Repository navigation
feat: add a 1920x1080 docked output mode - #1144
Conversation
Add 1920x1080 as a docked output size, with output_res=3 and HOS_OUTPUT_RES=3. Keep the saved identifiers for all existing modes and show the corresponding internal resolution in the Graphics menu. Add a game-free regression covering saved modes, the override, handheld fallback, and render-target budgets. On Windows, 119 checks passed and 12 were skipped. The four failures also reproduce on unmodified main. Smoke-tested the same runtime/menu patch on v2.0.7 on Windows with an RTX 4060 Laptop GPU (driver 617.42). The startup run reported 1920x1080 NVN window textures, a 1920x1080 scene at fixed 100%, and frame presents. No gameplay performance comparison was run. Refs NotRexed#1134 Assisted-by: OpenAI Codex Co-Authored-By: OpenAI Codex <codex@openai.com>
hckrman101
left a comment
hckrman101
left a comment
There was a problem hiding this comment.
Nice addition. I checked the other users of the output size, and the new mode needs nothing more. The game patches in nx2/oot_patches.py all go through hos_output_size(). The graphics-scale answer in nn_applet.c gives x1 for any chosen output. The render-target pool only grows at 4K, so 1080p keeps the shipped one. The ultrawide width in hos_main.c follows hos_output_size(1). Keeping the saved ids stable and sending an out-of-range value to 0 instead of clamping it to 1080p is the right call too.
One note on the motivation: the fps gain over 2048x1152 is mostly the ~13% fewer pixels, which the Resolution row can already trade. I think the stronger case is image quality on 1080p screens, the most common ones. Today the whole frame, HUD and text included, is composed at 2048x1152 and then scaled down to the window, which softens it, and 1080p output maps 1:1. Might be worth saying so in the readme row.
| struct Out { | ||
| uint32_t w, h; | ||
| }; | ||
| constexpr Out kOutputs[] = {{2048, 1152}, {2560, 1440}, {3840, 2160}, {1920, 1080}}; |
There was a problem hiding this comment.
With 1080p appended, the menu now cycles 2048x1152 → 2560x1440 → 3840x2160 → 1920x1080, so the smallest size comes after 4K. To keep the saved ids as they are (3 = 1080p) but show the sizes in order, the row could step through a display-order table, e.g. {3, 0, 1, 2} (with Left/Right mapping through it), or tx_output_res / the row could use a small id↔position mapping. Same for the readme table, which lists it last.
There was a problem hiding this comment.
Maybe we should break away from this int setting now before we add other of other resolutions too
There was a problem hiding this comment.
Thanks, the kOutputOrder table does it: Left/Right, wheel and the value halves all go through set_row, so they step in size order, and the saved output_res / HOS_OUTPUT_RES ids keep their meaning. Fine with keeping the int setting here, a different format can be its own PR if more sizes come.
|
|
||
| host = (ROOT / "host/nn_applet.c").read_text(encoding="utf-8") | ||
| menu = (ROOT / "gpu/vk/vk_menu.cpp").read_text(encoding="utf-8") | ||
| assert '"Output resolution", &opts.output_res, 0, kLastOutput, 1, tx_output_res' in menu |
There was a problem hiding this comment.
minor: this asserts on the exact text of the menu row, and the fixture below slices nn_applet.c / vk_menu.cpp between marker strings, so an unrelated rewording or reorder of those files breaks the test. The behavioural checks in the fixture are good. The row's range could be checked through kLastOutput (which the fixture already asserts) rather than the source text. Also a heads-up that #841 rewrites this same row (labels plus per-value help lines), so whichever lands second will need a small merge there.
There was a problem hiding this comment.
Thanks, dropping the row-text assert and testing the step order covers it. The marker slicing is still there but that is fine for now.
|
Thanks for the review, @hckrman101. I agree that sharper HUD and text on a 1080p display are the stronger reason to add this mode. I'll make that the main point in the README and keep the benchmark in the PR as a measured result. I'll put 1080p first in the menu and README while keeping the saved IDs stable. I'll also replace the exact menu-row text assertion with checks for the selectable modes and their order. I'd keep the integer setting for this PR to preserve compatibility with existing configs. If we expect to add more resolutions, I'm happy to discuss a more extensible format separately. Thanks for the heads-up about #841. |
Step through the sizes from 1080p to 4K, with the saved IDs intact. Describe native HUD and text scaling in the README and test navigation in both directions, including the endpoints. Assisted-by: OpenAI Codex Co-Authored-By: OpenAI Codex <codex@openai.com>
hckrman101
left a comment
hckrman101
left a comment
There was a problem hiding this comment.
LGTM. 1080p now comes first in the menu and readme with the saved ids unchanged, and the readme leads with the 1:1 HUD/text point.
One small nit, not blocking: the output row now stops at its ends rather than wrapping like the other multi-value rows, so at 3840x2160 Enter/A (which acts as Right) and a click on the right half do nothing. Since the arrow hides at the end that seems fine, just mentioning it in case it wasn't intended.
Brings in the 1920x1080 docked output (NotRexed#1144), Windows' UTF-8 code page (NotRexed#1191) and the saves guide (NotRexed#1190). Conflicts, resolved for both sides: - gpu/vk/vk_menu.cpp: the Output resolution row keeps this branch's place among the scaling rows, its short help and its restart flag, and takes main's four sizes (kLastOutput) and smallest-to-largest order (step_output_choice, after row_fixed and the Upscaler's steps in set_row). Its value no longer says "restart" (the restart rows show "(after restart)" and offer Restart now); 2048x1152 stays "(game's)". 1920x1080 has its own per-value hint. menu_init reads HOS_OUTPUT_RES through output_choice, then restart_init. DLSS's modes and DLAA at 1920x1080 need nothing else: their sizes are twelfths of the output (1280x720, 1120x630, 960x540, 640x360), and vk_dlss.cpp takes the output's size from the picture, not from a list. - gpu/vk/vk_menu.h: output_res lists 3 1920x1080 and still says it is read once, at the start. - readme.md: the Output resolution row with 1920x1080 in this branch's table; the scaling notes take main's 1920x1080 sentence and HOS_OUTPUT_RES values. - tests/run_all.py: both test_menu_restart.py and test_output_resolution.py. Also: - gpu/vk/vk_dlss.cpp: ngx_log no longer converts NGX's lines from the ANSI code page. With the manifest's UTF-8 code page it is UTF-8 (tests/test_utf8_paths.py forbids CP_ACP in gpu/vk). - tests/test_output_resolution.py: the labels without "restart", the menu's pieces it compiles taken around the Render resolution's hint and the Upscaler's rows (upscaler_rows.h, an Upscaler stand-in), DLSS's mode sizes and DLAA at 1920x1080 checked, and hos_wide_out defined (hos_rt_pool_size reads it; it did not link without it). - tests/test_rumble_menu.py: a stand-in for step_output_choice, which set_row now calls.
…, Windows paths in two tests) - test_output_resolution: hos_wide_out, which nn_applet.c reads since the ultrawide pool fix. - test_rumble_menu: step_output_choice, which set_row calls since #1144's Output row. - utf8_paths_test.c: the log is written through text-mode stderr, CR LF on Windows: compared without the CRs. - test_oot_saves: USERPROFILE redirected too, and skipped on Windows (its discovery is the Linux one). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Summary
Add 1920×1080 as a docked output size for the full game image and HUD. The new mode uses
output_res=3orHOS_OUTPUT_RES=3. The existing saved mode IDs and the default output are preserved.Testing
Compared the two output sizes in the same patched v2.0.7 executable on Windows with a Core i7-13700H and RTX 4060 Laptop GPU (driver 617.42). Both headless Vulkan runs loaded the same Hyrule Field save and held the same camera at fixed 100% render scale with frame generation off. Measured one 25-second window per mode after loading. The native present trace measures time between game present-call returns.
Verified the 1920×1080 render dimensions in the log and checked HUD placement in local gameplay captures. The regression test covers mode IDs, the environment override, handheld fallback, and render-target budgets. The full local source suite reported 119 passed, 12 skipped and four failures that also reproduced on unmodified
main(test_nup_paths.py,test_fps_overlay_order.py,test_precompile_compute_guard.py, andtest_program_answer_domains.py).Refs #1007 and #1134.