Xilinx ISE, EDK and some ground roules in software development
in [FPGA]
|
Prev: Uart problem, xapp223 + Spartan3A
Next: Convert Schematic Files (.sch) to Verilog Files (.v) for simulationin ModelSim
From: Franz Hollerer on 6 Jul 2007 02:44 Hi, I started to learn ISE and EDK 9.1i, but I spent much much time to get things running because they did not work as described in the manuals. Now I write this posting in the hope that someone from Xilinx will read it and consider it for further software development. I tracked down the problems to two sloppinesses which never should be done in software development: a) The GUI does not evaluate the return code of programs it calls. This leads to situations where the GUI shows 100% complete, 0 errors but nothing has been done at all, which makes it very hard to find the real cause for subsequent errors. b) Some command line programs like "simgen" return 0 even if errors occurred. Thus makes it impossible for "make" to stop on errors, and again, makes it very very hard to find the cause for subsequent errors. Thus dear Xilinx programmers please note: - NEVER ignore the return code of a program. - Always return 0 on success and a value between 1 and 255 if an error has occured. If you follow this hint you will save us (your customers) and your support department much time, money and nerves. Is there a Xilinx manager out there who have ever considered how much money Xilinx must have wasted by not following this simple ground rules in software programming? Best regards, Franz Hollerer
From: Antti on 6 Jul 2007 02:47 On Jul 6, 7:44 am, Franz Hollerer <franz.holle...(a)ims.co.at> wrote: > Hi, > > I started to learn ISE and EDK 9.1i, but I spent much much time to get > things running because they did not work as described in the manuals. > > Now I write this posting in the hope that someone from Xilinx will read > it and consider it for further software development. > > I tracked down the problems to two sloppinesses which never should be > done in software development: > > a) The GUI does not evaluate the return code of programs it calls. > This leads to situations where the GUI shows 100% complete, 0 errors but > nothing has been done at all, which makes it very hard to find the real > cause for subsequent errors. > > b) Some command line programs like "simgen" return 0 even if errors > occurred. Thus makes it impossible for "make" to stop on errors, and > again, makes it very very hard to find the cause for subsequent errors. > > Thus dear Xilinx programmers please note: > - NEVER ignore the return code of a program. > - Always return 0 on success and a value between 1 and 255 if an > error has occured. > > If you follow this hint you will save us (your customers) and your > support department much time, money and nerves. > > Is there a Xilinx manager out there who have ever considered how much > money Xilinx must have wasted by not following this simple ground rules > in software programming? > > Best regards, > > Franz Hollerer the amount of time Users of Xilin silicon devices spend "fighting" with Xilinx software is enorm. So far it really looks that Xilinx isnt listening. Things are not improving. Only new bug comes, and old ones re-apperar... Anti
From: PFC on 6 Jul 2007 05:32 > the amount of time Users of Xilin silicon devices spend "fighting" > with Xilinx software is enorm. > So far it really looks that Xilinx isnt listening. Things are not > improving. > Only new bug comes, and old ones re-apperar... Yeah, this sums up well my experience with ISE and especially EDK ! On my first FPGA project which lasted about a month I spent : - 3 weeks learning ISE+EDK then fighting it - 2 days learning Verilog - 1 week implementing the stuff Is Altera better ? EDK's nice part is that it isn't really expensive, includes a full Microblaze licence, and you can use a $20 "compatible" cable, plus it includes a huge number of IP cores for the same price. And the Spartan chips are really nice. I'd be willing to relearn the Altera tools if someone can confirm that you can get the full package for a decent price (ie the same or cheaper than EDK) ; full package meaning full license to use Nios (not time-limited or whatever), JTAG cpu debugging plus having a good core library. Cuz I saw nice board modules with Altera chips on them...
From: Laurent Pinchart on 6 Jul 2007 06:39 Antti wrote: > On Jul 6, 7:44 am, Franz Hollerer <franz.holle...(a)ims.co.at> wrote: >> Hi, >> >> I started to learn ISE and EDK 9.1i, but I spent much much time to get >> things running because they did not work as described in the manuals. >> >> Now I write this posting in the hope that someone from Xilinx will read >> it and consider it for further software development. >> >> I tracked down the problems to two sloppinesses which never should be >> done in software development: >> >> a) The GUI does not evaluate the return code of programs it calls. >> This leads to situations where the GUI shows 100% complete, 0 errors but >> nothing has been done at all, which makes it very hard to find the real >> cause for subsequent errors. >> >> b) Some command line programs like "simgen" return 0 even if errors >> occurred. Thus makes it impossible for "make" to stop on errors, and >> again, makes it very very hard to find the cause for subsequent errors. >> >> Thus dear Xilinx programmers please note: >> - NEVER ignore the return code of a program. >> - Always return 0 on success and a value between 1 and 255 if an >> error has occured. >> >> If you follow this hint you will save us (your customers) and your >> support department much time, money and nerves. >> >> Is there a Xilinx manager out there who have ever considered how much >> money Xilinx must have wasted by not following this simple ground rules >> in software programming? >> >> Best regards, >> >> Franz Hollerer > > the amount of time Users of Xilin silicon devices spend "fighting" > with Xilinx software is enorm. > So far it really looks that Xilinx isnt listening. Things are not > improving. > Only new bug comes, and old ones re-apperar... Xilinx seems to be listening (or at least to have been listening). Unfortunately, the decision makers are not in touch with the listeners whatsoever :-/ I've had some contacts with Xilinx a few months ago regarding the Xilinx Platform Cable USB and the horrible Windriver they use. I pointed out that there are clean, free and open-source solutions to interface USB devices from userspace that would solve many user issues with Windriver (not even talking about the numerous security holes). People listened, developers didn't act. A piece of the chain is probably broken in the middle. As long as companies will refuse offers such as "I can make your software work better, here's how, and I'm willing to spend time on this to help you for free", we won't get decent softwares. Just my 2 cents. Laurent Pinchart
From: pbFJKD on 6 Jul 2007 08:35
>I tracked down the problems to two sloppinesses which never should be >done in software development: >a) The GUI does not evaluate the return code of programs it calls. >This leads to situations where the GUI shows 100% complete, 0 errors but >nothing has been done at all, which makes it very hard to find the real >cause for subsequent errors. Add a wrapper that will pop up it's own "An error occoured" ..? >b) Some command line programs like "simgen" return 0 even if errors >occurred. Thus makes it impossible for "make" to stop on errors, and >again, makes it very very hard to find the cause for subsequent errors. Maybe one can workaround this as customer by renaming original program and then add a wrapper that will check that the original actually did work? >Thus dear Xilinx programmers please note: >- NEVER ignore the return code of a program. >- Always return 0 on success and a value between 1 and 255 if an > error has occured. This really makes me wonder. |