feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) #459

Open

nyannoying wants to merge 2 commits from nyannoying/Longshot:feat/xess-sr into main

pull from: nyannoying/Longshot:feat/xess-sr

merge into: MagicBeanDev:main

MagicBeanDev:main

Conversation 4 Commits 2 Files changed 33 +1937 -78

nyannoying commented 2026-11-14 15:02:45 +00:00

Contributor

• Copy link
• Copy Markdown

Type: feat · Risk: low · Tested on: Windows 11, Intel Arc B580 (driver 32.0.101.9154) · Needs: Windows with an Arc A-series card or an Intel iGPU; Windows with an AMD or NVIDIA GPU; a Linux build

Why

Intel Arc owners have no upscaler made for their GPU: DLSS needs an NVIDIA card and FSR 4 an RDNA 4 one. Intel's XeSS Super Resolution has a Vulkan API, so it can sit on Longshot's frame chain beside them.

No issue is linked: the description before this one listed none.

What changed

The Upscaler row gets two new values, "Intel XeSS" and "Intel XeSS Native AA". They run Intel's XeSS Super Resolution through its Vulkan API, from the libxess.dll the player puts beside the game; Longshot holds only Intel's MIT-licensed API headers. Without Intel's file, or on Linux, the game starts as before.

• vk: gpu/vk/vk_xess.cpp (new) loads the player's libxess.dll with LoadLibraryExA, resolves its functions by name, asks it for the Vulkan instance extensions, device extensions and features it needs before the instance and the device are made, and runs XeSS on the temporal chain. gpu/vk/xess_rules.h (new) chooses the preset for a render size. vk_temporal.cpp, vk_main.cpp and vk_draw.cpp call it; on Linux it is a stub that says why.
• menu: two new Upscaler values, "Intel XeSS" (7) and "Intel XeSS Native AA" (8), after Off, so saved options keep their meaning (upscaler_rows.h, vk_menu.cpp, auto_settings.h). When Intel's file or something it needs is missing, the row's hint says what; the game starts as before.
• third_party: third_party/xess holds Intel's MIT-licensed API headers only (xess.h, xess_vk.h, MIT.txt) with a PROVENANCE.md. libxess.dll is Intel's, under Intel's own licence, and is never in the repository or a build: the player puts it beside the game.
• build: nx2/build.py, nx2/pipeline_share.py and scripts/dev_packs.py know the new file; no shader-stamp input changed.
• tests: tests/test_xess_rules.py (new) and the upscaler rows, menu keys, temporal jitter, recorder sync and shader stamp checks updated for the two values.
• docs: the README's Intel XeSS section, the RX_VK_XESS row in docs/development.md, a line in CONTRIBUTING.md.

Reviewer notes:

• The licence line: only the MIT headers are vendored, from intel/xess de0fb9c1 (the v3.0.2 tag's headers still carry the proprietary notice; third_party/xess/PROVENANCE.md has the blob ids). libxess.dll is loaded by name at run time and the import library is not linked.
• Conventions, from Intel's XeSS-SR developer guide: motion vectors in pixels from the current frame to the previous one, jitter in pixels in [-0.5, 0.5], both scales 1.
• XeSS Frame Generation and XeLL are not part of this: Intel's SDK 3.0.2 has them for Direct3D 12 only (https://github.com/intel/xess/issues/60 is the open request).

Measurements

Verdict: it works on the Arc B580: Intel XeSS and its Native AA both run every frame with a right picture. XeSS costs about 1.7 ms of GPU time a frame more than the game's own upscaler at the same render size, Native AA 3.6 ms more; the frame rate is about the same three ways, because the CPU limits the frame at this spot. Not compared with main (see Baseline).

main when measured: cf605c85 (2026-11-14)
this PR when the description was written: feat/xess-sr @ 9cd76bec; the build that ran: b46daa73
Hardware: Intel Arc B580 12 GB, driver 32.0.101.9154; Core i5-13600K; 64 GB; Windows 11 Pro (26100-series Insider, 10.0.29704)
Scene: Kokiri Forest (child), the spawn spot on Link's balcony after the opening, standing still; 2048x1152 output, render size 1366x768 (67%), frame generation off, uncapped (HOS_FPS_MAX=1000); libxess.dll from Intel's XeSS SDK 3.0.2 (it reports XeSS 2.0.2)
Method: 6 runs on one build, the upscaler alternated (off, XeSS, Native AA, off, XeSS, Native AA) with RX_VK_XESS=0/1/2, each about 50 s standing after the load; fps from the log's vi: ... frames presented (... fps over the last 5.0 s) lines (last 8 windows, their median), GPU time from vk: GPU time ... ms a frame (last 5 lines, their median).
How to reproduce: put libxess.dll beside the game, choose Upscaler > Intel XeSS in the Graphics tab (or RX_VK_XESS=1), stand in Kokiri Forest; the log says vk: Intel XeSS: 1366x768 -> 2048x1152 (Ultra Quality).

Baseline: the before column is this PR's own build with RX_VK_XESS=0 (the game's own upscaler), XeSS never loaded. main itself was not the comparison because on the B580 it loses the device entering Kokiri Forest (#444). The build that ran is b46daa73, this change on main cf605c85, with three more changes in: #444's float-add fix, the uniform snapshot of #446 and the split rule of #443. The branch was then rebased onto main 26a2eab0 (9cd76bec; one conflict, in scripts/dev_packs.py's list of renderer-only folders, where main's new lines are kept and third_party/xess/ added) and was not run in the game again: a main-based build has another shader stamp than this card's prepared caches. It has since been rebased again, onto main 2b1d38c6: one conflict, in tests/test_shader_stamp_cache.py, both sides kept, no source line changed. The game was not run after either rebase, so the numbers below are from b46daa73.

	Before, RX_VK_XESS=0 (the game's own upscaler): this PR @ b46daa73	After, Intel XeSS (1366x768 -> 2048x1152): this PR @ b46daa73	After, Intel XeSS Native AA (2048x1152): this PR @ b46daa73
fps, median of each run	34.6, 38.0	35.3, 34.2	34.4, 34.2
GPU time a frame	11.7, 11.6 ms	13.3, 13.4 ms	15.1, 15.4 ms

X1 off  fps: 35.7 33.7 32.4 35.4 35.3 33.8 35.8 32.6  | GPU ms: 11.7 11.7 11.6 11.6 11.7
X2 xess fps: 35.1 35.5 35.3 35.6 35.3 35.0 35.7 34.7  | GPU ms: 13.4 13.3 13.3 13.3 13.3
X3 aa   fps: 34.0 34.6 34.3 34.7 33.9 35.0 34.1 34.4  | GPU ms: 15.2 15.1 15.2 15.1 15.1
X4 off  fps: 37.7 37.7 38.3 38.4 37.8 38.4 38.2 37.6  | GPU ms: 11.5 11.6 11.6 11.6 11.6
X5 xess fps: 35.3 34.9 35.5 33.8 31.0 32.2 33.6 34.6  | GPU ms: 13.4 13.5 13.5 13.3 13.4
X6 aa   fps: 33.1 34.4 34.2 34.5 34.2 33.3 33.7 34.3  | GPU ms: 15.4 15.4 15.3 15.3 15.4
[hos host] vk: Intel XeSS: XeSS 2.0.2
[hos Presentation Thread] vk: Intel XeSS: Ultra Quality made for outputs of 2048x1152, render sizes from 1366x768 (frame 5348, 19.1 ms of CPU; its temporary storage 86 MB)
[hos Presentation Thread] vk: Intel XeSS: 1366x768 -> 2048x1152 (Ultra Quality), jitter -0.000 +0.000 px (history reset)
[hos Presentation Thread] vk: Intel XeSS: Native AA made for outputs of 2048x1152, render sizes from 2048x1152 (frame 4535, 21.2 ms of CPU; its temporary storage 86 MB)

At this spot the frame is limited by the CPU (about 28 ms a frame against 12-15 ms of GPU time), so the frame rate is about the same three ways; what XeSS costs shows in the GPU time: about 1.7 ms a frame more than the game's own upscaler at the same render size, and 3.6 ms more for Native AA, which also renders at full size.

Screenshots: Kokiri Forest spawn, Arc B580, posted in the first comment (the game's own upscaler, Intel XeSS, Intel XeSS Native AA, from the measured runs).

Flags and settings

Two new values on the Upscaler menu row (saved under the options file's existing dlss key), three new switches for one run, and one build define.

Flag	What it is	Before this PR	After this PR	Effect	Helps / doesn't help / untested	Recommendation
Upscaler: Intel XeSS (dlss=7), Intel XeSS Native AA (dlss=8)	menu row values, saved in longshot_options.ini	not there: the row ended at Off (6)	there, not chosen: the Upscaler keeps the player's value, and Auto (this PC) never chooses XeSS	XeSS Super Resolution from the render size to the output, or at full size	Helps: Arc B580 (runs, see above). Doesn't help: Linux (Intel makes no XeSS library for it). Untested: Arc A-series, Intel iGPUs, AMD and NVIDIA GPUs (XeSS's own fallback path)	Off by default, the player's choice
RX_VK_XESS=0/1/2	switch, for one run	not there	unset	0 never, Intel's library not even loaded; 1 XeSS; 2 its Native AA	-	for testing
RX_VK_XESS_DIR=<folder>	switch	not there	unset: beside the exe, then the game folder	another folder for libxess.dll	-	for testing
RX_VK_XESS_HDR=0	switch	not there	unset: XeSS takes the HDR colour	with native HDR: the colour cut at the paper white, as the other upscalers had it before	Untested: HDR was not run	off unless the HDR picture is wrong
-DLONGSHOT_XESS	build define, set by nx2/build.py for vk_xess.cpp alone	not there	set in every Windows build, not on Linux	Windows: the code that loads the player's libxess.dll; Linux: the stubs that say XeSS is Windows only	-	as it is: nothing to set by hand

Testing

Tested: Arc B580, driver 32.0.101.9154, Windows 11, Kokiri Forest: the six runs above (files in docs/tested/); XeSS and Native AA both make their context, run every frame and show a right picture (first comment). With RX_VK_XESS=0 Intel's library is not loaded. python -I -B tests/run_all.py: on the head as it was after the first rebase (9cd76bec), 487 passed, 6 failed, 70 skipped; the failures were test_module_ids, test_pad_layout, test_tested_configurations, test_oot_state_tap, test_fpx_sticky and test_spawn_emit, which failed the same way on main here. On the head as it is now: 15 checks fail on this Windows PC and fail the same on main 2b1d38c6 alone (test_fpx_sticky, test_appimage_player_build, test_cpm_discovery, test_crash_dialog, test_env_on, test_lift_cleanup_error, test_make_game_cache, test_linux_preflight, test_oot_state_tap, test_readback_late, test_spawn_emit, test_tested_configurations, test_vk_pgo, test_vram_budget_row, test_walk_key). python -I scripts/check_third_party.py passes with the new third_party/xess.

Not tested: choosing the value in the menu in this session (the runs used RX_VK_XESS); native HDR with XeSS; the render sizes other than 67% and 100%; other areas; cutscenes; frame generation together with XeSS; GPUs other than the B580 (XeSS's path for non-Intel GPUs); a missing or old libxess.dll in the game; the rebased head in the game (see Baseline: the game was not run after either rebase); a Linux build (no Linux compiler on this PC; the Linux side is a stub, so the checklist's two-platform box is left out, not ticked).

Must be tested:

• Windows, an Arc A-series card or an Intel iGPU, with libxess.dll beside the game: choose Upscaler > Intel XeSS and then Intel XeSS Native AA from the menu, expecting a right picture and the vk: Intel XeSS: lines in the log (by hand: look at the picture and the log; test request T-UUJ4, open)
• Windows, an AMD or NVIDIA GPU: the same steps, expecting XeSS to run on its fallback path or the row's hint to say why not, and the game to start either way (by hand: look at the picture, the row's hint and the log; test request T-3X7L, open)
• Linux, any GPU: a build and a start, expecting the build to pass and Intel XeSS not to run there (the Upscaler row with its two XeSS choices: whether they are offered and what their help says; the README says a Linux build has no XeSS) (by hand: look at the build's result and the Upscaler row; test request T-3N56, open)

Risks

What could break: nothing until the player chooses Intel XeSS in the Upscaler row and has put Intel's libxess.dll in place. The file is looked for in RX_VK_XESS_DIR, beside the exe, then in the game folder (xess_rules.h library_dirs); without it the two values are not offered, the row's hint says the file was not found, and the game starts as before. With the file there, its Vulkan instance and device extensions and features are asked for at start even when another upscaler is chosen; when the driver lacks one, XeSS is left out with the reason in the log and the hint. Windows only: a Linux build compiles the stubs.

Turning it off: RX_VK_XESS=0 (Intel's library is then not even loaded), or another Upscaler choice, or taking libxess.dll away. Reverting the commit removes it.

Compatibility: the options file gets no new key: the existing dlss key takes two more values, 7 and 8, added after Off (6), so saved options keep their meaning. The shader cache stamp does not change: the diff touches none of the stamp's files (gpu/vk/eden, sirit, compat, vk_shader.cpp, vk_bank.cpp, vk_state.h), so nothing is translated again because of this PR. MKWVK_API_VERSION does not change: gpu/vk/vk_api.h is not in the diff. The builder's tools: nx2/build.py compiles vk_xess.cpp with Intel's headers on Windows, nx2/pipeline_share.py knows the two new Upscaler names, scripts/dev_packs.py counts third_party/xess/ as renderer-only. Saves: the diff touches no save code.

Checklist

Everyone

• One topic.
• The title is a Conventional Commit (type(scope): description) of the status line's type.
• Why gives the reason plainly and links the issues that fit (the tracker searched).
• Measurements has a verdict, or one line saying why there is nothing to measure.
• Testing names what was tested, what wasn't, and what must still be tested, where and how.
• Rebased on current main.
• python -I -B tests/run_all.py passes (any failure that also fails on main is named): 15 checks fail on this Windows PC and fail the same on main 2b1d38c6 alone (test_fpx_sticky, test_appimage_player_build, test_cpm_discovery, test_crash_dialog, test_env_on, test_lift_cleanup_error, test_make_game_cache, test_linux_preflight, test_oot_state_tap, test_readback_late, test_spawn_emit, test_tested_configurations, test_vk_pgo, test_vram_budget_row, test_walk_key).
• No game files, dumps, build output or generated files.
• third_party/ untouched (a new folder with a PROVENANCE.md instead).
• The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled.

When it applies

• AI assistance recorded with Assisted-by: / Co-Authored-By: trailers.
• Numbers from the same PC, scene and settings, at least 2 interleaved runs each, the median reported.
• Switches documented (docs/development.md; the README for a menu row) and shipped at their recommended value.
• Screenshots of the same spot, captioned like the Measurements columns, for a change that shows on screen.
• A file in docs/tested/ for each new measured run (format in docs/tested-configurations.md).
• A check under tests/ that reads changed source updated in this PR, not weakened.
• Docs updated.

🤖 Generated with Claude Code

**Type:** feat · **Risk:** low · **Tested on:** Windows 11, Intel Arc B580 (driver 32.0.101.9154) · **Needs:** Windows with an Arc A-series card or an Intel iGPU; Windows with an AMD or NVIDIA GPU; a Linux build ## Why Intel Arc owners have no upscaler made for their GPU: DLSS needs an NVIDIA card and FSR 4 an RDNA 4 one. Intel's XeSS Super Resolution has a Vulkan API, so it can sit on Longshot's frame chain beside them. No issue is linked: the description before this one listed none. ## What changed The Upscaler row gets two new values, "Intel XeSS" and "Intel XeSS Native AA". They run Intel's XeSS Super Resolution through its Vulkan API, from the `libxess.dll` the player puts beside the game; Longshot holds only Intel's MIT-licensed API headers. Without Intel's file, or on Linux, the game starts as before. - vk: `gpu/vk/vk_xess.cpp` (new) loads the player's `libxess.dll` with `LoadLibraryExA`, resolves its functions by name, asks it for the Vulkan instance extensions, device extensions and features it needs before the instance and the device are made, and runs XeSS on the temporal chain. `gpu/vk/xess_rules.h` (new) chooses the preset for a render size. `vk_temporal.cpp`, `vk_main.cpp` and `vk_draw.cpp` call it; on Linux it is a stub that says why. - menu: two new Upscaler values, "Intel XeSS" (7) and "Intel XeSS Native AA" (8), after Off, so saved options keep their meaning (`upscaler_rows.h`, `vk_menu.cpp`, `auto_settings.h`). When Intel's file or something it needs is missing, the row's hint says what; the game starts as before. - third_party: `third_party/xess` holds Intel's MIT-licensed API headers only (`xess.h`, `xess_vk.h`, `MIT.txt`) with a `PROVENANCE.md`. `libxess.dll` is Intel's, under Intel's own licence, and is never in the repository or a build: the player puts it beside the game. - build: `nx2/build.py`, `nx2/pipeline_share.py` and `scripts/dev_packs.py` know the new file; no shader-stamp input changed. - tests: `tests/test_xess_rules.py` (new) and the upscaler rows, menu keys, temporal jitter, recorder sync and shader stamp checks updated for the two values. - docs: the README's Intel XeSS section, the `RX_VK_XESS` row in `docs/development.md`, a line in `CONTRIBUTING.md`. Reviewer notes: - The licence line: only the MIT headers are vendored, from intel/xess de0fb9c1 (the v3.0.2 tag's headers still carry the proprietary notice; `third_party/xess/PROVENANCE.md` has the blob ids). `libxess.dll` is loaded by name at run time and the import library is not linked. - Conventions, from Intel's XeSS-SR developer guide: motion vectors in pixels from the current frame to the previous one, jitter in pixels in [-0.5, 0.5], both scales 1. - XeSS Frame Generation and XeLL are not part of this: Intel's SDK 3.0.2 has them for Direct3D 12 only (https://github.com/intel/xess/issues/60 is the open request). ## Measurements **Verdict:** it works on the Arc B580: Intel XeSS and its Native AA both run every frame with a right picture. XeSS costs about 1.7 ms of GPU time a frame more than the game's own upscaler at the same render size, Native AA 3.6 ms more; the frame rate is about the same three ways, because the CPU limits the frame at this spot. Not compared with main (see Baseline). main when measured: cf605c85 (2026-11-14) this PR when the description was written: feat/xess-sr @ 9cd76bec; the build that ran: b46daa73 Hardware: Intel Arc B580 12 GB, driver 32.0.101.9154; Core i5-13600K; 64 GB; Windows 11 Pro (26100-series Insider, 10.0.29704) Scene: Kokiri Forest (child), the spawn spot on Link's balcony after the opening, standing still; 2048x1152 output, render size 1366x768 (67%), frame generation off, uncapped (`HOS_FPS_MAX=1000`); `libxess.dll` from Intel's XeSS SDK 3.0.2 (it reports XeSS 2.0.2) Method: 6 runs on one build, the upscaler alternated (off, XeSS, Native AA, off, XeSS, Native AA) with `RX_VK_XESS=0/1/2`, each about 50 s standing after the load; fps from the log's `vi: ... frames presented (... fps over the last 5.0 s)` lines (last 8 windows, their median), GPU time from `vk: GPU time ... ms a frame` (last 5 lines, their median). How to reproduce: put `libxess.dll` beside the game, choose Upscaler > Intel XeSS in the Graphics tab (or `RX_VK_XESS=1`), stand in Kokiri Forest; the log says `vk: Intel XeSS: 1366x768 -> 2048x1152 (Ultra Quality)`. Baseline: the before column is this PR's own build with `RX_VK_XESS=0` (the game's own upscaler), XeSS never loaded. main itself was not the comparison because on the B580 it loses the device entering Kokiri Forest (https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/444). The build that ran is b46daa73, this change on main cf605c85, with three more changes in: #444's float-add fix, the uniform snapshot of https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/446 and the split rule of https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/443. The branch was then rebased onto main 26a2eab0 (9cd76bec; one conflict, in `scripts/dev_packs.py`'s list of renderer-only folders, where main's new lines are kept and `third_party/xess/` added) and was not run in the game again: a main-based build has another shader stamp than this card's prepared caches. It has since been rebased again, onto main 2b1d38c6: one conflict, in `tests/test_shader_stamp_cache.py`, both sides kept, no source line changed. The game was not run after either rebase, so the numbers below are from b46daa73. | | Before, `RX_VK_XESS=0` (the game's own upscaler): this PR @ b46daa73 | After, Intel XeSS (1366x768 -> 2048x1152): this PR @ b46daa73 | After, Intel XeSS Native AA (2048x1152): this PR @ b46daa73 | |---|---|---|---| | fps, median of each run | 34.6, 38.0 | 35.3, 34.2 | 34.4, 34.2 | | GPU time a frame | 11.7, 11.6 ms | 13.3, 13.4 ms | 15.1, 15.4 ms | ``` X1 off fps: 35.7 33.7 32.4 35.4 35.3 33.8 35.8 32.6 | GPU ms: 11.7 11.7 11.6 11.6 11.7 X2 xess fps: 35.1 35.5 35.3 35.6 35.3 35.0 35.7 34.7 | GPU ms: 13.4 13.3 13.3 13.3 13.3 X3 aa fps: 34.0 34.6 34.3 34.7 33.9 35.0 34.1 34.4 | GPU ms: 15.2 15.1 15.2 15.1 15.1 X4 off fps: 37.7 37.7 38.3 38.4 37.8 38.4 38.2 37.6 | GPU ms: 11.5 11.6 11.6 11.6 11.6 X5 xess fps: 35.3 34.9 35.5 33.8 31.0 32.2 33.6 34.6 | GPU ms: 13.4 13.5 13.5 13.3 13.4 X6 aa fps: 33.1 34.4 34.2 34.5 34.2 33.3 33.7 34.3 | GPU ms: 15.4 15.4 15.3 15.3 15.4 [hos host] vk: Intel XeSS: XeSS 2.0.2 [hos Presentation Thread] vk: Intel XeSS: Ultra Quality made for outputs of 2048x1152, render sizes from 1366x768 (frame 5348, 19.1 ms of CPU; its temporary storage 86 MB) [hos Presentation Thread] vk: Intel XeSS: 1366x768 -> 2048x1152 (Ultra Quality), jitter -0.000 +0.000 px (history reset) [hos Presentation Thread] vk: Intel XeSS: Native AA made for outputs of 2048x1152, render sizes from 2048x1152 (frame 4535, 21.2 ms of CPU; its temporary storage 86 MB) ``` At this spot the frame is limited by the CPU (about 28 ms a frame against 12-15 ms of GPU time), so the frame rate is about the same three ways; what XeSS costs shows in the GPU time: about 1.7 ms a frame more than the game's own upscaler at the same render size, and 3.6 ms more for Native AA, which also renders at full size. Screenshots: Kokiri Forest spawn, Arc B580, posted in the first comment (the game's own upscaler, Intel XeSS, Intel XeSS Native AA, from the measured runs). ### Flags and settings Two new values on the Upscaler menu row (saved under the options file's existing `dlss` key), three new switches for one run, and one build define. | Flag | What it is | Before this PR | After this PR | Effect | Helps / doesn't help / untested | Recommendation | |---|---|---|---|---|---|---| | Upscaler: Intel XeSS (`dlss=7`), Intel XeSS Native AA (`dlss=8`) | menu row values, saved in `longshot_options.ini` | not there: the row ended at Off (6) | there, not chosen: the Upscaler keeps the player's value, and Auto (this PC) never chooses XeSS | XeSS Super Resolution from the render size to the output, or at full size | Helps: Arc B580 (runs, see above). Doesn't help: Linux (Intel makes no XeSS library for it). Untested: Arc A-series, Intel iGPUs, AMD and NVIDIA GPUs (XeSS's own fallback path) | Off by default, the player's choice | | `RX_VK_XESS=0/1/2` | switch, for one run | not there | unset | 0 never, Intel's library not even loaded; 1 XeSS; 2 its Native AA | - | for testing | | `RX_VK_XESS_DIR=<folder>` | switch | not there | unset: beside the exe, then the game folder | another folder for `libxess.dll` | - | for testing | | `RX_VK_XESS_HDR=0` | switch | not there | unset: XeSS takes the HDR colour | with native HDR: the colour cut at the paper white, as the other upscalers had it before | Untested: HDR was not run | off unless the HDR picture is wrong | | `-DLONGSHOT_XESS` | build define, set by `nx2/build.py` for `vk_xess.cpp` alone | not there | set in every Windows build, not on Linux | Windows: the code that loads the player's `libxess.dll`; Linux: the stubs that say XeSS is Windows only | - | as it is: nothing to set by hand | ## Testing **Tested:** Arc B580, driver 32.0.101.9154, Windows 11, Kokiri Forest: the six runs above (files in `docs/tested/`); XeSS and Native AA both make their context, run every frame and show a right picture (first comment). With `RX_VK_XESS=0` Intel's library is not loaded. `python -I -B tests/run_all.py`: on the head as it was after the first rebase (9cd76bec), 487 passed, 6 failed, 70 skipped; the failures were `test_module_ids`, `test_pad_layout`, `test_tested_configurations`, `test_oot_state_tap`, `test_fpx_sticky` and `test_spawn_emit`, which failed the same way on main here. On the head as it is now: 15 checks fail on this Windows PC and fail the same on main 2b1d38c6 alone (test_fpx_sticky, test_appimage_player_build, test_cpm_discovery, test_crash_dialog, test_env_on, test_lift_cleanup_error, test_make_game_cache, test_linux_preflight, test_oot_state_tap, test_readback_late, test_spawn_emit, test_tested_configurations, test_vk_pgo, test_vram_budget_row, test_walk_key). `python -I scripts/check_third_party.py` passes with the new `third_party/xess`. **Not tested:** choosing the value in the menu in this session (the runs used `RX_VK_XESS`); native HDR with XeSS; the render sizes other than 67% and 100%; other areas; cutscenes; frame generation together with XeSS; GPUs other than the B580 (XeSS's path for non-Intel GPUs); a missing or old `libxess.dll` in the game; the rebased head in the game (see Baseline: the game was not run after either rebase); a Linux build (no Linux compiler on this PC; the Linux side is a stub, so the checklist's two-platform box is left out, not ticked). **Must be tested:** - Windows, an Arc A-series card or an Intel iGPU, with `libxess.dll` beside the game: choose Upscaler > Intel XeSS and then Intel XeSS Native AA from the menu, expecting a right picture and the `vk: Intel XeSS:` lines in the log (by hand: look at the picture and the log; test request T-UUJ4, open) - Windows, an AMD or NVIDIA GPU: the same steps, expecting XeSS to run on its fallback path or the row's hint to say why not, and the game to start either way (by hand: look at the picture, the row's hint and the log; test request T-3X7L, open) - Linux, any GPU: a build and a start, expecting the build to pass and Intel XeSS not to run there (the Upscaler row with its two XeSS choices: whether they are offered and what their help says; the README says a Linux build has no XeSS) (by hand: look at the build's result and the Upscaler row; test request T-3N56, open) ## Risks **What could break:** nothing until the player chooses Intel XeSS in the Upscaler row and has put Intel's `libxess.dll` in place. The file is looked for in `RX_VK_XESS_DIR`, beside the exe, then in the game folder (`xess_rules.h` `library_dirs`); without it the two values are not offered, the row's hint says the file was not found, and the game starts as before. With the file there, its Vulkan instance and device extensions and features are asked for at start even when another upscaler is chosen; when the driver lacks one, XeSS is left out with the reason in the log and the hint. Windows only: a Linux build compiles the stubs. **Turning it off:** `RX_VK_XESS=0` (Intel's library is then not even loaded), or another Upscaler choice, or taking `libxess.dll` away. Reverting the commit removes it. **Compatibility:** the options file gets no new key: the existing `dlss` key takes two more values, 7 and 8, added after Off (6), so saved options keep their meaning. The shader cache stamp does not change: the diff touches none of the stamp's files (`gpu/vk/eden`, `sirit`, `compat`, `vk_shader.cpp`, `vk_bank.cpp`, `vk_state.h`), so nothing is translated again because of this PR. `MKWVK_API_VERSION` does not change: `gpu/vk/vk_api.h` is not in the diff. The builder's tools: `nx2/build.py` compiles `vk_xess.cpp` with Intel's headers on Windows, `nx2/pipeline_share.py` knows the two new Upscaler names, `scripts/dev_packs.py` counts `third_party/xess/` as renderer-only. Saves: the diff touches no save code. ## Checklist ### Everyone - [x] One topic. - [x] The title is a Conventional Commit (`type(scope): description`) of the status line's type. - [x] Why gives the reason plainly and links the issues that fit (the tracker searched). - [x] Measurements has a verdict, or one line saying why there is nothing to measure. - [x] Testing names what was tested, what wasn't, and what must still be tested, where and how. - [x] Rebased on current `main`. - [x] `python -I -B tests/run_all.py` passes (any failure that also fails on `main` is named): 15 checks fail on this Windows PC and fail the same on main 2b1d38c6 alone (test_fpx_sticky, test_appimage_player_build, test_cpm_discovery, test_crash_dialog, test_env_on, test_lift_cleanup_error, test_make_game_cache, test_linux_preflight, test_oot_state_tap, test_readback_late, test_spawn_emit, test_tested_configurations, test_vk_pgo, test_vram_budget_row, test_walk_key). - [x] No game files, dumps, build output or generated files. - [x] `third_party/` untouched (a new folder with a `PROVENANCE.md` instead). - [x] The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled. ### When it applies - [x] AI assistance recorded with `Assisted-by:` / `Co-Authored-By:` trailers. - [x] Numbers from the same PC, scene and settings, at least 2 interleaved runs each, the median reported. - [x] Switches documented (docs/development.md; the README for a menu row) and shipped at their recommended value. - [x] Screenshots of the same spot, captioned like the Measurements columns, for a change that shows on screen. - [x] A file in `docs/tested/` for each new measured run (format in `docs/tested-configurations.md`). - [x] A check under `tests/` that reads changed source updated in this PR, not weakened. - [x] Docs updated. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

nyannoying added 2 commits 2026-11-14 15:02:45 +00:00

feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) ... c6df107451

Intel Arc owners had no upscaler built for their GPU: DLSS needs NVIDIA and FSR 4 needs RDNA 4.
XeSS-SR has a Vulkan API, so it can run on Longshot's frame chain beside DLSS and FSR 4. It
appears as two new Upscaler values, "Intel XeSS" (7) and "Intel XeSS Native AA" (8). They come
after Off, so saved options keep their meaning.

Intel's runtime (libxess.dll) is under Intel's own licence and is never bundled. Only the MIT API
headers are vendored, from intel/xess de0fb9c1 (the v3.0.2 tag's headers carry a proprietary
notice). The DLL is looked for in RX_VK_XESS_DIR, then beside the exe, then in the game folder.
It is loaded with LoadLibraryExA and its functions are resolved by name; the import library is not
linked. The Vulkan instance extensions, device extensions and device features XeSS needs are asked
of the library before the instance and device are made. If anything is missing, XeSS is left out
and the reason is written to the log and shown in the menu hint; startup does not fail.

Conventions, from Intel's XeSS-SR developer guide:
- Motion vectors are in pixels, pointing from the current frame to the previous one.
- Jitter is in pixels in [-0.5, 0.5], the same convention FSR 4 uses.
- Both scales are therefore 1.
Since XeSS 1.3 each preset accepts any input size from its ratio up to the output. Each Render
resolution stop runs under the first preset whose range holds it. LDR input uses
LDR_INPUT_COLOR; HDR input uses auto-exposure (RX_VK_XESS_HDR=0 turns HDR off).

The engine is Windows only; on Linux it is a stub with the reason. XeSS Frame Generation and XeLL
are not implemented because they are Direct3D 12 only. No shader-stamp inputs, MkwVkHost or
MkwVkRenderer changed.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

docs(tested): the XeSS Super Resolution runs on an Arc B580, six files ... 00458c6d1b

Co-Authored-By: Claude Fable 5.2 <[email protected]>

nyannoying commented 2026-11-14 15:03:05 +00:00

Author

Contributor

• Copy link
• Copy Markdown

Kokiri Forest, the spawn spot, Arc B580 (101.9154), 2048x1152 output, from runs X4 to X6 of the Measurements (one build, only RX_VK_XESS differs; the scene moves a little between runs).

Kokiri Forest, the spawn spot, Arc B580 (101.9154), 2048x1152 output, from runs X4 to X6 of the Measurements (one build, only `RX_VK_XESS` differs; the scene moves a little between runs).   

xess_off.png

2.6 MiB

xess_on.png

2.6 MiB

xess_aa.png

2.6 MiB

nyannoying commented 2026-11-14 15:03:31 +00:00

Author

Contributor

• Copy link
• Copy Markdown

The three test requests are open (T-UUJ4: Arc A-series or an Intel iGPU; T-3X7L: an AMD or NVIDIA GPU; T-3N56: a Linux build) and named in the description; they wait for a maintainer's testing wanted label.

The three test requests are open (T-UUJ4: Arc A-series or an Intel iGPU; T-3X7L: an AMD or NVIDIA GPU; T-3N56: a Linux build) and named in the description; they wait for a maintainer's testing wanted label.

nyannoying changed title from WIP: feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) to feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) 2026-11-14 15:03:34 +00:00

nyannoying force-pushed feat/xess-sr from 00458c6d1b to 3fdd4ed8ee 2026-11-14 15:30:45 +00:00 Compare

nyannoying commented 2026-11-14 15:30:47 +00:00

Author

Contributor

• Copy link
• Copy Markdown

Rebased onto main 26a2eab0 (it had moved; one conflict in scripts/dev_packs.py, main's new stamp lines kept and third_party/xess/ added to the renderer-only folders). The diff of the change is otherwise the same, so the measurements stand; the suite on the new head: 487 passed, 6 failed (the six that fail on main here), 70 skipped.

Rebased onto main 26a2eab0 (it had moved; one conflict in scripts/dev_packs.py, main's new stamp lines kept and third_party/xess/ added to the renderer-only folders). The diff of the change is otherwise the same, so the measurements stand; the suite on the new head: 487 passed, 6 failed (the six that fail on main here), 70 skipped.

mooki added the

area/upscaling

gpu/intel

kind

feature

priority

low

area/renderer

labels 2026-11-14 16:09:22 +00:00

nyannoying referenced this pull request 2026-11-14 16:52:47 +00:00

perf(vk): the Arc B-series compiles shaders side by side by default: a new area no longer freezes the picture for minutes #500

mooki added the

platform/linux

platform/windows

labels 2026-11-14 18:24:08 +00:00

nyannoying commented 2026-11-14 23:37:29 +00:00

Author

Contributor

• Copy link
• Copy Markdown

Rebased onto main 2b1d38c6 (head ea138e60); it had stopped merging.

• Conflict: tests/test_shader_stamp_cache.py only, where main's renderer profile tests (#569) and this PR's test_xess_reaches_only_vk_xess_on_windows were added at the same place. Both kept; no source line of the PR changed.
• Checked after the rebase: the Windows build of the renderer (llvm-mingw, 0 failed) and tests/run_all.py: 494 passed, 15 failed, 70 skipped. The same 15 fail on main 2b1d38c6 alone on this PC (run one by one there), test_shader_stamp_cache.py is not among them.
• Not run again: the game. The Measurements are from the commits named there (main cf605c85 with the open Arc fixes).

Rebased onto main 2b1d38c6 (head ea138e60); it had stopped merging. - **Conflict:** `tests/test_shader_stamp_cache.py` only, where main's renderer profile tests (#569) and this PR's `test_xess_reaches_only_vk_xess_on_windows` were added at the same place. Both kept; no source line of the PR changed. - **Checked after the rebase:** the Windows build of the renderer (llvm-mingw, 0 failed) and `tests/run_all.py`: 494 passed, 15 failed, 70 skipped. The same 15 fail on main 2b1d38c6 alone on this PC (run one by one there), `test_shader_stamp_cache.py` is not among them. - **Not run again:** the game. The Measurements are from the commits named there (main cf605c85 with the open Arc fixes).

nyannoying force-pushed feat/xess-sr from 3fdd4ed8ee to ea138e608a 2026-11-14 23:37:30 +00:00 Compare

ertomas referenced this pull request 2026-11-15 00:33:11 +00:00

Game: Arc B390 (Panther Lake iGPU, Windows): Kokiri Forest is washed out by a bright haze; an Arc B580 at the same spot is not #618

nyannoying referenced this pull request 2026-11-15 10:18:07 +00:00

perf(vk): Arc cards read the occlusion counts from a GPU copy: Castle Town's market from 34.8 to 45.8 fps on a B580 #761

HumanBean referenced this pull request from a commit 2026-11-15 10:22:55 +00:00

perf(vk): Arc cards read the occlusion counts from a GPU copy: Castle Town's market from 34.8 to 45.8 fps on a B580 (#761)

ertomas referenced this pull request 2026-11-15 15:09:25 +00:00

WIP: fix(vk): float adds native on Panther Lake iGPUs, where the loop lost the exposure's adds #816

ertomas referenced this pull request 2026-11-15 15:09:32 +00:00

WIP: fix(vk): float adds native on Panther Lake iGPUs, where the loop lost the exposure's adds #816

nyannoying referenced this pull request 2026-11-15 16:56:50 +00:00

WIP: fix(vk): the render thread may leave its core when another program takes it: no 30-40 ms stalls in Castle Town's market #838

Some checks are pending

Hide checks

CI / Conventional Commit title (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions

Details

CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions

Details

Third-party parts / As their notes say (pull_request) Blocked by required conditions

Details

Third-party parts / The same as upstream (pull_request) Blocked by required conditions

Details

Website / Website on Cloudflare (pull_request) Blocked by required conditions

Details

PR description / Description follows the template (pull_request_target) Successful in 17s

Details

This pull request has changes conflicting with the target branch.

• docs/development.md
• gpu/vk/upscaler_rows.h
• gpu/vk/vk_temporal.cpp
• nx2/build.py

Some workflows are waiting to be reviewed.

View command line instructions

Manual merge helper

Use this merge commit message when completing the merge manually.

Merge commit title feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) (#459)

Merge commit body Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/459

feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) (#459) Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/459 Copy merge commit message

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u 'feat/xess-sr:nyannoying-feat/xess-sr'

git switch 'nyannoying-feat/xess-sr'

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

area/accessibility

Barriers for people with disabilities

  

area/audio

Sound and music

  

area/input

Controllers, keyboard, mouse

  

area/install

Installing, updating, building from your game

  

area/menu

In-game menu and settings

  

area/renderer

Vulkan renderer, shaders, effects

  

area/saves

Save files

  

area/upscaling

DLSS, FSR, frame generation

  

beanstalk: approved @ 1697603

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 2c5a62e

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 4fd82b7

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 64d6d09

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 6f91fae

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 70a78f9

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 8d361dc

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 9d8356e

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ abf2227

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ d82f202

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ e5f1e40

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ ee8ab5f

Set by Beanstalk for show; the farm's database is what counts

  

ci: approved

A maintainer approved CircleCI for this fork PR (sticky; removed after 14 quiet days)

  

good first issue

A good start for a new contributor

  

gpu/amd

AMD GPUs

  

gpu/igpu

Integrated GPUs

  

gpu/intel

Intel GPUs

  

gpu/nvidia

NVIDIA GPUs

  

help wanted

Extra hands or knowledge wanted

  

kind

bug

Something doesn't work as it should

  

kind

chore

Build, CI, dependencies, tidying

  

kind

docs

Documentation

  

kind

feature

A new feature or an improvement

  

kind

performance

Frame rate, frame time, loading, memory

  

kind

question

A question rather than a change

  

needs measurements

Waiting for before/after measurements

  

platform/linux

Affects Linux (a change for every system carries both Windows and Linux)

  

platform/macos

macOS (Apple silicon or Intel Macs)

  

platform/steam-deck

Steam Deck

  

platform/windows

Affects Windows (a change for every system carries both Windows and Linux)

  

preview

Maintainers: deploy a preview of this PR's website (read its changes first). Removed after the run.

  

priority

critical

Crashes, data loss, or blocks most players

  

priority

high

Important: next in line

  

priority

low

Nice to have

  

regression

Worked in an earlier build, broken in a newer one: highest priority, above other bugs

  

status

blocked

Waits on something else first

  

status

confirmed

Reproduced or agreed on

  

status

duplicate

Already reported or proposed

  

status

invalid

Not a problem in this project

  

status

needs-info

Waiting for more information from the reporter

  

status

wontfix

Will not be worked on

  

testing wanted

Wanted: a test on other hardware (posts a test request in #test-requests)

  

ci: approved

No labels

area/accessibility

area/audio

area/input

area/install

area/menu

area/renderer

area/saves

area/upscaling

beanstalk: approved @ 1697603

beanstalk: approved @ 2c5a62e

beanstalk: approved @ 4fd82b7

beanstalk: approved @ 64d6d09

beanstalk: approved @ 6f91fae

beanstalk: approved @ 70a78f9

beanstalk: approved @ 8d361dc

beanstalk: approved @ 9d8356e

beanstalk: approved @ abf2227

beanstalk: approved @ d82f202

beanstalk: approved @ e5f1e40

beanstalk: approved @ ee8ab5f

ci: approved

good first issue

gpu/amd

gpu/igpu

gpu/intel

gpu/nvidia

help wanted

kind

bug

kind

chore

kind

docs

kind

feature

kind

performance

kind

question

needs measurements

platform/linux

platform/macos

platform/steam-deck

platform/windows

preview

priority

critical

priority

high

priority

low

regression

status

blocked

status

confirmed

status

duplicate

status

invalid

status

needs-info

status

wontfix

testing wanted

ci: approved

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

MagicBeanDev/Longshot!459

No description provided.