From 85d47a1931883f9529cb80923d9299aaab8fef50 Mon Sep 17 00:00:00 2001 From: Mikhail Katliar Date: Sun, 6 Oct 2024 12:18:54 +0200 Subject: [PATCH] Increased inlining threshold for Clang --- bench/common/CMakeLists.txt | 2 +- include/blast/math/algorithm/Tile.hpp | 5 ++--- include/blast/math/algorithm/arch/avx2/Tile.hpp | 2 -- 3 files changed, 3 insertions(+), 6 deletions(-) diff --git a/bench/common/CMakeLists.txt b/bench/common/CMakeLists.txt index 030ae1e..ddc5c5a 100644 --- 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 index 3b4efef..b15f2b5 100644 --- a/include/blast/math/algorithm/Tile.hpp +++ b/include/blast/math/algorithm/Tile.hpp @@ -13,8 +13,7 @@ #endif #include - -#include +#include namespace blast @@ -47,7 +46,7 @@ namespace blast * @param f_partial functor to call on partial tiles */ template - 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(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 index f7decfd..322dac2 100644 --- a/include/blast/math/algorithm/arch/avx2/Tile.hpp +++ b/include/blast/math/algorithm/arch/avx2/Tile.hpp @@ -10,8 +10,6 @@ #include -#include - namespace blast :: detail {