[FifteenPen]

This gets the game booting and fully playable, with Joy-Con 2 motion and calibration support (#2106).

What and why

Nintendo Switch Sports Resort (04008E001F098000, update v65536, SDK 22, NVN2, Nintendo EPD's sead / agl / gsys). On main alone it does not boot. With these it plays on the Mac to its title, Select User, the menus, golf and tennis at its own 60 fps:

• SDK 21 virtual output device properties and event-only stream open (Wonder Edition movie) #1561 (GetVirtualOutputDeviceProperties: without it the game aborts with 0xa99 after the safety screen);
• File calls on a mount only the SDK holds (a RAM disk) go to the SDK: Nintendo Switch Sports Resort boots past its VoiceText dictionaries #2105 (its VoiceText RAM disk), Switch 2 hid: controller icon colours and the controller IMU (samples, sensor fusion setup): Nintendo Switch Sports Resort boots, and its Joy-Con calibration completes #2106 (the SDK 22 hid calls it makes at boot, and the Joy-Cons' IMU samples: the SDK's sensor fusion calibrates them and gives the game their motion);
• The fragment's shading-rate size (system register 0x84) reads 1 x 1: Nintendo Switch Sports Resort's sea is no longer black #2103 (the shading-rate register: the black sea), A read-only mount of the game's own save data opens its committed save: Nintendo Switch Sports Resort no longer aborts on SaveDataErrorType::UpdateOrderForNotUsable #2104 (its own save read-only: the career abort);
• A task shader's payload past Metal's 16 KB goes through a device buffer: Switch Sports Resort's six refused task / mesh shaders build #2264 (the six task / mesh shaders refused on main, 6 Mac mesh-stage shaders failed to compile for Nintendo Switch Sports Resort (04008E001F098000, v65536) #1753; no draw with them seen yet);
• Integer reads of game uniforms are volatile loads: bare, they made Metal's backend take minutes and run out of memory on Switch Sports Resort's pipelines #2271 (integer uniform reads as volatile loads: on main its material pipelines take Metal's compiler minutes each and the prebuild runs it out of memory, so it compiles pipelines at every start);
• A colour clear under a partial channel mask leaves the other channels: Switch Sports Resort's outfits were black in its menus #2273 (a clear under a partial channel mask keeps the other channels: outfits black in the menus).

For Miis, #2274 with the player's own files: Mii head models (system data 0400000000000849) and the Mii database (system save 8000000000000030) from DM_SYSTEM_DIR. Without them the game runs with no Mii heads and only the default Miis.

#1609 and #1921 are no longer needed: main covers them as f02d047 and 30c9b16. A session on main with those and the PRs above (no nvn_draw_ctl) ran 815 s and exited cleanly.

The profile is the survey's guess and the NSP script's heap check (a 7.47 GB heap, depth_mode zero), unchanged after play. pso_manifest.tsv is the host's per-device list from the sessions played (5,934 rows).

Checked

• Tests, on this branch (main 5e1c1ae + this change): python3 -m unittest dm/tools/tests/test_docs.py dm/tools/tests/test_game_profile.py dm/tools/tests/test_architecture.py dm/tools/tests/test_nsp_chain.py; python3 dm/tools/docs_patches.py (no change: the game has no patches).
• Game run: a Mac app from this profile, on main 04811a6 with the pull requests above: boot, title, Select User, the menus, golf and tennis at 60 fps (frame interval 16.67 ms, GPU about 10 ms in a 2560x1440 window).
• Game run again on 2026-11-10, on main ced3612 with the pull requests above plus Translator: Ampere forms of PRAGMATA's shaders #1800 and the Ampere instructions checked against a Jetson Orin: what differed from the hardware, and the PRs that fix it #2040 series: past the safety screen, 815 s, exited cleanly; 11,059 of 11,059 shaders built.
• Game run on 2026-11-10 evening, on main ced3612 with the pull requests above, Integer reads of game uniforms are volatile loads: bare, they made Metal's backend take minutes and run out of memory on Switch Sports Resort's pipelines #2271, A colour clear under a partial channel mask leaves the other channels: Switch Sports Resort's outfits were black in its menus #2273, A player's own system data and Mii database from DM_SYSTEM_DIR: Switch Sports Resort draws Mii heads and lists the console's Miis #2274 (with a console's 0400000000000849 and a one-Mii database), Translator: Ampere forms of PRAGMATA's shaders #1800 and the Ampere instructions checked against a Jetson Orin: what differed from the hardware, and the PRs that fix it #2040 series: Select User and Customize with the Mii listed and lit outfits, a golf round; the second start after the shader rebuild took 8,446 of its prebuild's pipelines from the archive and built the rest with 0 failed compiles.
• Devices and OS: Apple M4 Pro, macOS 27.2.
• Not checked: iOS, online play, the other sports.

Docs

• The game note and its row in the games table; findings are in the reference entries of the pull requests it links
• No game files, keys or data extracted from a game in the change or the attachments

[FifteenPen]

Update: the issue with visuals in the Golf mode is now solved.

[FifteenPen]

This is now fully working and ready to review/merge.

[seedds]

Landed on main as f4b7e77 (profile, note, games table row and README line), with you credited as co-author: the note keeps your findings (the outfit material reading 0 from R and A of the blurred shadow target, the Mii system data, the six refused mesh shaders). The host fixes landed from the matching pull requests: #2103 (b581efc), and bymaxmarc's #2376, #2377, #2380, #2384 and #2385 for the RAM disk, the read-only save mount, the output device, the controller style and the IMU, which cover #2105, #2104, #1561 and #2106. The profile has render_cond on instead of #2273: see that pull request. Your pso_manifest.tsv was not taken: none of its render or mesh pipeline keys matches the current host's, so the profile has a list recorded on the current host. #2264, #2271 and #2274 are still to be judged on their own. Thank you!