On Monday 26 March 2007 1:14 pm, Axel Zeuner wrote: > Hi Avi, > On Sunday 25 March 2007 15:40, Avi Kivity wrote: > > Axel Zeuner wrote: > > > A full featured converter (cvtasm) has a lot of dependencies: it has to > > > support all hosts (M) (with all assembler dialects M') and all targets N, > > > i.e. in the worst case one would end with M'x N variants of it, or M x N > > > if one supports only one assembler dialect per host. It is clear, that > > > the number of variants is one of the biggest disadvantages of such an > > > approach. > > > > Perhaps a mixed approach can be made for gradual conversion: for > > combinations where cvtasm has been written, use that. Where that's > > still to be done, have dyngen generate call instructions to the ops > > instead of pasting the ops text directly. > Perhaps, but I am not sure, if the changes required for generating calls with > parameters to functions instead of copied code in dyngen are much smaller > than hand written code generators. Furthermore one would surely lose some > performance. > > Kind Regards > Axel
On a related note, I have this vague urge from time to time to get qemu to build with tcc. Haven't even come close to making it work yet, of course... :) Rob -- Penguicon 5.0 Apr 20-22, Linux Expo/SF Convention. Bruce Schneier, Christine Peterson, Steve Jackson, Randy Milholland, Elizabeth Bear, Charlie Stross...
