SMedia 3362 (Again... sorry)
In a previous e-mail it was confirmed that the driver for the SMedia accelerator found in the GTA02 will take the form of a KDrive driver. (Correct?) Given that the chip is OpenGL ES 1.2 compliant and the SMedia datasheet (At least for the 3370, no datasheet for the 3362) claim Embedded Linux software support, are we likely to see an OpenGL ES library for the GTA02? I realise OpenMoko is about open source, but if a closed source library is available, surely that's better than nothing, at least until open source drivers/libraries can be written? I think OpenGL ES support for the neo would be fantastic - The hardware's there, why not use it? Writing some clutter (http://www.clutter-project.org) based applications would be great fun! Cheers, Tom ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
On Mon, Sep 10, 2007 at 12:10:48PM -0700, Shawn Rutledge wrote: On 9/10/07, Harald Welte [EMAIL PROTECTED] wrote: So what exactly is not enough? You will get 100% free software drivers, down to the latest bit, no proprietary firmware whatsoever, plus hardware documentation that will be prepared by OpenMoko ? I didn't understand before that it would end up being 100% open source. We always said it would be entirely open source. Anything else is not acceptable! I thought you had to sign an NDA? But if the driver is to be completely open, and the documents are to be completely open too, what is the purpose of the NDA? The purpose of the NDA is to prevent OpenMoko from releasing the WinCE driver source code and/or the original documentation (or portions thereof), at least not without explicit approval from Smedia. The new free driver based on that infromation is explicitly excluded. It always depends on the exact wording of the NDA, what informaiton is not supposed to be disclosed, etc.! No, getting it working is the more important of the two, of course. But as someone else pointed out, maybe there is the possibility that someone outside your small team could help, if the NDA doesn't prevent it. no, that's impossible. What kind of driver are you planning on? (I don't think I saw that answered yet, sorry if I missed it) KDrive, DRI, etc... We don't disclose this information yet, sorry. As soon as there is something working, it will be in our subversion, though. Well the choices for open hardware are always slim, so far. I just thought some people might be putting more emphasis on OpenGL ES support than openness. To us, the openness is alwayts the primary issue at stake. -- - Harald Welte [EMAIL PROTECTED] http://openmoko.org/ Software for the world's first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
On Fri, Sep 07, 2007 at 10:23:26AM -0700, Shawn Rutledge wrote: On 9/6/07, Steve [EMAIL PROTECTED] wrote: It may be worth talking with ATI again. Since this announcement, I don't think it is too far fetched to at least get the same deal you currently have with SMedia. I wouldn't be surprised if the people you were talking to had no idea this sort of thing was being planned. ATI may even allow the release documentation at some point in the future. Or just use it for leverage to get more from SMedia. guys, as I indicated before, we already have the best possible support from Smedia. Not only have we some promises or statements, but we actually have signed a contract with them, binding them to support us to the utmost level. As indicated previously, this agreement includes a statement that OpenMoko will work on to-be-publicized documentation on the SMedia chip, which will be jointly released at some point. So what exactly is not enough? You will get 100% free software drivers, down to the latest bit, no proprietary firmware whatsoever, plus hardware documentation that will be prepared by OpenMoko ? Which part exactly are you missing? That there are no docs now? Well, there is no GTA02 hardware being shipped now either! And if the community rather wants us to finish the documentation first, and then write the driver: Please let us know. Do you really prefer to get a device that does not have any working driver at all, but with a thousand-page manual (rather than the other way around: first have FOSS Drivers, and then get the docs as soon as our incredibly small team finds time to do so)? Wrt: ATI/AMD Imageon: ATI's mobile processor diivsion is completely independent from their desktop graphics. It has totally different architecture, and the recent announcement by their desktop group doesn't have any maning about the mobile group. Also, ATI's mobile graphics are entirely focused on 2d and codec, plus they are 100% firmware based. So that means no 3D acceleration, and even if somebody ever was to write FOSS drivers, lots of code is hidden in the GPU firmware, rather than in those FOSS drivers. What I personally don't understand about this entire debate on our community list: You have very prominent people of the FOSS movement, particularly the Linux community in this project. Notably Werner and myself. Given my track history of clinging to every last word of the GPL, and my stance with regard to binary-only drivers or other abominations of the hardware industry: Why don't you trust us to do proper research and chose the vendor that works best for us, given all the circumstances? Do you think we would be foolish enough not to talk to all vendors of the respective components? I really feel personally very sad that anyone believes that I am in this project for anything else then to provide the highest level of freedom for both hardware and software that is possible. In GTA01, the only freedom related issue that we have is the Global Locate (now Broadcom). Given the start of OpenMoko (alternative software for a Windows smartphone that FIC was building) we didn't have any influence on that one. We have been trying hard to achieve a compromise with GL on the level of freedom that they're willing to provide. Unfortunately that compromise falls short of what many people in the FOSS community, including myself, deem acceptable. For GTA02, we evaluated all different A-GPS solutions on the market, and we took two of those actually in production. The graphics chip we ship will have FOSS drivers. We're working with NXP on publishing an open user manual for the PCF50633 PMU, and we already have their approval for it. We're staying with the publicly documented samsung s3c24xx CPU series. We use accelerometers with publicly available data sheets. We use a bluetooth chip with open data sheet. We use a WiFi module with GPL licensed free software driver. There is no other hardware vendor of devices with similar high level of integration that has taken openness to the degree that we are taking it. Starting with GTA02, we have a very firm openness policy for all our hardware components. Our future designs will follow the same line - and we're trying to continuously to push the borders any further. We make our position at chip manufacturers very clear. And we're having very fruitful discussions and results that I am proud of. -- - Harald Welte [EMAIL PROTECTED] http://openmoko.org/ Software for the world's first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
ma, 2007-09-10 kello 15:20 +0800, Harald Welte kirjoitti: guys, as I indicated before, we already have the best possible support from Smedia. Not only have we some promises or statements, but we actually have signed a contract with them, binding them to support us to the utmost level. As indicated previously, this agreement includes a statement that OpenMoko will work on to-be-publicized documentation on the SMedia chip, which will be jointly released at some point. This to-be-publicized documentation bit is something that I at least haven't noticed anywhere yet, and makes the deal better than I thought (and note that I've been pro-SMedia choice already, as long as you get to publish free drivers). I (as I think probably many others who've commented on the suboptimality of the deal) thought the free drivers would have to speak for themselves as for documentation to the wider audience. Do you really prefer to get a device that does not have any working driver at all, but with a thousand-page manual (rather than the other way around: first have FOSS Drivers, and then get the docs as soon as our incredibly small team finds time to do so)? Please do the driver first :] As said, I suspect people at this point just didn't know/realize that docs _are_ in the pipeline as well. What I personally don't understand about this entire debate on our community list: You have very prominent people of the FOSS movement, particularly the Linux community in this project. Notably Werner and myself. Given my track history of clinging to every last word of the GPL, and my stance with regard to binary-only drivers or other abominations of the hardware industry: Why don't you trust us to do proper research and chose the vendor that works best for us, given all the circumstances? Indeed, I can sympathize with this point. I think you're trustworthy guys especially as to providing the most freedom that you can, and I can see how it can become frustrating that every call you make is heavily questioned - from that spesific viewpoint where you should by all rights have the most credibility! So, please, members of the community, have a little confidence in the OM team. In GTA01, the only freedom related issue that we have is the Global Locate (now Broadcom). Given the start of OpenMoko (alternative software for a Windows smartphone that FIC was building) we didn't have any influence on that one. Good that you mentioned this, because I think this bit likely hasn't gotten the most publicity either. Starting with GTA02, we have a very firm openness policy for all our hardware components. Our future designs will follow the same line - and we're trying to continuously to push the borders any further. We make our position at chip manufacturers very clear. And we're having very fruitful discussions and results that I am proud of. And well you should. -- Mikko Rauhala - [EMAIL PROTECTED] - URL:http://www.iki.fi/mjr/ Transhumanist - WTA member - URL:http://www.transhumanism.org/ Singularitarian - SIAI supporter - URL:http://www.singinst.org/ ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
On Monday 10 September 2007 09:20:54 Harald Welte wrote: So what exactly is not enough? You will get 100% free software drivers, down to the latest bit, no proprietary firmware whatsoever, plus hardware documentation that will be prepared by OpenMoko ? Which part exactly are you missing? Some people seem to bemoan the lack of h.264 decoding. Personally, I'm not very fussed about it (watching video on 2.8 isn't much fun in my view)... Personally I far prefer true open source drivers to binary only that will have one more feature but then stop working at will (which occurs rather frequently with the currently shipping ATI fglrx drivers, BTW). signature.asc Description: This is a digitally signed message part. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
On 9/10/07, Harald Welte [EMAIL PROTECTED] wrote: On Fri, Sep 07, 2007 at 10:23:26AM -0700, Shawn Rutledge wrote: On 9/6/07, Steve [EMAIL PROTECTED] wrote: It may be worth talking with ATI again. Since this announcement, I don't think it is too far fetched to at least get the same deal you currently have with SMedia. I wouldn't be surprised if the people you were talking to had no idea this sort of thing was being planned. ATI may even allow the release documentation at some point in the future. Or just use it for leverage to get more from SMedia. guys, as I indicated before, we already have the best possible support from Smedia. Not only have we some promises or statements, but we actually have signed a contract with them, binding them to support us to the utmost level. This is excellent! As indicated previously, this agreement includes a statement that OpenMoko will work on to-be-publicized documentation on the SMedia chip, which will be jointly released at some point. First, I believe we need a working implementation, then a (very) rough draft of the docs released in the near future for the community. The docs don't have to be stable per se, if we can gain access sooner. But this all depends on the agreement between OM/SMedia. SMedia, I give you guys a lot of credit for taking a stand and provding open documentation. So what exactly is not enough? You will get 100% free software drivers, down to the latest bit, no proprietary firmware whatsoever, plus hardware documentation that will be prepared by OpenMoko ? did I... read that correctly? :) Which part exactly are you missing? That there are no docs now? Well, there is no GTA02 hardware being shipped now either! And if the community rather wants us to finish the documentation first, and then write the driver: Please let us know. Do you really prefer to get a device that does not have any working driver at all, but with a thousand-page manual (rather than the other way around: first have FOSS Drivers, and then get the docs as soon as our incredibly small team finds time to do so)? I think the best scenario would be to get the (simple/barebones?) driver/framework done, and release alpha docs if the situation allows. Wrt: ATI/AMD Imageon: ATI's mobile processor diivsion is completely independent from their desktop graphics. It has totally different architecture, and the recent announcement by their desktop group doesn't have any maning about the mobile group. Also, ATI's mobile graphics are entirely focused on 2d and codec, plus they are 100% firmware based. So that means no 3D acceleration, and even if somebody ever was to write FOSS drivers, lots of code is hidden in the GPU firmware, rather than in those FOSS drivers. What I personally don't understand about this entire debate on our community list: You have very prominent people of the FOSS movement, particularly the Linux community in this project. Notably Werner and myself. Given my track history of clinging to every last word of the GPL, and my stance with regard to binary-only drivers or other abominations of the hardware industry: Why don't you trust us to do proper research and chose the vendor that works best for us, given all the circumstances? I think we all don't want to see this little FOSS flame be extinguished, and that may have a lot to do with why the OM community is so concerned with all the details, down to the vendors. We should probably take a step back and let OM handle those details. There is a line where the community should end and OpenMoko begin. This (now unknown) line will need tweaking. I place a large amount of trust in the OM Team, as they are geeks just like us. For those in the community unfamiliar with the very prominent people within OM, google some of their names; you might be surprised. :) Do you think we would be foolish enough not to talk to all vendors of the respective components? I really feel personally very sad that anyone believes that I am in this project for anything else then to provide the highest level of freedom for both hardware and software that is possible. In GTA01, the only freedom related issue that we have is the Global Locate (now Broadcom). Given the start of OpenMoko (alternative software for a Windows smartphone that FIC was building) we didn't have any influence on that one. We have been trying hard to achieve a compromise with GL on the level of freedom that they're willing to provide. Unfortunately that compromise falls short of what many people in the FOSS community, including myself, deem acceptable. For GTA02, we evaluated all different A-GPS solutions on the market, and we took two of those actually in production. The graphics chip we ship will have FOSS drivers. We're working with NXP on publishing an open user manual for the PCF50633 PMU, and we already have their approval for it. We're staying with the
Re: ATI to provide specs (was: Re: SMedia 3362)
On 9/10/07, Harald Welte [EMAIL PROTECTED] wrote: guys, as I indicated before, we already have the best possible support from Smedia. Not only have we some promises or statements, but we actually have signed a contract with them, binding them to support us to the utmost level. As indicated previously, this agreement includes a statement that OpenMoko will work on to-be-publicized documentation on the SMedia chip, which will be jointly released at some point. So what exactly is not enough? You will get 100% free software drivers, down to the latest bit, no proprietary firmware whatsoever, plus hardware documentation that will be prepared by OpenMoko ? I didn't understand before that it would end up being 100% open source. I thought you had to sign an NDA? But if the driver is to be completely open, and the documents are to be completely open too, what is the purpose of the NDA? Or maybe I just misunderstood completely. Anyway it sounds like the best choice in that you do not get anywhere near this kind of support from the others. device that does not have any working driver at all, but with a thousand-page manual (rather than the other way around: first have FOSS Drivers, and then get the docs as soon as our incredibly small team finds time to do so)? No, getting it working is the more important of the two, of course. But as someone else pointed out, maybe there is the possibility that someone outside your small team could help, if the NDA doesn't prevent it. What kind of driver are you planning on? (I don't think I saw that answered yet, sorry if I missed it) KDrive, DRI, etc... Do you think we would be foolish enough not to talk to all vendors of the respective components? I really feel personally very sad that anyone believes that I am in this project for anything else then to provide the highest level of freedom for both hardware and software that is possible. Well the choices for open hardware are always slim, so far. I just thought some people might be putting more emphasis on OpenGL ES support than openness. But it sounds like you have really made the best choice. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
Dnia poniedziałek, 10 września 2007, Shawn Rutledge napisał: On 9/10/07, Harald Welte [EMAIL PROTECTED] wrote: I didn't understand before that it would end up being 100% open source. I thought you had to sign an NDA? But if the driver is to be completely open, and the documents are to be completely open too, what is the purpose of the NDA? Or maybe I just misunderstood completely. NDA can give all available documentation on which OpenMoko team will base open docs which will describe only those parts which vendor allows. As result community will get source code of drivers + documentation for chipset and OpenMoko team will know more about chipset internals then it is needed for writing drivers so if community will get problems with own code then there will be someone who can check does something more should be provided (and was not possible before). -- JID: hrw-jabber.org OpenEmbedded developer/consultant Someday I'm gonna die but it won't be from boredom [Pink] ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: ATI to provide specs (was: Re: SMedia 3362)
On 9/6/07, Steve [EMAIL PROTECTED] wrote: It may be worth talking with ATI again. Since this announcement, I don't think it is too far fetched to at least get the same deal you currently have with SMedia. I wouldn't be surprised if the people you were talking to had no idea this sort of thing was being planned. ATI may even allow the release documentation at some point in the future. Or just use it for leverage to get more from SMedia. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
RE: SMedia 3362
What exactly will the driver actually be? As far as I can tell there are several options: 1) A DRI/DRM kernel module associated mesa module 2) A hacked up KDrive with accelerated driver 3) An xorg EXA/XAA driver 4) A DirectFB kernel module 5) A bog-standard Linux frame buffer device Does anyone know the answer to this at all? I've had a quick look around the source and can't seem to find anything which might give us a clue... ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On Sat, Sep 01, 2007 at 10:44:20PM +0100, Ian Stirling wrote: Shawn Rutledge wrote: On 9/1/07, Raphael Jacquot [EMAIL PROTECTED] wrote: Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? depending on the timeframe, perhaps FIC could contact or help the opengraphics people with their project, the goal of which is to create a graphics controller asic that would be used for embedded systems. Using the opengraphics asic in the openmoko platform would be, IMNSHO the proper thing to do Or maybe use an FPGA so we can design our own hardware-accelerated graphics functions, and it can double for some other purposes too. OpenGraphics is starting that way (spartan 3 if memory serves). They are cheap. Not sure if they are low-power enough to use in a phone. Basically. FPGAs of equivalent size are _NOT_ cheap, or low power. I'd guess the SMedia chip is $20 or so. and it is multi chip package with 8MB sdram. Now, go and look for a low power FPGA with the thick end of a megabyte of embedded RAM, and many thousand gates. It'll be at least $100, maybe $200. 1MB is ridiculously small. oh, and not even thinking about the many man years it will take to write the verilog/vhdl for the graphics chip, as well as verification/testing/... This project is about building open source communications devices, not open source graphics chips. If we want to start building our own components, then we would have released GTA01 in 2010 and GTA02 in 2014. -- - Harald Welte [EMAIL PROTECTED] http://openmoko.org/ Software for the world's first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On Sat, Sep 01, 2007 at 05:19:22PM +0100, Giles Jones wrote: On 1 Sep 2007, at 17:11, Mikko Rauhala wrote: You're implying there are better choices... I wouldn't know since I've not looked into such things. But ATI have mobile GPUs and are open sourcing desktop drivers, maybe they would do the same for their mobile devices? Why do you think we have spent many weeks, if not months to meet with each and every graphics chip vendor? We're not that stupid, eh. It was a very painful and long process to finally find one company that was not fundamentally opposed to free software drivers. one in the entire industry. -- - Harald Welte [EMAIL PROTECTED] http://openmoko.org/ Software for the world's first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On Sat, Sep 01, 2007 at 04:31:00PM +0100, Giles Jones wrote: On 1 Sep 2007, at 16:07, Ian Stirling wrote: Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? There is as I understand it at the moment only a dumb driver for it, using it as a framebuffer. Unfortunately, documents are only available under NDA. This means that only FIC can write the drivers. Seems like an odd choice of unit then for an open source phone. So are you claiming the open source drivers that we are writing are not open source, merely by the fact that we are writing them? Using this argument, the entire openmoko software stack would not be open source, because we are writing it. -- - Harald Welte [EMAIL PROTECTED] http://openmoko.org/ Software for the world's first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Harald Welte [EMAIL PROTECTED] wrote : Why do you think we have spent many weeks, if not months to meet with each and every graphics chip vendor? We're not that stupid, eh. It was a very painful and long process to finally find one company that was not fundamentally opposed to free software drivers. one in the entire industry. Not stupid, just wondering why an open phone shouldn't be totally open, even if that means keeping certain things simple? Better optimising the 2D and having full control over the video hardware than having a 3D unit you can't hack. --- G O Jones ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Harald Welte [EMAIL PROTECTED] wrote : So are you claiming the open source drivers that we are writing are not open source, merely by the fact that we are writing them? Using this argument, the entire openmoko software stack would not be open source, because we are writing it. I misunderstand the announcement over these drivers then. It wasn't clear that these would be released as source or if they would be a binary like the GSM and GPS. If you're writing these but referring to NDA documents and the drivers will be open source then there's no problem at all. --- G O Jones ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Giles Jones wrote: Harald Welte [EMAIL PROTECTED] wrote : So are you claiming the open source drivers that we are writing are not open source, merely by the fact that we are writing them? Using this argument, the entire openmoko software stack would not be open source, because we are writing it. I misunderstand the announcement over these drivers then. It wasn't clear that these would be released as source or if they would be a binary like the GSM and GPS. Please don't mix real binary-only problems (GPS) versus including autonomous components that speak well-defined interfaces (GSM). -- - Michael Lauer [EMAIL PROTECTED] http://openmoko.org/ Software for the worlds' first truly open Free Software mobile phone ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Harald Welte writes: On Sat, Sep 01, 2007 at 04:31:00PM +0100, Giles Jones wrote: Seems like an odd choice of unit then for an open source phone. So are you claiming the open source drivers that we are writing are not open source, merely by the fact that we are writing them? Using this argument, the entire openmoko software stack would not be open source, because we are writing it. Well... not speaking for Giles, but the drivers I've seen for closed chipsets have generally involved a thin open-source wrapper around a closed binary driver. Letting you write a real open-source driver while demanding NDA to see the specs you're writing the driver to seems odd on their part (but much better than the norm!). ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
RE: SMedia 3362
So are you claiming the open source drivers that we are writing are not open source, merely by the fact that we are writing them? Using this argument, the entire openmoko software stack would not be open source, because we are writing it. So are you going to release the source code to the driver you write? I think the impression was that an NDA would include not disclosing any source code you wrote using the NDA'd docs (ala Gumstix Marvell wireless drivers). What exactly will the driver actually be? As far as I can tell there are several options: 1) A DRI/DRM kernel module associated mesa module 2) A hacked up KDrive with accelerated driver 3) An xorg EXA/XAA driver 4) A DirectFB kernel module 5) A bog-standard Linux frame buffer device If it's simply going to be a kernel framebuffer device, what's the point in including the smedia chip at all?!?!? If it's going to be an x-org EXA driver, how much memory is xorg going to use compared with kdrive? I am guessing it's going to be a modified KDrive, in which case I guess we can kiss accelerated OpenGL ES 3D graphics goodbye? Cheers, Tom ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Joe Pfeiffer [EMAIL PROTECTED] wrote : Well... not speaking for Giles, but the drivers I've seen for closed chipsets have generally involved a thin open-source wrapper around a closed binary driver. Letting you write a real open-source driver while demanding NDA to see the specs you're writing the driver to seems odd on their part (but much better than the norm!). It's odd and the only reason why I thought it was strange to go with this chip. I'm much happier now it's going to open source, seems like this is a good choice. It helps people maintain the driver if they have the code. --- G O Jones ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On Tuesday 04 September 2007, Joe Pfeiffer wrote: Well... not speaking for Giles, but the drivers I've seen for closed chipsets have generally involved a thin open-source wrapper around a closed binary driver. Letting you write a real open-source driver while demanding NDA to see the specs you're writing the driver to seems odd on their part (but much better than the norm!). Unusual but not unprecedented. Psion did the same for some of the documentation on the hardware for the Series 5 when linux was being ported to it. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Dnia niedziela, 2 września 2007, Shawn Rutledge napisał: What about the ATI chip used in the iPAQ hx4700? Did anybody figure out anything beyond plain framebuffer support for it? hx4700 use ATI W100 Imageon about which I already wrote. AFAIK it has accelerated framebuffer and X11 driver with XVideo and XRender funtions. It also allow to use VGA/QVGA resolution. -- JID: hrw-jabber.org OpenEmbedded developer/consultant Real programmers don't document. If it was hard to write, it should be hard to understand. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? There is as I understand it at the moment only a dumb driver for it, using it as a framebuffer. Unfortunately, documents are only available under NDA. This means that only FIC can write the drivers. It also means that they will be unlikely to be able to implement all the functions in the chip for at least several months after October, barring large increases in team size. It's also nearly impossible to reverse engineer functions missing in the driver into it. Simply as you don't have docs on what it should be able to do, register info, ... Still, in principle, they could make a driver which exploits all of the functionality of the accellerator. Things are worse (as I understand it) for potential upgrades to better devices. Some of them support DSPs, which are basically of sharply limited use if FIC are the only ones that can code modules for the DSP. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 1 Sep 2007, at 16:07, Ian Stirling wrote: Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? There is as I understand it at the moment only a dumb driver for it, using it as a framebuffer. Unfortunately, documents are only available under NDA. This means that only FIC can write the drivers. Seems like an odd choice of unit then for an open source phone. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
la, 2007-09-01 kello 16:31 +0100, Giles Jones kirjoitti: On 1 Sep 2007, at 16:07, Ian Stirling wrote: Unfortunately, documents are only available under NDA. This means that only FIC can write the drivers. Seems like an odd choice of unit then for an open source phone. You're implying there are better choices... With chip manufacturers being jealous of every bit of information on how to actually use their chips, we're unfortunately lucky that the OpenMoko guys at least are allowed to write free drivers themselves after (presumably) signing the NDA to get the specs. This is not really new for the Neo either; never mind GTA01's GPS chip, even the LCD screen doesn't have free docs. Signs of the times... -- Mikko Rauhala - [EMAIL PROTECTED] - URL:http://www.iki.fi/mjr/ Transhumanist - WTA member - URL:http://www.transhumanism.org/ Singularitarian - SIAI supporter - URL:http://www.singinst.org/ ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 1 Sep 2007, at 17:11, Mikko Rauhala wrote: You're implying there are better choices... I wouldn't know since I've not looked into such things. But ATI have mobile GPUs and are open sourcing desktop drivers, maybe they would do the same for their mobile devices? ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
la, 2007-09-01 kello 17:19 +0100, Giles Jones kirjoitti: I wouldn't know since I've not looked into such things. But ATI have mobile GPUs and are open sourcing desktop drivers, maybe they would do the same for their mobile devices? ATI's intentions to open source their GPU drivers have been greatly exaggarated. They have mostly made vague statements about supporting Linux / open source [OSes] better. And noise has been made for a while. I wouldn't hold my breath waiting for them to actually deliver. (Shame on AMD.) And of course, even if they freed such commodity drivers, some PHB could well be persuaded to keep a tight lid on their super-secret embedded stuff. But of course, it'd be a step forward. Oh, incidentally, I don't really know the market situation of free software friendly low-power GPUs either. It's just an educated assumption on my part that this is likely to be the best deal OM are likely to get in this area. I pretty much trust the OM guys to feel the same way, since, well, they picked the SMedia. -- Mikko Rauhala - [EMAIL PROTECTED] - URL:http://www.iki.fi/mjr/ Transhumanist - WTA member - URL:http://www.transhumanism.org/ Singularitarian - SIAI supporter - URL:http://www.singinst.org/ ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Dnia sobota, 1 września 2007, Giles Jones napisał: On 1 Sep 2007, at 17:11, Mikko Rauhala wrote: You're implying there are better choices... I wouldn't know since I've not looked into such things. But ATI have mobile GPUs and are open sourcing desktop drivers, maybe they would do the same for their mobile devices? ATI support is near null when it comes to documentation or drivers. In Zaurus machines ATI W100 Imageon is used and we had to use reverse engineering to get information how does chip works and to write proper kernel and X11 support. Each request from our ackers to ATI resulted in no answer or even mails such as we do not have docs for that chip. ATI desktop cards are crap under Linux. Propertiary driver (fglrx) lack AIGLX support (so goodbye beryl/compiz), free drivers hackers lack documentation how to get some stuff done. After few months of using ATI onboard graphics (under fglrx or free driver) I got into situation when only fglrx worked as free driver was unable to figure which port has connected monitor. So if you want to use Linux on anything avoid ATI. -- JID: hrw-jabber.org OpenEmbedded developer/consultant bloody internet there was once peace and then the internet came in :-) ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 1 Sep 2007, at 19:15, Marcin Juszkiewicz wrote: So if you want to use Linux on anything avoid ATI. Well Intel seem to be getting praise on the graphics front for their support of open source. They've licenced PowerVR for embedded chips, so who knows. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? depending on the timeframe, perhaps FIC could contact or help the opengraphics people with their project, the goal of which is to create a graphics controller asic that would be used for embedded systems. Using the opengraphics asic in the openmoko platform would be, IMNSHO the proper thing to do ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 9/1/07, Raphael Jacquot [EMAIL PROTECTED] wrote: Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? depending on the timeframe, perhaps FIC could contact or help the opengraphics people with their project, the goal of which is to create a graphics controller asic that would be used for embedded systems. Using the opengraphics asic in the openmoko platform would be, IMNSHO the proper thing to do Or maybe use an FPGA so we can design our own hardware-accelerated graphics functions, and it can double for some other purposes too. OpenGraphics is starting that way (spartan 3 if memory serves). They are cheap. Not sure if they are low-power enough to use in a phone. What about the XScale accelerators? There was one designed specifically as a companion to the PXA270. Are they more open? My Zaurus has a TC6393; it has line-drawing and rectangle-filling in hardware. I have used it successfully for those tasks. But that's not much acceleration. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Dnia sobota, 1 września 2007, Shawn Rutledge napisał: What about the XScale accelerators? There was one designed specifically as a companion to the PXA270. Are they more open? Intel 2700G? It is closed - the only source released was mess afaik. My Zaurus has a TC6393; it has line-drawing and rectangle-filling in hardware. I have used it successfully for those tasks. But that's not much acceleration. SL-6000? Too bad that it was so pricey - community lack hackers which want to get software support on them improved (OpenEmbedded project can even provide one or two SL-6000 for such tasks). -- JID: hrw-jabber.org OpenEmbedded developer/consultant Vi has two modes: the one in which it beeps, and the one in which it doesn't. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Shawn Rutledge wrote: On 9/1/07, Raphael Jacquot [EMAIL PROTECTED] wrote: Shawn Rutledge wrote: Is there any technical info available about this chip? Is there an X driver for it, or one in progress? depending on the timeframe, perhaps FIC could contact or help the opengraphics people with their project, the goal of which is to create a graphics controller asic that would be used for embedded systems. Using the opengraphics asic in the openmoko platform would be, IMNSHO the proper thing to do Or maybe use an FPGA so we can design our own hardware-accelerated graphics functions, and it can double for some other purposes too. OpenGraphics is starting that way (spartan 3 if memory serves). They are cheap. Not sure if they are low-power enough to use in a phone. Basically. FPGAs of equivalent size are _NOT_ cheap, or low power. I'd guess the SMedia chip is $20 or so. Now, go and look for a low power FPGA with the thick end of a megabyte of embedded RAM, and many thousand gates. It'll be at least $100, maybe $200. FPGAs are almost always the most expensive, highest power way to do things, if proper custom silicon is a possibility. (volumes of 1K, comparable processes) ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 9/1/07, Marcin Juszkiewicz [EMAIL PROTECTED] wrote: Dnia sobota, 1 września 2007, Shawn Rutledge napisał: What about the XScale accelerators? There was one designed specifically as a companion to the PXA270. Are they more open? Intel 2700G? It is closed - the only source released was mess afaik. Apparently it's also discontinued. That's too bad. I see that a Spartan 3L (low-power version) would take several times as much power as the 2700G. The million gate one (to cut it down a bit, in comparison to which the OpenGraphics project is using the 4 million gate version) has quiescent current ratings of 35 mA for internal supply plus 20 mA for AUX supply. So if I understand correctly that is the minimum, if it's turned on at all. I suppose if you are actively doing something, it would be several times as much. The OpenGraphics board has a heatsink, after all. http://wiki.duskglow.com/tiki-index.php?page=OGD1 That's a bummer. I think programmable hardware would be an excellent way for open software to innovate right around the barriers of proprietary graphics chips. But I don't have any experience at all programming them yet. There is the NVIDIA GoForce 4800 and I don't see datasheets for it either. SL-6000? Too bad that it was so pricey - community lack hackers which want to get software support on them improved (OpenEmbedded project can even provide one or two SL-6000 for such tasks). I'm still working with mine. Angstrom is working well enough to do the kind of hacking I want to do. I'm glad the 2.6 kernel is finally OK. On 9/1/07, Ian Stirling [EMAIL PROTECTED] wrote: FPGAs of equivalent size are _NOT_ cheap, or low power. I'd guess the SMedia chip is $20 or so. Now, go and look for a low power FPGA with the thick end of a megabyte of embedded RAM, and many thousand gates. It'll be at least $100, maybe $200. That's how it used to be. A million-gate Spartan IIIE ranges from about $20 down to $7.90 at Digikey depending on the number of IO pins (I think you can easily get under $5 in volume). Power would be the problem with those chips. I wonder who's the low-power leader for FPGAs. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
Shawn Rutledge wrote: On 9/1/07, Marcin Juszkiewicz [EMAIL PROTECTED] wrote: On 9/1/07, Ian Stirling [EMAIL PROTECTED] wrote: FPGAs of equivalent size are _NOT_ cheap, or low power. I'd guess the SMedia chip is $20 or so. Now, go and look for a low power FPGA with the thick end of a megabyte of embedded RAM, and many thousand gates. It'll be at least $100, maybe $200. That's how it used to be. A million-gate Spartan IIIE ranges from about $20 down to $7.90 at Digikey depending on the number of IO pins (I think you can easily get under $5 in volume). Power would be the problem with those chips. I wonder who's the low-power leader for FPGAs. Interesting (though I will note the 7.90 one is 100Kgate) Annoyingly, I managed to get mislead by the lack of sort by price on digikey. Time to re-run my script against the FPGA category. ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
Re: SMedia 3362
On 9/1/07, Shawn Rutledge [EMAIL PROTECTED] wrote: There is the NVIDIA GoForce 4800 and I don't see datasheets for it either. I found datasheets with real, substantial datasheets for these: Silicon Motion SM502 Phillips/NXP PNX17XX series Renesas HD64412 Epson S1D13513 (but it only has bitblt and sprites, no drawing functions) What about the ATI chip used in the iPAQ hx4700? Did anybody figure out anything beyond plain framebuffer support for it? ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community
SMedia 3362
Is there any technical info available about this chip? Is there an X driver for it, or one in progress? ___ OpenMoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community