Increased inlining threshold for Clang

mkatliar · Oct 6, 2024 · 85d47a1 · 85d47a1
1 parent 8092ab9
commit 85d47a1
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 6 deletions.
diff --git a/bench/common/CMakeLists.txt b/bench/common/CMakeLists.txt
@@ -16,6 +16,6 @@ target_link_libraries(bench-blast-common
 if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
     # More aggressive inlining with Clang
     target_compile_options(bench-blast-common
-        PUBLIC "-mllvm" "-inline-threshold=1000"
+        PUBLIC "-mllvm" "-inline-threshold=4000"
     )
 endif()
diff --git a/include/blast/math/algorithm/Tile.hpp b/include/blast/math/algorithm/Tile.hpp
@@ -13,8 +13,7 @@
 #endif
 
 #include <blast/math/StorageOrder.hpp>
-
-#include <cstdlib>
+#include <blast/util/Types.hpp>
 
 
 namespace blast
@@ -47,7 +46,7 @@ namespace blast
      * @param f_partial functor to call on partial tiles
      */
     template <typename ET, StorageOrder SO, typename FF, typename FP, typename Arch>
-    inline void tile(Arch arch, StorageOrder traversal_order, std::size_t m, std::size_t n, FF&& f_full, FP&& f_partial)
+    inline void tile(Arch arch, StorageOrder traversal_order, size_t m, size_t n, FF&& f_full, FP&& f_partial)
     {
         detail::tile<ET, SO>(arch, traversal_order, m, n, f_full, f_partial);
     }

diff --git a/include/blast/math/algorithm/arch/avx2/Tile.hpp b/include/blast/math/algorithm/arch/avx2/Tile.hpp
@@ -10,8 +10,6 @@
 
 #include <blast/math/Simd.hpp>
 
-#include <cstdlib>
-
 
 namespace blast :: detail
 {
-Original file line number
+Diff line change
@@ Expand Up / @@ -10,8 +10,6 @@ @@
     #include <blast/math/Simd.hpp>
-    #include <cstdlib>
     namespace blast :: detail
     {
@@ Expand Down @@