Hello Nathan, On Wed, 9 Sep 2015, Nathan Sidwell wrote: > I've applied this patch to port some cleanups, mainly formatting and loop > idioms from the gomp4 branch.
This patch that you committed to trunk in September 2015 forcefully disables generation of line number information, undoing a part of a patch by Bernd Schmidt that was committed to trunk by Thomas Schwinge in August 2015. I couldn't find a corresponding email for the gomp4 branch (apparently the same change was committed to the branch just a few hours prior to trunk), so I wonder if there was any technical reason to disable linenumber info, or just some kind of oversight? Would you like if I prepared a patch for trunk that backs out the override? It's useful not only for debugging, but also for instruction-level profiling (which is my main motivation at the moment). (relevant bit of the problematic patch quoted below) Thanks. Alexander > 2015-09-09 Nathan Sidwell <nat...@acm.org> > > * config/nvptx/nvptx.md (call_operation): Move bound out of loop. > (*cmp<mode>): Add assembler spacing. > (setcc_int<mode>, set_cc_float<mode>): Likewise. > * config/nvptx/nvptx.c (nvptx_option_override): Override debug > level. (the hunk for the above entry) > (write_func_decl_from_insn): Refactor argument loops & comma emission. > (nvptx_expand_call): Likewise. > (nvptx_output_call_insn): Likewise. > (nvptx_reorg_subreg): Add spacing. > > Index: src/gcc-mainline/gcc/config/nvptx/nvptx.c > =================================================================== > --- src/gcc-mainline/gcc/config/nvptx/nvptx.c (revision 227587) > +++ src/gcc-mainline/gcc/config/nvptx/nvptx.c (working copy) > @@ -102,6 +102,8 @@ nvptx_option_override (void) > flag_toplevel_reorder = 1; > /* Assumes that it will see only hard registers. */ > flag_var_tracking = 0; > + write_symbols = NO_DEBUG; > + debug_info_level = DINFO_LEVEL_NONE; > > declared_fndecls_htab = hash_table<tree_hasher>::create_ggc (17); > needed_fndecls_htab = hash_table<tree_hasher>::create_ggc (17);
