Hi Caleb, On Sat, 4 Jan 2025 at 14:54, Caleb Connolly <[email protected]> wrote: > > Hi Jerome, > > On 03/01/2025 12:13, Jerome Forissier wrote: > > Replace the init_sequence_r function array by direct calls. > > bloat-o-meter stats before/after on u-boot.bin for configuration > > xilinx_zynqmp_kria_defconfig are the following: > > > > - With LTO: > > add/remove: 87/128 grow/shrink: 8/23 up/down: 19743/-20923 (-1180) > > Total: Before=1067882, After=1066702, chg -0.11% > > > > - Without LTO: > > add/remove: 0/23 grow/shrink: 2/0 up/down: 795/-1112 (-317) > > Total: Before=1119494, After=1119177, chg -0.03% > > > > Signed-off-by: Jerome Forissier <[email protected]> > > --- > > [...] > > > + INITCALL(stdio_add_devices); > > + INITCALL(jumptable_init); > > + CONFIG_IS_ENABLED(API, (INITCALL(api_init);)) > > + INITCALL(console_init_r); /* fully init console as a device */ > > + CONFIG_IS_ENABLED(DISPLAY_BOARDINFO_LATE, > > + (INITCALL(console_announce_r); > > + INITCALL(show_board_info);)) > > Not that the array was extremely readable, but I think this is > significantly worse. Particularly as my IDE will no longer grey-out > sections for config options which are disabled. > > I would honestly rather still see pre-processor #if's here, or at least > something better than this semicolon-in-macro stuff.
Hmmm I think you have a point. I was actually thinking about having something like INIT_CALL_CONFIG(DISPLAY_BOARD_INFO_LATE, console_announce_r) i.e. a macro to combine the two things and the semicolon. But I decided it is really no better than this patch so didn't mention it. So another option would be to keep the #ifs as you say, with everything else about this patch left alone. I have the same issue with the IDE (I suspect some clever integration could be written to fix it). But I have changed my mind and think #ifs are better, in this case. Regards, Simon
