From 574b3c32937997eaf56700d06638cdfedb475846 Mon Sep 17 00:00:00 2001 From: cheesss Date: Sat, 20 Jun 2026 03:24:00 +0900 Subject: [PATCH] docs: add Windows SYCL install example --- README.md | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/README.md b/README.md index c02df9eed..89dc37485 100644 --- a/README.md +++ b/README.md @@ -232,6 +232,19 @@ To install with SYCL support, set the `GGML_SYCL=on` environment variable before source /opt/intel/oneapi/setvars.sh CMAKE_ARGS="-DGGML_SYCL=on -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx" pip install llama-cpp-python ``` + +On Windows, run the install from an Intel oneAPI command prompt, install Ninja if needed, and select the Ninja generator explicitly: + +```powershell +$env:CMAKE_ARGS = "-DGGML_SYCL=on -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -G Ninja" +python -m pip install llama-cpp-python +``` + +If you have multiple oneAPI devices, select the Level Zero GPU at runtime: + +```powershell +$env:ONEAPI_DEVICE_SELECTOR = "level_zero:gpu" +```