DDE

Is there any public (or private, under NDA) informations about jopcodes used by DDE?

Ah, I’ve just posted on your mailing list regarding this. Damn, I could have written it here in English. ;-)

To recap – a brief examination of parts of the DDE suggest that:

• objasm has no need for such things – there’s a 1:1 correspondence between input instructions and output opcodes.

• CMHG appears to contain parts of the output AOF within itself, so I’m guessing it builds an AOF directly from the input parameters.

• c++ and cfront perform a conversion of sorts from C++ into something that cc will accept. Sounds a bit of a hack…

This leaves:

• cc – it is known that the Norcroft compiler contains a front end, middle end, and back end. For understanding the C code and creating intermediate codes, for optimising the resultant code, and then for generating processor specific opcodes. While ROOL’s compiler is, obviously, aimed at the ARM; the fact that old timers refer to it as Norcroft gives away it’s heritage. It was developed by Codemist who have used the same technology to create rather impressive compilers for a range of processors. As such, it not only makes sense for the compiler to use an intermediate code format, but it also implies that this process is internal to the compiler itself.

Which says to me that unless there are some hidden command line options, even if the “jopcodes” were described and detailed, there would be no way to make use of them. It is simply a stage the compiler goes through between C and ARM code.

unless there are some hidden command line options

Nothing stands out. Through looking through the code in Zap, some of the messages are interesting: Dynamic initialisation too weird (structure). I wonder what triggers that? Or tentative definition confusion, that’s me on a Monday morning.

At the end of the cc (offset +&94F1C) there appears to be a list of VFP instructions. Is this for inline assembly? I’m asking because it would be nice if there was a switch in cc that permitted the use of VFP instead of the default FPE for FP ops in C.

Which says to me that unless there are some hidden command line options, even if the “jopcodes” were described and detailed, there would be no way to make use of them. It is simply a stage the compiler goes through between C and ARM code.

Arg, not cool :|