Ship portable CPU dispatch and verified shared runtime dependencies

2026-10-12 22:35:34 +02:00
1 parent 278b09f82c
commit 0bba1a0aa7
@@ -0,0 +1,98 @@
+--- a/ggml/src/ggml-cpu/arch/x86/cpu-feats.cpp
+++ b/ggml/src/ggml-cpu/arch/x86/cpu-feats.cpp
+@@ -261,10 +261,30 @@
+ #endif
+ 
+ static int ggml_backend_cpu_x86_score() {
+-    // FIXME: this does not check for OS support
+-
+     int score = 1;
+     cpuid_x86 is;
+    // CPUID describes hardware; AVX also requires OS-managed XMM/YMM state.
+    // Keep this selector compiled without architecture flags (upstream design).
+    unsigned long long xcr0 = 0;
+    if (is.XSAVE() && is.OSXSAVE()) {
+#ifdef _MSC_VER
+        xcr0 = _xgetbv(0);
+#else
+        unsigned int eax, edx;
+        __asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(0));
+        xcr0 = (static_cast<unsigned long long>(edx) << 32) | eax;
+#endif
+    }
+#if defined(GGML_AVX) || defined(GGML_AVX2) || defined(GGML_FMA) || defined(GGML_F16C) || defined(GGML_AVX_VNNI) || defined(GGML_AVX512) || defined(GGML_AVX512_VBMI) || defined(GGML_AVX512_BF16) || defined(GGML_AVX512_VNNI)
+    if (!is.AVX() || (xcr0 & 0x6) != 0x6) { return 0; }
+#endif
+#if defined(GGML_AVX512) || defined(GGML_AVX512_VBMI) || defined(GGML_AVX512_BF16) || defined(GGML_AVX512_VNNI)
+    if ((xcr0 & 0xe6) != 0xe6) { return 0; }
+#endif
+#ifdef GGML_AMX_INT8
+    // Tile state also needs OS permission. Searchi does not request it.
+    return 0;
+#endif
+ 
+ #ifdef GGML_FMA
+     if (!is.FMA()) { return 0; }
+--- a/ggml/src/ggml-backend-reg.cpp
+++ b/ggml/src/ggml-backend-reg.cpp
+@@ -582,6 +582,11 @@
+     bool silent = false;
+ #endif
+ 
+    const char * cpu_only = std::getenv("GGML_CPU_ONLY");
+    if (cpu_only && std::strcmp(cpu_only, "1") == 0) {
+        ggml_backend_load_best("cpu", silent, dir_path);
+        return;
+    }
+     ggml_backend_load_best("blas", silent, dir_path);
+     ggml_backend_load_best("zendnn", silent, dir_path);
+     ggml_backend_load_best("cann", silent, dir_path);
+@@ -590,7 +595,10 @@
+     ggml_backend_load_best("metal", silent, dir_path);
+     ggml_backend_load_best("rpc", silent, dir_path);
+     ggml_backend_load_best("sycl", silent, dir_path);
+-    ggml_backend_load_best("vulkan", silent, dir_path);
+    const char * disable_vulkan = std::getenv("GGML_DISABLE_VULKAN");
+    if (!disable_vulkan || std::strcmp(disable_vulkan, "1") != 0) {
+        ggml_backend_load_best("vulkan", silent, dir_path);
+    }
+     ggml_backend_load_best("virtgpu", silent, dir_path);
+     ggml_backend_load_best("opencl", silent, dir_path);
+     ggml_backend_load_best("hexagon", silent, dir_path);
+@@ -599,7 +607,7 @@
+     ggml_backend_load_best("cpu", silent, dir_path);
+     // check the environment variable GGML_BACKEND_PATH to load an out-of-tree backend
+     const char * backend_path = std::getenv("GGML_BACKEND_PATH");
+-    if (backend_path) {
+    if (backend_path && dir_path == nullptr) {
+         ggml_backend_load(backend_path);
+     }
+ }
+--- a/CMakeLists.txt
+++ b/CMakeLists.txt
+@@ -37,7 +37,9 @@
+ # Add path to modules
+ list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")
+ 
+-set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+if (NOT DEFINED CMAKE_RUNTIME_OUTPUT_DIRECTORY)
+    set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+endif()
+ set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+ 
+ if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
+--- a/ggml/src/ggml-cpu/CMakeLists.txt
+++ b/ggml/src/ggml-cpu/CMakeLists.txt
+@@ -13,7 +13,11 @@
+     # from inlining architecture-specific instructions into the score function.
+     # Without this, LTO can cause SIGILL when loading backends on older CPUs
+     # (e.g., loading power10 backend on power9 crashes before feature check runs).
+-    target_compile_options(${GGML_CPU_FEATS_NAME} PRIVATE -fno-lto)
+    if (MSVC)
+        target_compile_options(${GGML_CPU_FEATS_NAME} PRIVATE /GL-)
+    else()
+        target_compile_options(${GGML_CPU_FEATS_NAME} PRIVATE -fno-lto)
+    endif()
+     target_link_libraries(${cpu_name} PRIVATE ${GGML_CPU_FEATS_NAME})
+ endfunction()
+