src/cpu/x86/vm/stubGenerator_x86_64.cpp

@@ -260,10 +260,14 @@
     __ movptr(rbx_save, rbx);
     __ movptr(r12_save, r12);
     __ movptr(r13_save, r13);
     __ movptr(r14_save, r14);
     __ movptr(r15_save, r15);
+    if (UseAVX > 2) {
+      __ movl(rbx, 0xffff);
+      __ kmovql(k1, rbx);
+    }
 #ifdef _WIN64
     for (int i = 6; i <= 15; i++) {
       __ movdqu(xmm_save(i), as_XMMRegister(i));
     }
 

@@ -1316,11 +1320,14 @@
     __ align(OptoLoopAlignment);
     if (UseUnalignedLoadStores) {
       Label L_end;
       // Copy 64-bytes per iteration
       __ BIND(L_loop);
-      if (UseAVX >= 2) {
+      if (UseAVX > 2) {
+        __ evmovdqu(xmm0, Address(end_from, qword_count, Address::times_8, -56), Assembler::AVX_512bit);
+        __ evmovdqu(Address(end_to, qword_count, Address::times_8, -56), xmm0, Assembler::AVX_512bit);
+      } else if (UseAVX == 2) {
         __ vmovdqu(xmm0, Address(end_from, qword_count, Address::times_8, -56));
         __ vmovdqu(Address(end_to, qword_count, Address::times_8, -56), xmm0);
         __ vmovdqu(xmm1, Address(end_from, qword_count, Address::times_8, -24));
         __ vmovdqu(Address(end_to, qword_count, Address::times_8, -24), xmm1);
       } else {