Enable Vulkan cross-compilation for Windows via MinGW#1
Closed
Copilot wants to merge 27 commits into
Closed
Conversation
- Created platform-specific scripts: build_android.sh, build_apple.sh, build_linux.sh, build_windows.ps1 - Added verify_linux_docker.sh for container-based validation - Updated .github/workflows/build_native.yml to use modular scripts - Removed obsolete scripts and tool/build_ios.dart - Updated podspecs to link with new build scripts
…onfig - Adopted official llama.cpp build patterns for Windows, Android, Linux, and macOS. - Enabled runtime CPU detection (GGML_CPU_ALL_VARIANTS) and dynamic backend loading (GGML_BACKEND_DL). - Enhanced macOS build with Metal BF16 and embedded library support. - Streamlined iOS XCFramework generation via official llama.cpp scripts. - Configured .pubignore for zero-config pub.dev distribution. - Cleaned up build artifacts and temporary workspace files.
…clean artifacts - Removed source-build fallbacks from Android, Linux, macOS, and Windows build configurations. - Updated build scripts to clean destination directories and filter redundant versioned binaries. - Standardized use of pre-built binaries in 'jniLibs', 'Frameworks', and 'lib' directories. - Refined Windows loader logic to support multiple DLL naming conventions.
…x backend display reporting
- Optimize Android GPU offloading by increasing gpuLayers to 99 - Refactor application startup to disable automatic model loading - Remove default model paths to ensure manual selection - Implement auto-popup for model selection when no path is set - Refactor Android build scripts and add Zero-Patch CMake wrapper - Fix backend display reporting for BLAS/CPU
- Linux/Windows: Switch to static backend linking via top-level CMake - Refactor loader.dart to remove fallbacks and enforce strict path resolution - Add Linux verification scripts (Docker)
- Renamed package to llamadart across all platforms and metadata. - Removed CUDA support from documentation, scripts, and API. - Implemented Zero-Patch Strategy documentation. - Clarified supported architectures and web support (wllama). - Updated version to 0.2.0 in pubspec.yaml and CHANGELOG.md. - Performed final repository cleanup of stale binaries and logs.
- Updated README to highlight zero-config usage without local compilation. - Adjusted .pubignore to include Android jniLibs and exclude root src/.
Co-authored-by: leehack <1233094+leehack@users.noreply.github.com>
Co-authored-by: leehack <1233094+leehack@users.noreply.github.com>
Owner
|
Need vulkan support |
Co-authored-by: leehack <1233094+leehack@users.noreply.github.com>
Copilot
AI
changed the title
[WIP] Add build script for Windows binary
Add Windows binary build support via MinGW cross-compilation
Jan 29, 2026
Co-authored-by: leehack <1233094+leehack@users.noreply.github.com>
Co-authored-by: leehack <1233094+leehack@users.noreply.github.com>
Copilot
AI
changed the title
Add Windows binary build support via MinGW cross-compilation
Enable Vulkan cross-compilation for Windows via MinGW
Jan 29, 2026
leehack
force-pushed
the
feat/build-infra-refinement
branch
from
fb82e18 to
cbb8fc3
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Cross-compiling Windows binaries with Vulkan backend was failing due to missing Windows Vulkan SDK and incorrect linker configuration.
Changes
Build script (
scripts/build_windows_mingw.sh):vulkan-1.dllusingdlltoolVulkan-Headerssubmodule, library from generated stubglslcavailability for shader compilationCMake configuration (
src/native/CMakeLists.txt):--whole-archivelinker flag toWindowssystem name (MinGW cross-compilation)llamatarget was linked, excluding theggml-vulkanstatic libraryDocumentation:
mingw-w64,glslc,libvulkan-dev)Result
Shaders compile at build time using host
glslc. Runtime loadsvulkan-1.dllfrom target system.Warning
Firewall rules blocked me from connecting to one or more addresses (expand for details)
I tried to connect to the following addresses, but was blocked by firewall rules:
esm.ubuntu.com/usr/lib/apt/methods/https /usr/lib/apt/methods/https(dns block)sdk.lunarg.com/usr/bin/wget wget -q REDACTED --timeout=10 --tries=1 k/llamadart/llam-I rn m_cm2.comp.cpp.d-o _WARNINGS @CMakeFiles/llama.dir/includes_CXX.rsp -D_WIN32_WINNT=0x0601 -O3 -DNDEBUG -Wmissing-d m_cm2.comp.cpp.o-E -DACC_TYPE=floatcmake_echo_color 0x0601 -O3 -DNDE--switch= s rogress-dir=/hom--progress-dir=/home/REDACTED/work/llamadart/llamadart/build-windows-vulkan/CMakeF/usr/lib/gcc/x86_64-w64-mingw32/13-win32/collect2 /home/REDACTED/wor--progress-num=91 w64-mingw32-g++ dart/build-windo/usr/bin/x86_64-w64-mingw32-g++ -DACC_TYPE=float-DGGML_USE_CPU ma.dir/llama-qua-DGGML_USE_VULKAN w64-mingw32-g++(dns block)If you need me to access, download, or install something from one of these locations, you can either:
💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.