Xilinx 6.2i ISE WebPACK running under wine?

It does appear to run, though with some err... "features". Since I have the Linux version of ISE, I have not tested Webpack/Wine much.

First, Wine should be configured for a Version of "win2k". You also should get a copy of a native version of msvcrt.dll somewhere, and configure Wine to use it. If you are already running a previous version of Webpack under Wine, then I would guess you already meet those requirements.

When installing Webpack, you will get some error dialogs, and some Wine processes will even crash. Don't worry, Webpack is still installed ok ;) Really!

Finally, the biggest problem is that the GUI runs really slow with a current Wine, due to a named pipes bug. But this is also the same as in Webpack 5.1, so there will be no change here. If you are just using the command line tools, then they run fine. You could patch a version of Wine to run the GUI if you wanted; I have a patched version around that runs the GUI fine.

doesn't

Hello from Gregg C Levine Just for the sake of arguement, where did you find the Linux version of the tool? Every time I visit their webpages, all it talks about is the Windows version of the tool. And I've tried searching the site, its search engine does not properly return anything. Gregg C Levine drwho8 atsign att dot net "This signature says, "YACK!"."

....

I wonder how much Microsoft are paying Xilinx! After all Xilinx don't release ISE WebPACK for Linux. Has anyone heard a REAL reason for Xilinx not releasing WebPACK for Linux?

Andrew.

You should also be aware that it doesn't appear to work well if you have an Athlon64 processor. I've had Wine working fine under non-64 bit processors, but can't get it to compile on 64-bit linux for the AMD64.

There was some traffic on the Wine list saying that it now works perfectly, and if anyone has problems, it's an issue with their kernel or libraries, but I've tried the very latest Fedora-2 + all patches + latest kernel and still no joy :-(

This was after I'd 'upgraded' to something I thought would handle large amounts of memory well and synthesize/P&R faster... Oh well, back to windows :-( Xilinx now has the dubious honour of being the only company for which I maintain a Windows partition :-(

Simon.

The native Linux version of ISE runs on AMD64. Too bad there is no WebPack for native Linux though.

Petter

Well, annoying though it is, I don't think it's a conspiracy :-) It's probably a combination of:

o Most Linux WebPack users will never contribute sufficient funds back to Xilinx to make it worth their while.

o They pay a per-seat licence for their officially-supported foundation package on Linux to the GUI-library owners. They therefore can't offer it as a download for WebPack.

What it would probably take would be a large customer saying "we're going to move to Altera/whoever unless you do the port before XXX", with XXX being a realistic target in the future. At that point, there's a commercial pressure to do the port, and Xilinx can take a view on whether it's worth it. Even then, their view might be 'see that lake, run and jump' :-)

Another route for Xilinx to get their finger out would be if Altera/whoever did it first - Xilinx wouldn't be far behind then, I'd imagine, bragging rights and image are nowhere near as important as commercial pressures, but they still count :-)

Simon.

Equally applicable to Windows WebPACK users in my opinion. I'll buy FPGAs if the software is free, I'll buy nothing if the software isn't free.

How about the command line tools; xst, map, par, bitgen, etc?

At 110Kg I'm a large customer! Xilinx, I'm moving to Altera if you don't release the Linux WebPACK that you have hidden somewhere and are keeping quiet about!

Xilinx should retain their world leader status and should get in there first!

Andrew.

"andrewrogerstechcouk" wrote: : I have 5.1 ISE WebPACK running under wine but just discivered it doesn't : support Spartan3. From the archives I read about some difficulties : regarding 6.2 ISE running on wine, have these been solved?

:

In "70825" nothing about wine is said. In "70836" I tell about the success I have, your milage may vary.

: I urgently need to know if 6.2 ISE works under wine as I have just : ordered the $99 Spartan-3 Starter Kit. If not I shall need to cancel my : order.

: For those who haven't seen the Spartan 3 Kit:

:

: Looks like a superb bit of kit, I will be buying a number of these for : my University module. Having the RAM on board is great. The applications : I have already thought about - add an ADC and you have a digital storage : scope, Turbo decoder, LDPC codes and so on. We'll need to get the ISE : 6.2 WebPACK running on Linux and write our own programmer software.

Watch for MITOUJTAG

: ...

Sorry, but I paid for the "full" ISE, which is the only way to get the Linux version (the full ISE includes a couple of additional tools and supports more parts). Hopefully Webpack will eventually be made available in the Linux version. Possibly Xilinx, like many vendors, wants to use there paying customers for beta testing. I think that they realize the paying customers are probably a bit more experienced at debugging the vendor tools ;)

I had thought the application needs to be compiled for 64-bit operation, otherwise there will be little benefit to running it on a 64 bit processor. Are you seeing better performance running 32 bit apps on a 64 bit processor?

Well, I was initially hoping to compile Wine in 64-bit mode, figuring that because it "wasn't an emulator", then when the Windows calls finally made it through to the lower levels of Wine, it would be running at 64-bits, although of course the windows code itself would only be

Anyway, my hopes were dashed. Wine wouldn't work, even in 32-bit mode :-(

Simon

I think it's because they use that MainWin to port from Windows to Linux and there is a per-seat license charge they pay for it. They could have used Wine and Winelib to port from Windows to Linux for free, of course, and then we wouldn't be asking all these questions ;-)

While the big EDA companies (that cater to ASICs) now consider Linux to be a first-class citizen, the FPGA vendors (and EDA companies that cater to FPGAs) still don't seem to know much about Linux. At least they've started porting to Linux, but they're still developing on Windows and porting to Linux (and other Unices) whereas the big guys are developing on Linux and porting to Windows (I know this is true at Mentor Graphics for example). The interesting thing is that if you're coming from the Windows world (as Xilinx is) all you know about are proprietary tools whereas when you're coming from the Linux world you tend to use cross-platform, open source tools from the start. So porting from Linux to Windows is fairly easy because you're probably using a cross platform GUI toolkit like Tk, gtk or perhaps Qt. When you're locked into Windows as your development platform, porting to Linux can be a pain (especially if you don't know about Wine/Winelib, and you probably wouldn't if you're coming from the Windows world).

So, as someone else mentioned, it's not a conspiracy with Micro$oft so much as ignorance of the market and of the Linux world that leads to these problems. We need to educate the vendors about Linux and about what's possible (like we did the big EDA vendors several years ago). For example, Xilinx could save a lot of $$ if they didn't use MainWin for porting and used Wine/Winelib instead - perhaps they didn't know about that option - and as an added bonus their product would (of course) work under Wine.

BTW: The other issue with getting the Xilinx tools to work under Wine is the Jungo parallel port driver. You may be able to get the development tools to work under Wine, but you won't be able to program any parts with Impact (unless something has changed very recently).

Phil

It's interesting that for the big EDA companies (Synopsys, Mentor, Cadence) the switch to supporting Linux happened several years ago and now Linux is probably the majority platform for most of them. Sure, it was easier for them to make the switch because they already did all their development on some flavor of Unix, but there was still inertia to overcome to get them to move to Linux. When serious customers started asking for Linux, they bagan to take notice. Now, just as then, It will take a lot of customer pressure to get the FPGA companies (and the FPGA EDA companies) to properly support Linux.

As a Linux software developer who has done a bit of consulting I've seen several environments that were primarily Windows development houses. The people in these groups might be very good developers, but they have no idea of what exists in the Linux world (or the open source world in general). They tend to view it as if it were a dangerous foreign country that they want to keep their distance from. So what happens is that Linux developers have a hard time communicating with Windows developers (and vice-versa). Right now in Xilinx, I would bet that you've got a primarily Windows development culture. Sure they've made steps toward Linux-land, but they're very cautious steps. With time they'll discover that the Linux waters aren't shark-infested and they'll become more adventurous ;-)

Phil

That still won't help when it comes to programming devices. The Jungo parallel port driver doesn't work under Wine.

Phil

Phil Tomson wrote: ,,, : That still won't help when it comes to programming devices. The Jungo : parallel port driver doesn't work under Wine.

I had some success with

Bye

How about GtkJTAG from the same author? That runs on Linux, I have downloaded and compiled it. The downside is that in needs to run as root due to low level IO. But it should be possible to use the parport device driver. I am currently working on a modification to GtkJTAG (well, libjtag to be accurate) so that it uses the parport device driver.

As my Spartan 3 kit hasn't arrived yet I haven't been able to test GtkJTAG.

A seperate project aimed at Virtex:

As far as I can tell from XAPP188 this will also program Spartan2 devices and I can only guess that it may work for Spartan3. I can't find an XAPP for spartan3 configuration using JTAG.

Regards Andrew

Yes, but I suspect that their Solaris port is in the same shape as their Linux port. Also, I suspect that you can't program parts using Impact on Solaris either. My point was that they seem to be doing all of their initial development on Windows and then porting over to Linux (or Solaris) using MainWin and this, of course, shows.

Perhaps, but these days most designers coming to FPGAs from the ASIC world are used to running Linux (or perhaps Solaris) so they're not used to dealing with all of the virus problems that come with running Windows and they would prefer to not bother with them. Also, they may not want to have to take the time it takes to make their Windows box feel like 'home' (installing cygwin, scripting languages, etc.)

As stated above: the 'really useful' part of Linux/Unix that has not been ported to Windows yet is stability (and security). Freedom from viruses/worms etc. That's a big reason why people are not wanting to run Windows now and are looking at Linux.

You're very committed to the hobby! I can't afford such commitment at this time. However, I think it' also important to note that there are companies out there running Linux and doing ASIC development on Linux and as they look more and more at FPGAs they want tools that will run well on Linux.

Amen.

Phil

There is a libjtag?

Thanks for the links.

Phil

Mmm, not sure I agree with that. Without wishing to be an apologist for Xilinx (because basically I wish they'd get their act together and support Linux as a tier-1 platform!), I don't think it's a 'unix/linux is bad' attitude. They do support Suns, after all.

FWIW, I think it's a cost/reward thing. There's a large proportion of people for whom the platform on which you code VHDL or verilog is irrelevant. These people are more concerned with the end-product of their design than the method of getting there, as all good designers ought to be.

The issue comes when you have to change OS to do XYZ for a while rather than ABC. With the right environment, it's pretty hard to find any situation where it's impossible to do something on Windows that it is possible to do on Linux: all the 'really useful' things about Unix have been ported to Windows to fill the gaps. What then is the reason to still want a native Linux port ?

Well, for me (and it's a personal view), it's a matter of 'Linux is my work environment', and 'although there are workarounds (mingw, Wine, etc) none of them really cut the mustard. To whit, I've just had to go and buy Windows XP (and PC world, the only "nearby" computer shop, only had the 'professional' version!) So I'm already £270 ($400 or so for our US friends) down. Now I have to reboot into XP to do any experimenting:

o Hey! what if I use a blockram for the register file, and the high bit on the 6-bit register part of the opcode to select the 'switchable'

o Hey! What if I use the parity entries of the blockram to implement a 'this i-or-d-cache value is valid' flag. The CPU can increment it's current tag-is-valid 'count' on cache-reset, and only actually clear the cache (and hence introduce delay) when 'count' cycles around to 0. That's a 16x win :-)

o Hey! You get the idea. If it takes 15 mins to shut down all the programs I'm usually and normally running under Linux (since that's my work environment), and reboot into XP and try something, I'm far less likely to do it.

Now, I have an idea that I think could be worth some money. What I'm going to do over the next few months is evaluate whether it'll work, and if so, purchase Foundation, because for me, the $2500 (=£1800 after import, probably) will be worth it just for the time spent not switching OS all the time, oh yeah, and it allows me to access the larger chips :-) The critical thing though is that it works under Linux, and I'm willing to put that much money into a bet that I can get a return (hey, I'm only an amateur at this, for you professionals, it obviously makes sense to get Foundation :-)

I guess the take-home message of what I'm trying to say, is that although for some (perhaps a lot of) people the platform we work with is agnostic. However I'd be willing to guess that most of those people are already Windows users, and therefore 'agnostic' because it doesn't affect their existing choice. My PC (until recently) didn't have Windows on it. For me, the platform is an issue, to the point that I'm willing to spend serious cash to stop it from being such.

Xilinx, please take note: I'm using your devices despite your policy, not because I love it. I suspect that there are more like me, and I further suspect the number is growing. Please at least think about supporting the Linux platform more.

Simon