On Mon, 18 Dec 2023, Jakub Jelinek wrote: > On Mon, Dec 18, 2023 at 03:08:49PM +0100, Richard Biener wrote: > > The following improves the manual work needed to make a -gimple dump > > valid input to the GIMPLE FE. First of all it recognizes the 'sizetype' > > tree and dumps it as __SIZETYPE__, then it changes dumping vector types > > without name from 'vector(n) T' to 'T [[gnu::vector_size(n')]]' which > > we can parse in most relevant contexts (and that's shorter than > > using __attribute__). Third it avoids a NULL_TREE TMR_STEP when > > it would be one, an optimization that's re-done when generating RTL. > > Should we care that with -pedantic/-pedantic-errors it can be rejected > then before -std=c23?
Probably not, you need to use -fgimple when compiling anyway. > We could print also __extension__ before the element > type, but not sure if it would be the right spot in all cases. Adding __extension__ would make using __attribute__((vectors_size(N))) shorter. I'm still thinking of how to lessen the most annoying pain WRT -gimple dumping - the lack of global variable (and type) declarations. The current plan is to have a global hash-map with used/dumped non-local types and decls and incrementally handle them as preamble for dump_function_to_file. Of course that doesn't help with anonymous types which is what the vector patch is for. We also have quite some <unnamed-unsigned:17> _1; stuff that's even impossible to work around. But as always IMHO incrementally improving GIMPLE FE things makes most sense. > Otherwise LGTM. Thanks, Richard.
