Fix checking of label suffix (#2248)

* Update X86_64_PATTERN and X86_64_ADX_PATTERN

- Add missing patterns in X86_64_PATTERN from X86_64_ADX_PATTERN

- Add zen5 in X86_64_ADX_PATTERN

* Add Zen 5 recognition in config.(guess|sub)

* Check presence of install-sh instead of config.guess

During bootstrapping, config.guess is always present.

* Conditionally check for LABEL_SUFFIX

As it is only used for x86_64 and ARM64 in longlong_asm_gnu.h, only
check the suffix when host is either of these architectures.

* Set CFLAGS and temporary param_path for Zen 5

acinclude.m4

@@ -113,10 +113,10 @@ define(X86_PATTERN,
 [[i?86*-*-* | k[5-8]*-*-* | pentium*-*-* | athlon-*-* | viac3*-*-* | geode*-*-* | atom-*-*]])
 
 define(X86_64_PATTERN,
-[[athlon64-*-* | k8-*-* | k10-*-* | bobcat-*-* | jaguar*-*-* | bulldozer*-*-* | piledriver*-*-* | steamroller*-*-* | excavator*-*-* | zen*-*-* | pentium4-*-* | atom-*-* | silvermont-*-* | goldmont-*-* | tremont-*-* | core2-*-* | corei*-*-* | x86_64-*-* | nano-*-* | nehalem*-*-* | westmere*-*-* | sandybridge*-*-* | ivybridge*-*-* | haswell*-*-* | broadwell*-*-* | skylake*-*-* | kabylake*-*-* | cometlake*-*-* | icelake*-*-* | tigerlake*-*-* | rocketlake*-*-* | alderlake*-*-* | raptorlake*-*-* | meteorlake*-*-* | x86_64v[1234]-*-*]])
+[[athlon64-*-* | k8-*-* | k10-*-* | bobcat-*-* | jaguar-*-* | bulldozer-*-* | piledriver-*-* | steamroller-*-* | excavator-*-* | pentium4-*-* | atom-*-* | silvermont-*-* | goldmont-*-* | tremont-*-* | core2-*-* | corei*-*-* | x86_64-*-* | x86_64v[1234]-*-* | nano-*-* | nehalem-*-* | westmere-*-* | sandybridge-*-* | ivybridge-*-* | haswell-*-* | zen[12345]-*-* | broadwell-*-* | skylake-*-* | skylake_server-*-* | cannonlake-*-* | kabylake-*-* | cometlake-*-* | icelake-*-* | icelake_server-*-* | rocketlake-*-* | tigerlake-*-* | alderlake-*-* | raptorlake-*-* | meteorlake-*-* | knightslanding-*-* | sapphirerapids-*-*]])
 
 define(X86_64_ADX_PATTERN,
-[[zen[1234]-*-* | coreibwl-*-* | broadwell-*-* | skylake-*-* | skylake_server-*-* | cannonlake-*-* | kabylake-*-* | cometlake-*-* | icelake-*-* | icelake_server-*-* | rocketlake-*-* | tigerlake-*-* | alderlake-*-* | raptorlake-*-* | meteorlake-*-* | knightslanding-*-* | sapphirerapids-*-*]])
+[[zen[12345]-*-* | broadwell-*-* | skylake-*-* | skylake_server-*-* | cannonlake-*-* | kabylake-*-* | cometlake-*-* | icelake-*-* | icelake_server-*-* | rocketlake-*-* | tigerlake-*-* | alderlake-*-* | raptorlake-*-* | meteorlake-*-* | knightslanding-*-* | sapphirerapids-*-*]])
 
 define(ARM64_PATTERN,
 [[armcortexa53-*-* | armcortexa53neon-*-* | armcortexa55-*-* | armcortexa55neon-*-* | armcortexa57-*-* | armcortexa57neon-*-* | armcortexa7[2-9]-*-* | armcortexa7[2-9]neon-*-* | armexynosm1-*-* | armthunderx-*-* | armxgene1-*-* | aarch64*-*-* | applem[1-9]*-*-* | armv8*-*-*]])

bootstrap.sh

@@ -6,7 +6,7 @@ rm -rf autom4te.cache
 autoreconf -i -v -Wall
 
 # The following lines are from Semigroups/Semigroups, written by Max Horn.
-if ! test -x config/config.guess -a -x config/config.sub ;
+if ! test -x config/install-sh ;
 then
     automake -acf 2> /dev/null || :
 fi

config/config.guess

@@ -1064,6 +1064,10 @@ main ()
               modelstr = "zen4";
               break;
           }
+          break;
+        case 26:                /* Zen 5 */
+          cpu_64bit = 1, cpu_avx = 1, modelstr = "zen5";
+          break;
       }
   }
   else if (strcmp (vendor_string, "HygonGenuine") == 0)

config/config.sub

@@ -102,7 +102,7 @@ itanium | itanium2)
   test_cpu=ia64 ;;
 pentium | pentiummmx | pentiumpro | pentium[234m] | k[567] | k6[23] | geode | athlon | viac3*)
   test_cpu=i386 ;;
-athlon64 | atom | silvermont | goldmont | tremont | core2 | corei* | opteron | k[89] | k10 | bobcat | jaguar* | bulldozer* | piledriver* | steamroller* | excavator* | zen[1234] | nano | nehalem | westmere | sandybridge | ivybridge | haswell | broadwell | skylake | skylake_server | cannonlake | kabylake | icelake | icelake_server | rocketlake | tigerlake | alderlake | raptorlake | meteorlake | knightslanding | sapphirerapids | cometlake | x86_64v[1234])
+athlon64 | k8 | k10 | bobcat | jaguar | bulldozer | piledriver | steamroller | excavator | pentium4 | atom | silvermont | goldmont | tremont | core2 | corei* | x86_64 | x86_64v[1234] | nano | nehalem | westmere | sandybridge | ivybridge | haswell | zen[12345] | broadwell | skylake | skylake_server | cannonlake | kabylake | cometlake | icelake | icelake_server | rocketlake | tigerlake | alderlake | raptorlake | meteorlake | knightslanding | sapphirerapids)
   test_cpu=x86_64 ;;
 power[2-9] | power1[0-9] | power2sc)
   test_cpu=power ;;

configure.ac

@@ -685,7 +685,12 @@ Please report at <https://github.com/flintlib/flint/issues/>])
         param_path="x86_64/zen4"
         have_avx512="yes"
         ;;
-      coreibwl | broadwell)
+      zen5)
+        gcc_cflags_arch="-march=znver5 -march=znver4"
+        param_path="x86_64/zen4"  # fallback, to be refined
+        have_avx512="yes"
+        ;;
+      broadwell)
         gcc_cflags_arch="-march=broadwell"
         param_path="x86_64/broadwell"
         ;;
@@ -1396,12 +1401,16 @@ then
 
     GMP_FINISH
 else
-    # NOTE: LSYM_PREFIX is needed for longlong_asm_gnu.h
-    GMP_INIT(tmpconf)
-    GMP_ASM_TEXT
-    GMP_ASM_LABEL_SUFFIX
-    GMP_ASM_LSYM_PREFIX
-    rm -f $gmp_tmpconfigm4 $gmp_tmpconfigm4i $gmp_tmpconfigm4p
+    # LSYM_PREFIX is still needed for x86-64 and ARM64 in longlong_asm_gnu.h
+    case $host in
+        ARM64_PATTERN | X86_64_PATTERN)
+            GMP_INIT(tmpconf)
+            GMP_ASM_TEXT
+            GMP_ASM_LABEL_SUFFIX
+            GMP_ASM_LSYM_PREFIX
+            rm -f $gmp_tmpconfigm4 $gmp_tmpconfigm4i $gmp_tmpconfigm4p
+            ;;
+    esac
 fi
 
 if test "$enable_assembly" = "yes";