src/cpu/x86/vm/sharedRuntime_x86_64.cpp

@@ -45,10 +45,11 @@
 #include "opto/runtime.hpp"
 #endif
 #if INCLUDE_JVMCI
 #include "jvmci/jvmciJavaClasses.hpp"
 #endif
+#include "vm_version_x86.hpp"
 
 #define __ masm->
 
 const int StackAlignmentInSlots = StackAlignmentInBytes / VMRegImpl::stack_slot_size;
 

@@ -149,12 +150,12 @@
   if (UseAVX < 3) {
     num_xmm_regs = num_xmm_regs/2;
   }
 #if defined(COMPILER2) || INCLUDE_JVMCI
   if (save_vectors) {
-    assert(UseAVX > 0, "up to 512bit vectors are supported with EVEX");
-    assert(MaxVectorSize <= 64, "up to 512bit vectors are supported now");
+    assert(UseAVX > 0, "Vectors larger than 16 byte long are supported only with AVX");
+    assert(MaxVectorSize <= 64, "Only up to 64 byte long vectors are supported");
   }
 #else
   assert(!save_vectors, "vectors are generated only by C2 and JVMCI");
 #endif
 

@@ -204,10 +205,13 @@
       for (int n = 16; n < num_xmm_regs; n++) {
         __ movsd(Address(rsp, base_addr+(off++*64)), as_XMMRegister(n));
       }
     }
   }
+  if (VM_Version::supports_avx()) {
+    __ vzeroupper();
+  }
   if (frame::arg_reg_save_area_bytes != 0) {
     // Allocate argument register save area
     __ subptr(rsp, frame::arg_reg_save_area_bytes);
   }
 

@@ -320,17 +324,21 @@
     __ addptr(rsp, frame::arg_reg_save_area_bytes);
   }
 
 #if defined(COMPILER2) || INCLUDE_JVMCI
   if (restore_vectors) {
-    assert(UseAVX > 0, "up to 512bit vectors are supported with EVEX");
-    assert(MaxVectorSize <= 64, "up to 512bit vectors are supported now");
+    assert(UseAVX > 0, "Vectors larger than 16 byte long are supported only with AVX");
+    assert(MaxVectorSize <= 64, "Only up to 64 byte long vectors are supported");
   }
 #else
   assert(!restore_vectors, "vectors are generated only by C2");
 #endif
 
+  if (VM_Version::supports_avx()) {
+    __ vzeroupper();
+  }
+
   // On EVEX enabled targets everything is handled in pop fpu state
   if (restore_vectors) {
     // Restore upper half of YMM registers (0..15)
     int base_addr = XSAVE_AREA_YMM_BEGIN;
     for (int n = 0; n < 16; n++) {

@@ -527,10 +535,13 @@
 
   // align stack so push_CPU_state doesn't fault
   __ andptr(rsp, -(StackAlignmentInBytes));
   __ push_CPU_state();
 
+  if (VM_Version::supports_avx()) {
+    __ vzeroupper();
+  }
   // VM needs caller's callsite
   // VM needs target method
   // This needs to be a long call since we will relocate this adapter to
   // the codeBuffer and it may not reach
 

@@ -545,10 +556,13 @@
   // De-allocate argument register save area
   if (frame::arg_reg_save_area_bytes != 0) {
     __ addptr(rsp, frame::arg_reg_save_area_bytes);
   }
 
+  if (VM_Version::supports_avx()) {
+    __ vzeroupper();
+  }
   __ pop_CPU_state();
   // restore sp
   __ mov(rsp, r13);
   __ bind(L);
 }

@@ -1463,11 +1477,10 @@
 
   __ reset_last_Java_frame(false);
 
   save_or_restore_arguments(masm, stack_slots, total_in_args,
                             arg_save_area, NULL, in_regs, in_sig_bt);
-
   __ bind(cont);
 #ifdef ASSERT
   if (StressCriticalJNINatives) {
     // Stress register saving
     OopMap* map = new OopMap(stack_slots * 2, 0 /* arg_slots*/);

@@ -2483,10 +2496,13 @@
     // and never return here preventing us from clearing _last_native_pc down below.
     // Also can't use call_VM_leaf either as it will check to see if rsi & rdi are
     // preserved and correspond to the bcp/locals pointers. So we do a runtime call
     // by hand.
     //
+    if (VM_Version::supports_avx()) {
+      __ vzeroupper();
+    }
     save_native_result(masm, ret_type, stack_slots);
     __ mov(c_rarg0, r15_thread);
     __ mov(r12, rsp); // remember sp
     __ subptr(rsp, frame::arg_reg_save_area_bytes); // windows
     __ andptr(rsp, -16); // align stack as required by ABI

@@ -2659,10 +2675,13 @@
     // BEGIN Slow path unlock
     __ bind(slow_path_unlock);
 
     // If we haven't already saved the native result we must save it now as xmm registers
     // are still exposed.
+    if (VM_Version::supports_avx()) {
+      __ vzeroupper();
+    }
 
     if (ret_type == T_FLOAT || ret_type == T_DOUBLE ) {
       save_native_result(masm, ret_type, stack_slots);
     }
 

@@ -2705,10 +2724,13 @@
   } // synchronized
 
   // SLOW PATH Reguard the stack if needed
 
   __ bind(reguard);
+  if (VM_Version::supports_avx()) {
+    __ vzeroupper();
+  }
   save_native_result(masm, ret_type, stack_slots);
   __ mov(r12, rsp); // remember sp
   __ subptr(rsp, frame::arg_reg_save_area_bytes); // windows
   __ andptr(rsp, -16); // align stack as required by ABI
   __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::reguard_yellow_pages)));