took out prefetch and added commented out pragmas - they not help

magyarsort.h

@@ -9,6 +9,12 @@
  * LICENCE: CC3 - look it up, you need to mention me but that is all
  */
 
+/*
+ * Does not help much:
+// #pragma GCC target ("avx2")
+// #pragma GCC optimization ("unroll-loops")
+*/
+
 #include <cstdio>
 #include <cstdint>
 #include <cstring> // memset
@@ -90,8 +96,7 @@ namespace MagyarSort {
 		#pragma GCC unroll 64
 		for(COUNTER_TYP i = 0; i < size; ++i) {
 			// Prefetch caches
-			__builtin_prefetch(&arr[i]); // TODO: is good?
-			__builtin_prefetch(&arr[i + 64]);
+			//__builtin_prefetch(&arr[i + 64]);
 			// Creates no object, struct is empty
 			OccurenceMagic<DIGITS - 1, COUNTER_TYP>(arr, i, radicsOut);
 		}
@@ -143,8 +148,10 @@ namespace MagyarSort {
 			#pragma GCC unroll 64
 			for(COUNTER_TYP i = size; i > 0; --i) { // right-to-left to ensure already sorted digits order we keep for iterations
 				// Prefetch caches
+				/*
 				__builtin_prefetch(&from[i]); // TODO: is good?
 				if(i >= 64) { __builtin_prefetch(&from[i - 64]); } // TODO: manually unroll?
+				*/
 				// Get num and its new offset / location
 				auto num = from[i - 1];
 				auto digVal = getDigit<DIGIT>(num);