Skip to content

docs(server): document Strix Halo ROCm 10.1 build #836

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
RPHllc wants to merge 1 commit into from
Closed

docs(server): document Strix Halo ROCm 10.1 build #836

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 25 additions & 0 deletions server/DEVELOPER.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,31 @@ The binary lands at `server/build/test_dflash`.

---

### Native AMD HIP build

The C++ runtime also supports AMD GPUs through HIP. This native configuration was verified on a Ryzen AI MAX+ 395 (Strix Halo, `gfx1151`) with Ubuntu 26.04.1 LTS, ROCm 10.1.0 (HIP 7.16.26385), GCC 15.2, and CMake 4.2.3, using Lucebox commit `ae039d0acc2f03cd0e8cff33836ad05e4f64d310`.

Initialize the vendored dependencies, then configure and build the server:

```bash
git submodule update --init --recursive
cmake -S server -B server/build-hip -G Ninja \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_CXX_FLAGS="-isystem /opt/rocm/core-10.1/include" \
-DCMAKE_HIP_COMPILER=/opt/rocm/core-10.1/lib/llvm/bin/clang++ \
-DCMAKE_HIP_COMPILER_ROCM_ROOT=/opt/rocm/core-10.1 \
-DCMAKE_HIP_FLAGS="-isystem /opt/rocm/core-10.1/include -isystem /opt/rocm/core-10.1/include/hip" \
-DLUCE_GPU_BACKEND=hip \
-DLUCE_HIP_ARCHITECTURES=gfx1151 \
-DGGML_HIP_MMQ_MFMA=ON \
-DGGML_HIP_NO_VMM=ON
cmake --build server/build-hip --target luce_server -j4
```

Use your system ROCm paths if they differ. For Strix Halo, set `LUCE_HIP_ARCHITECTURES` to `gfx1151`. Keep a separate build directory when switching between CUDA and HIP so CMake does not reuse a cache configured for another backend.

---

## Model files

Download models before running the server:
Expand Down