Re: RedHat Toolchain for small chips

On Tue, Jan 14, 2014 at 02:08:53PM +0100, Emil Renner Berthing wrote:
> Right, so the workaround is something like
> $  msp430-elf-gcc -mmcu=msp430g2231 -Os -static -nostdlib -Wall -Wextra
> simple.c -o simple.elf `msp430-elf-gcc gcc -print-libgcc-file-name`
> ..and then add my own assembly to clear .bss etc. and jump to main().
> 
> I guess I had just hoped there would be a standard way of doing this, so
> everyone doesn't have to implement that bit again.

An objdump of the newlib .o could tell where the bss is eaten up, then
selective surgery on newlib/libgloss can be done (do the MSP *really*
need stdio?). Other cruft (the C++ and java one) can be ifdeffed out in
crtstuff.c (in libgcc). And that printf is pure evil :D (not that
I often use printf on MCUs...)

Compiling the target app with LTO (verify that ISR are actually
emitted! attribute used helps) or better yet do a single compilation
unit with the world declared static (essentially what Byte Craft does
from forever) gain *huge* code space with -Os. Yes, I'm aggressively
spacing optimizing the code :D

Also unsigned int is obviously easier to process than an unsigned char,
most of the time: the MSP is almost pure 16 bit (except for load/store),
the stack is slotted on even addresses and registers needs to be anded
with FFh to ensure 8-bit semantic. At 4 bytes (IIRC) for each unsigned
char parameter used, the gain is significant.

-- 
Lorenzo Marcantonio
Logos Srl

------------------------------------------------------------------------------
CenturyLink Cloud: The Leader in Enterprise Cloud Services.
Learn Why More Businesses Are Choosing CenturyLink Cloud For
Critical Workloads, Development Environments & Everything In Between.
Get a Quote or Start a Free Trial Today. 
http://pubads.g.doubleclick.net/gampad/clk?id=119420431&iu=/4140/ostg.clktrk