CFV1 with CW6.1 floating point support for printf

[KAR]

Hello,

I've looked at several posts talking about shrinking the footprint of printf-like functions by setting #define switches in header files.  There is also a "Tips & Tricks" segment about this discussed here.  I wish to do the opposite.  I want full floating point support for printf and scanf, but I'm running into problems making it happen.

First of all, here is what I have:
CodeWarrior 6.1 Special Eddition (Under "Help->About" it states the version is 5.9.0, Build 2404)
MCF51QE64 for my target MPU

Some other related threads that I have found useful are:
CW 6.3 for ColdFire and floating point
Simple Printf on CF question

As it stands by default, I can compile printf functions with the "%f" term, but nothing will be displayed.  (I'm actually using sprintf and whatever putchar function I choose.)  Also, printf works fine for printing integer values (%d, etc.).

I have tried adding print.c from "C:\Program Files\Freescale\CodeWarrior for Microcontrollers V6.1\lib\hc08c\src" and modifying libdefs.h from "C:\Program Files\Freescale\CodeWarrior for Microcontrollers V6.1\lib\hc08c\include", but I getting a lot of errors.  Specifically in libdefs.h I include the switch "#define LIBDEF_PRINTF_FLOATING   1".

I think others have had success with this process before, but maybe there is a quirk for the Coldfire V1 that I am not aware of.  I also wonder if I should be grabing printf.c from the HC08 source directory.

Thank you in advance for any guidance you can provide.

Kevin

[CompilerGuru]

Be sure not to mix the tips for HC08 and for CF. The way to go for the CF V1 is to use another library,
I think the second thread you mentions tells you which one (sorry, I don't now thouse lib names out of my memory).

Using the printf.c for the HC08 (the out of of lib/hc08c/src) or libdefs.h (out of lib/hc08c/include) is not applicable for the CF V1, instead use the provided MSL library for ColdFire.

Daniel


BTW: I still think that not using printf for small the devices is the right thing to do, so maybe a specialized float printing routing and a simple string to UART output is the way to go.

[KAR]

Thank you for your reply CompilerGuru.  I thought the HC08 sources might not be completely compatible with the V1, but that was the only place I could find printf.c.  There isn't a subdirectory for that source for the V1 that I could find.

The default libs that were added to my project are:

fp_coldfire_nodiv.a
C_4i_CF_RegABI_MSL.a
C_4i_CF_RegABI_Runtime.a

I found information about the MSL from here:

{install}\Help\PDF\ColdFireV1_Build_Tools_Reference.pdf

Then under "ColdFire V1 Tools --> Using MSL Libraries".  I replaced the C_4i_CF_RegABI_MSL.a with the C_4i_CF_RegABI_MATH_MSL.a choice because it contained floating point manipulation.  This didn't change anything.  I then tried replacing the other two libraries with the various other available libraries without any luck.  Is there a specific one I need to choose?  Is there anything special I need to do besides changing the libraries in the project and recompiling?

I understand that printf functions are bloated and I am not concerned in this application.  I just wanted to save a little time instead of writing any specialized functions.

Thanks,

Kevin

[kef · Accepted solution]

Did you see these files:

{install|}\ColdFire_Support\msl\MSL_C\MSL_Common\Src\printf.c

{install|}\ColdFire_Support\msl\MSL_C\MSL_Common\Src\printf_tiny_IO.c

I don't know what's the right way. I added printf_tiny_IO.c to the project and defined

#define LIBDEF_PRINTF_FLOATING 1

also I had to unsuppress code for Ten() routine, there's a #if 0 at line 197.

[KAR]

Thanks a lot kef!  That did the trick.

I had to also "#define LIBDEF_PRINTF_PREC 1" as well in order for it to compile.  That was fine as I needed the precision feature anyway.

I also found that I did need the C_4i_CF_RegABI_MATH_MSL.a library instead of the default or I would get a link error for "frexp".

Thanks again.  I hope this may be of use to others.  For a heads up though, adding the floating point support added about 8 kB to my code.  Ouch

[kef]

You are right, LIBDEF_PRINTF_PREC 1 is necessary to make it compileable.