Skip to content
Snippets Groups Projects
Commit 7e23a15d authored by Peter W. Draper's avatar Peter W. Draper
Browse files

Add optimizations for the AMD genoa and bergamo chips 

These have AVX512. With the Intel compilers you need to be careful not to use -x optimizations so it isn't straight-forward to get these optimizations, faking skylake-avx512 seems to work.
parent 5b8ee090
No related branches found
No related tags found
3 merge requests!1887Updating . . .,!1878updating working branch,!1805Add optimizations for the AMD genoa and bergamo chips
Hide whitespace changes
Inline Side-by-side
Showing
with 10 additions and 3 deletions
m4/ax_cc_maxopt.m4
+ 10
− 3
View file @ 7e23a15d
......@@ -125,11 +125,11 @@ if test "$ac_test_CFLAGS" != "set"; then
*2?6[[ad]]?:*:*:*) icc_flags="-xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Sandy-bridge
*3?6[[ae]]?:*:*:*) icc_flags="-xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; #Ivy-bridge
*3?6[[cf]]?:*:*:*|*4?6[[56]]?:*:*:*|*4?6[[ef]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Haswell
*3?6d?:*:*:*|*4?6[[7f]]?:*:*:*|*5?66?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Broadwell
*3?6d?:*:*:*|*4?6[[7f]]?:*:*:*|*5?66?:*:*:*) icc_flags=" -xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Broadwell
*4?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Skylake
*5?6[[56]]?:*:*:*) icc_flags="-xCORE-AVX512 -xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Skylake-AVX512
*5?67?:*:*:*) icc_flags="-xMIC-AVX512 -xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;; # Knights-Landing
*8?6[[de]]?:*:*:*|*9?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;;# Kabylake
*8?6[[de]]?:*:*:*|*9?6[[de]]?:*:*:*) icc_flags="-xCORE-AVX2 -xCORE-AVX-I -xAVX -SSE4.2 -xS -xT -xB -xK" ;;# Kabylake
*000?f[[346]]?:*:*:*|?f[[346]]?:*:*:*|f[[346]]?:*:*:*) icc_flags="-xSSE3 -xP -xO -xN -xW -xK" ;;
*00??f??:*:*:*|??f??:*:*:*|?f??:*:*:*|f??:*:*:*) icc_flags="-xSSE2 -xN -xW -xK" ;;
esac ;;
......@@ -139,7 +139,14 @@ if test "$ac_test_CFLAGS" != "set"; then
*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="$CFLAGS -fma -ftz -fomit-frame-pointer";; # EPYC
CFLAGS="$CFLAGS -fma -ftz -fomit-frame-pointer";; # ROME
a0?f??:*:*:*) icc_flags="-march=core-avx2"
CFLAGS="$CFLAGS -fma -ftz -fomit-frame-pointer";; # MILAN
a1?f??:*:*:*) icc_flags="-axCORE-AVX512"
CFLAGS="$CFLAGS -march=skylake-avx512 -fma -ftz -fomit-frame-pointer";; # GENOA
aa?f??:*:*:*) icc_flags="-axCORE-AVX512"
CFLAGS="$CFLAGS -march=skylake-avx512 -fma -ftz -fomit-frame-pointer";; # BERGAMO
esac ;;
esac ;;
esac
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment