FPGA Soft Core CPUs
in [Embedded]
|
Prev: PID autotuning - not working for heating application
Next: auto-detecting the resolution of HD44780-based LCD modules ?
From: Jack Klein on 5 Nov 2009 23:00 Parts obsolescence is biting again, and at least one of my EEs is insisting that we spin replacement boards with soft-core processors in FPGAs. There are several things I don't like about that. First, how robust are these systems. What if power is lost during updating the configuration device? Second, I don't like the holy wars among our EEs. Most of them use Altera and wouldn't touch Xilinx to save their lives. But one guy uses only Xilinx and won't give the Altera rep the time of day. This hasn't bothered me up until now, when we put together a peripheral that is accessed to a microcontroller or DSP, because we define an interface and they meet it. I don't need to care who the FPGA source is. But if we're talking soft core, I'm not going to allow our embedded software department get saddled with supporting both NIOS and Micro Blaze. So third, I'd be grateful for any experiences with either soft core, or even better both, with a comparison between the two. And does anybody have any experience with the synthesizable ARM soft core for FPGA? Unless there's a serious handicap, I'd prefer ARM if we do soft core, we've already good tools and programmer experience. Any and all comments, real or delusional, welcome. -- Jack Klein http://JK-Technology.Com FAQs for news:comp.lang.c http://c-faq.com/ news:comp.lang.c++ http://www.parashift.com/c++-faq-lite/ news:alt.comp.lang.learn.c-c++ http://www.club.cc.cmu.edu/~ajo/docs/FAQ-acllc.html
From: Mark McDougall on 6 Nov 2009 01:36 Jack Klein wrote: > So third, I'd be grateful for any experiences with either soft core, > or even better both, with a comparison between the two. Do you already have an FPGA on the board? If not, I think you'd be bloody crazy to go soft-core. Your EE's just want new toys to play with if they're pushing to stick an FPGA on there solely for that reason... If you already have an FPGA, how much headroom do you have in the device? How much bigger will you need to go to add a soft-core in there? How much grunt do you need? I'm guessing a little if you're considering ARM. What's the price difference in FPGA silicon? I've had quite a bit of experience with NIOS in a number of different projects. Overall it's been relatively successful, though one project ended up in the trash-can because of robustness issues that we couldn't pin down - using uCOS-II and 3rd-party IP stack and we couldn't prove it wasn't our code, though a port of our code to a micro ran without problems. So to be clear, I'm pointing my finger at the software libraries, not the NIOS-II itself. Oh yeah, the IDE is Eclipse (which I hate with a passion) and the debugger is so buggy it's next-to-useless. So bear that in mind. They really come into their own when you need to glue logic/peripherals to the core that can also go inside the FPGA. Having said that, a lot of the stuff might already be tacked onto a micro anyway - SPI, I2C etc. For the record, I wouldn't touch Xilinx with a 10-foot clown pole either. Regards, -- Mark McDougall, Engineer Virtual Logic Pty Ltd, <http://www.vl.com.au> 21-25 King St, Rockdale, 2216 Ph: +612-9599-3255 Fax: +612-9599-3266
From: Antti on 6 Nov 2009 03:26 On Nov 6, 6:00 am, Jack Klein <jackkl...(a)spamcop.net> wrote: > Parts obsolescence is biting again, and at least one of my EEs is > insisting that we spin replacement boards with soft-core processors in > FPGAs. > > There are several things I don't like about that. > > First, how robust are these systems. What if power is lost during > updating the configuration device? > > Second, I don't like the holy wars among our EEs. Most of them use > Altera and wouldn't touch Xilinx to save their lives. But one guy > uses only Xilinx and won't give the Altera rep the time of day. This > hasn't bothered me up until now, when we put together a peripheral > that is accessed to a microcontroller or DSP, because we define an > interface and they meet it. I don't need to care who the FPGA source > is. > > But if we're talking soft core, I'm not going to allow our embedded > software department get saddled with supporting both NIOS and Micro > Blaze. > > So third, I'd be grateful for any experiences with either soft core, > or even better both, with a comparison between the two. > > And does anybody have any experience with the synthesizable ARM soft > core for FPGA? Unless there's a serious handicap, I'd prefer ARM if > we do soft core, we've already good tools and programmer experience. > > Any and all comments, real or delusional, welcome. > > -- > Jack Kleinhttp://JK-Technology.Com > FAQs for > news:comp.lang.chttp://c-faq.com/ > news:comp.lang.c++http://www.parashift.com/c++-faq-lite/ > news:alt.comp.lang.learn.c-c++http://www.club.cc.cmu.edu/~ajo/docs/FAQ-acllc.html Hi Jack, I am one who loves to play around with soft-cores, yet I would also stay unless you have REAL valid reasons to use a soft-core do not go that path. I bet your EE guys can not give valid reasons (or that I would be able to invalidate their reasons) 1) obsolete - dont fool your self, FPGA vendor soft-core do get obsoleted as well! 2) FPGA vendor soft-core is actually a client-trap vendor lock-in now, Cortex-M3 RTL code is really beatiful, small and easy, basically you just select ITCM/DTCM sizes, number NVIC interrupts, and you are done and it useable on any FPGA with enough resources, so truly vendor independant optionally there free Coldfire (original freescale) core available for Cyclone-3 (for other FPGAs there is license fee) Antti
From: -jg on 6 Nov 2009 04:35 On Nov 6, 5:00 pm, Jack Klein <jackkl...(a)spamcop.net> wrote: > Parts obsolescence is biting again, and at least one of my EEs is > insisting that we spin replacement boards with soft-core processors in > FPGAs. ?? Do you think FPGA's don't go obsolete as well ? There is a lot more to go wrong in a fpga flow, so unless you NEED the fpga for another reason, this is not a good basis for a decision.? What parts have gone obsolete, and why ? There are things you can do, to mitigate obsolescence - like chosing multi sourced parts, and designing your code and PCBs, so that a Physical part swap, is not a high-impact step. -jg
From: Joseph Yiu on 6 Nov 2009 06:53
Antti wrote: > On Nov 6, 6:00 am, Jack Klein <jackkl...(a)spamcop.net> wrote: >> Parts obsolescence is biting again, and at least one of my EEs is >> insisting that we spin replacement boards with soft-core processors in >> FPGAs. >> >> There are several things I don't like about that. >> >> First, how robust are these systems. What if power is lost during >> updating the configuration device? >> >> Second, I don't like the holy wars among our EEs. Most of them use >> Altera and wouldn't touch Xilinx to save their lives. But one guy >> uses only Xilinx and won't give the Altera rep the time of day. This >> hasn't bothered me up until now, when we put together a peripheral >> that is accessed to a microcontroller or DSP, because we define an >> interface and they meet it. I don't need to care who the FPGA source >> is. >> >> But if we're talking soft core, I'm not going to allow our embedded >> software department get saddled with supporting both NIOS and Micro >> Blaze. >> >> So third, I'd be grateful for any experiences with either soft core, >> or even better both, with a comparison between the two. >> >> And does anybody have any experience with the synthesizable ARM soft >> core for FPGA? Unless there's a serious handicap, I'd prefer ARM if >> we do soft core, we've already good tools and programmer experience. >> >> Any and all comments, real or delusional, welcome. >> >> -- >> Jack Kleinhttp://JK-Technology.Com >> FAQs for >> news:comp.lang.chttp://c-faq.com/ >> news:comp.lang.c++http://www.parashift.com/c++-faq-lite/ >> news:alt.comp.lang.learn.c-c++http://www.club.cc.cmu.edu/~ajo/docs/FAQ-acllc.html > > Hi Jack, > > I am one who loves to play around with soft-cores, > yet I would also stay unless you have REAL valid reasons > to use a soft-core do not go that path. > I bet your EE guys can not give valid reasons > (or that I would be able to invalidate their reasons) > > 1) obsolete - dont fool your self, FPGA vendor soft-core do get > obsoleted as well! > 2) FPGA vendor soft-core is actually a client-trap vendor lock-in > > now, Cortex-M3 RTL code is really beatiful, small and easy, basically > you just > select ITCM/DTCM sizes, number NVIC interrupts, and you are done > and it useable on any FPGA with enough resources, so truly vendor > independant > > optionally there free Coldfire (original freescale) core available for > Cyclone-3 > (for other FPGAs there is license fee) > > Antti > > Hi Antti, Do you mean Cortex-M1? (Cortex-M3 do not have TCM interface :-) ) Hi Jack, Currently you can access Cortex-M1 with : 1) Actel - Design using Libero and SoftConcole - Library of AMBA components available - Royality free - Allow easy reuse of IP blocks with AMBA interface Note: The early version of SoftConcole does not support TCM, please check with distributor to see if this has been improved. 2) Altera Cortex-M1 Altera kit (contact Arrow Electronics) - Design using SOPC builder - Library of peripheral components available - Free royal for 1000 units (according to http://www.altera.com/products/ip/processors/32_16bit/m-arm-cortex-m1.html) Note: Use Avalon interface with SOPC builder rather than AMBA 3) Synplicity used to have a program called ReadyIP. Not sure if they are still doing it. 4) License it directly from ARM - access to RTL so you can retarget to any FPGA - include scripts for major FPGA tools and FPGA devices - Use AMBA interface, and TCM support - peripheral IP not included (you can possible find those in your FPGA development tools, on the Internet, books, or develop them on your own). Hope this helps. regards, Joseph |