Lichii

Émulateur build-228

Émulateur Nintendo Switch 2 pour Mac Apple Silicon.

Lichii est un émulateur Nintendo Switch 2 en cours de développement, présenté par son auteur comme très expérimental. Chaque build est publié en DMG et en ZIP pour Apple Silicon.

Dernière mise à jour :

Plateformes

Nintendo Switch 2

Architecture

Mac OS X - macOS Mac OS X - macOS Apple Silicon Apple Silicon 64 bits 64 bits

Téléchargement et liens

Historique des versions

228
  • Nouveauté majeure : ajout d’un mode 60 FPS pour Mario Kart World (actif uniquement sur les cutscènes par défaut), avec un basculement automatique vers 30 FPS en cas de ralentissement.
  • Correction critique : suppression définitive du flash blanc à l’écran complet dans Ocarina of Time (lié à une optimisation de FPS pour Mario Kart World).
  • Correction macOS : compatibilité corrigée avec MoltenVK (sous-ensemble de portabilité activé, gestion des lectures de buffer en sécurité, réutilisation des captures pour les captures d’écran).
  • Correction GPU : suppression des artefacts de texture (reflets métalliques déformés) sur les GPU sans support natif des textures ASTC (ex. : RX 6650 XT).
  • Correction AMD : élimination des blocages GPU sur RDNA 3/4 (Linux) avec Mesa < 26.2.5, et optimisation des pipelines pour éviter les reconstructions inutiles au démarrage.
  • Amélioration performance : réduction des latences lors des changements de zone (Windows) et optimisation des calculs matriciels logiciels sur GPU sans cœurs dédiés (ex. : Steam Deck).
  • Nouveauté utilisateur : intégration de SteamGridDB pour les icônes de jeu (optionnel, nécessite une clé API) et rumbles personnalisables sur les déclencheurs de manettes.

Changes since build 177

Built from nx2 9fbffe253b plus 11 pull requests that merged into nx2 after it (the test/all-in-one-3 branch at be1f4d9c84). Build 177 was 87d84d513c. That's 152 merged changes, 69 of them from community contributors.

The first launch of each game after updating takes longer. A lot of shader code changed in this release, so caches are partly rebuilt once: the compute precache runs its pass once more (#1620), pipelines that an older cache saved with the wrong texture type are rebuilt (#1443), NVIDIA on Linux fills a new, larger driver cache folder (#1301), and the LSFG shader cache rebuilds (#1625). Later launches should be fast again. If a second launch of the same build still rebuilds most shaders, send the log from Help → Report a Bug (Ctrl+F12): we're looking into reports of this.

Big fixes

  • Ocarina of Time: the full-screen white flashing from build 177 is gone. Build 177's groundwork for higher frame rates in Mario Kart World let shaders read stale data with Skip CPU Inner Invalidation on, and the screen flashed white. Players traced it to that change on an RX 6900 XT, and it showed up on an RTX 3090 too. The change is reverted; it gained Mario Kart World only about 0.5-1.2 fps (#1651).
  • Ocarina of Time no longer freezes for good after an ocarina warp. Warping to a place you'd visited (Temple of Time, Desert Colossus) froze picture and sound within a minute or two, and a forced stop then showed error 2010-0212. A game thread, the GPU thread and its second half were waiting on each other's locks. GPU memory updates are now handed off after the locks are released. On the author's RX 7900 GRE (RADV) it froze almost every time before; afterwards, warp loops and a long session ran without one. From Nanoc6 (#1218).
  • Mario Kart World: steady resolution and a higher frame rate. The game sizes its dynamic resolution from GPU timestamps, and in the form Lichii reported them the GPU looked almost out of time, so the resolution flipped every few frames and render targets were rebuilt over and over. Mario Kart World gets real GPU ticks again; other games are unchanged. On Mario Bros. Circuit with an RDNA 2 GPU on RADV it went from 18.6 to 22.1 fps, and image creations dropped from about 40 to about 1.3 per frame. Ported from Rhys Denno's upstream Eglantine change (#1817).
  • Donkey Kong Bananza no longer crashes on AMD right after the golden-banana loading screen. The game's voxel-world pipeline uses a geometry shader feature AMD doesn't have, and Lichii's emulated version declared inputs it can't pass and more outputs than AMD's limit of 128. AMD's Windows driver doesn't report that; it kills the process, so the log just stopped. The emulated shader now passes only what the previous stage writes. On an RX 6700 XT the unpatched build crashed every time. From sopedound (#1110).
  • AMD on Linux: no more GPU hang on RDNA 3 and RDNA 4 with Mesa older than 26.2.5. A RADV bug kept stale state when binding a mesh shader that doesn't read its grid size, and none of Lichii's native mesh shaders read it, so the GPU hung a few seconds into Ocarina of Time's intro. On those drivers mesh shaders now read the grid size once, which avoids the bug, and native mesh shaders stay on. On an RX 9070 XT with RADV 26.2.4 it hung 2 out of 2 times before and not at all after. From Molier (#1203).
  • Ocarina of Time on RX 6000 cards (Linux): no more GPU crash on mesh draws. RDNA 2 cards on RADV lost the device with a GPU read fault at a fixed address, on several players' RX 6600, 6650 XT and 6950 XT machines, always on native mesh draws. RDNA 2 on RADV now uses the slower but safer compute fallback for mesh draws. The root cause isn't known yet; RDNA 3 and newer, AMD's Windows driver and other GPUs are unchanged (#1732). That fallback also draws grass and bush shadows again: it only drew triangles, so the point program that fills the foliage shadow map was dropped on every GPU that takes it. From nemiward (#863).
  • Ocarina of Time: a likely fix for the GPU hang on NVIDIA. Two compute shaders could hang NVIDIA GPUs a few minutes into play. The way they pick the buffer behind each memory access was the one unusual thing in them. First every branch was merged into a single return (#1681), which still hung after about 390 s on an RTX 5090; now, on NVIDIA's driver only, the shaders pick the buffer with one select instead of a branch per buffer. rtrevisanuto found and tested that approach: a 529 s run had no hang, with falling leaves and particles drawn right. It was tested on Linux, so players still need to confirm it on Windows (#1757).
  • Ocarina of Time on Intel Linux (Mesa): no more black screen after the intro movie. One compute shader never finished compiling, because each memory access checked all 32 buffers the shader binds and the compiler spilled without end. Each access now checks only the buffers its address can come from, and the shader builds in about 9 s. The reporter confirmed the compile hang is gone (#1807).
  • AMD on Windows: boots no longer rebuild thousands of pipelines. AMD's own driver cache doesn't keep the kind of pipelines Lichii builds, so every boot rebuilt about 5,800 of them at about 120 ms each. Lichii's own pipeline binary store now stays on for AMD's Windows driver: on an RX 6650 XT the time to the graphics device went from 867-876 s to 17-30 s from the second boot on. From Reiixan (#1406). Closing a game no longer trims that store hard enough to drop a third of a precache's pipelines: the boot after a precache went from about 14 minutes of rebuilding to a first frame in about 30 s. From Reiixan (#1407). And Loading Shaders now waits only for compute pipelines and the last two sessions' pipelines, and builds older ones in the background, so a returning player no longer waits for every pipeline of the last 16 sessions; "Build all cached shaders" still builds everything first (#1803).
  • Yoshi and the Mysterious Book: the book page no longer slows to 2 fps and crashes. On the Chapter 1 book page one GPU buffer kept swallowing its neighbours, grew past 3.5 GiB and was copied whole each time, until memory ran out. Large buffers that only share a page with a new one are no longer joined to it. On an RTX 5070 Ti the page went from 2-12 fps and 21 GB of memory to 57-61 fps at 6.6-7.8 GB. From JohnsGoated (#1534).
  • Donkey Kong Bananza: a likely fix for the random freeze. Presenting a frame could block while holding a lock the rest of the GPU syncing needed, and the picture stopped. Those actions now run with the lock released. It isn't confirmed in the game yet. Ported from MythosOfs's banachii fix, which adapts Andrew's Eglantine change (#1700).
  • Mario Kart World on Intel: no more GPU crash while loading on Linux. Two of the game's loading shaders run a barrier that part of the workgroup never reaches. NVIDIA lets that through, but on Intel's Linux driver the rest of the workgroup waited forever and the device was lost about 40 s into loading. Lichii now finds the barriers only part of the workgroup reaches and loosens only those, so the barriers the game's lighting needs stay (#1677). On Intel's Windows driver, barriers are now kept whenever the whole group takes the same path, a likely fix for the washed-out, overexposed picture there. Neither is tested on Intel hardware yet (#1742).
  • Ocarina of Time: no more green and red speckles on RTX 20 cards. With Ocarina of Time anti-aliasing set to Game, Turing GPUs (RTX 2060 to 2080 Ti) showed speckles because NVIDIA's compiler miscompiles the cooperative-matrix path on Turing, which NVIDIA has confirmed. Turing now runs those operations in shader emulation, which on an RTX 2070 removed the speckles at about the same fps. RTX 30, 40 and 50 keep the faster path. From Maximel3105's patch and A/B test, with pamontoya-coder's bisect (#1694).
  • Kirby Air Riders: no more GPU thread crash a few minutes in. Textures with no memory behind them get placeholder addresses past the end of device memory, and a memory hint read far out of bounds at those addresses. Such hints are now ignored. On a Steam Deck, the route that crashed about four minutes into the first lesson ran into a race for over twelve minutes. From cerroverb (#1568). A crash while translating one of its shaders is fixed too: an analysis walked deep loops recursively and ran out of stack. From kdegonia-star (#1616).
  • Closing a game during "Loading Shaders" no longer freezes Lichii. The close stopped the shader load, but the guest cores still started, and the services starting on them deadlocked with the services being shut down. That's fixed, along with what it left behind, so launching a game again in the same window boots normally. From PantACRO4life (#1417).

New features

  • Mario Kart World: 60 FPS mode at normal game speed. A new "Mario Kart World: 60 FPS mode" setting, off by default, runs races at 60 fps instead of only cutscenes, one 1/60 s step per frame, so the game never runs fast. "Drop to 30 FPS when slow", on by default with it, switches to the game's 30 fps mode at normal speed after three slow frames in a row and tries 60 again later, so a slower PC gets 30 fps instead of slow motion. On an RTX 4070 Laptop GPU, races ran at 96-100% of normal speed with the fallback on. It applies after restarting the game. From thatrandomalt (#1704).
  • Report a Bug. Help → Report a Bug, or Ctrl+F12 (rebindable), writes one zip under 19 MiB with your description, logs, system and GPU details, settings, mods, the last 10 s of frame times, a screenshot and the next frame with its shaders. User names are removed, but the zip holds the game's textures and shaders, so send it privately to the maintainers and never in a public channel (#1647).
  • Solid NSZ files play. Their sections are unpacked through a small temporary file, and a failure is logged with the file's name. Ported from n0ctbyte (#1734). Idle solid NSZ files give their decoder memory back, so a base game with several updates no longer fails to open (#1737), and the game list opens a file's parts only when needed: two solid NSZ games with 2 GiB programs went from about 44 s and 8.5 GB of temp files to under 2 s to scan (#1784).
  • Custom artwork and names for every game format, without rewriting the game. Changing a game's icon or name used to rewrite the whole NUP file, about 25 GB for Ocarina of Time, and froze the window. Both are now stored per program ID in the data folder and can be changed while a game runs. From PantACRO4life (#1401).
  • Game artwork from SteamGridDB, with your own API key. Off by default. With a key in Configure → UI, games with no icon get artwork, saved as their custom artwork. Ported from PantACRO4life (#1717).
  • Trigger rumble on controllers with trigger motors. A new Trigger rumble option in Controls → Vibration (off by default, strength 1-100%) plays the light-motor part of each side's rumble on that trigger. On Windows it needs Controls → Advanced → "Enable XInput 8 player support". Tested on an Xbox Wireless Controller over Bluetooth. From samaBR85 (#1583).

Performance

  • High GPU accuracy keeps the faster split GPU thread. High is the desktop default, and it turned the GPU thread split off. The split now runs at every accuracy level and keeps the safety checks High adds. danieljesus measured Ocarina of Time on a Steam Deck at High: 16.61 to 18.57 fps standing and 13.32 to 15.01 fps panning. Built on danieljesus's tests, compared against sof006's split-at-High branch (#1808).
  • Linux AMD (RADV): Mesa keeps its own shader cache. Lichii no longer keeps a second pipeline cache on top of Mesa's, as it already did on NVIDIA and AMD's own driver. The original PR measured Ocarina of Time's Castle Town going from 34 to 51 fps. Setting MESA_SHADER_CACHE_DISABLE brings the old behaviour back. Ported from jschert's port of upstream Eglantine PR 714 (#1653).
  • Integrated GPUs and Steam Deck: the texture cache no longer gets stuck clearing itself. It compared all of Lichii's GPU memory, not just textures, against a limit meant for textures, so it stayed in its most aggressive mode and kept deleting and re-creating thousands of images. Now only textures count. On a Steam Deck OLED in Kokiri Forest, panning went from 11.62 to 14.74 fps and peak GPU memory from 2.3-2.9 GB to about 1.7-1.8 GB. From janvitos (#823).
  • Older AMD cards on Linux: no full image decompress after every copy. On RADV with Polaris and Vega GPUs, every copy into or out of an image made the driver decompress the whole image, which cost Ocarina of Time about 40 ms a frame on an RX 580. Colour images are now created uncompressed on those GPUs. On its own this took Kokiri Forest from 9.5 to 13.8 fps on an RX 580. From blackxzin (#950).
  • Windows: fewer stutters after zone changes. Thousands of memory protection changes a second waited on one lock while the game remapped memory in bulk, so the GPU thread stalled after every zone change. They now try without the lock first. In Ocarina of Time on an RX 6650 XT, frames over 100 ms went from 37 to 10 on the same route, and the worst 10 s window from 25 to 35 fps. From Reiixan (#1780).
  • Faster with High GPU accuracy or a single GPU thread. Draw-state reuse now also works when the GPU runs on one thread. In Donkey Kong Bananza at High it went from 39.10 to 42.33 fps, measured with some of the author's other changes in both runs. From DuskDarf1 (#1641).
  • Ocarina of Time: fewer waits when the game reads GPU results back. The game reads about 220 GPU-written pages every 5 to 40 frames, more than the early-download limits allowed, so most reads waited behind the GPU's queued work. The limits are raised. jpkto measured Castle Town's heavy route going from 19.10 to 17.75 ms per frame, with p99 from 31.7 to 28.2 ms. Ported from jpkto (#1795).
  • Ocarina of Time: no more 2-3 second freezes when a compute shader rebuilds. When a compute program had to be translated again mid-game (Hyrule Field grass culling), every GPU channel waited for the driver to compile it; the three worst freezes of a 66-minute session were these rebuilds. The rebuild now compiles on a worker while the old version keeps running. From VirusEx (#995).
  • Guest writes skip the texture cache lock on pages that never held a texture. In The Duskbloods on a Steam Deck, standing still in the City Streets went from 119-125 to 134-142 frames per 10 s. From cerroverb (#1789).
  • Faster software matrix maths on GPUs without matrix cores. The results are rounded to half precision with one native conversion: on a Steam Deck in The Duskbloods those passes went from 21.1 to 14.7 ms of GPU time per frame, with the same picture. From cerroverb (#1283).
  • Smaller cuts: waiting for a frame's GPU fence no longer blocks a guest CPU core, from DuskDarf1 (#1639); GPU memory range lookups for large bindings use data kept at map time, ported from Kurebayashi (#1820); Ocarina of Time's new pipelines translate on the shader workers instead of the GPU thread, from PantACRO4life, found by zmckenney (#1155); and on Linux the JIT skips perf-map work when you aren't profiling, from ne1i (#1265).

Ocarina of Time

  • FPS Unlock fixes:
    • Sound and music no longer cut out above 60 fps. The sound engine got less than a full step per update, so some sounds were never released and piled up until sound stopped; above 60 fps the sound manager now runs at 60 Hz. Ported from auz466 (#1660).
    • Above 100 fps, moving platforms such as the Deku Tree raft no longer slide away under Link, and above 60 fps Gohma no longer turns more and more slowly. Ported from atk3001 (#1679).
    • Movie cutscenes stay in sync at 30 or 120 fps. They timed voices and overlays by counting frames, so the sages' lines played over each other; the count now follows the real frame step. The approach follows NotRexed's Hookshot (#1743).
    • The Scarecrow's Song plays past its first note: recording rounded each note's length down to nothing (#1745).
    • Water ripples keep moving with FPS Unlock and Dynamic FPS, from auz466 (#1180).
    • Four more counters keep real speed above 60 fps: the pickup count-up and ocarina playback no longer stall, and the step rate countdown and ambient waves no longer run fast. Ported from atk3001 (#1663).
    • Clothes and hair sway at normal speed above 60 fps, instead of about twice as fast at 120 fps. From atk3001 (#1793).
  • Two compute programs that blacked out the world now run. Lichii didn't know four instructions write a single register and rejected the programs, which left the world black on an RX 6700 XT. From kafasam (#1008).
  • Steam Deck and small AMD cards: far less memory for some textures. AMD GPUs kept ASTC textures with uneven mip sizes uncompressed, at 4-16x the size: on a Steam Deck that took about 1.9 GB for 113 MB of game data, and the system killed the emulator while walking out of Link's house. Integrated GPUs and cards with less than 6 GiB of VRAM now keep them compressed. From danieljesus (#948).
  • A huge texture array no longer fails over and over when video memory is full. Decoding one 90-layer ASTC texture needed an 890 MiB image in one piece; on a nearly full card that failed and was retried until the emulator exited. It's now decoded a few layers at a time when it doesn't fit. From marcochiusso-lgtm (#1083). Integrated GPUs use batches of about 230 MiB for it (#1823).
  • No ripple band along the right and bottom edges with Ocarina FSR on AMD's driver. AMD's driver cuts the viewport to whole pixels, so FSR's camera jitter dropped the last column and row on half the frames and the water refraction spread that edge inward (#1744).
  • No sharpening works on version 1.0.0. It left the HUD over a black scene because the base game's output shaders were rejected; they're now matched, also with dynamic resolution. From simonMettler2000 (#1137, #1778).
  • Two rejected mesh shaders now run, so their draws are no longer skipped. From snoozedDev (#1181).

Other games

  • Pokémon Legends Z-A: the picture no longer goes flat after the train station. The tonemap shader was first built while its texture slot was empty, an empty slot was compiled as a 1D texture, and that wrong type was saved in the shader cache. Empty slots now compile as 2D textures and the affected cached pipelines are rebuilt once, so you don't need to delete your cache. From JohnsGoated (#1443).
  • Orbitals has sound on 5.1 and 7.1 audio devices. On a device with six channels the game asks for 5.1 output, which Lichii refused, so it stayed silent. Lichii now plays 5.1, downmixed for stereo devices. ghzeni's test with a 7.1 headset went from no sound to normal sound. Ported from ghzeni (#1676).
  • New Super Mario Bros. U Deluxe: no black screen at Start without Mii system data. Lichii's stand-in files were empty headers and the game crashed reading them. They're now valid blank parts, so the game plays; Mii faces stay blank. From DuskDarf1, ported from ScarletNX (#1695).
  • Pokémon Pokopia and Yoshi and the Mysterious Book should hold 60 fps instead of 58.7. Lichii's timer thread sleeps through the OS, which wakes it about 0.4 ms late on Windows, and the games' timers carried that delay into every frame. Just before a game timer is due, the thread now waits the last stretch in hardware. It isn't measured on hardware yet (#1733).
  • Star Fox: no more brief audio dropouts, from myshadowww (#1298); text no longer goes missing with native async draws set to Bounded or Skip, from myshadowww, porting upstream Eglantine PR 785 (#1296); no black rectangles in the docked opening cutscene, ported from myshadowww (#1684); and the GameChat camera and chat services answer at boot, ported from Takoroka's upstream work with layouts from MythosOfs (#1750).
  • The Duskbloods: no black and purple blocks in the bloom, ported from kurebeyashi-noah (#1714), and the game keeps its own DLSS upscale instead of dropping it minutes into play, ported from kurebeyashi-noah (#1689).
  • Hyrule Warriors: Age of Imprisonment: movies no longer drop frames. The game paces its own movie frames, so Lichii now drops frames only for games that don't. Upstream measured 843 dropped frames to 0 on the title movie. Ported from Rhys Denno's upstream Eglantine change (#1818).

Shader cache and loading

  • Linux NVIDIA: the driver's shader cache no longer fills up. The driver's disk cache is capped at 1 GiB, so a large Ocarina of Time cache (about 20,000 pipelines) spilled over and every boot compiled the rest again. Lichii now gives the driver its own cache folder with a 10 GiB limit, unless you set those variables yourself. On an RTX 3050 Laptop the shader load went from 541 s plus 106 s of compute precache to 39 s from the second boot on. From espaker (#1301).
  • Cached pipelines the boot skips are built in the background, newest first. The background prebuild stopped at a memory cap and went oldest first, so on large Ocarina of Time caches 10,800 to 31,500 pipelines were left to build during play. Now every skipped pipeline goes to the prebuild, read back from the cache file instead of held in RAM. Built on Frizti's Fritchii work (#1619).
  • Pipelines a game needs right now no longer wait behind the boot prebuild (#1621).
  • The compute shader precache reruns only when it has to. It now goes by the file's contents, the shader generator and the settings baked into it, so an update that changes the shaders reruns it once on the loading screen, and an unchanged file isn't redone (Reiixan measured 508 s for 2,948 programs). Ported from Reiixan and PantACRO4life (#1620).
  • Upgrading an old shader cache no longer eats huge amounts of RAM: a legacy Vulkan cache is converted in pieces instead of held three times over. Ported from puppywang (#1654).

Graphics and GPU drivers

  • Cards out of video memory free textures instead of spilling them to system RAM. The texture cache now forces up to four clean-ups and then keeps a margin free. The author's RTX 4050 Laptop had spilled 290-420 MiB for minutes in Castle Town, dropping from 50-57 to 20-26 fps. Ported from Reiixan (#1686).
  • No more shiny, wrong-coloured textures for a few seconds on GPUs without native ASTC. The temporary buffer of an ASTC conversion could be handed to the next texture too early, so two textures' data mixed. Seen on an RX 6650 XT in Hyrule Field, and gone on the same route after the fix. From Reiixan (#1796).
  • Shaders that use matrix cores no longer race on shared memory on RTX 30, 40 and 50 and RDNA 3 and 4 (#1698).
  • Intel on Windows: the driver cache save at the end of loading, which can crash Intel's driver while background builds run, is skipped; the cache is still written when the game stops. From GoatTech-42 (#1685).
  • Snapdragon (Adreno): no 8x8 holes where the sky shows through in Ocarina of Time. Ported from Terobyte (#1682).
  • The texture cache clean-up no longer ends a pass with nothing deleted when the oldest images can't be evicted. From moonzzeyy (#1731).
  • GPU command lists aren't read stale after DMA copies and clears (#1736).

Upscaling and frame generation

  • LSFG frame generation runs on AMD and Intel GPUs on Windows. It needed 16-bit floats, which Lichii turns off on those drivers, so AMD sessions reported "unsupported GPU". Lichii now loads the 32-bit LSFG shaders from Lossless.dll when 16-bit floats aren't available. It needs a Lossless.dll that carries them (the 3.2.2 one, about 7 MB); the one Lossless Scaling ships today doesn't. Restores the path from Eden (#1625).
  • DLSS super resolution works without the NVIDIA app. The NVIDIA DLSS panel downloaded only two of the four files DLSS needs, so DLSS fell back to FSR 3.1 while the panel showed a green check. It now installs the missing plugins and only shows ready when everything is there. Found by marcochiusso-lgtm (#1753).
  • Ocarina of Time: no black squares around far birds and particles with DLSS. NaN and out-of-range pixels in the colour sent to DLSS spread into black blocks; they're now zeroed and clamped, which removed the squares on an RTX 5080. From darknagel's patch (#1672).
  • Ocarina of Time: FSR 4 and DLSS no longer drop out when you turn the camera in the Water Temple. From njmdietrich (#1659).
  • Frame generation no longer drops into slow motion at high refresh rates. With FSR3 4x at 240 Hz, Ocarina of Time fell to 11-17 fps after the menu, because each generated frame's wait was added on top of the wait for the screen. The spacing now counts from each frame's deadline (#1779).
  • FSR 3.1 updated to 3.1.6. With sharpening on, it can no longer push pixels negative, which should mean fewer black specks and dark halos next to bright edges (#1624).

Crashes and stability

  • A stop that gets stuck no longer freezes Lichii. When a game's teardown stopped making progress, Lichii killed the emulation thread, which could leave a lock held: 2 of 3 Ocarina of Time closes froze in testing. It now saves play time and settings and exits. Settings and play time are written to a new file first, so a crash mid-save keeps the old ones, and one more kind of Windows crash leaves a log line and a crash dump. Ported from pabCMX (#1728).
  • No crash when a render target gets no GPU memory. On a ROG Ally with a 512 MB graphics memory carve-out, the driver crashed on the empty attachment it was handed. The draw now runs without that target (#1766).
  • A GPU reset the Linux driver recovers from silently no longer hangs Lichii. Occlusion query reads waited up to 20 s each, for minutes. Lichii now reports a device loss after 1 s, with the driver's fault address in the log (#1809).
  • Crashes inside JIT code say what happened. A fault that wasn't a fast memory access ended in "FastmemCallback: unreachable"; the log and crash dump now carry the real fault and the code bytes around it (#1751).
  • Logs keep their last lines after a hang or kill: the log is flushed once a second (#1771).
  • Guest memory spans are checked page by page, so a span can't point into the wrong host memory after a partial remap. Ported from k1r4ct's Eden change (#1783).
  • Two places that could submit GPU work in the middle of a draw no longer do (#1812); no crash when a game moves a missing file to another folder, ported from Marssvoodoo (#1658); Install keys no longer crashes when you pick keys already in the keys folder, from danieljesus (#891); worker threads finish cleaning up a task before they report it done, from jfhutchi, a port of Marssvoodoo's fix (#1105).

Audio and input

  • GameCube controllers rumble in HD rumble games. On a GameCube adapter the faint, short HD rumble turned into pulses the motor can't spin up on; the motor now runs fully for any rumble above a small cut-off. From davidgith1 (#543).
  • Motion controls stop when the controller stops. After a swing, the orientation kept turning for 1-4 seconds and then snapped still. In a replay of recorded motion it now turned 8° at rest instead of 1,544°. From Lancito01 (#1396).
  • Linux: a warning when a controller connects but sends no input and its HID device can't be opened, with a udev rule to try (#1747).

Smaller fixes

  • Linux: a 16 GB Steam Deck no longer gets the low-RAM warning on every boot, from danieljesus (#1506); on Wayland the picture follows the window to another screen, from goof787lord (#1615); fast guest memory turns on on ARM Linux (#1669); the Nix shell is pinned to nixpkgs 26.05 and loads every Mesa Vulkan driver, from zibiax (#1250); builds against newer cpp-httplib, from bclaud (#1610).
  • Windows: the loading screen no longer freezes in borderless fullscreen, from BrunoSilvestre00 (#1558); the programs carry version resources and the updater isn't built with the PGO profile, which lowers the risk of antivirus false positives (#1617); the MSVC build compiles again, from sanchodecuba11-del (#1596).
  • macOS: the network interface list is filled, from krim404 (#617); MoltenVK's portability subset is turned on as Vulkan requires, from krim404 (#1368); buffer reads near a buffer's end stay in bounds on MoltenVK, from krim404, taken from dylkw's Macchii (#1375); screenshots reuse one capture frame, from krim404 (#1374); the build compiles again after warnings from the GPU trace code, one from auz466 (#1701, #1702).
  • Settings and UI: importing an old profile folder that is a link to another drive works again, from PantACRO4life (#1393); the Configure dialog fits the Controls page, from PantACRO4life (#1394); each Lichii Dark status bar chip has its own colour, from PantACRO4life (#1395); loose dump games show their size, from BrunoSilvestre00 (#1408); "Limit Speed Percent" stays as you set it after a game closes, ported from BrunoSilvestre00 (#1687); F2 opens Load/Remove Amiibo again, ported from Velken (#1716); game titles in the grid wrap to the cover's width, ported from inxomnyaa (#1718); saved room bans are no longer erased at startup, from goof787lord (#1788); About shows the commit a build was made from, from StiensWout (#1603).
  • Game files: short reads from game files and failed install copies are reported as errors, and an incomplete install is removed, from n0ctbyte (#1674); system services no longer hand games leftover bytes they didn't fill (#1802).
  • Security: the bundled certificate list drops eMudhra's emSign roots, from ctborg (#1498).
  • Android (Lichii ships no Android build): the Ocarina of Time anti-aliasing option shows in settings, ported from marinsalinas (#1720).

Codebase

  • GPU traces: a trace format with headless replay (eden-cli --replay, any Vulkan device, lavapipe included) (#1648), a recorder in the debug server and EDEN_NX2_GPU_TRACE (#1655), and replay checks with a lavapipe CI step (#1656).
  • An offline AMD (RADV) shader compiler that builds Lichii's shaders for six RDNA 2-4 chips and checks the SPIR-V for driver differences (#1650).
  • Diagnostics, all off unless their variable is set: EDEN_NX2_TC_TRACE for texture cache image life, from liomaroliveira (#1130); EDEN_NX2_VULKAN_CACHE_FLUSH, ported from Gunzmo (#1712); GPU frame time by source line, ported from josh-wirick (#1723); JIT blocks named by guest address in the Windows sampler, ported from Xiphias1337 (#1724), which no longer crashes on a bad stack walk, ported from hellasleeper108 (#1722); emulated mesh pipeline waits in the stutter trace, ported from myshadowww (#1721); debug server input for Switch 1 games, from DuskDarf1 (#1696); warm-cache tools, ported from PantACRO4life (#1713).
  • Default logs say what still builds on the GPU thread mid-game and why (#1626), and the startup log names the Qt platform and Wayland or XWayland, from goof787lord (#1268).
  • Fixes: SPIR-V dumps are no longer spliced together, from myshadowww (#1052); the verify tools give no false mismatches, from 6c696e6b (#1081); PGO training builds load Ocarina of Time again, from liomaroliveira (#1204); tests link with OpenGL off, ported from jschert (#1715); an nx2 build break is fixed (#1668); the sockets service answers numeric IPv4 name lookups, ported from n0ctbyte (#1666).
  • CI and repository: one GitHub bug report form that asks for the log, GPU driver and a video (#1644); CircleCI build checks were added and then turned off again (#1652, #1669, #1705, #1735); the AMD notes mark the FP16 switch as closed (#1726).

Known issues

  • Linux: the packages still bundle the Vulkan loader (libvulkan.so.1), which can cause Vulkan start-up problems. Affected: linux-x86_64, linux-x86_64-pgo and linux-aarch64-pgo. The workaround is to delete it so the system loader is used: for the tar.gz, delete usr/lib/libvulkan.so.1 from the extracted lichii/ folder; for the AppImage, run it with --appimage-extract, delete squashfs-root/usr/lib/libvulkan.so.1 and start squashfs-root/AppRun. A packaging fix is coming (issue #1767).
  • The first shader load after updating is slow, because many shader changes landed in this build (see the note at the top).
  • Intel Arc on Windows: pipeline builds are slow. Being worked on.
  • Ocarina of Time: purple speckles reported on some AMD GPUs.
  • Donkey Kong Bananza: black stripes on distant landscape on some AMD Linux setups.
  • Fire Emblem: Fortune's Weave: crashes on Linux with RDNA 3 GPUs (RADV).
  • Splatoon Raiders: black flickering at the development stations and in hideout cutscenes since build 130.
  • Mario Tennis Fever: characters and the court are missing on some setups.
  • Ocarina of Time on RX 6000 cards (Linux): #1732 avoids the GPU crash, but a separate freeze is still reported.

Downloads

Built from test/all-in-one-3 at be1f4d9c8432eb9e4b5944ff568304f3fd62953a: nx2 9fbffe253b plus 11 fixes that have since been merged into nx2 as well.

Linux

Linux packages come as an AppImage and a tar.gz. Both are self-contained and need glibc 2.39 or newer (Ubuntu 24.04, Fedora 40, Debian 13, Arch). The tar.gz holds the same files unpacked, for systems without FUSE: extract it and run lichii/lichii.

Build Type Standard PGO (Recommended) Notes
amd64 AppImage · tar.gz AppImage · tar.gz Make the AppImage executable (chmod +x) and run it
ARM (aarch64) AppImage · tar.gz AppImage · tar.gz

Windows

Windows packages are in-place zip files: extract and run lichii.exe, or eden-cli.exe from a terminal.

Build Type Standard PGO (Recommended) Notes
amd64 / x86_64 zip PGO zip Both built with clang-cl and ThinLTO; the PGO zip adds a profile trained on Ocarina of Time

macOS

Build Type Standard PGO (Recommended) Notes
arm64 (Apple Silicon) zip macOS 15 or newer. Not notarized: right-click and choose Open the first time, or run xattr -dr com.apple.quarantine lichii.app. eden-cli is inside the app, in Contents/MacOS

The PGO builds are the same programs compiled with profile-guided optimization. Linux uses clang with Eglantine's profile; Windows uses clang-cl and ThinLTO with a profile trained in Lichii on Ocarina of Time, so other games gain less. Prefer them when your CPU supports it.

177
  • Changement de nom officiel : suppression définitive de la référence "yuzu", incluant le bundle ID macOS (io.github.zortos293.lichii), nécessitant une nouvelle autorisation pour les permissions caméra et entrée.
  • Correction majeure pour Ocarina of Time : suppression des artefacts visuels (forêt flottante, framerate instable) et des ombres déformées sur l’eau, grâce à des ajustements dans les shaders et les maps de lumière.
  • Stabilité améliorée : suppression des plantages lors du passage d’un jeu à l’autre (ex. Donkey Kong Bananza après Ocarina of Time), des freezes après le logo Unreal (Orbitals), et des pannes GPU après des sessions longues.
  • Optimisations matérielles : support du DLSS Frame Generation (RTX 40/50) pour Ocarina of Time (Windows uniquement), et amélioration des vibrations DualSense HD via USB (rumble côté gauche/droite séparé).
  • Corrections spécifiques macOS : builds ARM64 (Apple Silicon) compilées à nouveau après corrections de type et de références, avec un bundle ID mis à jour.
  • Améliorations pour Hyrule Warriors : ajout d’un mode 30 FPS dynamique et suppression des bugs de chargement (hangs, ennemis invisibles, incompatibilité avec les mises à jour obsolètes).
  • Performances globales : gains de FPS pour Mario Kart World (jusqu’à +8 FPS sur Mario Circuit) via optimisations JIT, cache GPU et gestion mémoire.

Changes since build 160

Built from nx2 87d84d513c. Build 160 was 2eb5f08. That's 86 merged changes.

The yuzu name is gone. The build system, source folders, Qt frontend, translations, emulator libraries and docs now say Lichii instead of the name inherited from yuzu. Your settings, saves and shader caches are kept: config keys, data folders and cache formats don't change, and the few on-disk names that still carry the old name (the shader cache's file magic, the save-size file inside saves, the import path for yuzu data) stay as they are. The program files are still called lichii, eden-cli and eden-room. On macOS the app's bundle id changed, so macOS asks for camera and input permission again (#1541, #1550, #1551, #1552, #1553, #1544, #1582).

Big fixes

  • Ocarina of Time: the far forest no longer floats, and Hyrule Field's frame rate is back. In test builds after build 160, the far pine forest in Hyrule Field and in Kakariko at night was drawn unculled and partly above the ridge, and a spot that ran at 40 fps dropped to 27 fps (RTX 3080). An upstream change that came in with the Eglantine sync (#1462) generated the shader helpers for these foliage programs in a form NVIDIA's compiler mishandles. They're generated the previous way again, which also gives up the cold-compile time upstream saved on them (#1539).
  • Orbitals no longer freezes after the Unreal logo. It froze at 0 fps on every Lichii build so far, because vi2 command 53, which Unreal uses to see which frames were shown, had no handler. This covers Orbitals and the Global Friends Pass. Ported from ghzeni's #742 (#1590).
  • Booting a second game without restarting Lichii no longer crashes. A GPU channel still waiting on a semaphore when a game stopped left a stale entry behind, and the next game's GPU thread read the stopped game's freed memory. It was seen when booting Donkey Kong Bananza after stopping Ocarina of Time (#1592).
  • Stopping a game no longer ends in a "Userspace PANIC". Guest threads kept running while services stopped replying, so about half of all stopped Ocarina of Time sessions logged a panic (2010-0212) that looked like a crash. Guest threads are now suspended before the system powers off (#1450).
  • Ocarina of Time: the dark grid over water is gone. Boxy dark squares followed Link over water whenever a light such as Navi, an enemy or a rupee was near. Shadow maps from the game's light table are now bound through descriptor arrays. Testers on an RX 7800 XT, RTX 4090 and RTX 5080 confirmed it. Port of jpkto's #354, with his commits kept (#1485).
  • Ocarina of Time: a likely fix for the GPU hang after long play. 38 OoT sessions from 28 players ended in a device loss after long play, on AMD, RADV and NVIDIA. A native task shader could launch a leftover, huge number of mesh workgroups; it now sets that count to zero first. Ported from SethRamey's upstream fix. It can't be confirmed yet: the hang shows up about 0.05 to 0.2 times per hour of play (#1492).
  • Mario Kart World: the hang while loading is fixed. Two GPU threads and a guest core deadlocked. The GPU threads no longer wait on kernel locks for sparse memory: a background thread prepares memory ahead of time, and writes are kept until it's ready. The memory reserve follows godpow's upstream design (#1413).
  • Yoshi and the Mysterious Book (demo): no more memory runaway. The demo remaps the same memory every frame, each remap got a new device address, and one buffer kept growing until the process ran out of memory. Freed device ranges of the same size are now reused. In a 150 s run the buffer cache peaked at 308 MiB instead of 2,025 MiB, and the process at 4.7 GB instead of 14.7 GB (#1528). A follow-up gives the kept ranges back when device space runs out, so allocations of other sizes don't fail (#1532).
  • Games no longer vanish from the list when keys are missing. Switch 1 NSP, NSZ, XCI and XCZ files that can't be decrypted are listed with "Decryption keys missing" and a tooltip that says how to install keys. Keys that your key file can produce (master keys, key area keys and titlekeks for newer generations) are now derived, ported from MeownsterHunter's Eden PR 4607 (#1503). A missing key now gives an error that names the key and tells you to install keys, instead of "bad header, please redump" (#1457).
  • Switch 1 games packed as .nx run as Switch 1 again. Every .nx archive was treated as a Switch 2 title, so a Switch 1 game in one got the Switch 2 memory layout and a guest clock 1.63x too fast (#1456).

New features

  • NVIDIA DLSS Frame Generation for Ocarina of Time. It's a new Frame generation type and includes Multi Frame Generation: up to 2x on RTX 40 and up to 4x on RTX 50. It needs "Use NVIDIA DLSS" on, and the NVIDIA DLSS panel downloads NVIDIA's Streamline files; nothing of NVIDIA's ships with Lichii. Other games and GPUs fall back to FSR3 frame generation, and the log says why. Windows only, and not yet tested on RTX 40 or RTX 50 hardware. Built on Dankann's #1348 (based on Frizti's Fritchii#38), with Multi Frame Generation from Revens2's #1291 (#1508).
  • DualSense HD haptics over USB. A wired DualSense plays Switch HD rumble on its voice coils, with each side on its own coil, instead of plain rumble. Turn it on under Controls → Vibration → "Enable DualSense HD Haptics"; it's on by default on Windows. Bluetooth still uses normal rumble. Not tested on hardware yet. Ported from pentalex's Eglantine-Community #624, whose synthesis is adapted from Three-Dola's RenHDrum (#1497).
  • Better rumble on two-motor controllers. On one controller, left and right rumble no longer overwrite each other: each motor plays the stronger side. A new "Frequency-aware rumble" option (off by default) sends each band to the motor that matches its frequency, so a low thud lands on the heavy motor (#1499).

Performance

  • Guest reads skip the GPU cache locks on pages nothing on the GPU touched. In Mario Kart World on Mario Circuit this went from 30.1 to 33.5 fps, with p50 frame time down from 32.7 to 29.4 ms. Read half of Stephen Jayakar's fast path, compared against Frizti's version (#1479).
  • Faster x86-64 guest code. The JIT now uses host call and return, live host flags, address folding and a separate region for cold code. Mario Kart World went from 32.0 to 34.2 fps, and the guest threads spend about a quarter less time on a core. Ocarina of Time in Hyrule Field at night ran at 32.6/32.8 fps against 31.3/32.1, measured with two other candidates on. ARM and Paranoid CPU accuracy keep their old code generation. From Julius Bairaktaris's #831, with Stephen Jayakar's back-edge halt checks from #717 (#1531).
  • Faster JIT compiles. The register allocator tracks its slots in bitmasks instead of scanning all 102 after every instruction. The generated code is identical. liomaroliveira measured 27.9% less JIT compile time in Ocarina of Time and the title-logo frame spike dropping from 60-67 ms to 49-50 ms. Ported from liomaroliveira's fork (#1517).
  • Guest cores wait for work in hardware instead of spinning. In Mario Kart World the old yield loop took 20% of the guest cores' CPU time, about 0.8 of a host core. Idle cores now wait with MONITORX/MWAITX on AMD or UMONITOR/UMWAIT on Intel, for a time that adapts per core (#1373).
  • The GPU thread is woken instead of polling suspended channels. Polling cost about 10 ms of GPU-thread CPU per frame in Mario Kart World. The frame rate doesn't change, but together with #1373 Lichii uses about 0.43 of a CPU core less. The idea comes from Serki's Eglantine-mkw60 and upstream (#1376).
  • Cheaper constant buffer writes. Constant buffer data cost Mario Kart World about 6 ms of the GPU back thread per frame. Each bind is translated once, writes are merged into one invalidation, and small uniform buffers are kept across draws until something writes them (#1441).
  • Data the GPU's engines write keeps its pages tracked, which saves two page-protection changes per page on every write cycle. From lucastaonline's #1161 (#1436).
  • Ocarina of Time water work from #1161. lucastaonline's Kokiri water work is ported in parts, each with his commits kept and without its test switches:
    • Cheaper query cache lookups, frees and locks (#1515).
    • Cheaper reports and fences on the GPU back thread (#1516).
    • Guest writes change page protection after the buffer cache lock is released, so the GPU back thread no longer waits for it. A hazard in the original that could lose an invalidation after a remap is fixed (#1518).
    • Fast draws for pipelines with computed texture handles, word-exact handle checks, and vertex formats as draw parameters with dynamic vertex input (#1519).
    • Command list copies come from a pool, and re-copies at resumes move off the GPU thread. He measured 62.4 to 63.8 fps at the OoT water. A race in the original when the copy queue was full is fixed (#1520).
    • Fast draws copy only the parts of the draw state in use, and kept uniform buffers end at every write and submit they used to miss (#1521).
  • Uploaded guest memory is pushed out of RAM only when RAM is short. Lichii used to drop it from memory every 600 frames and then fault it back in. In Mario Kart World the 1% low went from 15.1/14.3 to 17.1/16.6 fps. It uses more RAM while there's plenty: 6.8-7.0 GB resident instead of 4.9-5.0 GB. Ported from Julius Bairaktaris's change in ThatGuyMcd/Lichii-TGM, based on Frizti's Fritchii idea (#1581).
  • The Windows zero-page pass reads less and runs less often. It stops at the first non-zero line of a page, waits out bulk mapping, and backs off to 5 minutes when it frees little. A pass in Ocarina of Time went from 1,405 ms to about 1,080 ms, and Mario Kart World's 1% low went from 15.1/14.3 to 16.7/17.0 fps. Ported from Frizti (#1580).

Mario Kart World

  • Races went from about 30 to about 38 fps on Mario Circuit across this release, on the owner's Ryzen 9 5900X and RTX 3080. The steps were the cache-read fast path (#1479) and the faster JIT code (#1531) above, then two back-thread changes measured together: 3D state writes sent to the back as register blocks (#1585) and five small cuts from forks (#1586). The last two went from 36.5/36.8 to 37.8/38.8 fps, close to the run-to-run spread, and the owner found the run with both on the best. #1585 is ported from Damfortx with stephenjayakar's encoding; #1586 takes cuts from jpkto, DahBard, skolvolt and lucastaonline and fixes a GPU write that crossed a page and was never collected.
  • Groundwork for higher frame rates. A compute macro that feeds grid sizes from GPU-written command lists is now taken without the GPU thread waiting for those words. By its own estimate this saves nothing at today's 30 fps but removes 4-7 ms of round trips per frame at 60 fps. It builds on DjouWish's #1336 and jpkto's #805 (#1486).
  • No more launches stuck at about 17 fps. Some launches ran at 17.5 fps instead of 28 because the texture cache deleted images the race was still using. Images used in the last 150 frames are now kept while the GPU has room: the slow launches went from 13.7-18.0 fps to 26.4-29.9 fps on a 10 GB RTX 3080 (#1435).
  • The local multiplayer crash is fixed. A process whose program failed to load stayed linked in the kernel's object list, and the next object created at that address crashed. LAN play reaches this path. Ported from Dan Chamberlain's fix, diagnosed by twobasb (#1574).
  • No hang with asynchronous GPU emulation off. Map and run requests are queued without waiting, as they already were with it on. Ported from kraibse (#1575).
  • Particle effects are precompiled. Mario Kart World's .esetb effect sets and Fire Emblem: Fortune's Weave's .ptcl particle files are now read by the shader precache, so their compute programs no longer build mid-race. Upstream measured 66 of 168 compute pipelines missing from MKW's precache before and all 168 covered after. An existing precache file is rebuilt once at the next launch. Ported from kraibse (#1577).

Ocarina of Time

  • No more "boiling" noise on a shader cache from an older build. A pipeline cached while the light-table pointer was empty bound a random texture as a shadow map. It now binds none, and such cache records are rebuilt live (#1523).
  • Fewer one-frame white flashes of the ground, moss and wood. A deleted buffer is now cleared from every GPU channel's bindings, not just the current one. Ported from AManWithAPlan's upstream change and godpow's upstream PR 503 (#1459).
  • A likely fix for device losses on GTX 970-class cards. Mesh-emulation buffers could be reused while the GPU was still writing them. The players affected still have to confirm it (#1452).
  • No more silent exits on PCs near their Windows commit limit. A failed memory-trim pass ended the emulator; it's now skipped, and the log shows the commit numbers (#1451).

Donkey Kong Bananza

  • Faster on the same scene: 17.5 fps against 15.5 with the cache-read fast path (#1479) and another candidate on, and p99 down from about 81 to 69 ms. Most of the gain is probably #1479.

Hyrule Warriors: Age of Imprisonment

  • 30 FPS mode and dynamic FPS. Below 60 fps in single player the game runs in slow motion. "30 FPS mode" runs it at 30 fps at normal speed, and dynamic FPS switches to that only while the game can't hold about 50 fps. Both are off by default, in Graphics → Advanced, and can be set per game. Ported from Rhys Denno's upstream #844 (#1513).
  • The boot hang is fixed. Storage reads with no delay made the game hang at its first loading screen on some PCs; its reads are now throttled at 100 MB/s. Other titles keep their load speed. Ported from Rhys Denno (#1500).
  • Invisible enemies fixed. A shader form the game uses was rejected; upstream also reports invisible roads in Mario Kart World from it. Ported from Rhys Denno, with Hydro74000's test (#1511).
  • Old installed updates no longer stop it from booting. A NUP or tree .nx game used the program of an installed update left in the data folder, with the container's own data. It now ignores the update and logs that it did, and enabled mods are logged (#1504).

Yoshi and the Mysterious Book

  • No more movie abort (2162-0001). It came after 1-4 minutes of movies, and now and then in Age of Imprisonment. A woken client now always reads the current syncpoint value. Ported from Rhys Denno (#1512).

Other games

  • Star Fox: Switch 2 vibration through commands 213 and 220. Titles that rumble through those had none. It's not confirmed in game yet. The 220 handler is ported from Takoroka (#1491).

Crashes and stability

  • Windows: crashes now leave a "Host crash" line and a minidump. Before, a host crash left a log that just stopped, often missing its last lines. The line names the exception, thread and stack, and eden_crash.dmp is written next to the log (#1453).
  • A forced stop no longer hangs Lichii on Windows at "Closing software...". Ported from Quixomatic (#1449).
  • A slow stop is waited out instead of killed. Writing the pipeline cache and destroying tens of thousands of pipelines can take more than 10 s. Lichii now keeps waiting as long as the teardown makes progress, for up to 3 minutes, so it no longer freezes with unsaved caches. Ported from Frizti and Quincy Forbes (#1576).
  • A boot that crashed while building cached shaders doesn't crash every boot. If the last boot never finished loading the shader cache, the next one builds nothing from it at boot and builds pipelines as the game draws. This was one cause of games not starting after reusing an old data folder (#1505).
  • Long sessions no longer run out of SMMU space (invisible geometry). Releasing memory frees every device range inside it. Ported from Ca1nM's upstream PR 824 (#1458).
  • Game writes are no longer lost when several land on one page. Ported from AmateurWizrd's #1328 (#1460).
  • No crash from freed shader programs after a remap. Ported from Anthony Loukinas (#1573).
  • The JIT no longer fails on SM4 instructions in Windows and position-independent builds. From Julius Bairaktaris's #831 (#1570).
  • A leftover EDEN_NX2_TC_MIN_AGE no longer fills GPU memory. It's ignored while memory is short; this was the #1290 crash. Based on RomainROCH's #1311 (#1439).

Upscaling and frame generation

  • Ocarina of Time: the DLSS quality setting now works. It used to run at Quality with a 2560x1440 input whatever you chose. Balanced, Performance and Ultra Performance now lower the game's render resolution (2276x1280, 1920x1080, 1280x720) and DLSS upscales that to 4K. It needs docked mode, applies at the next game start, and uses FSR 3.1 if DLSS turns out to be unavailable. Ported from ProfessorTee's #903 and QuincyForbes/Lichii (#1589).
  • Frame generation works with HDR, and 2x actually looks like 2x. HDR used to switch frame generation off without saying so; FSR3 and LSFG now generate HDR10 and scRGB frames. Generated frames are spaced evenly instead of going out right before the real frame. The log says whether frame generation runs and why not. Includes Zarigis's #783 and ne1i's #1224 (45.9 to 57.3 fps in their OoT test), and the HDR store from felt-mx/fritchii (#1493).

Linux

  • FSR 4 in the x86-64 packages through fsr4-vk by Astyyyyy, a community Vulkan port of FSR 4.1.1. It is not AMD's official FSR 4. It's offered for RDNA 2 or newer Radeons on RADV and replaces FSR 3.1 in Ocarina of Time; when it can't run, the log says why and FSR 3.1 is used. It hasn't upscaled a frame on AMD hardware yet, so it needs RDNA 3 or RDNA 4 testers. The aarch64 packages don't include it (#1496).

macOS

  • The arm64 builds compile again after type and shadowing errors (#1502, #1569, #1591).
  • New bundle id io.github.zortos293.lichii, so macOS asks for camera and input permission again (#1551).

Settings and UI

  • "Use recommended settings" also resets per-game overrides of those settings, and Help → Restore My Previous Settings puts them back. Other per-game settings stay as they are (#1536).
  • A warning before booting a Switch 2 game with less than 16 GB of RAM, with "Continue anyway" and "Don't show this again". The CLI prints it too, and the log marks it (#1478).
  • Custom RTC Date works again, globally and per game, including dates before today. Ported from Amnesia1000's upstream PR 808 (#1465).

From upstream Eglantine

  • Upstream nx2 merged up to 6be10b5c12 (184 upstream commits reviewed, 16 taken). It brings (#1462):
    • More shader forms: ULDC with 8 and 16-bit loads, I2F.S16 from a uniform register and F2I with constant-buffer abs and neg (Duck912), and HSETP2 with a constant buffer (rdenno).
    • Intel Arc fixes from SethRamey: the guest warp size for compute pipelines that lost a barrier (Mario Kart World banding on Intel's Windows driver), no driver pipeline cache reads during builds on Intel's Windows driver, and channel state cleared on teardown. Plus kraibse's matching warp check.
    • A buffer cache fix: GPU writes on some pages weren't downloaded for a guest read (Jonas).
    • Switch 2 games release their guest RAM (up to 12 GiB) when they stop (SmileyTatsu).
    • The font service Super Mario Bros. Wonder - Nintendo Switch 2 Edition needs (Alleexxi).
    • A loose dump without control data is named after its folder, and rich presence can show the game's name, cover and version (MushyMeow). Presence stays off in Lichii's release builds.
    • Readable colours in the "Closing software..." dialog (PastaLaPate).
    • The split check reports on Linux, offline shader tools, upstream's tests for the lowering it brings, and small fixes to the debug server scripts.
    • Its global-memory helper change was rolled back by #1539 (see Big fixes).
  • Three shader precache fixes: precache files from compatible older cache versions load again, loose dumps can be precached, and .zst shader archives are read. From kraibse and Aidan Kinzett (#1463).
  • ARM64 native execution fixes: register 31 reads as zero in the fallback interpreter (which hung Paper Mario: The Thousand-Year Door upstream), and an illegal instruction in guest code stops the guest thread instead of the emulator. From linyesh (#1464).
  • Two AMD fixes: Mario Tennis Fever no longer closes at the start of a match on AMD, and a bogus sparse block size reported for Mario Kart World's terrain is rejected. From MythosOfs (#1466).
  • Faster shader translation. The SSA pass creates phis only when needed and drops unused identities. Upstream measured 19-23% less translation time on Mario Kart World with identical output. From Wael3rd (#1468).
  • Module names for Switch 2 programs, so debugger library lists and backtraces of Switch 2 games are symbolized. From quixomatic (#1469).

Other upstream and fork ports are listed with their game or topic above: #1413, #1458, #1459, #1492, #1500, #1511, #1512, #1513, #1574, #1575, #1577.

Codebase

  • yuzu name removal. A rename codemod (tools/rename) applies the rules per area and reports any yuzu name an upstream sync brings back. The build system, CMake targets and LICHII_* options (old YUZU_* options still carry over), CI, packaging and source folders are renamed (#1541). Then the emulator libraries (#1550), the Qt frontend and translations (#1551), the Android frontend's package (io.github.zortos293.lichii; Lichii ships no Android build) (#1552) and the docs (#1553). The yuzu SPDX copyright lines are removed (#1544), and so are the Android help links to the yuzu quickstart guide (#1582).
  • Diagnostics, all off unless their variable is set:
    • Present fence, GPU back-thread idle and draw reuse traces (#1437).
    • A trace of GPU-written command lists and the native macros they feed (#1446).
    • Per-guest-thread, JIT and GPU front wait stats (#1472).
    • The list macro trace follows data RAM words to their readers, and its verify mode no longer runs loops to the step limit (#1482).
    • A buffer cache growth trace, the one that found the Yoshi demo runaway (#1530).
    • A log line when a game asks for a socket type other than IPv4, plus new program ids in the developer notes (#1352).

Known issues

  • Ocarina of Time: purple speckles on AMD GPUs. Being investigated.
  • Donkey Kong Bananza: black stripes on distant landscape on some AMD Linux setups.
  • Ocarina of Time: can run out of memory while decoding one very large texture.
  • Fire Emblem: Fortune's Weave: crashes on Linux with RDNA 3 GPUs (RADV).

Downloads

Automatic build of nx2 at 87d84d513c67275ad0d6ae50ba51cce76e7e3ded.

Linux

Linux packages come as an AppImage and a tar.gz. Both are self-contained and need glibc 2.39 or newer (Ubuntu 24.04, Fedora 40, Debian 13, Arch). The tar.gz holds the same files unpacked, for systems without FUSE: extract it and run lichii/lichii.

Build Type Standard PGO (Recommended) Notes
amd64 AppImage · tar.gz AppImage · tar.gz Make the AppImage executable (chmod +x) and run it
ARM (aarch64) AppImage · tar.gz AppImage · tar.gz

Windows

Windows packages are in-place zip files: extract and run lichii.exe, or eden-cli.exe from a terminal.

Build Type Standard PGO (Recommended) Notes
amd64 / x86_64 zip PGO zip Both built with clang-cl and ThinLTO; the PGO zip adds a profile trained on Ocarina of Time

macOS

Build Type Standard PGO (Recommended) Notes
arm64 (Apple Silicon) zip macOS 15 or newer. Not notarized: right-click and choose Open the first time, or run xattr -dr com.apple.quarantine lichii.app. eden-cli is inside the app, in Contents/MacOS

The PGO builds are the same programs compiled with profile-guided optimization. Linux uses clang with Eglantine's profile; Windows uses clang-cl and ThinLTO with a profile trained in Lichii on Ocarina of Time, so other games gain less. Prefer them when your CPU supports it.

160
  • Nouveau thème UI "Lichii Dark" avec icônes intégrées, liste de jeux affichant les comptes de dossiers et titres IDs complets, et un système de mise en avant des paramètres modifiés.
  • Prise en charge native du HDR avec options HDR10/scRGB et réglages automatique/manuel pour l’output.
  • Mode LAN pour les jeux locaux sans fil, accessible via Configure → Network → LAN play (désactivé par défaut).
  • Correction majeure du rendu 3D dans Yoshi and the Mysterious Book, fixant des problèmes de shaders manquants.
  • Build macOS Apple Silicon (arm64) compatible avec macOS 15 ou supérieur, nécessitant une ouverture forcée via xattr -dr com.apple.quarantine.
  • Intégration de nxformat 1.0.3 dans l’émulateur, évitant les crashes liés à la suppression de nxformat.dll par les antivirus.

Changes since build 150

Built from nx2 2eb5f081c8. Build 150 was 17baed7. That's 30 merged changes.

New look. Lichii now opens in the new Lichii Dark theme, and a short What's new window explains what moved. You can reopen it from Help → What's New, and the old themes are still in the settings.

Big fixes

  • Yoshi and the Mysterious Book: the 3D world renders. Three shader forms the game uses were rejected, which left the world black. Built on ggonzalospineli-ops' #1189 and FrsrG's #1284, with ALD/OUT from MythosOfs (#1363).
  • The Duskbloods network test: boots to its menus. It no longer aborts on memory, and its shader forms are supported. By cerroverb (#552, #851). It still stops at its online checks. It also always uses reactive flushing, gets the DLSS replacement (#828, #637), and has a 3D colour-target clear fix (#638).
  • Ocarina of Time: the guard house freeze is fixed by a built-in Texture Pool x2 patch (#1313).
  • Mario Kart World: characters and karts are lit correctly. Read-only compute bindings are sized from their size word (#1333).
  • No more crash when antivirus deletes nxformat.dll. nxformat 1.0.3 is now built into Lichii. Converting an XCI to NX also keeps its update partition. By fab918 (#1232).

New features

  • New UI: Lichii Dark. It has icons throughout. Settings dialogs now fit the screen at any scale, and every tab scrolls. A per-game setting you changed gets a pink outline and a reset button. The game list shows folder game counts, full title IDs and favourite stars. Grid and carousel get their own 160 px cover size (#1365, #1371).
  • Native HDR output with HDR10/scRGB passthrough and an Off/Auto/On setting (#1326).
  • LAN play for local-wireless games: Configure → Network → LAN play, off by default (#1359).
  • Frame stats overlay redesign. A big fps number, CPU and GPU %, and a 5-second frame-time graph. It's still movable (#1345).
  • Mario Kart World dynamic FPS skips the game's own second-vblank wait (#1332).

Performance

  • Shader stutter cut down a lot. Pipelines are kept up to a RAM limit instead of 9,000 frames, and an evicted pipeline is rebuilt from its kept SPIR-V instead of being translated again. Revisiting an area went from 360–770 ms hitches to about 55–75 ms. There's also an option to build all cached shaders at boot (#1334, #1337).
  • Less shader compilation work (#1320).
  • CPU: plain loads for LDAR/LDAPR/LDAXR and no barrier for DMB LD/ST on x86, each processor's exclusive reservation on its own cache line, and no empty submits in FlushCommands (#1327, #1331, #1329, a port of Fritchii EXP61). The GPU front/back cache lines are split too (#1332).
  • 3D state registers are written in one call (#1095).

Crash and stability fixes

  • AMD: device memory overallocation is allowed (#1335).
  • Textures over the texel limit that fit in 1 GiB now get backed, which fixes Metroid Prime 4 NS2 Edition (#1325).
  • Looping timers measure their lateness from the end of a pause (#852).
  • The previous log is no longer lost when log rotation fails (#889).
  • Pokémon Pokopia: the furniture-guide shader clamp is folded correctly, ported from Eglantine #781 (#1347).
  • Fixes for the built-in profiler (#1349) and quieter network logs (#1384).

Downloads

Automatic build of nx2 at 2eb5f081c8734affd272466f89133aa4e3e56e18.

Linux

Linux packages come as an AppImage and a tar.gz. Both are self-contained and need glibc 2.39 or newer (Ubuntu 24.04, Fedora 40, Debian 13, Arch). The tar.gz holds the same files unpacked, for systems without FUSE: extract it and run lichii/lichii.

Build Type Standard PGO (Recommended) Notes
amd64 AppImage · tar.gz AppImage · tar.gz Make the AppImage executable (chmod +x) and run it
ARM (aarch64) AppImage · tar.gz AppImage · tar.gz

Windows

Windows packages are in-place zip files: extract and run lichii.exe, or eden-cli.exe from a terminal.

Build Type Standard PGO (Recommended) Notes
amd64 / x86_64 zip PGO zip Both built with clang-cl and ThinLTO; the PGO zip adds a profile trained on Ocarina of Time

macOS

Build Type Standard PGO (Recommended) Notes
arm64 (Apple Silicon) zip macOS 15 or newer. Not notarized: right-click and choose Open the first time, or run xattr -dr com.apple.quarantine lichii.app. eden-cli is inside the app, in Contents/MacOS

The PGO builds are the same programs compiled with profile-guided optimization. Linux uses clang with Eglantine's profile; Windows uses clang-cl and ThinLTO with a profile trained in Lichii on Ocarina of Time, so other games gain less. Prefer them when your CPU supports it.

Réagir sur le forum

Une question, un retour d’expérience ? Ouvrez la discussion sur le forum.

Réagir sur le forum