Repository navigation
RetroRecomp v1.2.4 — scan levels per console and options in the main window
·
725 commits
to main
since this release
RetroRecomp v1.2.4 — scan levels per console and options in the main window
- Discovery scan levels. Every console tab in Options now sets how long
discovery runs: Quick (one short pass to check a setting), Standard
(the qualified default, unchanged from before), Extended and
Thorough, which add one or two reproducible pseudo-random play routes of
18,000 frames seeded from the exact ROM and compared with the reference
before any learning. Each tab also offers its own passes and frames (empty
fields follow the shared values) and a short note on what is recommended and
why. On Mega Drive and Super Nintendo, Extended runs a 12,000-frame attract
test with one route and Thorough the former advanced scan; on NES, Extended
is the former advanced scan and Thorough adds the routes; Master System,
Game Gear, Game Boy and Neo Geo gain the routes for the first time. On the
command line,--scan-levelapplies to a whole run; the former switches keep
their meaning. - Sharp X68000. The default scan is now Standard: main route plus mouse
and keyboard exploration. Thorough keeps the three 36,000-frame routes that
qualified the library, and the exploration routes now replay side by side
within the game's share of the processor: two games converted together in
6 minutes instead of 21 on the test machine. - Options in the main window. The Options button no longer opens a second
window: the tabs take the place of the queue, like the compiled-games view,
and close with "Save and close the options". - Finished means finished. When a batch ends, the log shows a framed
"RECOMPILATION FINISHED" line with the totals and the elapsed time, the
window rings and changes its title for twenty seconds, the status line
starts with "Finished", and the command line prints the same summary.
The generated games do not change: regenerating is not needed, and quick
states, settings, libraries and cartridge saves are unaffected. Because the
converter code changed for every console, it may report existing games as made
by an earlier converter; regenerating them produces the same executables. The
pseudo-random routes broaden the tested paths of the first level; they do not
explore whole games, and none of the levels certifies hardware fidelity.
Linux remains qualified on Debian 13 / glibc 2.41; other baselines need testing.