Configuring lsfg‐vk - Pahheb/lsfg-vk GitHub Wiki
Configuring lsfg-vk is done by environment variables. Let's quickly go through setting an environment variable:
- If you're running from the command line, put the environment variables in this format before the command you are running:
$ VARIABLE_NAME="value" EXAMPLE_PATH="/path/to/something" NUMBER_VARIABLE=5 command
- If you're running through Steam, add the environment variables followed by `%COMMAND% into the launch arguments:
SOME_VARIABLE=1 %COMMAND%
Supported environment variables
The following is a table of supported environment variables, their usage and recommended values, as well as some notes.
| Variable | Recommended Value | Description | Notes |
|---|---|---|---|
LSFG_LEGACY |
1 | Enables the frame generation layer. | |
LSFG_MULTIPLIER |
2-4 | Traditional FPS multiplier value. | If you set this value higher than your hardware supports, your game will crash. Try lowering this if you encounter any issues. |
LSFG_DLL_PATH |
Override where lsfg-vk looks for Lossless Scaling. Specify the Lossless.dll file here, if you installed LS in a different location. | ||
LSFG_LOG_DEBUG |
List of modules to enable debugging for. | This only works on Debug builds. | |
LSFG_LOG_FILE |
Write all messages to a file in addition to the console | ||
LSFG_HDR_MODE |
1 when using HDR | Switches the shaders to HDR mode (scRGB not supported) | Do not set this flag if you're not using HDR |
LSFG_FLOW_SCALE |
1.0 | Lowers the flow scale, where 1.0 equals 100% and 0.25 equals 25%. This can improve performance | Use the benchmark to figure out what your system can handle. |
LSFG_PERFORMANCE_MODE |
Enable performance mode, which can as much as quintuple the performance. | ||
LSFG_EXPERIMENTAL_PRESENT_MODE |
fifo / vsync / immediate / mailbox |
Overrides the present mode without altering synchronization logic. Very experimental. | |
LSFG_EXPERIMENTAL_FPS_LIMIT |
Try to override the FPS limit on DXVK and VKD3D games BEFORE applying lsfg-vk |
Usually, specifying LSFG_LEGACY=1 and LSFG_MULTIPLIER is enough:
$ LSFG_LEGACY=1 LSFG_MULTIPLIER=4 vkcube
Benchmark
If you want to see how well lsfg-vk performs on your system, you can use the integrated benchmark. Run the process like this:
$ LD_PRELOAD=~/.local/lib/liblsfg-vk.so LSFG_BENCHMARK=1 (...) /bin/echo -n
Instead of (...) you can specify the flags above as well as LSFG_EXTENT_WIDTH and LSFG_EXTENT_HEIGHT which change the resolution of the test. If you're having issues with the device selection, use LSFG_DEVICE_UUID to override the device used in the benchmark (The device id is 0x(vendorid)(deviceid) in decimal)
The output of the test should look something like this:
$ LD_PRELOAD=/home/pancake/.local/lib/liblsfg-vk.so LSFG_BENCHMARK=1 LSFG_EXTENT_WIDTH=2560 LSFG_EXTENT_HEIGHT=1440 LSFG_MULTIPLIER=3 LSFG_FLOW_SCALE=0.70 /bin/echo -n
lsfg-vk(bench): Running 3x benchmark with 2560x1440 extent and flow scale of 0.7 without HDR
lsfg-vk(bench): Benchmark context created, ready to run
lsfg-vk(bench): 0.00% done (1/4001)
lsfg-vk(bench): 12.50% done (501/4001)
lsfg-vk(bench): 24.99% done (1001/4001)
lsfg-vk(bench): 37.49% done (1501/4001)
lsfg-vk(bench): 49.99% done (2001/4001)
lsfg-vk(bench): 62.48% done (2501/4001)
lsfg-vk(bench): 74.98% done (3001/4001)
lsfg-vk(bench): 87.48% done (3501/4001)
lsfg-vk(bench): 99.98% done (4001/4001)
lsfg-vk(bench): Benchmark completed in 9591 ms
lsfg-vk(bench): Time per iteration: 2.40 ms
lsfg-vk(bench): Generation FPS: 834.32
lsfg-vk(bench): Final FPS: 1251.49
lsfg-vk(bench): Benchmark finished, exiting
It's recommended to have a game running at locked FPS to see just how fast lsfg-vk can run while a game is also running.