kevin diggs <diggskevin38@xxxxxxxxx> writes: > On Wed, Apr 27, 2011 at 7:25 PM, Ian Lance Taylor <iant@xxxxxxxxxx> wrote: >> kevin diggs <diggskevin38@xxxxxxxxx> writes: >> >>> Does anyone know the reason -fdump-tree-original does not work for C >>> in the 3.4.6 compilogical era? Compiling the file as C++ (-xc++) seems >>> to produce useful output. >> >> As far as I can recall, it was invented for C++ in the first place. ÂIt >> was only supported for C with the introduction of tree-ssa in gcc 4.0. >> > Is 'it' the dump-tree option or the tree representation? The -fdump-tree-original option. > I have jumped to the conclusion that the absence of this option > implies that there are significant differences in the intermediate > tree representations for C and C++. There are differences, but they are not large. > Since you mention 4.x, The 'dumped representation' in 4.x is quite > different than what is printed in 3.4. Is this due to the internal > representation being different (if any response involves gimple vs > generic, please give some details - i.e. does the 4.x dump represent > generic or gimple?)? Or is it just a different 'design decision' that > the simplified C provides a better representation? The internal representation is different. The -fdump-tree-original dump represents the frontend specific tree representation. It is similar to GENERIC but includes frontend specific tree codes. > To try to clarify my confusion, could 3.4 do a tree dump for C but it > is just not implemented? I think so but I don't really remember and I don't bother to keep gcc 3 online. That was a really long time ago. Ian
