On Thu, Nov 13, 2014 at 4:50 PM, Jan Hubicka <hubi...@ucw.cz> wrote: >> gcc/ChangeLog: >> >> 2014-11-12 Martin Liska <mli...@suse.cz> >> >> * Makefile.in: New object file is added. >> * cgraph.h (symbol_table::allocate_cgraph_symbol): Summary UID >> is filled up. >> * cgraph_summary.c: New file. >> * cgraph_summary.h: New file. > > Since I am trying to get rid of the cgraph prefixes for symbols (keep it for > the graph only) and the summaries can be annotated to variables too. Even if > it > not necessarily supported by your current implementation, lets keep API > prepared for it. So I would call it symtab-summary.* for source files and > symtab_summary for base type (probably function_summary for annotating > functions/cgraph_edge_summary for annotating edges?)
Also please don't use underscores in filenames but dashes, thus cgraph-summary.[ch], not cgraph_summary.[ch]. Richard. >> diff --git a/gcc/cgraph.h b/gcc/cgraph.h >> index e2becb9..588b6d5 100644 >> --- a/gcc/cgraph.h >> +++ b/gcc/cgraph.h >> @@ -1225,6 +1225,8 @@ public: >> int count_materialization_scale; >> /* Unique id of the node. */ >> int uid; >> + /* Summary unique id of the node. */ >> + int summary_uid; > > What makes summary_uid better than uid? > >> diff --git a/gcc/cgraph_summary.c b/gcc/cgraph_summary.c >> new file mode 100644 >> index 0000000..9af1d7e >> --- /dev/null >> +++ b/gcc/cgraph_summary.c > > And why do we need this file? It will need license header if really needed. > > The implementation seems sane - I will check the actual uses :) > Please send the updated patch though. > > Honza
