Merge

src/cpu/sparc/vm/c1_LIRGenerator_sparc.cpp

@@ -956,7 +956,8 @@ void LIRGenerator::do_NewMultiArray(NewMultiArray* x) {
     size->load_item();
     store_stack_parameter (size->result(),
                            in_ByteSize(STACK_BIAS +
-                                       (i + frame::memory_parameter_word_sp_offset) * wordSize));
+                                       frame::memory_parameter_word_sp_offset * wordSize +
+                                       i * sizeof(jint)));
   }
 
   // This instruction can be deoptimized in the slow path : use

src/cpu/sparc/vm/relocInfo_sparc.cpp

@@ -204,3 +204,9 @@ void Relocation::pd_swap_out_breakpoint(address x, short* instrs, int instrlen)
   NativeInstruction* ni = nativeInstruction_at(x);
   ni->set_long_at(0, u.l);
 }
+
+void poll_Relocation::fix_relocation_after_move(const CodeBuffer* src, CodeBuffer* dest) {
+}
+
+void poll_return_Relocation::fix_relocation_after_move(const CodeBuffer* src, CodeBuffer* dest) {
+}

src/cpu/sparc/vm/sharedRuntime_sparc.cpp

@@ -465,9 +465,7 @@ int SharedRuntime::java_calling_convention(const BasicType *sig_bt,
 
     case T_LONG:
       assert(sig_bt[i+1] == T_VOID, "expecting VOID in other half");
-#ifdef COMPILER2
 #ifdef _LP64
-        // Can't be tiered (yet)
         if (int_reg < int_reg_max) {
           Register r = is_outgoing ? as_oRegister(int_reg++) : as_iRegister(int_reg++);
           regs[i].set2(r->as_VMReg());
@@ -476,11 +474,12 @@ int SharedRuntime::java_calling_convention(const BasicType *sig_bt,
           stk_reg_pairs += 2;
         }
 #else
+#ifdef COMPILER2
         // For 32-bit build, can't pass longs in O-regs because they become
         // I-regs and get trashed.  Use G-regs instead.  G1 and G4 are almost
         // spare and available.  This convention isn't used by the Sparc ABI or
         // anywhere else. If we're tiered then we don't use G-regs because c1
-        // can't deal with them as a "pair".
+        // can't deal with them as a "pair". (Tiered makes this code think g's are filled)
         // G0: zero
         // G1: 1st Long arg
         // G2: global allocated to TLS
@@ -500,7 +499,6 @@ int SharedRuntime::java_calling_convention(const BasicType *sig_bt,
           regs[i].set2(VMRegImpl::stack2reg(stk_reg_pairs));
           stk_reg_pairs += 2;
         }
-#endif // _LP64
 #else // COMPILER2
         if (int_reg_pairs + 1 < int_reg_max) {
           if (is_outgoing) {
@@ -514,6 +512,7 @@ int SharedRuntime::java_calling_convention(const BasicType *sig_bt,
           stk_reg_pairs += 2;
         }
 #endif // COMPILER2
+#endif // _LP64
       break;
 
     case T_FLOAT:
@@ -699,17 +698,16 @@ Register AdapterGenerator::next_arg_slot(const int st_off){
 // Stores long into offset pointed to by base
 void AdapterGenerator::store_c2i_long(Register r, Register base,
                                       const int st_off, bool is_stack) {
-#ifdef COMPILER2
 #ifdef _LP64
   // In V9, longs are given 2 64-bit slots in the interpreter, but the
   // data is passed in only 1 slot.
   __ stx(r, base, next_arg_slot(st_off));
 #else
+#ifdef COMPILER2
   // Misaligned store of 64-bit data
   __ stw(r, base, arg_slot(st_off));    // lo bits
   __ srlx(r, 32, r);
   __ stw(r, base, next_arg_slot(st_off));  // hi bits
-#endif // _LP64
 #else
   if (is_stack) {
     // Misaligned store of 64-bit data
@@ -721,6 +719,7 @@ void AdapterGenerator::store_c2i_long(Register r, Register base,
     __ stw(r             , base, next_arg_slot(st_off)); // hi bits
   }
 #endif // COMPILER2
+#endif // _LP64
   tag_c2i_arg(frame::TagCategory2, base, st_off, r);
 }
 
@@ -1637,7 +1636,7 @@ static void long_move(MacroAssembler* masm, VMRegPair src, VMRegPair dst) {
     }
   } else if (dst.is_single_phys_reg()) {
     if (src.is_adjacent_aligned_on_stack(2)) {
-      __ ld_long(FP, reg2offset(src.first()) + STACK_BIAS, dst.first()->as_Register());
+      __ ldx(FP, reg2offset(src.first()) + STACK_BIAS, dst.first()->as_Register());
     } else {
       // dst is a single reg.
       // Remember lo is low address not msb for stack slots
@@ -1811,7 +1810,6 @@ nmethod *SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
                                                 VMRegPair *in_regs,
                                                 BasicType ret_type) {
 
-
   // Native nmethod wrappers never take possesion of the oop arguments.
   // So the caller will gc the arguments. The only thing we need an
   // oopMap for is if the call is static

src/cpu/x86/vm/assembler_x86_64.inline.hpp → src/cpu/x86/vm/assembler_x86.inline.hpp

 /*
- * Copyright 2003-2005 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2005 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -22,12 +22,6 @@
  *
  */
 
-inline void Assembler::emit_long64(jlong x) {
-  *(jlong*) _code_pos = x;
-  _code_pos += sizeof(jlong);
-  code_section()->set_end(_code_pos);
-}
-
 inline void MacroAssembler::pd_patch_instruction(address branch, address target) {
   unsigned char op = branch[0];
   assert(op == 0xE8 /* call */ ||
@@ -69,18 +63,25 @@ inline void MacroAssembler::pd_print_patched_instruction(address branch) {
 }
 #endif // ndef PRODUCT
 
-inline void MacroAssembler::movptr(Address dst, intptr_t src) {
-#ifdef _LP64
-  Assembler::mov64(dst, src);
-#else
-  Assembler::movl(dst, src);
-#endif // _LP64
-}
+#ifndef _LP64
+inline int Assembler::prefix_and_encode(int reg_enc, bool byteinst) { return reg_enc; }
+inline int Assembler::prefixq_and_encode(int reg_enc) { return reg_enc; }
 
-inline void MacroAssembler::movptr(Register dst, intptr_t src) {
-#ifdef _LP64
-  Assembler::mov64(dst, src);
+inline int Assembler::prefix_and_encode(int dst_enc, int src_enc, bool byteinst) { return dst_enc << 3 | src_enc; }
+inline int Assembler::prefixq_and_encode(int dst_enc, int src_enc) { return dst_enc << 3 | src_enc; }
+
+inline void Assembler::prefix(Register reg) {}
+inline void Assembler::prefix(Address adr) {}
+inline void Assembler::prefixq(Address adr) {}
+
+inline void Assembler::prefix(Address adr, Register reg,  bool byteinst) {}
+inline void Assembler::prefixq(Address adr, Register reg) {}
+
+inline void Assembler::prefix(Address adr, XMMRegister reg) {}
 #else
-  Assembler::movl(dst, src);
-#endif // _LP64
+inline void Assembler::emit_long64(jlong x) {
+  *(jlong*) _code_pos = x;
+  _code_pos += sizeof(jlong);
+  code_section()->set_end(_code_pos);
 }
+#endif // _LP64

src/cpu/x86/vm/assembler_x86_32.inline.hpp

-/*
- * Copyright 1997-2005 Sun Microsystems, Inc.  All Rights Reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- */
-
-inline void MacroAssembler::pd_patch_instruction(address branch, address target) {
-  unsigned char op = branch[0];
-  assert(op == 0xE8 /* call */ ||
-         op == 0xE9 /* jmp */ ||
-         op == 0xEB /* short jmp */ ||
-         (op & 0xF0) == 0x70 /* short jcc */ ||
-         op == 0x0F && (branch[1] & 0xF0) == 0x80 /* jcc */,
-         "Invalid opcode at patch point");
-
-  if (op == 0xEB || (op & 0xF0) == 0x70) {
-    // short offset operators (jmp and jcc)
-    char* disp = (char*) &branch[1];
-    int imm8 = target - (address) &disp[1];
-    guarantee(this->is8bit(imm8), "Short forward jump exceeds 8-bit offset");
-    *disp = imm8;
-  } else {
-    int* disp = (int*) &branch[(op == 0x0F)? 2: 1];
-    int imm32 = target - (address) &disp[1];
-    *disp = imm32;
-  }
-}
-
-#ifndef PRODUCT
-inline void MacroAssembler::pd_print_patched_instruction(address branch) {
-  const char* s;
-  unsigned char op = branch[0];
-  if (op == 0xE8) {
-    s = "call";
-  } else if (op == 0xE9 || op == 0xEB) {
-    s = "jmp";
-  } else if ((op & 0xF0) == 0x70) {
-    s = "jcc";
-  } else if (op == 0x0F) {
-    s = "jcc";
-  } else {
-    s = "????";
-  }
-  tty->print("%s (unresolved)", s);
-}
-#endif // ndef PRODUCT

src/cpu/x86/vm/c1_CodeStubs_x86.cpp

@@ -43,11 +43,12 @@ void ConversionStub::emit_code(LIR_Assembler* ce) {
     __ comisd(input()->as_xmm_double_reg(),
               ExternalAddress((address)&double_zero));
   } else {
-    __ pushl(rax);
+    LP64_ONLY(ShouldNotReachHere());
+    __ push(rax);
     __ ftst();
     __ fnstsw_ax();
     __ sahf();
-    __ popl(rax);
+    __ pop(rax);
   }
 
   Label NaN, do_return;
@@ -61,7 +62,7 @@ void ConversionStub::emit_code(LIR_Assembler* ce) {
 
   // input is NaN -> return 0
   __ bind(NaN);
-  __ xorl(result()->as_register(), result()->as_register());
+  __ xorptr(result()->as_register(), result()->as_register());
 
   __ bind(do_return);
   __ jmp(_continuation);
@@ -139,7 +140,7 @@ NewInstanceStub::NewInstanceStub(LIR_Opr klass_reg, LIR_Opr result, ciInstanceKl
 void NewInstanceStub::emit_code(LIR_Assembler* ce) {
   assert(__ rsp_offset() == 0, "frame size should be fixed");
   __ bind(_entry);
-  __ movl(rdx, _klass_reg->as_register());
+  __ movptr(rdx, _klass_reg->as_register());
   __ call(RuntimeAddress(Runtime1::entry_for(_stub_id)));
   ce->add_call_info_here(_info);
   ce->verify_oop_map(_info);
@@ -306,10 +307,10 @@ void PatchingStub::emit_code(LIR_Assembler* ce) {
     assert(_obj != noreg, "must be a valid register");
     Register tmp = rax;
     if (_obj == tmp) tmp = rbx;
-    __ pushl(tmp);
+    __ push(tmp);
     __ get_thread(tmp);
-    __ cmpl(tmp, Address(_obj, instanceKlass::init_thread_offset_in_bytes() + sizeof(klassOopDesc)));
-    __ popl(tmp);
+    __ cmpptr(tmp, Address(_obj, instanceKlass::init_thread_offset_in_bytes() + sizeof(klassOopDesc)));
+    __ pop(tmp);
     __ jcc(Assembler::notEqual, call_patch);
 
     // access_field patches may execute the patched code before it's
@@ -434,7 +435,7 @@ void ArrayCopyStub::emit_code(LIR_Assembler* ce) {
     VMReg r_1 = args[i].first();
     if (r_1->is_stack()) {
       int st_off = r_1->reg2stack() * wordSize;
-      __ movl (Address(rsp, st_off), r[i]);
+      __ movptr (Address(rsp, st_off), r[i]);
     } else {
       assert(r[i] == args[i].first()->as_Register(), "Wrong register for arg ");
     }
@@ -449,7 +450,7 @@ void ArrayCopyStub::emit_code(LIR_Assembler* ce) {
   ce->add_call_info_here(info());
 
 #ifndef PRODUCT
-  __ increment(ExternalAddress((address)&Runtime1::_arraycopy_slowcase_cnt));
+  __ incrementl(ExternalAddress((address)&Runtime1::_arraycopy_slowcase_cnt));
 #endif
 
   __ jmp(_continuation);

src/cpu/x86/vm/c1_Defs_x86.hpp

@@ -36,27 +36,34 @@ enum {
 
 // registers
 enum {
-  pd_nof_cpu_regs_frame_map = 8,  // number of registers used during code emission
-  pd_nof_fpu_regs_frame_map = 8,  // number of registers used during code emission
-  pd_nof_xmm_regs_frame_map = 8,  // number of registers used during code emission
-  pd_nof_caller_save_cpu_regs_frame_map = 6,  // number of registers killed by calls
-  pd_nof_caller_save_fpu_regs_frame_map = 8,  // number of registers killed by calls
-  pd_nof_caller_save_xmm_regs_frame_map = 8,  // number of registers killed by calls
+  pd_nof_cpu_regs_frame_map = RegisterImpl::number_of_registers,       // number of registers used during code emission
+  pd_nof_fpu_regs_frame_map = FloatRegisterImpl::number_of_registers,  // number of registers used during code emission
+  pd_nof_xmm_regs_frame_map = XMMRegisterImpl::number_of_registers,    // number of registers used during code emission
 
-  pd_nof_cpu_regs_reg_alloc = 6,  // number of registers that are visible to register allocator
+#ifdef _LP64
+  #define UNALLOCATED 4    // rsp, rbp, r15, r10
+#else
+  #define UNALLOCATED 2    // rsp, rbp
+#endif // LP64
+
+  pd_nof_caller_save_cpu_regs_frame_map = pd_nof_cpu_regs_frame_map - UNALLOCATED,  // number of registers killed by calls
+  pd_nof_caller_save_fpu_regs_frame_map = pd_nof_fpu_regs_frame_map,  // number of registers killed by calls
+  pd_nof_caller_save_xmm_regs_frame_map = pd_nof_xmm_regs_frame_map,  // number of registers killed by calls
+
+  pd_nof_cpu_regs_reg_alloc = pd_nof_caller_save_cpu_regs_frame_map,  // number of registers that are visible to register allocator
   pd_nof_fpu_regs_reg_alloc = 6,  // number of registers that are visible to register allocator
 
-  pd_nof_cpu_regs_linearscan = 8, // number of registers visible to linear scan
-  pd_nof_fpu_regs_linearscan = 8, // number of registers visible to linear scan
-  pd_nof_xmm_regs_linearscan = 8, // number of registers visible to linear scan
+  pd_nof_cpu_regs_linearscan = pd_nof_cpu_regs_frame_map, // number of registers visible to linear scan
+  pd_nof_fpu_regs_linearscan = pd_nof_fpu_regs_frame_map, // number of registers visible to linear scan
+  pd_nof_xmm_regs_linearscan = pd_nof_xmm_regs_frame_map, // number of registers visible to linear scan
   pd_first_cpu_reg = 0,
-  pd_last_cpu_reg = 5,
+  pd_last_cpu_reg = NOT_LP64(5) LP64_ONLY(11),
   pd_first_byte_reg = 2,
   pd_last_byte_reg = 5,
   pd_first_fpu_reg = pd_nof_cpu_regs_frame_map,
   pd_last_fpu_reg =  pd_first_fpu_reg + 7,
   pd_first_xmm_reg = pd_nof_cpu_regs_frame_map + pd_nof_fpu_regs_frame_map,
-  pd_last_xmm_reg =  pd_first_xmm_reg + 7
+  pd_last_xmm_reg =  pd_first_xmm_reg + pd_nof_xmm_regs_frame_map - 1
 };
 
 

src/cpu/x86/vm/c1_FrameMap_x86.cpp

@@ -39,10 +39,15 @@ LIR_Opr FrameMap::map_to_opr(BasicType type, VMRegPair* reg, bool) {
     opr = LIR_OprFact::address(new LIR_Address(rsp_opr, st_off, type));
   } else if (r_1->is_Register()) {
     Register reg = r_1->as_Register();
-    if (r_2->is_Register()) {
+    if (r_2->is_Register() && (type == T_LONG || type == T_DOUBLE)) {
       Register reg2 = r_2->as_Register();
+#ifdef _LP64
+      assert(reg2 == reg, "must be same register");
+      opr = as_long_opr(reg);
+#else
       opr = as_long_opr(reg2, reg);
-    } else if (type == T_OBJECT) {
+#endif // _LP64
+    } else if (type == T_OBJECT || type == T_ARRAY) {
       opr = as_oop_opr(reg);
     } else {
       opr = as_opr(reg);
@@ -88,18 +93,39 @@ LIR_Opr FrameMap::rax_oop_opr;
 LIR_Opr FrameMap::rdx_oop_opr;
 LIR_Opr FrameMap::rcx_oop_opr;
 
-LIR_Opr FrameMap::rax_rdx_long_opr;
-LIR_Opr FrameMap::rbx_rcx_long_opr;
+LIR_Opr FrameMap::long0_opr;
+LIR_Opr FrameMap::long1_opr;
 LIR_Opr FrameMap::fpu0_float_opr;
 LIR_Opr FrameMap::fpu0_double_opr;
 LIR_Opr FrameMap::xmm0_float_opr;
 LIR_Opr FrameMap::xmm0_double_opr;
 
+#ifdef _LP64
+
+LIR_Opr  FrameMap::r8_opr;
+LIR_Opr  FrameMap::r9_opr;
+LIR_Opr FrameMap::r10_opr;
+LIR_Opr FrameMap::r11_opr;
+LIR_Opr FrameMap::r12_opr;
+LIR_Opr FrameMap::r13_opr;
+LIR_Opr FrameMap::r14_opr;
+LIR_Opr FrameMap::r15_opr;
+
+// r10 and r15 can never contain oops since they aren't available to
+// the allocator
+LIR_Opr  FrameMap::r8_oop_opr;
+LIR_Opr  FrameMap::r9_oop_opr;
+LIR_Opr FrameMap::r11_oop_opr;
+LIR_Opr FrameMap::r12_oop_opr;
+LIR_Opr FrameMap::r13_oop_opr;
+LIR_Opr FrameMap::r14_oop_opr;
+#endif // _LP64
+
 LIR_Opr FrameMap::_caller_save_cpu_regs[] = { 0, };
 LIR_Opr FrameMap::_caller_save_fpu_regs[] = { 0, };
 LIR_Opr FrameMap::_caller_save_xmm_regs[] = { 0, };
 
-XMMRegister FrameMap::_xmm_regs [8] = { 0, };
+XMMRegister FrameMap::_xmm_regs [] = { 0, };
 
 XMMRegister FrameMap::nr2xmmreg(int rnr) {
   assert(_init_done, "tables not initialized");
@@ -113,18 +139,39 @@ XMMRegister FrameMap::nr2xmmreg(int rnr) {
 void FrameMap::init() {
   if (_init_done) return;
 
-  assert(nof_cpu_regs == 8, "wrong number of CPU registers");
-  map_register(0, rsi);  rsi_opr = LIR_OprFact::single_cpu(0);  rsi_oop_opr = LIR_OprFact::single_cpu_oop(0);
-  map_register(1, rdi);  rdi_opr = LIR_OprFact::single_cpu(1);  rdi_oop_opr = LIR_OprFact::single_cpu_oop(1);
-  map_register(2, rbx);  rbx_opr = LIR_OprFact::single_cpu(2);  rbx_oop_opr = LIR_OprFact::single_cpu_oop(2);
-  map_register(3, rax);  rax_opr = LIR_OprFact::single_cpu(3);  rax_oop_opr = LIR_OprFact::single_cpu_oop(3);
-  map_register(4, rdx);  rdx_opr = LIR_OprFact::single_cpu(4);  rdx_oop_opr = LIR_OprFact::single_cpu_oop(4);
-  map_register(5, rcx);  rcx_opr = LIR_OprFact::single_cpu(5);  rcx_oop_opr = LIR_OprFact::single_cpu_oop(5);
-  map_register(6, rsp);  rsp_opr = LIR_OprFact::single_cpu(6);
-  map_register(7, rbp);  rbp_opr = LIR_OprFact::single_cpu(7);
-
-  rax_rdx_long_opr = LIR_OprFact::double_cpu(3 /*eax*/, 4 /*edx*/);
-  rbx_rcx_long_opr = LIR_OprFact::double_cpu(2 /*ebx*/, 5 /*ecx*/);
+  assert(nof_cpu_regs == LP64_ONLY(16) NOT_LP64(8), "wrong number of CPU registers");
+  map_register(0, rsi);  rsi_opr = LIR_OprFact::single_cpu(0);
+  map_register(1, rdi);  rdi_opr = LIR_OprFact::single_cpu(1);
+  map_register(2, rbx);  rbx_opr = LIR_OprFact::single_cpu(2);
+  map_register(3, rax);  rax_opr = LIR_OprFact::single_cpu(3);
+  map_register(4, rdx);  rdx_opr = LIR_OprFact::single_cpu(4);
+  map_register(5, rcx);  rcx_opr = LIR_OprFact::single_cpu(5);
+
+#ifndef _LP64
+  // The unallocatable registers are at the end
+  map_register(6, rsp);
+  map_register(7, rbp);
+#else
+  map_register( 6, r8);    r8_opr = LIR_OprFact::single_cpu(6);
+  map_register( 7, r9);    r9_opr = LIR_OprFact::single_cpu(7);
+  map_register( 8, r11);  r11_opr = LIR_OprFact::single_cpu(8);
+  map_register( 9, r12);  r12_opr = LIR_OprFact::single_cpu(9);
+  map_register(10, r13);  r13_opr = LIR_OprFact::single_cpu(10);
+  map_register(11, r14);  r14_opr = LIR_OprFact::single_cpu(11);
+  // The unallocatable registers are at the end
+  map_register(12, r10);  r10_opr = LIR_OprFact::single_cpu(12);
+  map_register(13, r15);  r15_opr = LIR_OprFact::single_cpu(13);
+  map_register(14, rsp);
+  map_register(15, rbp);
+#endif // _LP64
+
+#ifdef _LP64
+  long0_opr = LIR_OprFact::double_cpu(3 /*eax*/, 3 /*eax*/);
+  long1_opr = LIR_OprFact::double_cpu(2 /*ebx*/, 2 /*ebx*/);
+#else
+  long0_opr = LIR_OprFact::double_cpu(3 /*eax*/, 4 /*edx*/);
+  long1_opr = LIR_OprFact::double_cpu(2 /*ebx*/, 5 /*ecx*/);
+#endif // _LP64
   fpu0_float_opr   = LIR_OprFact::single_fpu(0);
   fpu0_double_opr  = LIR_OprFact::double_fpu(0);
   xmm0_float_opr   = LIR_OprFact::single_xmm(0);
@@ -137,6 +184,15 @@ void FrameMap::init() {
   _caller_save_cpu_regs[4] = rdx_opr;
   _caller_save_cpu_regs[5] = rcx_opr;
 
+#ifdef _LP64
+  _caller_save_cpu_regs[6]  = r8_opr;
+  _caller_save_cpu_regs[7]  = r9_opr;
+  _caller_save_cpu_regs[8]  = r11_opr;
+  _caller_save_cpu_regs[9]  = r12_opr;
+  _caller_save_cpu_regs[10] = r13_opr;
+  _caller_save_cpu_regs[11] = r14_opr;
+#endif // _LP64
+
 
   _xmm_regs[0] = xmm0;
   _xmm_regs[1] = xmm1;
@@ -147,18 +203,51 @@ void FrameMap::init() {
   _xmm_regs[6] = xmm6;
   _xmm_regs[7] = xmm7;
 
+#ifdef _LP64
+  _xmm_regs[8]   = xmm8;
+  _xmm_regs[9]   = xmm9;
+  _xmm_regs[10]  = xmm10;
+  _xmm_regs[11]  = xmm11;
+  _xmm_regs[12]  = xmm12;
+  _xmm_regs[13]  = xmm13;
+  _xmm_regs[14]  = xmm14;
+  _xmm_regs[15]  = xmm15;
+#endif // _LP64
+
   for (int i = 0; i < 8; i++) {
     _caller_save_fpu_regs[i] = LIR_OprFact::single_fpu(i);
+  }
+
+  for (int i = 0; i < nof_caller_save_xmm_regs ; i++) {
     _caller_save_xmm_regs[i] = LIR_OprFact::single_xmm(i);
   }
 
   _init_done = true;
 
+  rsi_oop_opr = as_oop_opr(rsi);
+  rdi_oop_opr = as_oop_opr(rdi);
+  rbx_oop_opr = as_oop_opr(rbx);
+  rax_oop_opr = as_oop_opr(rax);
+  rdx_oop_opr = as_oop_opr(rdx);
+  rcx_oop_opr = as_oop_opr(rcx);
+
+  rsp_opr = as_pointer_opr(rsp);
+  rbp_opr = as_pointer_opr(rbp);
+
+#ifdef _LP64
+  r8_oop_opr = as_oop_opr(r8);
+  r9_oop_opr = as_oop_opr(r9);
+  r11_oop_opr = as_oop_opr(r11);
+  r12_oop_opr = as_oop_opr(r12);
+  r13_oop_opr = as_oop_opr(r13);
+  r14_oop_opr = as_oop_opr(r14);
+#endif // _LP64
+
   VMRegPair regs;
   BasicType sig_bt = T_OBJECT;
   SharedRuntime::java_calling_convention(&sig_bt, &regs, 1, true);
   receiver_opr = as_oop_opr(regs.first()->as_Register());
-  assert(receiver_opr == rcx_oop_opr, "rcvr ought to be rcx");
+
 }
 
 

src/cpu/x86/vm/c1_FrameMap_x86.hpp

@@ -38,8 +38,13 @@
     nof_xmm_regs = pd_nof_xmm_regs_frame_map,
     nof_caller_save_xmm_regs = pd_nof_caller_save_xmm_regs_frame_map,
     first_available_sp_in_frame = 0,
+#ifndef _LP64
     frame_pad_in_bytes = 8,
     nof_reg_args = 2
+#else
+    frame_pad_in_bytes = 16,
+    nof_reg_args = 6
+#endif // _LP64
   };
 
  private:
@@ -65,17 +70,49 @@
   static LIR_Opr rax_oop_opr;
   static LIR_Opr rdx_oop_opr;
   static LIR_Opr rcx_oop_opr;
+#ifdef _LP64
 
-  static LIR_Opr rax_rdx_long_opr;
-  static LIR_Opr rbx_rcx_long_opr;
+  static LIR_Opr  r8_opr;
+  static LIR_Opr  r9_opr;
+  static LIR_Opr r10_opr;
+  static LIR_Opr r11_opr;
+  static LIR_Opr r12_opr;
+  static LIR_Opr r13_opr;
+  static LIR_Opr r14_opr;
+  static LIR_Opr r15_opr;
+
+  static LIR_Opr  r8_oop_opr;
+  static LIR_Opr  r9_oop_opr;
+
+  static LIR_Opr r11_oop_opr;
+  static LIR_Opr r12_oop_opr;
+  static LIR_Opr r13_oop_opr;
+  static LIR_Opr r14_oop_opr;
+
+#endif // _LP64
+
+  static LIR_Opr long0_opr;
+  static LIR_Opr long1_opr;
   static LIR_Opr fpu0_float_opr;
   static LIR_Opr fpu0_double_opr;
   static LIR_Opr xmm0_float_opr;
   static LIR_Opr xmm0_double_opr;
 
+#ifdef _LP64
+  static LIR_Opr as_long_opr(Register r) {
+    return LIR_OprFact::double_cpu(cpu_reg2rnr(r), cpu_reg2rnr(r));
+  }
+  static LIR_Opr as_pointer_opr(Register r) {
+    return LIR_OprFact::double_cpu(cpu_reg2rnr(r), cpu_reg2rnr(r));
+  }
+#else
   static LIR_Opr as_long_opr(Register r, Register r2) {
     return LIR_OprFact::double_cpu(cpu_reg2rnr(r), cpu_reg2rnr(r2));
   }
+  static LIR_Opr as_pointer_opr(Register r) {
+    return LIR_OprFact::single_cpu(cpu_reg2rnr(r));
+  }
+#endif // _LP64
 
   // VMReg name for spilled physical FPU stack slot n
   static VMReg fpu_regname (int n);

src/cpu/x86/vm/c1_LIRAssembler_x86.hpp

@@ -36,13 +36,20 @@
   address float_constant(float f);
   address double_constant(double d);
 
+  bool is_literal_address(LIR_Address* addr);
+
+  // When we need to use something other than rscratch1 use this
+  // method.
+  Address as_Address(LIR_Address* addr, Register tmp);
+
+
 public:
 
   void store_parameter(Register r, int offset_from_esp_in_words);
   void store_parameter(jint c,     int offset_from_esp_in_words);
   void store_parameter(jobject c,  int offset_from_esp_in_words);
 
-  enum { call_stub_size = 15,
+  enum { call_stub_size = NOT_LP64(15) LP64_ONLY(28),
          exception_handler_size = DEBUG_ONLY(1*K) NOT_DEBUG(175),
-         deopt_handler_size = 10
+         deopt_handler_size = NOT_LP64(10) LP64_ONLY(17)
        };

src/cpu/x86/vm/c1_LIRGenerator_x86.cpp

@@ -77,7 +77,7 @@ LIR_Opr LIRGenerator::result_register_for(ValueType* type, bool callee) {
   switch (type->tag()) {
     case intTag:     opr = FrameMap::rax_opr;          break;
     case objectTag:  opr = FrameMap::rax_oop_opr;      break;
-    case longTag:    opr = FrameMap::rax_rdx_long_opr; break;
+    case longTag:    opr = FrameMap::long0_opr;        break;
     case floatTag:   opr = UseSSE >= 1 ? FrameMap::xmm0_float_opr  : FrameMap::fpu0_float_opr;  break;
     case doubleTag:  opr = UseSSE >= 2 ? FrameMap::xmm0_double_opr : FrameMap::fpu0_double_opr;  break;
 
@@ -117,12 +117,14 @@ bool LIRGenerator::can_store_as_constant(Value v, BasicType type) const {
 
 
 bool LIRGenerator::can_inline_as_constant(Value v) const {
+  if (v->type()->tag() == longTag) return false;
   return v->type()->tag() != objectTag ||
     (v->type()->is_constant() && v->type()->as_ObjectType()->constant_value()->is_null_object());
 }
 
 
 bool LIRGenerator::can_inline_as_constant(LIR_Const* c) const {
+  if (c->type() == T_LONG) return false;
   return c->type() != T_OBJECT || c->as_jobject() == NULL;
 }
 
@@ -155,6 +157,13 @@ LIR_Address* LIRGenerator::emit_array_address(LIR_Opr array_opr, LIR_Opr index_o
     addr = new LIR_Address(array_opr,
                            offset_in_bytes + index_opr->as_jint() * elem_size, type);
   } else {
+#ifdef _LP64
+    if (index_opr->type() == T_INT) {
+      LIR_Opr tmp = new_register(T_LONG);
+      __ convert(Bytecodes::_i2l, index_opr, tmp);
+      index_opr = tmp;
+    }
+#endif // _LP64
     addr =  new LIR_Address(array_opr,
                             index_opr,
                             LIR_Address::scale(type),
@@ -164,7 +173,7 @@ LIR_Address* LIRGenerator::emit_array_address(LIR_Opr array_opr, LIR_Opr index_o
     // This store will need a precise card mark, so go ahead and
     // compute the full adddres instead of computing once for the
     // store and again for the card mark.
-    LIR_Opr tmp = new_register(T_INT);
+    LIR_Opr tmp = new_pointer_register();
     __ leal(LIR_OprFact::address(addr), tmp);
     return new LIR_Address(tmp, 0, type);
   } else {
@@ -174,9 +183,8 @@ LIR_Address* LIRGenerator::emit_array_address(LIR_Opr array_opr, LIR_Opr index_o
 
 
 void LIRGenerator::increment_counter(address counter, int step) {
-  LIR_Opr temp = new_register(T_INT);
-  LIR_Opr pointer = new_register(T_INT);
-  __ move(LIR_OprFact::intConst((int)counter), pointer);
+  LIR_Opr pointer = new_pointer_register();
+  __ move(LIR_OprFact::intptrConst(counter), pointer);
   LIR_Address* addr = new LIR_Address(pointer, 0, T_INT);
   increment_counter(addr, step);
 }
@@ -481,7 +489,7 @@ void LIRGenerator::do_ArithmeticOp_Long(ArithmeticOp* x) {
     left.load_item();
     right.load_item();
 
-    LIR_Opr reg = FrameMap::rax_rdx_long_opr;
+    LIR_Opr reg = FrameMap::long0_opr;
     arithmetic_op_long(x->op(), reg, left.result(), right.result(), NULL);
     LIR_Opr result = rlock_result(x);
     __ move(reg, result);
@@ -690,10 +698,10 @@ void LIRGenerator::do_AttemptUpdate(Intrinsic* x) {
   LIRItem new_value (x->argument_at(2), this);  // replace field with new_value if it matches cmp_value
 
   // compare value must be in rdx,eax (hi,lo); may be destroyed by cmpxchg8 instruction
-  cmp_value.load_item_force(FrameMap::rax_rdx_long_opr);
+  cmp_value.load_item_force(FrameMap::long0_opr);
 
   // new value must be in rcx,ebx (hi,lo)
-  new_value.load_item_force(FrameMap::rbx_rcx_long_opr);
+  new_value.load_item_force(FrameMap::long1_opr);
 
   // object pointer register is overwritten with field address
   obj.load_item();
@@ -720,7 +728,10 @@ void LIRGenerator::do_CompareAndSwap(Intrinsic* x, ValueType* type) {
   LIRItem val   (x->argument_at(3), this);  // replace field with val if matches cmp
 
   assert(obj.type()->tag() == objectTag, "invalid type");
-  assert(offset.type()->tag() == intTag, "invalid type");
+
+  // In 64bit the type can be long, sparc doesn't have this assert
+  // assert(offset.type()->tag() == intTag, "invalid type");
+
   assert(cmp.type()->tag() == type->tag(), "invalid type");
   assert(val.type()->tag() == type->tag(), "invalid type");
 
@@ -735,8 +746,8 @@ void LIRGenerator::do_CompareAndSwap(Intrinsic* x, ValueType* type) {
     cmp.load_item_force(FrameMap::rax_opr);
     val.load_item();
   } else if (type == longType) {
-    cmp.load_item_force(FrameMap::rax_rdx_long_opr);
-    val.load_item_force(FrameMap::rbx_rcx_long_opr);
+    cmp.load_item_force(FrameMap::long0_opr);
+    val.load_item_force(FrameMap::long1_opr);
   } else {
     ShouldNotReachHere();
   }
@@ -833,12 +844,33 @@ void LIRGenerator::do_ArrayCopy(Intrinsic* x) {
   // operands for arraycopy must use fixed registers, otherwise
   // LinearScan will fail allocation (because arraycopy always needs a
   // call)
+
+#ifndef _LP64
   src.load_item_force     (FrameMap::rcx_oop_opr);
   src_pos.load_item_force (FrameMap::rdx_opr);
   dst.load_item_force     (FrameMap::rax_oop_opr);
   dst_pos.load_item_force (FrameMap::rbx_opr);
   length.load_item_force  (FrameMap::rdi_opr);
   LIR_Opr tmp =           (FrameMap::rsi_opr);
+#else
+
+  // The java calling convention will give us enough registers
+  // so that on the stub side the args will be perfect already.
+  // On the other slow/special case side we call C and the arg
+  // positions are not similar enough to pick one as the best.
+  // Also because the java calling convention is a "shifted" version
+  // of the C convention we can process the java args trivially into C
+  // args without worry of overwriting during the xfer
+
+  src.load_item_force     (FrameMap::as_oop_opr(j_rarg0));
+  src_pos.load_item_force (FrameMap::as_opr(j_rarg1));
+  dst.load_item_force     (FrameMap::as_oop_opr(j_rarg2));
+  dst_pos.load_item_force (FrameMap::as_opr(j_rarg3));
+  length.load_item_force  (FrameMap::as_opr(j_rarg4));
+
+  LIR_Opr tmp =           FrameMap::as_opr(j_rarg5);
+#endif // LP64
+
   set_no_result(x);
 
   int flags;
@@ -857,7 +889,7 @@ LIR_Opr fixed_register_for(BasicType type) {
     case T_FLOAT:  return FrameMap::fpu0_float_opr;
     case T_DOUBLE: return FrameMap::fpu0_double_opr;
     case T_INT:    return FrameMap::rax_opr;
-    case T_LONG:   return FrameMap::rax_rdx_long_opr;
+    case T_LONG:   return FrameMap::long0_opr;
     default:       ShouldNotReachHere(); return LIR_OprFact::illegalOpr;
   }
 }
@@ -1161,9 +1193,13 @@ void LIRGenerator::do_If(If* x) {
 
 
 LIR_Opr LIRGenerator::getThreadPointer() {
+#ifdef _LP64
+  return FrameMap::as_pointer_opr(r15_thread);
+#else
   LIR_Opr result = new_register(T_INT);
   __ get_thread(result);
   return result;
+#endif //
 }
 
 void LIRGenerator::trace_block_entry(BlockBegin* block) {

src/cpu/x86/vm/c1_LinearScan_x86.hpp

@@ -23,18 +23,29 @@
  */
 
 inline bool LinearScan::is_processed_reg_num(int reg_num) {
+#ifndef _LP64
   // rsp and rbp (numbers 6 ancd 7) are ignored
   assert(FrameMap::rsp_opr->cpu_regnr() == 6, "wrong assumption below");
   assert(FrameMap::rbp_opr->cpu_regnr() == 7, "wrong assumption below");
   assert(reg_num >= 0, "invalid reg_num");
 
   return reg_num < 6 || reg_num > 7;
+#else
+  // rsp and rbp, r10, r15 (numbers 6 ancd 7) are ignored
+  assert(FrameMap::r10_opr->cpu_regnr() == 12, "wrong assumption below");
+  assert(FrameMap::r15_opr->cpu_regnr() == 13, "wrong assumption below");
+  assert(FrameMap::rsp_opr->cpu_regnrLo() == 14, "wrong assumption below");
+  assert(FrameMap::rbp_opr->cpu_regnrLo() == 15, "wrong assumption below");
+  assert(reg_num >= 0, "invalid reg_num");
+
+  return reg_num < 12 || reg_num > 15;
+#endif // _LP64
 }
 
 inline int LinearScan::num_physical_regs(BasicType type) {
   // Intel requires two cpu registers for long,
   // but requires only one fpu register for double
-  if (type == T_LONG) {
+  if (LP64_ONLY(false &&) type == T_LONG) {
     return 2;
   }
   return 1;

src/cpu/x86/vm/c1_MacroAssembler_x86.cpp

@@ -26,18 +26,17 @@
 #include "incls/_c1_MacroAssembler_x86.cpp.incl"
 
 int C1_MacroAssembler::lock_object(Register hdr, Register obj, Register disp_hdr, Register scratch, Label& slow_case) {
-  const int aligned_mask = 3;
+  const int aligned_mask = BytesPerWord -1;
   const int hdr_offset = oopDesc::mark_offset_in_bytes();
   assert(hdr == rax, "hdr must be rax, for the cmpxchg instruction");
   assert(hdr != obj && hdr != disp_hdr && obj != disp_hdr, "registers must be different");
-  assert(BytesPerWord == 4, "adjust aligned_mask and code");
   Label done;
   int null_check_offset = -1;
 
   verify_oop(obj);
 
   // save object being locked into the BasicObjectLock
-  movl(Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()), obj);
+  movptr(Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()), obj);
 
   if (UseBiasedLocking) {
     assert(scratch != noreg, "should have scratch register at this point");
@@ -47,16 +46,16 @@ int C1_MacroAssembler::lock_object(Register hdr, Register obj, Register disp_hdr
   }
 
   // Load object header
-  movl(hdr, Address(obj, hdr_offset));
+  movptr(hdr, Address(obj, hdr_offset));
   // and mark it as unlocked
-  orl(hdr, markOopDesc::unlocked_value);
+  orptr(hdr, markOopDesc::unlocked_value);
   // save unlocked object header into the displaced header location on the stack
-  movl(Address(disp_hdr, 0), hdr);
+  movptr(Address(disp_hdr, 0), hdr);
   // test if object header is still the same (i.e. unlocked), and if so, store the
   // displaced header address in the object header - if it is not the same, get the
   // object header instead
   if (os::is_MP()) MacroAssembler::lock(); // must be immediately before cmpxchg!
-  cmpxchg(disp_hdr, Address(obj, hdr_offset));
+  cmpxchgptr(disp_hdr, Address(obj, hdr_offset));
   // if the object header was the same, we're done
   if (PrintBiasedLockingStatistics) {
     cond_inc32(Assembler::equal,
@@ -76,11 +75,11 @@ int C1_MacroAssembler::lock_object(Register hdr, Register obj, Register disp_hdr
   //
   // assuming both the stack pointer and page_size have their least
   // significant 2 bits cleared and page_size is a power of 2
-  subl(hdr, rsp);
-  andl(hdr, aligned_mask - os::vm_page_size());
+  subptr(hdr, rsp);
+  andptr(hdr, aligned_mask - os::vm_page_size());
   // for recursive locking, the result is zero => save it in the displaced header
   // location (NULL in the displaced hdr location indicates recursive locking)
-  movl(Address(disp_hdr, 0), hdr);
+  movptr(Address(disp_hdr, 0), hdr);
   // otherwise we don't care about the result and handle locking via runtime call
   jcc(Assembler::notZero, slow_case);
   // done
@@ -90,35 +89,34 @@ int C1_MacroAssembler::lock_object(Register hdr, Register obj, Register disp_hdr
 
 
 void C1_MacroAssembler::unlock_object(Register hdr, Register obj, Register disp_hdr, Label& slow_case) {
-  const int aligned_mask = 3;
+  const int aligned_mask = BytesPerWord -1;
   const int hdr_offset = oopDesc::mark_offset_in_bytes();
   assert(disp_hdr == rax, "disp_hdr must be rax, for the cmpxchg instruction");
   assert(hdr != obj && hdr != disp_hdr && obj != disp_hdr, "registers must be different");
-  assert(BytesPerWord == 4, "adjust aligned_mask and code");
   Label done;
 
   if (UseBiasedLocking) {
     // load object
-    movl(obj, Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()));
+    movptr(obj, Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()));
     biased_locking_exit(obj, hdr, done);
   }
 
   // load displaced header
-  movl(hdr, Address(disp_hdr, 0));
+  movptr(hdr, Address(disp_hdr, 0));
   // if the loaded hdr is NULL we had recursive locking
-  testl(hdr, hdr);
+  testptr(hdr, hdr);
   // if we had recursive locking, we are done
   jcc(Assembler::zero, done);
   if (!UseBiasedLocking) {
     // load object
-    movl(obj, Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()));
+    movptr(obj, Address(disp_hdr, BasicObjectLock::obj_offset_in_bytes()));
   }
   verify_oop(obj);
   // test if object header is pointing to the displaced header, and if so, restore
   // the displaced header in the object - if the object header is not pointing to
   // the displaced header, get the object header instead
   if (os::is_MP()) MacroAssembler::lock(); // must be immediately before cmpxchg!
-  cmpxchg(hdr, Address(obj, hdr_offset));
+  cmpxchgptr(hdr, Address(obj, hdr_offset));
   // if the object header was not pointing to the displaced header,
   // we do unlocking via runtime call
   jcc(Assembler::notEqual, slow_case);
@@ -141,13 +139,14 @@ void C1_MacroAssembler::initialize_header(Register obj, Register klass, Register
   assert_different_registers(obj, klass, len);
   if (UseBiasedLocking && !len->is_valid()) {
     assert_different_registers(obj, klass, len, t1, t2);
-    movl(t1, Address(klass, Klass::prototype_header_offset_in_bytes() + klassOopDesc::klass_part_offset_in_bytes()));
-    movl(Address(obj, oopDesc::mark_offset_in_bytes()), t1);
+    movptr(t1, Address(klass, Klass::prototype_header_offset_in_bytes() + klassOopDesc::klass_part_offset_in_bytes()));
+    movptr(Address(obj, oopDesc::mark_offset_in_bytes()), t1);
   } else {
-    movl(Address(obj, oopDesc::mark_offset_in_bytes ()), (int)markOopDesc::prototype());
+    // This assumes that all prototype bits fit in an int32_t
+    movptr(Address(obj, oopDesc::mark_offset_in_bytes ()), (int32_t)(intptr_t)markOopDesc::prototype());
   }
 
-  movl(Address(obj, oopDesc::klass_offset_in_bytes()), klass);
+  movptr(Address(obj, oopDesc::klass_offset_in_bytes()), klass);
   if (len->is_valid()) {
     movl(Address(obj, arrayOopDesc::length_offset_in_bytes()), len);
   }
@@ -160,25 +159,27 @@ void C1_MacroAssembler::initialize_body(Register obj, Register len_in_bytes, int
   assert(obj != len_in_bytes && obj != t1 && t1 != len_in_bytes, "registers must be different");
   assert((hdr_size_in_bytes & (BytesPerWord - 1)) == 0, "header size is not a multiple of BytesPerWord");
   Register index = len_in_bytes;
-  subl(index, hdr_size_in_bytes);
+  // index is positive and ptr sized
+  subptr(index, hdr_size_in_bytes);
   jcc(Assembler::zero, done);
   // initialize topmost word, divide index by 2, check if odd and test if zero
   // note: for the remaining code to work, index must be a multiple of BytesPerWord
 #ifdef ASSERT
   { Label L;
-    testl(index, BytesPerWord - 1);
+    testptr(index, BytesPerWord - 1);
     jcc(Assembler::zero, L);
     stop("index is not a multiple of BytesPerWord");
     bind(L);
   }
 #endif
-  xorl(t1, t1);      // use _zero reg to clear memory (shorter code)
+  xorptr(t1, t1);    // use _zero reg to clear memory (shorter code)
   if (UseIncDec) {
-    shrl(index, 3);  // divide by 8 and set carry flag if bit 2 was set
+    shrptr(index, 3);  // divide by 8/16 and set carry flag if bit 2 was set
   } else {
-    shrl(index, 2);  // use 2 instructions to avoid partial flag stall
-    shrl(index, 1);
+    shrptr(index, 2);  // use 2 instructions to avoid partial flag stall
+    shrptr(index, 1);
   }
+#ifndef _LP64
   // index could have been not a multiple of 8 (i.e., bit 2 was set)
   { Label even;
     // note: if index was a multiple of 8, than it cannot
@@ -186,16 +187,17 @@ void C1_MacroAssembler::initialize_body(Register obj, Register len_in_bytes, int
     //       => if it is even, we don't need to check for 0 again
     jcc(Assembler::carryClear, even);
     // clear topmost word (no jump needed if conditional assignment would work here)
-    movl(Address(obj, index, Address::times_8, hdr_size_in_bytes - 0*BytesPerWord), t1);
+    movptr(Address(obj, index, Address::times_8, hdr_size_in_bytes - 0*BytesPerWord), t1);
     // index could be 0 now, need to check again
     jcc(Assembler::zero, done);
     bind(even);
   }
+#endif // !_LP64
   // initialize remaining object fields: rdx is a multiple of 2 now
   { Label loop;
     bind(loop);
-    movl(Address(obj, index, Address::times_8, hdr_size_in_bytes - 1*BytesPerWord), t1);
-    movl(Address(obj, index, Address::times_8, hdr_size_in_bytes - 2*BytesPerWord), t1);
+    movptr(Address(obj, index, Address::times_8, hdr_size_in_bytes - 1*BytesPerWord), t1);
+    NOT_LP64(movptr(Address(obj, index, Address::times_8, hdr_size_in_bytes - 2*BytesPerWord), t1);)
     decrement(index);
     jcc(Assembler::notZero, loop);
   }
@@ -227,30 +229,30 @@ void C1_MacroAssembler::initialize_object(Register obj, Register klass, Register
   const Register index = t2;
   const int threshold = 6 * BytesPerWord;   // approximate break even point for code size (see comments below)
   if (var_size_in_bytes != noreg) {
-    movl(index, var_size_in_bytes);
+    mov(index, var_size_in_bytes);
     initialize_body(obj, index, hdr_size_in_bytes, t1_zero);
   } else if (con_size_in_bytes <= threshold) {
     // use explicit null stores
     // code size = 2 + 3*n bytes (n = number of fields to clear)
-    xorl(t1_zero, t1_zero); // use t1_zero reg to clear memory (shorter code)
+    xorptr(t1_zero, t1_zero); // use t1_zero reg to clear memory (shorter code)
     for (int i = hdr_size_in_bytes; i < con_size_in_bytes; i += BytesPerWord)
-      movl(Address(obj, i), t1_zero);
+      movptr(Address(obj, i), t1_zero);
   } else if (con_size_in_bytes > hdr_size_in_bytes) {
     // use loop to null out the fields
     // code size = 16 bytes for even n (n = number of fields to clear)
     // initialize last object field first if odd number of fields
-    xorl(t1_zero, t1_zero); // use t1_zero reg to clear memory (shorter code)
-    movl(index, (con_size_in_bytes - hdr_size_in_bytes) >> 3);
+    xorptr(t1_zero, t1_zero); // use t1_zero reg to clear memory (shorter code)
+    movptr(index, (con_size_in_bytes - hdr_size_in_bytes) >> 3);
     // initialize last object field if constant size is odd
     if (((con_size_in_bytes - hdr_size_in_bytes) & 4) != 0)
-      movl(Address(obj, con_size_in_bytes - (1*BytesPerWord)), t1_zero);
+      movptr(Address(obj, con_size_in_bytes - (1*BytesPerWord)), t1_zero);
     // initialize remaining object fields: rdx is a multiple of 2
     { Label loop;
       bind(loop);
-      movl(Address(obj, index, Address::times_8,
-        hdr_size_in_bytes - (1*BytesPerWord)), t1_zero);
-      movl(Address(obj, index, Address::times_8,
-        hdr_size_in_bytes - (2*BytesPerWord)), t1_zero);
+      movptr(Address(obj, index, Address::times_8, hdr_size_in_bytes - (1*BytesPerWord)),
+             t1_zero);
+      NOT_LP64(movptr(Address(obj, index, Address::times_8, hdr_size_in_bytes - (2*BytesPerWord)),
+             t1_zero);)
       decrement(index);
       jcc(Assembler::notZero, loop);
     }
@@ -269,17 +271,17 @@ void C1_MacroAssembler::allocate_array(Register obj, Register len, Register t1,
   assert_different_registers(obj, len, t1, t2, klass);
 
   // determine alignment mask
-  assert(BytesPerWord == 4, "must be a multiple of 2 for masking code to work");
+  assert(!(BytesPerWord & 1), "must be a multiple of 2 for masking code to work");
 
   // check for negative or excessive length
-  cmpl(len, max_array_allocation_length);
+  cmpptr(len, (int32_t)max_array_allocation_length);
   jcc(Assembler::above, slow_case);
 
   const Register arr_size = t2; // okay to be the same
   // align object end
-  movl(arr_size, header_size * BytesPerWord + MinObjAlignmentInBytesMask);
-  leal(arr_size, Address(arr_size, len, f));
-  andl(arr_size, ~MinObjAlignmentInBytesMask);
+  movptr(arr_size, (int32_t)header_size * BytesPerWord + MinObjAlignmentInBytesMask);
+  lea(arr_size, Address(arr_size, len, f));
+  andptr(arr_size, ~MinObjAlignmentInBytesMask);
 
   try_allocate(obj, arr_size, 0, t1, t2, slow_case);
 
@@ -305,12 +307,13 @@ void C1_MacroAssembler::inline_cache_check(Register receiver, Register iCache) {
   // check against inline cache
   assert(!MacroAssembler::needs_explicit_null_check(oopDesc::klass_offset_in_bytes()), "must add explicit null check");
   int start_offset = offset();
-  cmpl(iCache, Address(receiver, oopDesc::klass_offset_in_bytes()));
+  cmpptr(iCache, Address(receiver, oopDesc::klass_offset_in_bytes()));
   // if icache check fails, then jump to runtime routine
   // Note: RECEIVER must still contain the receiver!
   jump_cc(Assembler::notEqual,
           RuntimeAddress(SharedRuntime::get_ic_miss_stub()));
-  assert(offset() - start_offset == 9, "check alignment in emit_method_entry");
+  const int ic_cmp_size = LP64_ONLY(10) NOT_LP64(9);
+  assert(offset() - start_offset == ic_cmp_size, "check alignment in emit_method_entry");
 }
 
 
@@ -364,7 +367,7 @@ void C1_MacroAssembler::verify_stack_oop(int stack_offset) {
 void C1_MacroAssembler::verify_not_null_oop(Register r) {
   if (!VerifyOops) return;
   Label not_null;
-  testl(r, r);
+  testptr(r, r);
   jcc(Assembler::notZero, not_null);
   stop("non-null oop required");
   bind(not_null);
@@ -373,12 +376,12 @@ void C1_MacroAssembler::verify_not_null_oop(Register r) {
 
 void C1_MacroAssembler::invalidate_registers(bool inv_rax, bool inv_rbx, bool inv_rcx, bool inv_rdx, bool inv_rsi, bool inv_rdi) {
 #ifdef ASSERT
-  if (inv_rax) movl(rax, 0xDEAD);
-  if (inv_rbx) movl(rbx, 0xDEAD);
-  if (inv_rcx) movl(rcx, 0xDEAD);
-  if (inv_rdx) movl(rdx, 0xDEAD);
-  if (inv_rsi) movl(rsi, 0xDEAD);
-  if (inv_rdi) movl(rdi, 0xDEAD);
+  if (inv_rax) movptr(rax, 0xDEAD);
+  if (inv_rbx) movptr(rbx, 0xDEAD);
+  if (inv_rcx) movptr(rcx, 0xDEAD);
+  if (inv_rdx) movptr(rdx, 0xDEAD);
+  if (inv_rsi) movptr(rsi, 0xDEAD);
+  if (inv_rdi) movptr(rdi, 0xDEAD);
 #endif
 }
 

src/cpu/x86/vm/c1_MacroAssembler_x86.hpp

@@ -94,16 +94,17 @@
   // Note: NEVER push values directly, but only through following push_xxx functions;
   //       This helps us to track the rsp changes compared to the entry rsp (->_rsp_offset)
 
-  void push_jint (jint i)     { _rsp_offset++; pushl(i); }
+  void push_jint (jint i)     { _rsp_offset++; push(i); }
   void push_oop  (jobject o)  { _rsp_offset++; pushoop(o); }
-  void push_addr (Address a)  { _rsp_offset++; pushl(a); }
-  void push_reg  (Register r) { _rsp_offset++; pushl(r); }
-  void pop       (Register r) { _rsp_offset--; popl (r); assert(_rsp_offset >= 0, "stack offset underflow"); }
+  // Seems to always be in wordSize
+  void push_addr (Address a)  { _rsp_offset++; pushptr(a); }
+  void push_reg  (Register r) { _rsp_offset++; push(r); }
+  void pop_reg   (Register r) { _rsp_offset--; pop(r); assert(_rsp_offset >= 0, "stack offset underflow"); }
 
   void dec_stack (int nof_words) {
     _rsp_offset -= nof_words;
     assert(_rsp_offset >= 0, "stack offset underflow");
-    addl(rsp, wordSize * nof_words);
+    addptr(rsp, wordSize * nof_words);
   }
 
   void dec_stack_after_call (int nof_words) {

src/cpu/x86/vm/dump_x86_32.cpp

@@ -98,24 +98,24 @@ void CompactingPermGenGen::generate_vtable_methods(void** vtbl_list,
   // table.
 
 #ifdef WIN32
-  __ pushl(rcx);                        // save "this"
+  __ push(rcx);                         // save "this"
 #endif
-  __ movl(rcx, rax);
-  __ shrl(rcx, 8);                      // isolate vtable identifier.
-  __ shll(rcx, LogBytesPerWord);
+  __ mov(rcx, rax);
+  __ shrptr(rcx, 8);                    // isolate vtable identifier.
+  __ shlptr(rcx, LogBytesPerWord);
   Address index(noreg, rcx,  Address::times_1);
   ExternalAddress vtbl((address)vtbl_list);
   __ movptr(rdx, ArrayAddress(vtbl, index)); // get correct vtable address.
 #ifdef WIN32
-  __ popl(rcx);                         // restore "this"
+  __ pop(rcx);                          // restore "this"
 #else
-  __ movl(rcx, Address(rsp, 4));        // fetch "this"
+  __ movptr(rcx, Address(rsp, BytesPerWord));   // fetch "this"
 #endif
-  __ movl(Address(rcx, 0), rdx);        // update vtable pointer.
+  __ movptr(Address(rcx, 0), rdx);      // update vtable pointer.
 
-  __ andl(rax, 0x00ff);                 // isolate vtable method index
-  __ shll(rax, LogBytesPerWord);
-  __ addl(rax, rdx);                    // address of real method pointer.
+  __ andptr(rax, 0x00ff);                       // isolate vtable method index
+  __ shlptr(rax, LogBytesPerWord);
+  __ addptr(rax, rdx);                  // address of real method pointer.
   __ jmp(Address(rax, 0));              // get real method pointer.
 
   __ flush();

src/cpu/x86/vm/frame_x86.cpp

@@ -217,7 +217,8 @@ bool frame::safe_for_sender(JavaThread *thread) {
 
 void frame::patch_pc(Thread* thread, address pc) {
   if (TracePcPatching) {
-    tty->print_cr("patch_pc at address  0x%x [0x%x -> 0x%x] ", &((address *)sp())[-1], ((address *)sp())[-1], pc);
+    tty->print_cr("patch_pc at address" INTPTR_FORMAT " [" INTPTR_FORMAT " -> " INTPTR_FORMAT "] ",
+                  &((address *)sp())[-1], ((address *)sp())[-1], pc);
   }
   ((address *)sp())[-1] = pc;
   _cb = CodeCache::find_blob(pc);

src/cpu/x86/vm/frame_x86.inline.hpp

@@ -159,7 +159,7 @@ inline intptr_t** frame::interpreter_frame_locals_addr() const {
 
 inline intptr_t* frame::interpreter_frame_bcx_addr() const {
   assert(is_interpreted_frame(), "must be interpreted");
-  return (jint*) &(get_interpreterState()->_bcp);
+  return (intptr_t*) &(get_interpreterState()->_bcp);
 }
 
 
@@ -179,7 +179,7 @@ inline methodOop* frame::interpreter_frame_method_addr() const {
 
 inline intptr_t* frame::interpreter_frame_mdx_addr() const {
   assert(is_interpreted_frame(), "must be interpreted");
-  return (jint*) &(get_interpreterState()->_mdx);
+  return (intptr_t*) &(get_interpreterState()->_mdx);
 }
 
 // top of expression stack