Vojkovich wrote: > 3) No preallocated colors by default. > > It might be reasonable to assume users stuck in PseudoColor are there > because they have antiquated hardware, in which case, they don't have > very high expectations for things like anti-aliasing. Either that > or they are in PseudoColor for a the palettes, in which case, they > might resent preallocations of any type. That's equivalent to disabling the Render extension, which should probably be an option. As Render operates strictly with RGBA values, not providing a mapping for pixels on the screen means there's no way to draw anything. I'd argue that this is not a reasonable default though, I hope that most applications will be able to count on the existance of the Render extension in the future and avoid significant client-side complexity as a result. The question then is what the right default is; I think Render should be enabled, but probably with fewer colors than it consumes right now. What do others think about making read-only PseudoColor allocations a bit less strict and permitting them to match nearby entries with not exactly the same RGB values? That would reduce color fidelity while also avoiding application failure in many cases. [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sat, 15 Sep 2001, Keith Packard wrote: > I see two easy solutions for this problem: > > 1) Have Render use fewer colors from the default colormap > 2) Have XFree86 promote StaticColor as the usual default visual 3) No preallocated colors by default. It might be reasonable to assume users stuck in PseudoColor are there because they have antiquated hardware, in which case, they don't have very high expectations for things like anti-aliasing. Either that or they are in PseudoColor for a the palettes, in which case, they might resent preallocations of any type. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sat, 15 Sep 2001, Sidik Isani wrote: > Keith Packard wrote: > | > | > |Around 10 o'clock on Sep 15, Sidik Isani wrote: > | > |> Unfortunately, there is. We look at astronomy CCD images, and > |> the visualization tools like to manipulate a colormap to give a > |> quick contrast adjustment. > | > |If you have source, you could consider fixing these to use DirectColor > |instead; that would permit adjusting the luminence curves on the fly. > |But, I suspect you also like to switch colors around, which this would not > |do. > > It may be enough. Usually we look at our images with simple > grey scales, and just need a way to exaggerate contrast to bring > out features. I'd like to learn more about DirectColor. How much > control does it give over luminence curves? Which XFree86 drivers > support it (at depths > 8bpp)? Most drivers support this for depths > 8bpp. NVIDIA and Matrox at least, probably alot of others. TrueColor is essentially three palette lookups with the indicies for the 3 channels (Red, Green and Blue) being extracted from the pixel according to the RGB masks. The difference between TrueColor and DirectColor is that you can change the lookup tables with DirectColor. Still, there is only one hardware palette so you get flashing when changing focus between TrueColor and a DirectColor window. If you just want to change the contrast a little this is not so annoying. Switching colors around or heavy thresholding, while possible with DirectColor, will suffer the same flashing consequences as private PseudoColor maps. > > |I assume these tools are creating their own colormaps; if this is the > |case, you might consider using StaticColor for the default visual, this > |would avoid flashing among the other windows while still permitting the > |display to run at 8 bits. > > Thanks. I'll give that a try too. But of course, the StaticColor palette is a simple color cube and might not give you the range of colors you need. Certainly not a good number of greys. And in the overlay mode only PseudoColor and GrayScale visuals are offered, not StaticColor. That is because the transparency key steals a color and only in the read/write maps can the key be preallocated. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 11 o'clock on Sep 15, Sidik Isani wrote: > It may be enough. Usually we look at our images with simple > grey scales, and just need a way to exaggerate contrast to bring > out features. I'd like to learn more about DirectColor. How much > control does it give over luminence curves? Which XFree86 drivers > support it (at depths > 8bpp)? DirectColor provides a LUT for each component -- the 256 possible values for each component are mapped to any of 256 possible inputs to the DAC. Traditionally, this is used to compensate for the monitor response curve while exposing linear intensity to applications. Most PC graphics hardware supports this mode; although the i128 driver doesn't currently expose it. I don't know if this is a hardware limitation or just missing code in the driver though. The apm, ati, mga, nv, s3, savage drivers look like they do support DirectColor; I don't know about the others. > Would "x2x" plus 2 screen layouts do the trick? I don't know; I'm not very familiar with x2x. > That sounds like a reasonable solution. Do any popular programs > require Render? Can I just leave this extension out completely > for this application? As Render is still new, most applications can work without it. It provides for anti-aliased graphics and RGBA image compositing, and on your digital monitors performs sub-pixel optimized rendering to make text really sharp looking. I'll see about adding a configuration option to tune the number of pixels used; it will work with only black & white which are always available. [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Keith Packard wrote: | | |Around 10 o'clock on Sep 15, Sidik Isani wrote: | |> Unfortunately, there is. We look at astronomy CCD images, and |> the visualization tools like to manipulate a colormap to give a |> quick contrast adjustment. | |If you have source, you could consider fixing these to use DirectColor |instead; that would permit adjusting the luminence curves on the fly. |But, I suspect you also like to switch colors around, which this would not |do. It may be enough. Usually we look at our images with simple grey scales, and just need a way to exaggerate contrast to bring out features. I'd like to learn more about DirectColor. How much control does it give over luminence curves? Which XFree86 drivers support it (at depths > 8bpp)? |I assume these tools are creating their own colormaps; if this is the |case, you might consider using StaticColor for the default visual, this |would avoid flashing among the other windows while still permitting the |display to run at 8 bits. Thanks. I'll give that a try too. |> Since we had some Matrox cards lying around, |> I gave the Overlay feature a try and our applications work well with that. |> We'd have to order some PCI versions to be able to plug three in, and we'd |> have to use multi-link converters on each monitor. | |Analog signals to your digital monitors. Yuck. Indeed. But Matrox has a digital version (or add-on?) to the card. It would still require the use of the multi-link to go from one digital interface to another though. Yuck. |> If we run only one screen in Overlay mode and exclude it from the |> Xinerama desktop, I guess that could work, but we'd prefer one |> logical screen, of course, even if dragging a pseudocolor window |> out of the Overlay'd screen made it appear black/garbage on the |> other monitors which don't support Overlay. | |Xinerama exposes the intersection of the screen capabilities, not the |union. It would be a significant undertaking to change that. It also |doesn't allow you to split things so that some screens are xinerama'd and |others are not. Would "x2x" plus 2 screen layouts do the trick? |In the long term, however, the RandR extension will allow unsupported |visual classes to be emulated in software. This, in conjunction with |Xinerama, would allow you to use one card with hardware overlays while |emulating the overlay in software on the other two cards. I'm still |struggling over how to make this work with the existing XFree86 |architecture though; don't expect news on this front anytime soon. Ok, good to know though. |I think that for now, you should whack the number of colors reserved for |the Render extension; eventually that will probably make it into the |configuration file. That sounds like a reasonable solution. Do any popular programs require Render? Can I just leave this extension out completely for this application? Thanks, - Sidik ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 10 o'clock on Sep 15, Sidik Isani wrote: > Unfortunately, there is. We look at astronomy CCD images, and > the visualization tools like to manipulate a colormap to give a > quick contrast adjustment. If you have source, you could consider fixing these to use DirectColor instead; that would permit adjusting the luminence curves on the fly. But, I suspect you also like to switch colors around, which this would not do. I assume these tools are creating their own colormaps; if this is the case, you might consider using StaticColor for the default visual, this would avoid flashing among the other windows while still permitting the display to run at 8 bits. > Since we had some Matrox cards lying around, > I gave the Overlay feature a try and our applications work well with that. > We'd have to order some PCI versions to be able to plug three in, and we'd > have to use multi-link converters on each monitor. Analog signals to your digital monitors. Yuck. > If we run only one screen in Overlay mode and exclude it from the > Xinerama desktop, I guess that could work, but we'd prefer one > logical screen, of course, even if dragging a pseudocolor window > out of the Overlay'd screen made it appear black/garbage on the > other monitors which don't support Overlay. Xinerama exposes the intersection of the screen capabilities, not the union. It would be a significant undertaking to change that. It also doesn't allow you to split things so that some screens are xinerama'd and others are not. In the long term, however, the RandR extension will allow unsupported visual classes to be emulated in software. This, in conjunction with Xinerama, would allow you to use one card with hardware overlays while emulating the overlay in software on the other two cards. I'm still struggling over how to make this work with the existing XFree86 architecture though; don't expect news on this front anytime soon. I think that for now, you should whack the number of colors reserved for the Render extension; eventually that will probably make it into the configuration file. [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
|Around 9 o'clock on Sep 15, Sidik Isani wrote: | ||> With 8 bpp pseudocolor, I'm finding that simply starting the ||> X-Server allocates almost all the color cells in the default ||> colormap. There are only about 12 free cells, and most programs ||> either fail or install a private colormap. Is there an extension ||> which may be doing this? | |Yes, the Render extension allocates most of your colormap now. I was |planning on making the number it consumed configurable but I got |sidetracked by other work temporarily. If you build your own X server, |you can easily hack this number in the source; look at | | programs/Xserver/render/miindex.c:miInitIndexed | |What do others think we should do about this? One possibility would be to |bend the protocol and match incoming requests to this palette when the |colormap was otherwise filled; that would eliminate color flashing while |still providing reasonable color matching. The protocol is rather vague |on when this kind of sharing will occur; perhaps some less stringent |metric could be applied that allowed more distant colors to be used. | |The other option is to use StaticColor as your default visual; read-only |allocations will always succeed. Is there some reason you can't run your |display at 16 or 24 bits? Hello Keith - Unfortunately, there is. We look at astronomy CCD images, and the visualization tools like to manipulate a colormap to give a quick contrast adjustment. Many will not even run without a pseudocolor visual. We are using 3 1600SW's for the task, and it is a real shame to run them at 8-bits! Since we had some Matrox cards lying around, I gave the Overlay feature a try and our applications work well with that. We'd have to order some PCI versions to be able to plug three in, and we'd have to use multi-link converters on each monitor. (The power strip with all the AC adapters plugged in for this will become a fire hazard. ;-) Andrew Aitchison just posted that the Oxygen VX1-1600SW also has the Overlay feature (thanks!) so an interesting, but pricey solution for us might be to buy six of those (since we have two workstations which we'd like to keep the same). This would allow direct digital connections to our 1600SWs, like we have now with the i128. If we run only one screen in Overlay mode and exclude it from the Xinerama desktop, I guess that could work, but we'd prefer one logical screen, of course, even if dragging a pseudocolor window out of the Overlay'd screen made it appear black/garbage on the other monitors which don't support Overlay. I know it's a hack (but one which saves us about $1400) ... is it possible? Thanks for the help, - Sidik ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 12 o'clock on Sep 15, Mark Vojkovich wrote: >Can render allocate the colors as needed and use whatever it > gets? This would be more consistent with traditional PsuedoColor > behavior. That is, first come first serve and if you don't like that > use a private map. Hmm. Not with the current protocol. Nor do I think it possible in general. As you know, pixels don't have RGB values in the core protocol. Render imposes RGB values through PictFormats. For indexed PictFormats, it provides for a static pixel<->RGB mapping. For each indexed visual (i.e. those other than Direct/True), Render creates an indexed PictFormat and initializes the static mapping by allocating colors from an appropriate colormap. For the default visual, it uses the default colormap. For non-default visuals, it creates a private colormap that can be shared among all windows using that visual. As any RGB value can be generated as a result of an image composition operation, Render needs to be able to select an appropriate pixel value with the assurance that it will be available. Attempting to do this dynamically would require windows to switch colormaps on the fly -- if Render was unable to allocate the desired color, it would have little choice but to force the window to use a different colormap. There isn't any provision to update the mapping dynamically; the data are provided to applications without any asychronous notification of change. Given the ever-shrinking number of pseudo-color displays, I felt it was more important that applications developed on TrueColor hardware work without error on PseudoColor than that the PseudoColor hardware be used to its best advantage. We've learned the hard way that developers make incautious assumptions about how X visuals work; many PseudoColor applications still fail on TrueColor displays. I was attempting to avoid the reverse problem when designing this new extension. The extension could be changed to allocate the color cube when the first application requested information about it's contents; that would make environments not using the Render extension work just like they did before. However, I think that's a band-aid solution; users would have to avoid most modern applications (all of KDE and Gnome) to make this work. I see two easy solutions for this problem: 1) Have Render use fewer colors from the default colormap 2) Have XFree86 promote StaticColor as the usual default visual The second would eliminate flashing for the vast bulk of applications while causing strange application failures where assumptions about the default visual class were made. The first would provide for Render applications to run at reduced color fidelity, that may be acceptable for people continuing to use older hardware incapable of running deeper than 8 bits. [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sat, 15 Sep 2001, Keith Packard wrote: > > Around 9 o'clock on Sep 15, Sidik Isani wrote: > > |> With 8 bpp pseudocolor, I'm finding that simply starting the > |> X-Server allocates almost all the color cells in the default > |> colormap. There are only about 12 free cells, and most programs > |> either fail or install a private colormap. Is there an extension > |> which may be doing this? > > Yes, the Render extension allocates most of your colormap now. I was > planning on making the number it consumed configurable but I got > sidetracked by other work temporarily. If you build your own X server, > you can easily hack this number in the source; look at > > programs/Xserver/render/miindex.c:miInitIndexed > > What do others think we should do about this? One possibility would be to > bend the protocol and match incoming requests to this palette when the > colormap was otherwise filled; that would eliminate color flashing while > still providing reasonable color matching. The protocol is rather vague > on when this kind of sharing will occur; perhaps some less stringent > metric could be applied that allowed more distant colors to be used. Can render allocate the colors as needed and use whatever it gets? This would be more consistent with traditional PsuedoColor behavior. That is, first come first serve and if you don't like that use a private map. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Sat, 15 Sep 2001, Sidik Isani wrote: > Mark Vojkovich wrote: > |On Fri, 14 Sep 2001, Sidik Isani wrote: > | > |> Hello - > |> > |> With 8 bpp pseudocolor, I'm finding that simply starting the > |> X-Server allocates almost all the color cells in the default > |> colormap. There are only about 12 free cells, and most programs > |> either fail or install a private colormap. Is there an extension > |> which may be doing this? > | > | Not that I know of. Run xcmap to visualize the colormap. > |When the server starts up you will get approximately 2 colors > |(black and white). I don't think xterm should be allocating > |very much of the palette. If you find differently, this is > |something that should be addressed. > > Hello Mark - > > 2 colors (black and white) is indeed what I get when I run 4.1.0. > But if I swap that server out for the CVS one (which I need in order > to have Xinerama working on the i128 driver) I get black, white, and > exactly 242 colors in a standard-looking colormap pattern. > > I wonder if the fact that I've compiled my X-Server without modules > is causing some extension to initialize itself? I tried again > without BuildCup, but same problem. Is there a call, such as > AllocColor(), to which I could add some debugging to track this > down? > Hmmm. This appears to be caused by the Render extension. Have PictureInitIndexedFormats() in xc/programs/Xserver/render/miindex.c return immediately without doing anything to see if that fixes it. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Around 9 o'clock on Sep 15, Sidik Isani wrote: |> With 8 bpp pseudocolor, I'm finding that simply starting the |> X-Server allocates almost all the color cells in the default |> colormap. There are only about 12 free cells, and most programs |> either fail or install a private colormap. Is there an extension |> which may be doing this? Yes, the Render extension allocates most of your colormap now. I was planning on making the number it consumed configurable but I got sidetracked by other work temporarily. If you build your own X server, you can easily hack this number in the source; look at programs/Xserver/render/miindex.c:miInitIndexed What do others think we should do about this? One possibility would be to bend the protocol and match incoming requests to this palette when the colormap was otherwise filled; that would eliminate color flashing while still providing reasonable color matching. The protocol is rather vague on when this kind of sharing will occur; perhaps some less stringent metric could be applied that allowed more distant colors to be used. The other option is to use StaticColor as your default visual; read-only allocations will always succeed. Is there some reason you can't run your display at 16 or 24 bits? [EMAIL PROTECTED]XFree86 Core Team SuSE, Inc. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
Mark Vojkovich wrote: |On Fri, 14 Sep 2001, Sidik Isani wrote: | |> Hello - |> |> With 8 bpp pseudocolor, I'm finding that simply starting the |> X-Server allocates almost all the color cells in the default |> colormap. There are only about 12 free cells, and most programs |> either fail or install a private colormap. Is there an extension |> which may be doing this? | | Not that I know of. Run xcmap to visualize the colormap. |When the server starts up you will get approximately 2 colors |(black and white). I don't think xterm should be allocating |very much of the palette. If you find differently, this is |something that should be addressed. Hello Mark - 2 colors (black and white) is indeed what I get when I run 4.1.0. But if I swap that server out for the CVS one (which I need in order to have Xinerama working on the i128 driver) I get black, white, and exactly 242 colors in a standard-looking colormap pattern. I wonder if the fact that I've compiled my X-Server without modules is causing some extension to initialize itself? I tried again without BuildCup, but same problem. Is there a call, such as AllocColor(), to which I could add some debugging to track this down? Thanks, - Sidik ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Fri, 14 Sep 2001, Sidik Isani wrote: > Hello - > > With 8 bpp pseudocolor, I'm finding that simply starting the > X-Server allocates almost all the color cells in the default > colormap. There are only about 12 free cells, and most programs > either fail or install a private colormap. Is there an extension > which may be doing this? Not that I know of. Run xcmap to visualize the colormap. When the server starts up you will get approximately 2 colors (black and white). I don't think xterm should be allocating very much of the palette. If you find differently, this is something that should be addressed. Mark. ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]pre-allocated colormap?
On Fri, 14 Sep 2001, Sidik Isani wrote: > P.S.: Other than mga, does anything else support "Overlay"? The glint driver supports overlay on appropriate 3Dlabs cards (Oxygen VX1-1600SW is the only one I tried, and it works). *Some* S3 cards have suitable hardware (TVP3026 or RGB526 ramdacs, possibly other members of those chip families); I don't think that the drivers support the feature yet. > What happens if I include one mga screen with "overlay" > in a xinerama configuration where the other screens are > different cards not supporting overlay? xinerama only exports features available on all screens, so you would lose overlay in this case. -- Dr. Andrew C. Aitchison Computer Officer, DPMMS, Cambridge [EMAIL PROTECTED] http://www.dpmms.cam.ac.uk/~werdna ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
[Xpert]pre-allocated colormap?
Hello - With 8 bpp pseudocolor, I'm finding that simply starting the X-Server allocates almost all the color cells in the default colormap. There are only about 12 free cells, and most programs either fail or install a private colormap. Is there an extension which may be doing this? This also happens when starting the server with: xinit /usr/X11R6/bin/xterm -- :0 I'm running XFree4 and the i128 driver from yesterday's CVS. Thanks for your help, - Sidik P.S.: Other than mga, does anything else support "Overlay"? What happens if I include one mga screen with "overlay" in a xinerama configuration where the other screens are different cards not supporting overlay? ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert