diff --git a/docs/settings.md b/docs/settings.md index a2cbd89b..670c69f2 100644 --- a/docs/settings.md +++ b/docs/settings.md @@ -22,7 +22,7 @@ Many rows below are developer tools that only act together with another setting | `bind_stats` | `1` counts bindings and descriptors per draw of each stage and how much of the bound state changes between draws. | Adds per-draw bookkeeping to the render thread and extra log output; frames get slower. | Caution | | `contact_debug` | `1` draws the contact-shadow factor in grey in place of the picture. | The frame shows a grey debug image while it is set; has no effect when contact shadows are off. | Caution | | `core_diag` | `1` records why ready guest threads wait for a virtual core and for how long, with a histogram every 60 presents. | Turns on host-call and SVC depth tracking and writes many log lines (one per present over 25 ms); frames get slower. | Caution | -| `barrier_merge` (default `0`) | Experimental. `1` keeps a render pass open through the game's Barrier calls that ask only for fragment ordering and texture invalidation, when no draw of the pass can write a storage buffer or image; the next draw that samples a written target gets a by-region barrier in the pass (the feedback layout). In scenes with about a thousand such barriers a frame this cuts the passes from about 1200 to 130 and the GPU time from 24.5 to 7.8 ms. | Edges differ by about a pixel in screenshots; not fully explained. | Caution | +| `barrier_merge` (default `0`) | Experimental. `1` keeps a render pass open through the game's Barrier calls that ask only for fragment ordering and texture invalidation, when no draw of the pass can write a storage buffer or image; the next draw that samples a written target gets a by-region barrier in the pass (the feedback layout). In scenes with about a thousand such barriers a frame this cuts the passes from about 1200 to 130 and the GPU time from 24.5 to 7.8 ms. | The game's barrier orders every pixel, the in-pass by-region barrier only the pixel a draw writes: a driver that runs it per screen tile can show flicker or squares at the edges of water, decals or particles. | Caution | | `menu_cursor_log` | `1` logs what the menu cursor code reads at every screen change. | Only adds log lines. | Safe | | `perf_slow_ms` (default `50`) | Frames at least this many milliseconds long get a `perf: slow frame` log line with the time blocked per cause. | Changes how many frames are logged; a small value fills the log. | Safe | | `replay_check` | `1` reads every texture read-back a second time after a full drain of the replay queue and compares the two. | Stalls the replay queue at every read-back, so frames get much slower; mismatches are only counted and logged. | Caution | diff --git a/totk.ini.example b/totk.ini.example index 560f6797..61f74cf4 100644 --- a/totk.ini.example +++ b/totk.ini.example @@ -509,9 +509,9 @@ # none of the pass's draws can write a storage buffer or image, making the writes visible to the next draw with a by-region # barrier instead of ending the pass. Scenes where the game issues about a thousand such barriers a frame (the water in # the Nachoyah Shrine, terrain decals) go from about 1200 passes and 10000 layout changes to about 130 passes and 150; -# the GPU time there fell from 24.5 to 7.8 ms a frame. The picture is the same to the eye and differs at edges by about a -# pixel in screenshots (not fully explained: report any glitch in water, decals or particles). Needs the feedback layout -# (the default). Read at start. Default: 0. +# the GPU time there fell from 24.5 to 7.8 ms a frame. The game's barrier orders every pixel, the one inside the pass only +# the pixel a draw writes: a driver that runs it per screen tile can show flicker or squares at the edges of water, decals +# or particles (report any). Needs the feedback layout (the default). Read at start. Default: 0. # barrier_merge = 0 # Extra lines in the log, comma separated: svc, host, ipc, fs, gfx, perf, sched, audio, present or all. The menu's # Performance log sets perf (frame timings). Default: none.