From f7b4f98cb0c642d747539980d258343e7882562f Mon Sep 17 00:00:00 2001 From: "Peter W. Draper" <p.w.draper@durham.ac.uk> Date: Thu, 21 Sep 2023 13:59:07 +0100 Subject: [PATCH] Add ICC Milan optimization and change architecture to corei7-avx as test show that to be 20% faster for gravity only and the same speed for hydro. --- m4/ax_cc_maxopt.m4 | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/m4/ax_cc_maxopt.m4 b/m4/ax_cc_maxopt.m4 index 282668af72..8e33cd869b 100644 --- a/m4/ax_cc_maxopt.m4 +++ b/m4/ax_cc_maxopt.m4 @@ -129,17 +129,21 @@ if test "$ac_test_CFLAGS" != "set"; then *4?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Skylake *5?6[[56]]?:*:*:*) icc_flags="-xCORE-AVX512 -xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Skylake-AVX512 *5?67?:*:*:*) icc_flags="-xMIC-AVX512 -xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Knights-Landing - *8?6[[de]]?:*:*:*|*9?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;;# Kabylake + *8?6[[de]]?:*:*:*|*9?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;;# Kabylake *000?f[[346]]?:*:*:*|?f[[346]]?:*:*:*|f[[346]]?:*:*:*) icc_flags="-xSSE3 -xP -xO -xN -xW -xK" ;; *00??f??:*:*:*|??f??:*:*:*|?f??:*:*:*|f??:*:*:*) icc_flags="-xSSE2 -xN -xW -xK" ;; esac ;; *:68747541:444d4163:69746e65) # AMDs with AVX2 support. + # SWIFT gravity prefers AVX so + # we used corei7-avx not core-avx2 case $ax_cv_gcc_x86_cpuid_1 in - *061?f??:*:*:*|61?f??:*:*:*) icc_flags="-march=core-avx2" ;; - *06??f??:*:*:*|6??f??:*:*:*) icc_flags="-march=core-avx2" ;; - *070?f??:*:*:*|70?f??:*:*:*) icc_flags="-march=core-avx2" ;; - 83?f??:*:*:*) icc_flags="-march=core-avx2" - CFLAGS="$CFLAGS -fma -ftz -fomit-frame-pointer";; # EPYC + *061?f??:*:*:*|61?f??:*:*:*) icc_flags="-march=corei7-avx" ;; + *06??f??:*:*:*|6??f??:*:*:*) icc_flags="-march=corei7-avx" ;; + *070?f??:*:*:*|70?f??:*:*:*) icc_flags="-march=corei7-avx" ;; + 83?f??:*:*:*) icc_flags="-march=corei7-avx" + CFLAGS="$CFLAGS -fma -ftz -fomit-frame-pointer";; # EPYC ROME + a0?f??:*:*:*) icc_flags="-march=corei7-avx" + CFLAGS="$CFLAGS -fma -ftz -fomit-frame-pointer";; # MILAN esac ;; esac ;; esac -- GitLab