Is KEIL better or IAR for STM32 ?

[fgokcegoz]

Posted on October 21, 2013 at 16:37

Hi Friends...

Is KEIL better or IAR for STM32 ? Code optimization, speed compiling, Debugging !!! What do you think about this ?

[Tesla DeLorean]

Posted on October 21, 2013 at 16:39

What do you think about this ?

Yes, but it's very subjective.

[jj2]

Posted on October 21, 2013 at 16:58

As our leader states - ''it is subjective.''   And - throwing (again) caution to the wind - my vote with IAR...

Only you know your (present!) taste, sensibilities, goals. (likely beyond the factors you listed)  Price often intrudes upon this process... Our group evaluated both - 7 years past - chose 256KB ''paid'' IAR. Both vendors offer, ''code-size limited'' - free applications for your trial/familiarity.  Suspect such use - by you/your team (and then vote) proves best means...

[Tesla DeLorean]

Posted on October 21, 2013 at 19:05

My preference list is Keil, Rowley, then IAR.

GNU/GCC is fantastically cheap, and will trump most arguments on that basis alone, if you've got a lot of cash to burn GHS comes with a unicorn, or at least you'd expect it too. As processors have gotten faster, most compilers seems to have gotten significantly slower. Pity there isn't a ''Borland C'' for ARM, current versions of Keil seem to have ingested a sloth.

[jpeacock2399]

Posted on October 21, 2013 at 21:24

If you can afford it Keil would be the way to go.  For us poor folk GCC and a good JTAG works just about as well, providing you are far enough along the learning curve.  My total tools budget was less than what Keil costs so the decision was easy.

I've used GCC on STR750 and STM32 parts for about 7 years for various commercial projects.  GCC is fairly easy to use (Eclipse and Yagarto gives you 90% of the setup).  The hassle comes with the JTAG.  OpenOCD does work but you need a LOT of patience.  I switched to the pricier Segger because the time lost using OpenOCD was too expensive.   Jack Peacock

[Tesla DeLorean]

Posted on October 21, 2013 at 22:10

+1 On Jack's Segger and Yagarto observations. Segger's stuff is good, tool-chain agnostic, and they have a trade in program. The J-Link commander has a very low level debug/step function. Yagarto+Make function are extremely serviceable. I've posted a couple of projects using them here, and also used them to build boot loaders for ATMEL parts.

If you're ''old school'' embedded you can probably live with Serial or SWV based debugging and diagnostics. If you need source level debugging, and stepping, you've got to look at some of the more professional IDE. I haven't used GDB, so don't let that colour you options. The ST-Link used on many of ST boards is also quite serviceable. The Utilities permit BIN/HEX flashing, and output of diagnostics/telemetry from the SWV channel.

[tonyyu]

Posted on October 23, 2013 at 21:00

I use Keil, and happy with that. However, it seems to me that ST DSP and/or other example code are compiled with IAR first before port to Keil. Not sure if this is a rule, but I found that some ST code have IAR proj setup and no Keil's. It's not difficult to port but it may saves you some time if you go with IAR directly.

Sink.

[Tesla DeLorean]

Posted on October 23, 2013 at 22:05

However, it seems to me that ST DSP and/or other example code are compiled with IAR first before port to Keil. Not sure if this is a rule, but I found that some ST code have IAR proj setup and no Keil's.

This is certainly true of one of the STM32L-Discovery demo, if you're familiar with your tools it's not a huge issue. More's the pity these tools can't import a text based project from another chain.

[emalund]

Posted on October 25, 2013 at 00:53

having used both with ST I will give Keil six points and IAR half a dozen points

Erik