Use eigen blas for blas routines (#2672)

Summary:
Pull Request resolved: #2672

In previous diffs we added eigen_blas for accelerating blas routines. In this
diff we enable linking it for llama runner
ghstack-source-id: 220378204
exported-using-ghexport

Pre-existing lint failures
bypass-github-export-checks
bypass-github-pytorch-ci-checks

Reviewed By: digantdesai, mcr229

Differential Revision: D55344019

fbshipit-source-id: e76c71c36d8f0044c7c26364ec136aa7ed963544

Author: kimishpatel

build/cmake_deps.toml

@@ -275,7 +275,7 @@ filters = [
 excludes = [
   "^codegen",
   # Exclude blas, since it's built as a separate target.
-  # "^kernels/optimized/blas",
+  "^kernels/optimized/blas",
 ]
 deps = [
   "executorch",

examples/models/llama2/custom_ops/CMakeLists.txt

@@ -49,6 +49,8 @@ set(_common_include_directories ${EXECUTORCH_ROOT}/..)
 set(custom_ops_libs extension_module)
 list(APPEND custom_ops_libs pthreadpool)
 list(APPEND custom_ops_libs cpuinfo)
+list(APPEND custom_ops_libs cpublas)
+list(APPEND custom_ops_libs eigen_blas)
 
 # Generate C++ bindings to register kernels into both PyTorch (for AOT) and
 # Executorch (for runtime). Here select all ops in optimized.yaml
@@ -65,7 +67,7 @@ add_library(custom_ops ${_custom_ops__srcs})
 target_include_directories(custom_ops PUBLIC "${_common_include_directories}")
 target_include_directories(custom_ops PRIVATE
   "${CMAKE_CURRENT_BINARY_DIR}/../../../../include")
-target_link_libraries(custom_ops PRIVATE ${custom_ops_libs})
+target_link_libraries(custom_ops PUBLIC ${custom_ops_libs})
 
 target_compile_options(custom_ops PUBLIC ${_common_compile_options}
   -DET_USE_THREADPOOL)

kernels/optimized/CMakeLists.txt

@@ -46,7 +46,7 @@ if(NOT PYTHON_EXECUTABLE)
 endif()
 # Build cpublas.
 list(TRANSFORM _optimized_cpublas__srcs PREPEND "${EXECUTORCH_ROOT}/")
-add_library(cpublas ${_optimized_cpublas__srcs})
+add_library(cpublas STATIC ${_optimized_cpublas__srcs})
 target_link_libraries(cpublas PRIVATE executorch eigen_blas)
 target_compile_options(cpublas PUBLIC ${_common_compile_options})
 
@@ -72,3 +72,5 @@ gen_operators_lib(
   DEPS executorch)
 
 install(TARGETS cpublas optimized_kernels optimized_ops_lib DESTINATION lib)
+
+install(TARGETS cpublas DESTINATION lib)