From: Howard M. <ho...@fi...> - 2005-09-28 17:52:16
|
On Wednesday 28 September 2005 08:17, Beau E. Cox wrote: > Hi Howard- > > ======= At 2005-09-26, 14:38:02 you wrote: ======= > > >On Saturday 24 September 2005 10:54, you wrote: > > > >Hi Beau, > > I've worked a little more with the files. > > > > First, a question in the devices.inc file. The structure has a ram > > size field. Should this be actual ram size, or highest address? (e.g. > > 0x1000 vs 0xfff). Perhaps Raphael can speak to that. > > It should be actual ram size; My conversion is wrong - for some reason I > always spit out 0x1000 - I will fix asap! Good catch Howard ;) > > > I find myself having to reinstall the virgin cvs to compare things. I > > put forth the suggestion of adding the include line in the stock sdcc > > device.c file, but got no responses. My thinking is still that we should > > add the line, but keep the existing structures in the device.c file to > > ease the transition. In the interim, perhaps you could make a backup of > > the original device.c file as part of the script. > > I would personally vote for individual include files (one for each > >processor), collected by into the devices.inc file. > > I can go either way - if the project ever gets to the state where all pics > are converted, would not one file be better? Ideally, but as Microchip will be adding new devices, it will (hopefully) never be 'done'. It seems to me we should have device entries somewhat classified as to there tested/untested status. If Microchip changes file formats, or copyright becomes an issue, the situation could change. In the short term, I'd like to see an evolutionary path that doesn't break the CVS, but lets people integrate your work easily. > > > Forgive my newbieness, but regarding the mnemonic definitions, don't > >these have to match the gputils header files? I asked the group if people > > use sdcc without gputils, and there was 1 response that non-pic > > programmers frequently don't. So maybe I should have put a finer point on > > it. > > I guess I solved this one by building the gpasm include files with my > mnemonics; doing a 'make install-asm' from the PIC/devices dir > replaces the gpasm headers. > > I am planning to 'hand convert' the bit names to the 'standard' my > introducing a XML input file that specifies them for the conversion; more > work, but there seems to be no way around it. Hmmm, the script I'm working on only uses the include files from gputils. I didn't know the .dev files existed until your script. I wonder if I could output a file of legacy/gputils mnemonics, keyed by address that could be integrated into your XML files by address, bit number etc. I don't know where the legacy code will come from, presumably ports from windows systems using the same .inc files in gputils. There seem to be discrepencies between the Microchip .inc files and the Microchip documentation, so I doubt there will ever be a definitive list. It would be great if the system allows for multiple names sanely. I had just started playing with the intermediate file idea when you announced the Allpics project, but if you want to tell me what such a file should look like, I can try to provide it. I'll send you an updated version to play with on the next debug cycle. > > > I'll keep after it as time permits. Me too ;) > >Thanks > >Howard > > Aloha => Beau; > be...@ha... > 2005-09-28 Thanks, Howard |