make : use -lfto=auto to avoid warnings and maintain perf

.github/workflows/build.yml

@@ -33,17 +33,17 @@ jobs:
         id: depends
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential gcc-8
+          sudo apt-get install build-essential gcc-8 g++-8
 
       - name: Build
         id: make_build
         run: |
-          CC=gcc-8 make -j $(nproc)
+          CC=gcc-8 CXX=g++-8 make -j $(nproc)
 
       - name: Test
         id: make_test
         run: |
-          CC=gcc-8 make tests -j $(nproc)
+          CC=gcc-8 CXX=g++-8 make tests -j $(nproc)
           make test -j $(nproc)
 
   ubuntu-latest-cmake:

.github/workflows/code-coverage.yml

@@ -15,13 +15,13 @@ jobs:
       - name: Dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install build-essential gcc-8 lcov
+          sudo apt-get install build-essential gcc-9 g++-9 lcov
 
       - name: Build
-        run: CC=gcc-8 make -j LLAMA_CODE_COVERAGE=1 tests
+        run: CC=gcc-9 CXX=g++-9 make -j LLAMA_CODE_COVERAGE=1 tests
 
       - name: Run tests
-        run: CC=gcc-8 make test
+        run: CC=gcc-9 CXX=g++-9 make test
 
       - name: Generate coverage report
         run: |

CMakeLists.txt

@@ -45,7 +45,7 @@ endif()
 # general
 option(LLAMA_STATIC                     "llama: static link libraries"                          OFF)
 option(LLAMA_NATIVE                     "llama: enable -march=native flag"                      ON)
-option(LLAMA_LTO                        "llama: enable link time optimization"                  OFF)
+option(LLAMA_LTO                        "llama: enable link time optimization"                  ON)
 
 # debug
 option(LLAMA_ALL_WARNINGS               "llama: enable all compiler warnings"                   ON)

Makefile

@@ -120,11 +120,11 @@ MK_CXXFLAGS = -std=c++11 -fPIC
 
 # -Ofast tends to produce faster code, but may not be available for some compilers.
 ifdef LLAMA_FAST
-MK_CFLAGS        += -Ofast
+MK_CFLAGS        += -Ofast -flto=auto
 MK_HOST_CXXFLAGS += -Ofast
 MK_CUDA_CXXFLAGS += -O3
 else
-MK_CFLAGS        += -O3
+MK_CFLAGS        += -O3 -flto=auto
 MK_CXXFLAGS      += -O3
 endif