MSP430 GCC - does it support 20 bit addressing?

[tml 19]

Hi,

I got tired fighting with inlined assemblies to get the access to ROM over 64kB and after thorough considerations decided to move to CCSv6. But, I noticed that CCSv6 brought cooperation with MSP430 GCC (http://www.ti.com/tool/msp430-gcc-opensource) into CCSv6 and I wonder is this compiler will support 20-bit addressing natively (with memory model) as opposed to inline assembly.

 

Do you have any experience with it?

 

Best Regards,

tml

[spirilis 1,265]

The old mspgcc didn't except a last dev release @@pabigot made, but after that RedHat took on the project at TI's funding to re-do the msp430 gcc port (there were legal issues with having Peter's mspgcc supported by the official GCC project). This new RedHat port is still a bit... rough around the edges, but it does support 20-bit addressing. Use -mlarge to enable it (I think you need to add this manually to the CCS project properties).

 

Sent from my Galaxy Note 10.1

[tml 19]

The old mspgcc didn't except a last dev release @@pabigot made, but after that RedHat took on the project at TI's funding to re-do the msp430 gcc port (there were legal issues with having Peter's mspgcc supported by the official GCC project). This new RedHat port is still a bit... rough around the edges, but it does support 20-bit addressing. Use -mlarge to enable it (I think you need to add this manually to the CCS project properties).

 

Sent from my Galaxy Note 10.1

As for the mspgcc 4.6.3 I painfully got to know it doesn't support it :-) Possibly after turning on the optimization (don't know yet, didn't check), the far rom access function based on inlined assemblies stopped working properly and I thought CCS will be a better solution. But somehow I still feel gcc would be better for me if only it supported access to far rom from the C level.

 

As for the memory model for the sake of efficiency I would rather stick to small memory model and just declare some pointers as 20-bit ones, like mentioned at http://sourceforge.net/apps/mediawiki/mspgcc/index.php?title=Gcc47:20-Bit_Design#Data_type_attribute___a20__

[spirilis 1,265]

As for the mspgcc 4.6.3 I painfully got to know it doesn't support it :-) Possibly after turning on the optimization (don't know yet, didn't check), the far rom access function based on inlined assemblies stopped working properly and I thought CCS will be a better solution. But somehow I still feel gcc would be better for me if only it supported access to far rom from the C level.

 

As for the memory model for the sake of efficiency I would rather stick to small memory model and just declare some pointers as 20-bit ones, like mentioned at http://sourceforge.net/apps/mediawiki/mspgcc/index.php?title=Gcc47:20-Bit_Design#Data_type_attribute___a20__

Ok keep in mind that page and wiki references @@pabigot 's mspgcc, and any talk of 20-bit refers to his last dev release (4.7.0-based), so it won't work in the new RedHat msp430-elf-gcc. I'm not sure if what you're thinking about is possible with msp430-elf-gcc's current codebase.

 

Sent from my Galaxy Note II with Tapatalk 4

[pabigot 355]

The 4.7 dev version of mspgcc which supports 20-bit addresses tried to support mixed memory models so code could be small with data large, or vice-versa, but GCC wasn't architected for that sort of thing so there can be issues.

 

AFAIK the Red Hat port supports only everything-16-bit and everything-32-bit, with -mlarge having a pretty big impact on code and data size (all pointers and size_t become 32-bit). I don't have a lot of experience with it, though, since a feature I need that's in the CCS6 version still hasn't been made available in the upstream source repositories.