Replaced intrinsic by builtin for popcount

include/bmat8_impl.hpp

@@ -276,10 +276,10 @@ inline void BMat8::row_space_bitset(epu8 &res0, epu8 &res1) const {
 inline uint64_t BMat8::row_space_size_bitset() const {
     epu8 res0 {}, res1 {};
     row_space_bitset(res0, res1);
-    return (_mm_popcnt_u64(_mm_extract_epi64(res0, 0)) +
-            _mm_popcnt_u64(_mm_extract_epi64(res1, 0)) +
-            _mm_popcnt_u64(_mm_extract_epi64(res0, 1)) +
-            _mm_popcnt_u64(_mm_extract_epi64(res1, 1)));
+    return (__builtin_popcountll(_mm_extract_epi64(res0, 0)) +
+            __builtin_popcountll(_mm_extract_epi64(res1, 0)) +
+            __builtin_popcountll(_mm_extract_epi64(res0, 1)) +
+            __builtin_popcountll(_mm_extract_epi64(res1, 1)));
 }
 
 inline uint64_t BMat8::row_space_size_incl1() const {
@@ -292,7 +292,7 @@ inline uint64_t BMat8::row_space_size_incl1() const {
             orincl |= ((in | block) == block) & in;
             in = permuted(in, rotboth);
         }
-        res += _mm_popcnt_u64(_mm_movemask_epi8(block == orincl));
+        res += __builtin_popcountll(_mm_movemask_epi8(block == orincl));
         block += Epu8(16);
     }
     return res;
@@ -308,7 +308,7 @@ inline uint64_t BMat8::row_space_size_incl() const {
             in = permuted(in, rotboth);
             orincl |= ((in | block) == block) & in;
         }
-        res += _mm_popcnt_u64(_mm_movemask_epi8(block == orincl));
+        res += __builtin_popcountll(_mm_movemask_epi8(block == orincl));
         block += Epu8(16);
     }
     return res;
@@ -399,7 +399,7 @@ inline std::vector<uint8_t> BMat8::rows() const {
 
 inline size_t BMat8::nr_rows() const {
     epu8 x = _mm_set_epi64x(_data, 0);
-    return _mm_popcnt_u64(_mm_movemask_epi8(x != epu8 {}));
+    return __builtin_popcountll(_mm_movemask_epi8(x != epu8 {}));
 }
 
 static HPCOMBI_CONSTEXPR epu8

include/epu_impl.hpp

@@ -404,7 +404,7 @@ inline epu8 eval16_cycle(epu8 v) {
 inline epu8 eval16_popcount(epu8 v) {
     epu8 res{};
     for (size_t i = 0; i < 16; i++) {
-        res[i] = _mm_popcnt_u32(_mm_movemask_epi8(v == Epu8(uint8_t(i))));
+        res[i] = __builtin_popcountl(_mm_movemask_epi8(v == Epu8(uint8_t(i))));
     }
     return res;
 }

include/perm16_impl.hpp

@@ -76,7 +76,7 @@ inline uint32_t PTransf16::rank_ref() const {
     return std::accumulate(tmp.begin(), tmp.end(), uint8_t(0));
 }
 inline uint32_t PTransf16::rank() const {
-    return _mm_popcnt_u32(image_bitset());
+    return __builtin_popcountl(image_bitset());
 }
 
 inline epu8 PTransf16::fix_points_mask(bool complement) const {
@@ -107,7 +107,7 @@ inline uint8_t PTransf16::largest_moved_point() const {
 }
 /** Returns the number of fix points of \c *this */
 inline uint8_t PTransf16::nb_fix_points() const {
-    return _mm_popcnt_u32(fix_points_bitset());
+    return __builtin_popcountl(fix_points_bitset());
 }
 
 inline static HPCOMBI_CONSTEXPR uint8_t hilo_exchng_fun(uint8_t i) {
@@ -307,7 +307,7 @@ inline uint8_t Perm16::nb_descents_ref() const {
     return res;
 }
 inline uint8_t Perm16::nb_descents() const {
-    return _mm_popcnt_u32(_mm_movemask_epi8(v < shifted_right(v)));
+    return __builtin_popcountl(_mm_movemask_epi8(v < shifted_right(v)));
 }
 
 inline uint8_t Perm16::nb_cycles_ref() const {
@@ -338,7 +338,7 @@ inline epu8 Perm16::cycles_partition() const {
 
 inline uint8_t Perm16::nb_cycles_unroll() const {
     epu8 res = (epu8id == cycles_partition());
-    return _mm_popcnt_u32(_mm_movemask_epi8(res));
+    return __builtin_popcountl(_mm_movemask_epi8(res));
 }
 
 inline bool Perm16::left_weak_leq_ref(Perm16 other) const {

list_intrin.txt

@@ -11,8 +11,8 @@ _mm_max_epu8;__m128i(),__m128i()
 _mm_min_epi8;__m128i(),__m128i()
 _mm_min_epu8;__m128i(),__m128i()
 _mm_movemask_epi8;__m128i()
-_mm_popcnt_u32;1
-_mm_popcnt_u64;1
+__builtin_popcountl;1
+__builtin_popcountll;1
 _mm_set_epi64x;1,1
 _mm_shuffle_epi8;__m128i(),__m128i()
 _mm_slli_epi32;__m128i(),1