GDB-compatible Debugger for the Freescale Coldfire MCF52235CAL60 V2 Ethernet Processor

Why not use directly the Freescale Toolchain (that is based on Eclipse)?

Bye Jack

PEEDI from Ronetix matches the specs, but it is in the same price range as Abatron's device.

There are a number of open-source BDM interfaces (eg. OSBDM from P&E[1], USBDM[2]), but they are pretty much only supported with CodeWarrior. TBLCF, the originator of all these open-source interfaces, had some degree of support in the BDM tools[3] package, but AFAIK it hasn't been maintained in a long while.

-a

You mean the one that costs $5000?

...and that only runs under Windoze?

Thanks much, Anders, for these specific references!

Dunno about the first part, but if it's based upon Eclipse why would it only run under windows?

tim

Eclipse is not the whole show. It is using other tools, like compiler, assembler, linker and debugger, which all need to run under the OS with the Eclipse setup. Any of the necessary tools may be sufficient to ruin running under something else than Windows.

tim,

I'm just going by what they say on their site. As Tauno said, it's probably because one or more of the components (e.g., the compiler toolchain, or even more likely, the driver for the debugger) only runs under Windoze.

The special edition is completely functional and free (until 64KB of compiled code for the Coldfire). At a certain point (last time I checked was 6 month ago),there was a version for Linux, but it seems that there isn't anymore.

Bye Jack

(I don't have any secret information from Freescale here - this is just what I have understood from tool availability, forum posts, etc.)

Freescale CodeWarrior (Eclipse version) used to run under Linux, but support was dropped several versions ago. There was never any clear explanation given - my guess is that some PHB went cost-cutting based on figures of the moment rather than trends and predictions. It is not uncommon for testing and support to cost more for a Linux port, simply because there is a much wider variety of Linux distros and setups, and the user base was smaller than for Windows (I heard figures of 15-20%). They also quoted figures showing that many who downloaded the Linux version also downloaded the Windows version (I certainly did that - like many places, we have a mixed environment and it is good practice to download releases for multiple platforms at the same time) - so they "reasoned" that Linux users could just switch to Windows. This is despite the growing trend for technical people, such as embedded developers, to drop Windows as fast as Windows is dropping them (we want operating systems, not overgrown telephones).

There were no good technical reasons for the decision to drop Linux - almost all the parts were fully cross-platform (the editor, the debugger, both CodeWarrior's own compiler and the gcc ports used for several targets, most wizards and "beans", and even P&E's debugger interface). There were a few parts that were not included in the Linux version, but nothing important.

As for the $5000 price tag, that's basically for C++ support. (It's actually for the full "professional" version, but as that's the only version with C++, the C++ support is the main reason for buying it.) This is something that has always annoyed me with CodeWarrior. For small or medium projects, the free size-limited version is often fine - but if you want C++, it's $5K. I can well understand that they want to make a bit of money from professional tool users, though the main purpose of CW is to help sell Freescale chips, but this jump is a bit steep.

Freescale have now changed this philosophy considerably for their Kinetis line. The new Kinetis Design Suite is free of charge on Windows and Linux, with debugger support for OpenOCD and P&E Micro, C and C++ (just gcc - they have dropped their own compiler), and lots of other goodies.

Unfortunately for you, Coldfire is a dead-end architecture. Freescale haven't killed it yet - they tend to continue support for a long time after a family has left the mainstream. But they only have room to concentrate on two architectures - ARM and PPC. And it is the ARM users that are growing, and where they need good, cheap, cross-platform tools to compete with other ARM vendors (especially in the Cortex-M space).

David,

Thank you for this thorough discussion. As it turns out, I proposed a Kinetis K60 as an alternative based on other considerations (basically, more MIPS, more memory, and less $$$) during the proposal phase of this project, but they shot it down because "they already have a basic system running on the V2." I sensed the dead-end of the Coldfire (I think it's basically a 68K with some special peripherals, no?), and really didn't want to work on this processor. I also don't like the looks of the \ uTasker OS they're using, but I admittedly know little about it.

The Coldfire was a complete from-scratch re-implementation of the 68k ISA. At the time it was designed, the 68k was a well-loved ISA, widely supported by tools, OS, and other software. Many people still consider it the best designed ISA ever to be used in mainstream processors. But the 68k family were getting long in the tooth - there were lots of inefficiences, it was hard to take advantage of newer cpu features, and the existing cores were big, slow, and power-hungry. So Freescale made a few modifications to the ISA (such as to cut out the more complex addressing modes and to add a few new instructions), then made a modern cpu core family design using up-to-date ideas and covering a number of size/power/cost/performance points (the v1..v4 cores). The result was a rather elegant cpu - the Coldfire has a very nice core and ISA, is power and space efficient, and can be implemented in modern high-density microcontrollers.

The only problem is that Freescale was competing with itself with the PPC core, the M-core (remember them?), their DSP cores (especially their hybrid core), and their newly licensed ARM cores. For a good while, Coldfire had a solid place in the market - for several years it was the top choice for NAT routers and it was the key reason for the existence of ucLinux (now the MMU-less architectures in mainline Linux). But MIPS stole the small network market, and ARM7 and later Cortex M took the high-end microcontroller market, and Freescale's own PPC was a better choice for high-end networking and automotive. Coldfire found a new niche in a family of low-power devices that were pin and peripheral compatible with a family of 8-bit devices - letting users design the one board and choose low-end or high-end cpu cores. But there is nothing Coldfire can do there that can't be done cheaper, faster, and lower-power with Kinetis Cortex M4.

Coldfire isn't going anywhere. It is not going away, but it is not getting any (significant) attention from either Freescale or third parties. But as long as you are happy with what you have got, and don't need a path to bigger/faster/smaller/cheaper future devices, then it is a solid if unexciting choice.