Ship portable CPU dispatch and verified shared runtime dependencies

This commit is contained in:
karl committed 2026-10-12 22:35:34 +02:00
1 parent 278b09f82c
commit 0bba1a0aa7
19 files changed
+675 -17

No files matched your search

+24 -4
View File
@@ -4,10 +4,28 @@ if(NOT DEFINED LLAMA_CPP_DIR OR NOT EXISTS "${LLAMA_CPP_DIR}/include/llama.h")
message(FATAL_ERROR "Supply -DLLAMA_CPP_DIR=<pinned patched llama.cpp b11461 source>")
endif()
set(CMAKE_CXX_STANDARD 17)
if(MSVC)
set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>" CACHE STRING "Standalone worker static MSVC runtime" FORCE)
# CPU-only builds remain small internal VM tools. Normal Windows x64 builds
# ship baseline and optimized CPU modules in the same GPU-capable package.
set(SEARCHI_ACCELERATION "auto" CACHE STRING "auto; explicit developer overrides: cpu, metal, vulkan, cuda")
set(SEARCHI_CPU_DISPATCH OFF)
if(WIN32 AND CMAKE_SIZEOF_VOID_P EQUAL 8 AND
(CMAKE_SYSTEM_PROCESSOR MATCHES "^(AMD64|amd64|x86_64|X86_64)$" OR CMAKE_GENERATOR_PLATFORM STREQUAL "x64") AND
NOT SEARCHI_ACCELERATION STREQUAL "cpu")
set(SEARCHI_CPU_DISPATCH ON)
endif()
set(BUILD_SHARED_LIBS OFF CACHE BOOL "Static local worker runtime" FORCE)
if(MSVC)
# llama and ggml exchange FILE* and other CRT-owned state across DLLs.
# Every shared-runtime component must use the same dynamic CRT.
if(SEARCHI_CPU_DISPATCH)
set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>DLL" CACHE STRING "Shared worker CRT across runtime DLLs" FORCE)
else()
set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>" CACHE STRING "Internal static worker CRT" FORCE)
endif()
endif()
set(BUILD_SHARED_LIBS ${SEARCHI_CPU_DISPATCH} CACHE BOOL "Shared runtime for Windows CPU dispatch" FORCE)
set(GGML_BACKEND_DL ${SEARCHI_CPU_DISPATCH} CACHE BOOL "Load compatible packaged CPU backend" FORCE)
set(GGML_CPU_ALL_VARIANTS ${SEARCHI_CPU_DISPATCH} CACHE BOOL "Baseline and optimized CPU variants" FORCE)
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}")
set(GGML_OPENMP OFF CACHE BOOL "Use bounded native llama thread pools without external OpenMP runtime" FORCE)
set(GGML_NATIVE OFF CACHE BOOL "Portable CPU kernels")
# Normal distribution builds include GPU support and CPU fallback. CPU-only
@@ -15,7 +33,6 @@ set(GGML_NATIVE OFF CACHE BOOL "Portable CPU kernels")
set(GGML_CUDA OFF CACHE BOOL "Production CUDA backend")
set(GGML_VULKAN OFF CACHE BOOL "Production Vulkan backend")
set(GGML_METAL OFF CACHE BOOL "Production Metal backend")
set(SEARCHI_ACCELERATION "auto" CACHE STRING "auto; explicit developer overrides: cpu, metal, vulkan, cuda")
if(SEARCHI_ACCELERATION STREQUAL "cpu")
set(GGML_CUDA OFF CACHE BOOL "CPU test build" FORCE)
set(GGML_VULKAN OFF CACHE BOOL "CPU test build" FORCE)
@@ -46,5 +63,8 @@ add_subdirectory("${LLAMA_CPP_DIR}" llama-build)
add_executable(searchi-gemma-worker worker.cpp)
target_include_directories(searchi-gemma-worker PRIVATE "${LLAMA_CPP_DIR}/vendor")
target_link_libraries(searchi-gemma-worker PRIVATE llama mtmd)
if(SEARCHI_CPU_DISPATCH)
target_compile_definitions(searchi-gemma-worker PRIVATE SEARCHI_CPU_DISPATCH=1)
endif()
target_compile_definitions(searchi-gemma-worker PRIVATE SEARCHI_LLAMA_TAG="b11461-no-logits-v1")
install(TARGETS searchi-gemma-worker RUNTIME DESTINATION bin)