llm: build with GGML_STATIC=on for static lib

.github/workflows/release.yaml

@@ -304,11 +304,6 @@ jobs:
           write-host "Installing plugin"
           & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
           write-host "plugin installed"
-      - name: remove unwanted mingw dll.a files
-        run: |
-          Get-ChildItem -Path "C:\mingw64" -Recurse -Filter "libpthread.dll.a" -File | Remove-Item -Force
-          Get-ChildItem -Path "C:\mingw64" -Recurse -Filter "libwinpthread.dll.a" -File | Remove-Item -Force
-          Get-ChildItem -Path "C:\mingw64" -Recurse -Filter "libstdc++.dll.a" -File | Remove-Item -Force
       - uses: actions/setup-go@v5
         with:
           go-version-file: go.mod

gpu/gpu_darwin.go

@@ -56,7 +56,7 @@ func GetCPUInfo() GpuInfoList {
 func GetCPUMem() (memInfo, error) {
 	return memInfo{
 		TotalMemory: uint64(C.getPhysicalMemory()),
-		FreeMemory:  uint64(C.getFreeMemory()),
+		FreeMemory:  0,
 	}, nil
 }
 

gpu/gpu_info_darwin.h

@@ -2,4 +2,3 @@
 #include <stdint.h>
 uint64_t getRecommendedMaxVRAM();
 uint64_t getPhysicalMemory();
-uint64_t getFreeMemory();

gpu/gpu_info_darwin.m

@@ -1,5 +1,4 @@
-#import <Foundation/Foundation.h>
-#import <mach/mach.h>
+// go:build darwin
 #include "gpu_info_darwin.h"
 
 uint64_t getRecommendedMaxVRAM() {
@@ -9,27 +8,6 @@ uint64_t getRecommendedMaxVRAM() {
   return result;
 }
 
-// getPhysicalMemory returns the total physical memory in bytes
 uint64_t getPhysicalMemory() {
-  return [NSProcessInfo processInfo].physicalMemory;
-}
-
-// getFreeMemory returns the total free memory in bytes, including inactive
-// memory that can be reclaimed by the system.
-uint64_t getFreeMemory() {
-  mach_port_t host_port = mach_host_self();
-  mach_msg_type_number_t host_size = sizeof(vm_statistics64_data_t) / sizeof(integer_t);
-  vm_size_t pagesize;
-  vm_statistics64_data_t vm_stat;
-
-  host_page_size(host_port, &pagesize);
-  if (host_statistics64(host_port, HOST_VM_INFO64, (host_info64_t)&vm_stat, &host_size) != KERN_SUCCESS) {
-    return 0;
-  }
-
-  uint64_t free_memory = (uint64_t)vm_stat.free_count * pagesize;
-  free_memory += (uint64_t)vm_stat.speculative_count * pagesize;
-  free_memory += (uint64_t)vm_stat.inactive_count * pagesize;
-
-  return free_memory;
+  return [[NSProcessInfo processInfo] physicalMemory];
 }

llm/generate/gen_linux.sh

@@ -77,7 +77,7 @@ if [ -z "${OLLAMA_SKIP_CPU_GENERATE}" ]; then
     if [ -n "${OLLAMA_CUSTOM_CPU_DEFS}" ]; then
         init_vars
         echo "OLLAMA_CUSTOM_CPU_DEFS=\"${OLLAMA_CUSTOM_CPU_DEFS}\""
-        CMAKE_DEFS="${OLLAMA_CUSTOM_CPU_DEFS} -DBUILD_SHARED_LIBS=off -DCMAKE_POSITION_INDEPENDENT_CODE=on ${CMAKE_DEFS}"
+        CMAKE_DEFS="${OLLAMA_CUSTOM_CPU_DEFS} -DCMAKE_POSITION_INDEPENDENT_CODE=on ${CMAKE_DEFS}"
         BUILD_DIR="../build/linux/${ARCH}/cpu"
         echo "Building custom CPU"
         build
@@ -93,7 +93,7 @@ if [ -z "${OLLAMA_SKIP_CPU_GENERATE}" ]; then
         # -DGGML_AVX512_VBMI -- 2018 Intel Cannon Lake
         # -DGGML_AVX512_VNNI -- 2021 Intel Alder Lake
 
-        COMMON_CPU_DEFS="-DBUILD_SHARED_LIBS=off -DCMAKE_POSITION_INDEPENDENT_CODE=on -DGGML_NATIVE=off -DGGML_OPENMP=off"
+        COMMON_CPU_DEFS="-DCMAKE_POSITION_INDEPENDENT_CODE=on -DGGML_NATIVE=off -DGGML_OPENMP=off"
         if [ -z "${OLLAMA_CPU_TARGET}" -o "${OLLAMA_CPU_TARGET}" = "cpu" ]; then
             #
             # CPU first for the default library, set up as lowest common denominator for maximum compatibility (including Rosetta)

llm/generate/gen_windows.ps1

@@ -204,6 +204,7 @@ function build_static() {
             "-DCMAKE_C_COMPILER=gcc.exe",
             "-DCMAKE_CXX_COMPILER=g++.exe",
             "-DBUILD_SHARED_LIBS=off",
+            "-DGGML_STATIC=on",
             "-DGGML_NATIVE=off",
             "-DGGML_AVX=off",
             "-DGGML_AVX2=off",

llm/llm.go

@@ -1,7 +1,7 @@
 package llm
 
 // #cgo CFLAGS: -Illama.cpp -Illama.cpp/include -Illama.cpp/ggml/include
-// #cgo LDFLAGS: -lllama -lggml -lstdc++ -lpthread
+// #cgo LDFLAGS: -lllama -lggml -lstdc++
 // #cgo darwin,arm64 LDFLAGS: -L${SRCDIR}/build/darwin/arm64_static -L${SRCDIR}/build/darwin/arm64_static/src -L${SRCDIR}/build/darwin/arm64_static/ggml/src -framework Accelerate -framework Metal
 // #cgo darwin,amd64 LDFLAGS: -L${SRCDIR}/build/darwin/x86_64_static -L${SRCDIR}/build/darwin/x86_64_static/src -L${SRCDIR}/build/darwin/x86_64_static/ggml/src
 // #cgo windows,amd64 LDFLAGS: -L${SRCDIR}/build/windows/amd64_static -L${SRCDIR}/build/windows/amd64_static/src -L${SRCDIR}/build/windows/amd64_static/ggml/src