From Hamsterworks Wiki!

Revision as of 23:31, 27 September 2011 by User (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

If you follow these steps through, you will be well on your way with uClinux. You will have

  • installed the complete tool chain
  • created a system from source
  • tested it
  • added a custom I/O port
  • added a custom executable to the boot image
  • successfully interfaced to the outside world

The system genreated will have approximately these attributes

  • 100MHz CPU, no floating point, no hardware divide, no MMU
  • 32MB RAM, with about 29MB free
  • The console running over the JTAG cable
  • No persistent storage
  • A little under 20% of the FPGA resources dedicated to the processor, leaving 80% for your own logic.


The use-case for a DE0-nano + uClinux

Let me start by saying that running Linux on your DE0-nano doesn't make much sense - but don't let that stop you.

Why does it not make sense?

  • You will need to license a full NIOS II processor core to allow it to run untethered.
  • You will need to configure the device every time power is applied or develop a boot loader to load the boot from the EPCS flash
  • The development environment is very heavy
  • The big benefit is networking, and the DE0 doesn't have any Ethernet connectivity out of the box.
  • Writing Linux device drivers is a big undertaking (but you don't really need to).
  • Without a MMU to keep things in check bugs cause the system to fall to bits really quick.
  • Offchip SDRAM is a lot slower than using embedded RAM blocks
  • If you what o use Linux as you are super-committed to open source, using the closed-source FPGA tools might be not to your liking.

When does it make sense?

  • If your device needs a good reliable TCP/IP stack
  • If you need a multitasking OS
  • If you want to use this as a development platform for custom hardware
  • If your development team are Linux bigots, or want to develop a complex applications on PCs before moving to the FPGA platform
  • If you are trying to remain platform agnostic - today NIOS II, tomorrow ARM or PowerPC. For example, proof of concepts or if the hardware platform doesn't yet exist
  • In a R&D or lab environment, where time and ease of implementation is more important than creating a marketable product

What is the alternative?

If you are committed to the Altera/NIOS II platform then the better option is using the standard Altera "Eclipse for NIOS II" environment.

  • You can use on-chip memory for your code, which is configured on startup - no boot loader required
  • You can still use the SDRAM for data
  • It is fully integrate into the IDE workflow
  • It is very compact, with only the features you need.
  • Provides enough hardware abstraction, but not too much.

With that said, running uClinux on a DE0-nano isn't too hard, especially with a bit of guidance.

Building the base uClinux environment

See UClinux/Building for the Terasic DE0-nano on Fedora 14

Creating custom hardware on NIOS2 based systems

See uClinux/Expanding your DE0-NIOS2 system for how to add a simple parallel output port to the base system. This example drives LEDs on the DE0-nano, which isn't configured in the 'simple' hardware build.

Hacking a custom program into the distribution

See uClinux/Hacking a app for a program to drive the LEDs added above.

Personal tools