Merge

hotspot/src/cpu/sparc/vm/c1_CodeStubs_sparc.cpp

@@ -298,7 +298,7 @@ void PatchingStub::emit_code(LIR_Assembler* ce) {
     for (int i = 0; i < _bytes_to_copy; i++) {
       address ptr = (address)(_pc_start + i);
       int a_byte = (*ptr) & 0xFF;
-      __ a_byte (a_byte);
+      __ emit_int8 (a_byte);
     }
   }
 
@@ -340,10 +340,10 @@ void PatchingStub::emit_code(LIR_Assembler* ce) {
   int being_initialized_entry_offset = __ offset() - being_initialized_entry + sizeof_patch_record;
 
   // Emit the patch record.  We need to emit a full word, so emit an extra empty byte
-  __ a_byte(0);
-  __ a_byte(being_initialized_entry_offset);
-  __ a_byte(bytes_to_skip);
-  __ a_byte(_bytes_to_copy);
+  __ emit_int8(0);
+  __ emit_int8(being_initialized_entry_offset);
+  __ emit_int8(bytes_to_skip);
+  __ emit_int8(_bytes_to_copy);
   address patch_info_pc = __ pc();
   assert(patch_info_pc - end_of_patch == bytes_to_skip, "incorrect patch info");
 

hotspot/src/cpu/sparc/vm/macroAssembler_sparc.cpp

@@ -100,34 +100,6 @@ const char* Argument::name() const {
 bool AbstractAssembler::pd_check_instruction_mark() { return false; }
 #endif
 
-
-void MacroAssembler::print_instruction(int inst) {
-  const char* s;
-  switch (inv_op(inst)) {
-  default:         s = "????"; break;
-  case call_op:    s = "call"; break;
-  case branch_op:
-    switch (inv_op2(inst)) {
-      case fb_op2:     s = "fb";   break;
-      case fbp_op2:    s = "fbp";  break;
-      case br_op2:     s = "br";   break;
-      case bp_op2:     s = "bp";   break;
-      case cb_op2:     s = "cb";   break;
-      case bpr_op2: {
-        if (is_cbcond(inst)) {
-          s = is_cxb(inst) ? "cxb" : "cwb";
-        } else {
-          s = "bpr";
-        }
-        break;
-      }
-      default:         s = "????"; break;
-    }
-  }
-  ::tty->print("%s", s);
-}
-
-
 // Patch instruction inst at offset inst_pos to refer to dest_pos
 // and return the resulting instruction.
 // We should have pcs, not offsets, but since all is relative, it will work out

hotspot/src/cpu/sparc/vm/macroAssembler_sparc.hpp

@@ -603,7 +603,6 @@ class MacroAssembler : public Assembler {
   friend class Label;
 
  protected:
-  static void print_instruction(int inst);
   static int  patched_branch(int dest_pos, int inst, int inst_pos);
   static int  branch_destination(int inst, int pos);
 
@@ -759,9 +758,6 @@ class MacroAssembler : public Assembler {
   // Required platform-specific helpers for Label::patch_instructions.
   // They _shadow_ the declarations in AbstractAssembler, which are undefined.
   void pd_patch_instruction(address branch, address target);
-#ifndef PRODUCT
-  static void pd_print_patched_instruction(address branch);
-#endif
 
   // sethi Macro handles optimizations and relocations
 private:

hotspot/src/cpu/sparc/vm/macroAssembler_sparc.inline.hpp

@@ -43,14 +43,6 @@ inline void MacroAssembler::pd_patch_instruction(address branch, address target)
   stub_inst = patched_branch(target - branch, stub_inst, 0);
 }
 
-#ifndef PRODUCT
-inline void MacroAssembler::pd_print_patched_instruction(address branch) {
-  jint stub_inst = *(jint*) branch;
-  print_instruction(stub_inst);
-  ::tty->print("%s", " (unresolved)");
-}
-#endif // PRODUCT
-
 // Use the right loads/stores for the platform
 inline void MacroAssembler::ld_ptr( Register s1, Register s2, Register d ) {
 #ifdef _LP64

hotspot/src/cpu/sparc/vm/sparc.ad

@@ -10224,7 +10224,7 @@ instruct array_equals(o0RegP ary1, o1RegP ary2, g3RegI tmp1, notemp_iRegI result
 
 //---------- Zeros Count Instructions ------------------------------------------
 
-instruct countLeadingZerosI(iRegI dst, iRegI src, iRegI tmp, flagsReg cr) %{
+instruct countLeadingZerosI(iRegIsafe dst, iRegI src, iRegI tmp, flagsReg cr) %{
   predicate(UsePopCountInstruction);  // See Matcher::match_rule_supported
   match(Set dst (CountLeadingZerosI src));
   effect(TEMP dst, TEMP tmp, KILL cr);
@@ -10321,7 +10321,7 @@ instruct countLeadingZerosL(iRegIsafe dst, iRegL src, iRegL tmp, flagsReg cr) %{
   ins_pipe(ialu_reg);
 %}
 
-instruct countTrailingZerosI(iRegI dst, iRegI src, flagsReg cr) %{
+instruct countTrailingZerosI(iRegIsafe dst, iRegI src, flagsReg cr) %{
   predicate(UsePopCountInstruction);  // See Matcher::match_rule_supported
   match(Set dst (CountTrailingZerosI src));
   effect(TEMP dst, KILL cr);
@@ -10364,19 +10364,21 @@ instruct countTrailingZerosL(iRegIsafe dst, iRegL src, flagsReg cr) %{
 
 //---------- Population Count Instructions -------------------------------------
 
-instruct popCountI(iRegI dst, iRegI src) %{
+instruct popCountI(iRegIsafe dst, iRegI src) %{
   predicate(UsePopCountInstruction);
   match(Set dst (PopCountI src));
 
-  format %{ "POPC   $src, $dst" %}
+  format %{ "SRL    $src, G0, $dst\t! clear upper word for 64 bit POPC\n\t"
+            "POPC   $dst, $dst" %}
   ins_encode %{
-    __ popc($src$$Register, $dst$$Register);
+    __ srl($src$$Register, G0, $dst$$Register);
+    __ popc($dst$$Register, $dst$$Register);
   %}
   ins_pipe(ialu_reg);
 %}
 
 // Note: Long.bitCount(long) returns an int.
-instruct popCountL(iRegI dst, iRegL src) %{
+instruct popCountL(iRegIsafe dst, iRegL src) %{
   predicate(UsePopCountInstruction);
   match(Set dst (PopCountL src));
 

hotspot/src/cpu/sparc/vm/templateInterpreter_sparc.cpp

@@ -434,7 +434,7 @@ void TemplateInterpreterGenerator::generate_stack_overflow_check(Register Rframe
 
   // the frame is greater than one page in size, so check against
   // the bottom of the stack
-  __ cmp_and_brx_short(SP, Rscratch, Assembler::greater, Assembler::pt, after_frame_check);
+  __ cmp_and_brx_short(SP, Rscratch, Assembler::greaterUnsigned, Assembler::pt, after_frame_check);
 
   // the stack will overflow, throw an exception
 

hotspot/src/cpu/x86/vm/c1_CodeStubs_x86.cpp

@@ -313,10 +313,10 @@ void PatchingStub::emit_code(LIR_Assembler* ce) {
 #endif
   } else {
     // make a copy the code which is going to be patched.
-    for ( int i = 0; i < _bytes_to_copy; i++) {
+    for (int i = 0; i < _bytes_to_copy; i++) {
       address ptr = (address)(_pc_start + i);
       int a_byte = (*ptr) & 0xFF;
-      __ a_byte (a_byte);
+      __ emit_int8(a_byte);
       *ptr = 0x90; // make the site look like a nop
     }
   }
@@ -363,11 +363,11 @@ void PatchingStub::emit_code(LIR_Assembler* ce) {
   // emit the offsets needed to find the code to patch
   int being_initialized_entry_offset = __ pc() - being_initialized_entry + sizeof_patch_record;
 
-  __ a_byte(0xB8);
-  __ a_byte(0);
-  __ a_byte(being_initialized_entry_offset);
-  __ a_byte(bytes_to_skip);
-  __ a_byte(_bytes_to_copy);
+  __ emit_int8((unsigned char)0xB8);
+  __ emit_int8(0);
+  __ emit_int8(being_initialized_entry_offset);
+  __ emit_int8(bytes_to_skip);
+  __ emit_int8(_bytes_to_copy);
   address patch_info_pc = __ pc();
   assert(patch_info_pc - end_of_patch == bytes_to_skip, "incorrect patch info");
 

hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp

@@ -1023,7 +1023,7 @@ void MacroAssembler::lea(Address dst, AddressLiteral adr) {
 
 void MacroAssembler::leave() {
   // %%% is this really better? Why not on 32bit too?
-  emit_byte(0xC9); // LEAVE
+  emit_int8((unsigned char)0xC9); // LEAVE
 }
 
 void MacroAssembler::lneg(Register hi, Register lo) {
@@ -2112,11 +2112,11 @@ void MacroAssembler::fat_nop() {
   if (UseAddressNop) {
     addr_nop_5();
   } else {
-    emit_byte(0x26); // es:
-    emit_byte(0x2e); // cs:
-    emit_byte(0x64); // fs:
-    emit_byte(0x65); // gs:
-    emit_byte(0x90);
+    emit_int8(0x26); // es:
+    emit_int8(0x2e); // cs:
+    emit_int8(0x64); // fs:
+    emit_int8(0x65); // gs:
+    emit_int8((unsigned char)0x90);
   }
 }
 
@@ -2534,12 +2534,12 @@ void MacroAssembler::jump_cc(Condition cc, AddressLiteral dst) {
     int offs = (intptr_t)dst.target() - ((intptr_t)pc());
     if (dst.reloc() == relocInfo::none && is8bit(offs - short_size)) {
       // 0111 tttn #8-bit disp
-      emit_byte(0x70 | cc);
-      emit_byte((offs - short_size) & 0xFF);
+      emit_int8(0x70 | cc);
+      emit_int8((offs - short_size) & 0xFF);
     } else {
       // 0000 1111 1000 tttn #32-bit disp
-      emit_byte(0x0F);
-      emit_byte(0x80 | cc);
+      emit_int8(0x0F);
+      emit_int8((unsigned char)(0x80 | cc));
       emit_long(offs - long_size);
     }
   } else {
@@ -3085,7 +3085,8 @@ void MacroAssembler::xorps(XMMRegister dst, AddressLiteral src) {
 
 void MacroAssembler::pshufb(XMMRegister dst, AddressLiteral src) {
   // Used in sign-bit flipping with aligned address.
-  assert((UseAVX > 0) || (((intptr_t)src.target() & 15) == 0), "SSE mode requires address alignment 16 bytes");
+  bool aligned_adr = (((intptr_t)src.target() & 15) == 0);
+  assert((UseAVX > 0) || aligned_adr, "SSE mode requires address alignment 16 bytes");
   if (reachable(src)) {
     Assembler::pshufb(dst, as_Address(src));
   } else {

hotspot/src/cpu/x86/vm/macroAssembler_x86.hpp

@@ -126,25 +126,6 @@ class MacroAssembler: public Assembler {
     }
   }
 
-#ifndef PRODUCT
-  static void 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
-
   // The following 4 methods return the offset of the appropriate move instruction
 
   // Support for fast byte/short loading with zero extension (depending on particular CPU)

hotspot/src/cpu/x86/vm/stubGenerator_x86_32.cpp

@@ -2174,13 +2174,13 @@ class StubGenerator: public StubCodeGenerator {
   //   c_rarg2   - K (key) in little endian int array
   //
   address generate_aescrypt_encryptBlock() {
-    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    assert(UseAES, "need AES instructions and misaligned SSE support");
     __ align(CodeEntryAlignment);
     StubCodeMark mark(this, "StubRoutines", "aescrypt_encryptBlock");
     Label L_doLast;
     address start = __ pc();
 
-    const Register from        = rsi;      // source array address
+    const Register from        = rdx;      // source array address
     const Register to          = rdx;      // destination array address
     const Register key         = rcx;      // key array address
     const Register keylen      = rax;
@@ -2189,47 +2189,74 @@ class StubGenerator: public StubCodeGenerator {
     const Address  key_param (rbp, 8+8);
 
     const XMMRegister xmm_result = xmm0;
-    const XMMRegister xmm_temp   = xmm1;
-    const XMMRegister xmm_key_shuf_mask = xmm2;
+    const XMMRegister xmm_key_shuf_mask = xmm1;
+    const XMMRegister xmm_temp1  = xmm2;
+    const XMMRegister xmm_temp2  = xmm3;
+    const XMMRegister xmm_temp3  = xmm4;
+    const XMMRegister xmm_temp4  = xmm5;
 
-    __ enter(); // required for proper stackwalking of RuntimeStub frame
-    __ push(rsi);
-    __ movptr(from , from_param);
-    __ movptr(to   , to_param);
-    __ movptr(key  , key_param);
+    __ enter();   // required for proper stackwalking of RuntimeStub frame
+    __ movptr(from, from_param);
+    __ movptr(key, key_param);
 
+    // keylen could be only {11, 13, 15} * 4 = {44, 52, 60}
     __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
-    // keylen = # of 32-bit words, convert to 128-bit words
-    __ shrl(keylen, 2);
-    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
 
     __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
     __ movdqu(xmm_result, Address(from, 0));  // get 16 bytes of input
+    __ movptr(to, to_param);
 
     // For encryption, the java expanded key ordering is just what we need
 
-    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
-    __ pxor(xmm_result, xmm_temp);
-    for (int offset = 0x10; offset <= 0x90; offset += 0x10) {
-      aes_enc_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
-    }
-    load_key  (xmm_temp, key, 0xa0, xmm_key_shuf_mask);
-    __ cmpl(keylen, 0);
-    __ jcc(Assembler::equal, L_doLast);
-    __ aesenc(xmm_result, xmm_temp);                   // only in 192 and 256 bit keys
-    aes_enc_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
-    load_key(xmm_temp, key, 0xc0, xmm_key_shuf_mask);
-    __ subl(keylen, 2);
-    __ jcc(Assembler::equal, L_doLast);
-    __ aesenc(xmm_result, xmm_temp);                   // only in 256 bit keys
-    aes_enc_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
-    load_key(xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+    load_key(xmm_temp1, key, 0x00, xmm_key_shuf_mask);
+    __ pxor(xmm_result, xmm_temp1);
+
+    load_key(xmm_temp1, key, 0x10, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0x20, xmm_key_shuf_mask);
+    load_key(xmm_temp3, key, 0x30, xmm_key_shuf_mask);
+    load_key(xmm_temp4, key, 0x40, xmm_key_shuf_mask);
+
+    __ aesenc(xmm_result, xmm_temp1);
+    __ aesenc(xmm_result, xmm_temp2);
+    __ aesenc(xmm_result, xmm_temp3);
+    __ aesenc(xmm_result, xmm_temp4);
+
+    load_key(xmm_temp1, key, 0x50, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0x60, xmm_key_shuf_mask);
+    load_key(xmm_temp3, key, 0x70, xmm_key_shuf_mask);
+    load_key(xmm_temp4, key, 0x80, xmm_key_shuf_mask);
+
+    __ aesenc(xmm_result, xmm_temp1);
+    __ aesenc(xmm_result, xmm_temp2);
+    __ aesenc(xmm_result, xmm_temp3);
+    __ aesenc(xmm_result, xmm_temp4);
+
+    load_key(xmm_temp1, key, 0x90, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xa0, xmm_key_shuf_mask);
+
+    __ cmpl(keylen, 44);
+    __ jccb(Assembler::equal, L_doLast);
+
+    __ aesenc(xmm_result, xmm_temp1);
+    __ aesenc(xmm_result, xmm_temp2);
+
+    load_key(xmm_temp1, key, 0xb0, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xc0, xmm_key_shuf_mask);
+
+    __ cmpl(keylen, 52);
+    __ jccb(Assembler::equal, L_doLast);
+
+    __ aesenc(xmm_result, xmm_temp1);
+    __ aesenc(xmm_result, xmm_temp2);
+
+    load_key(xmm_temp1, key, 0xd0, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xe0, xmm_key_shuf_mask);
 
     __ BIND(L_doLast);
-    __ aesenclast(xmm_result, xmm_temp);
+    __ aesenc(xmm_result, xmm_temp1);
+    __ aesenclast(xmm_result, xmm_temp2);
     __ movdqu(Address(to, 0), xmm_result);        // store the result
     __ xorptr(rax, rax); // return 0
-    __ pop(rsi);
     __ leave(); // required for proper stackwalking of RuntimeStub frame
     __ ret(0);
 
@@ -2245,13 +2272,13 @@ class StubGenerator: public StubCodeGenerator {
   //   c_rarg2   - K (key) in little endian int array
   //
   address generate_aescrypt_decryptBlock() {
-    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    assert(UseAES, "need AES instructions and misaligned SSE support");
     __ align(CodeEntryAlignment);
     StubCodeMark mark(this, "StubRoutines", "aescrypt_decryptBlock");
     Label L_doLast;
     address start = __ pc();
 
-    const Register from        = rsi;      // source array address
+    const Register from        = rdx;      // source array address
     const Register to          = rdx;      // destination array address
     const Register key         = rcx;      // key array address
     const Register keylen      = rax;
@@ -2260,51 +2287,76 @@ class StubGenerator: public StubCodeGenerator {
     const Address  key_param (rbp, 8+8);
 
     const XMMRegister xmm_result = xmm0;
-    const XMMRegister xmm_temp   = xmm1;
-    const XMMRegister xmm_key_shuf_mask = xmm2;
+    const XMMRegister xmm_key_shuf_mask = xmm1;
+    const XMMRegister xmm_temp1  = xmm2;
+    const XMMRegister xmm_temp2  = xmm3;
+    const XMMRegister xmm_temp3  = xmm4;
+    const XMMRegister xmm_temp4  = xmm5;
 
     __ enter(); // required for proper stackwalking of RuntimeStub frame
-    __ push(rsi);
-    __ movptr(from , from_param);
-    __ movptr(to   , to_param);
-    __ movptr(key  , key_param);
+    __ movptr(from, from_param);
+    __ movptr(key, key_param);
 
+    // keylen could be only {11, 13, 15} * 4 = {44, 52, 60}
     __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
-    // keylen = # of 32-bit words, convert to 128-bit words
-    __ shrl(keylen, 2);
-    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
 
     __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
     __ movdqu(xmm_result, Address(from, 0));
+    __ movptr(to, to_param);
 
     // for decryption java expanded key ordering is rotated one position from what we want
     // so we start from 0x10 here and hit 0x00 last
     // we don't know if the key is aligned, hence not using load-execute form
-    load_key(xmm_temp, key, 0x10, xmm_key_shuf_mask);
-    __ pxor  (xmm_result, xmm_temp);
-    for (int offset = 0x20; offset <= 0xa0; offset += 0x10) {
-      aes_dec_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
-    }
-    __ cmpl(keylen, 0);
-    __ jcc(Assembler::equal, L_doLast);
-    // only in 192 and 256 bit keys
-    aes_dec_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
-    aes_dec_key(xmm_result, xmm_temp, key, 0xc0, xmm_key_shuf_mask);
-    __ subl(keylen, 2);
-    __ jcc(Assembler::equal, L_doLast);
-    // only in 256 bit keys
-    aes_dec_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
-    aes_dec_key(xmm_result, xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+    load_key(xmm_temp1, key, 0x10, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0x20, xmm_key_shuf_mask);
+    load_key(xmm_temp3, key, 0x30, xmm_key_shuf_mask);
+    load_key(xmm_temp4, key, 0x40, xmm_key_shuf_mask);
+
+    __ pxor  (xmm_result, xmm_temp1);
+    __ aesdec(xmm_result, xmm_temp2);
+    __ aesdec(xmm_result, xmm_temp3);
+    __ aesdec(xmm_result, xmm_temp4);
+
+    load_key(xmm_temp1, key, 0x50, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0x60, xmm_key_shuf_mask);
+    load_key(xmm_temp3, key, 0x70, xmm_key_shuf_mask);
+    load_key(xmm_temp4, key, 0x80, xmm_key_shuf_mask);
+
+    __ aesdec(xmm_result, xmm_temp1);
+    __ aesdec(xmm_result, xmm_temp2);
+    __ aesdec(xmm_result, xmm_temp3);
+    __ aesdec(xmm_result, xmm_temp4);
+
+    load_key(xmm_temp1, key, 0x90, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xa0, xmm_key_shuf_mask);
+    load_key(xmm_temp3, key, 0x00, xmm_key_shuf_mask);
+
+    __ cmpl(keylen, 44);
+    __ jccb(Assembler::equal, L_doLast);
+
+    __ aesdec(xmm_result, xmm_temp1);
+    __ aesdec(xmm_result, xmm_temp2);
+
+    load_key(xmm_temp1, key, 0xb0, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xc0, xmm_key_shuf_mask);
+
+    __ cmpl(keylen, 52);
+    __ jccb(Assembler::equal, L_doLast);
+
+    __ aesdec(xmm_result, xmm_temp1);
+    __ aesdec(xmm_result, xmm_temp2);
+
+    load_key(xmm_temp1, key, 0xd0, xmm_key_shuf_mask);
+    load_key(xmm_temp2, key, 0xe0, xmm_key_shuf_mask);
 
     __ BIND(L_doLast);
-    // for decryption the aesdeclast operation is always on key+0x00
-    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
-    __ aesdeclast(xmm_result, xmm_temp);
+    __ aesdec(xmm_result, xmm_temp1);
+    __ aesdec(xmm_result, xmm_temp2);
 
+    // for decryption the aesdeclast operation is always on key+0x00
+    __ aesdeclast(xmm_result, xmm_temp3);
     __ movdqu(Address(to, 0), xmm_result);  // store the result
-
     __ xorptr(rax, rax); // return 0
-    __ pop(rsi);
     __ leave(); // required for proper stackwalking of RuntimeStub frame
     __ ret(0);
 
@@ -2340,7 +2392,7 @@ class StubGenerator: public StubCodeGenerator {
   //   c_rarg4   - input length
   //
   address generate_cipherBlockChaining_encryptAESCrypt() {
-    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    assert(UseAES, "need AES instructions and misaligned SSE support");
     __ align(CodeEntryAlignment);
     StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_encryptAESCrypt");
     address start = __ pc();
@@ -2393,7 +2445,7 @@ class StubGenerator: public StubCodeGenerator {
     __ jcc(Assembler::notEqual, L_key_192_256);
 
     // 128 bit code follows here
-    __ movptr(pos, 0);
+    __ movl(pos, 0);
     __ align(OptoLoopAlignment);
     __ BIND(L_loopTop_128);
     __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
@@ -2423,15 +2475,15 @@ class StubGenerator: public StubCodeGenerator {
     __ leave();                                  // required for proper stackwalking of RuntimeStub frame
     __ ret(0);
 
-  __ BIND(L_key_192_256);
-  // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
+    __ BIND(L_key_192_256);
+    // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
     __ cmpl(rax, 52);
     __ jcc(Assembler::notEqual, L_key_256);
 
     // 192-bit code follows here (could be changed to use more xmm registers)
-    __ movptr(pos, 0);
-  __ align(OptoLoopAlignment);
-  __ BIND(L_loopTop_192);
+    __ movl(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_192);
     __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
     __ pxor  (xmm_result, xmm_temp);                                // xor with the current r vector
 
@@ -2452,11 +2504,11 @@ class StubGenerator: public StubCodeGenerator {
     __ jcc(Assembler::notEqual, L_loopTop_192);
     __ jmp(L_exit);
 
-  __ BIND(L_key_256);
+    __ BIND(L_key_256);
     // 256-bit code follows here (could be changed to use more xmm registers)
-    __ movptr(pos, 0);
-  __ align(OptoLoopAlignment);
-  __ BIND(L_loopTop_256);
+    __ movl(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_256);
     __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
     __ pxor  (xmm_result, xmm_temp);                                // xor with the current r vector
 
@@ -2495,7 +2547,7 @@ class StubGenerator: public StubCodeGenerator {
   //
 
   address generate_cipherBlockChaining_decryptAESCrypt() {
-    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    assert(UseAES, "need AES instructions and misaligned SSE support");
     __ align(CodeEntryAlignment);
     StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_decryptAESCrypt");
     address start = __ pc();
@@ -2556,9 +2608,9 @@ class StubGenerator: public StubCodeGenerator {
 
 
     // 128-bit code follows here, parallelized
-    __ movptr(pos, 0);
-  __ align(OptoLoopAlignment);
-  __ BIND(L_singleBlock_loopTop_128);
+    __ movl(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_128);
     __ cmpptr(len_reg, 0);           // any blocks left??
     __ jcc(Assembler::equal, L_exit);
     __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
@@ -2597,7 +2649,7 @@ class StubGenerator: public StubCodeGenerator {
     __ jcc(Assembler::notEqual, L_key_256);
 
     // 192-bit code follows here (could be optimized to use parallelism)
-    __ movptr(pos, 0);
+    __ movl(pos, 0);
     __ align(OptoLoopAlignment);
     __ BIND(L_singleBlock_loopTop_192);
     __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
@@ -2622,7 +2674,7 @@ class StubGenerator: public StubCodeGenerator {
 
     __ BIND(L_key_256);
     // 256-bit code follows here (could be optimized to use parallelism)
-    __ movptr(pos, 0);
+    __ movl(pos, 0);
     __ align(OptoLoopAlignment);
     __ BIND(L_singleBlock_loopTop_256);
     __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input

hotspot/src/cpu/x86/vm/vm_version_x86.cpp

@@ -489,8 +489,8 @@ void VM_Version::get_processor_features() {
   }
 
   // The AES intrinsic stubs require AES instruction support (of course)
-  // but also require AVX and sse3 modes for instructions it use.
-  if (UseAES && (UseAVX > 0) && (UseSSE > 2)) {
+  // but also require sse3 mode for instructions it use.
+  if (UseAES && (UseSSE > 2)) {
     if (FLAG_IS_DEFAULT(UseAESIntrinsics)) {
       UseAESIntrinsics = true;
     }

hotspot/src/cpu/zero/vm/assembler_zero.cpp

@@ -56,15 +56,9 @@ void Assembler::pd_patch_instruction(address branch, address target) {
   ShouldNotCallThis();
 }
 
-#ifndef PRODUCT
-void Assembler::pd_print_patched_instruction(address branch) {
-  ShouldNotCallThis();
-}
-#endif // PRODUCT
-
 void MacroAssembler::align(int modulus) {
   while (offset() % modulus != 0)
-    emit_byte(AbstractAssembler::code_fill_byte());
+    emit_int8(AbstractAssembler::code_fill_byte());
 }
 
 void MacroAssembler::bang_stack_with_offset(int offset) {
@@ -72,8 +66,7 @@ void MacroAssembler::bang_stack_with_offset(int offset) {
 }
 
 void MacroAssembler::advance(int bytes) {
-  _code_pos += bytes;
-  sync();
+  code_section()->set_end(code_section()->end() + bytes);
 }
 
 RegisterOrConstant MacroAssembler::delayed_value_impl(

hotspot/src/cpu/zero/vm/assembler_zero.hpp

@@ -37,9 +37,6 @@ class Assembler : public AbstractAssembler {
 
  public:
   void pd_patch_instruction(address branch, address target);
-#ifndef PRODUCT
-  static void pd_print_patched_instruction(address branch);
-#endif // PRODUCT
 };
 
 class MacroAssembler : public Assembler {

hotspot/src/os_cpu/solaris_x86/vm/assembler_solaris_x86.cpp

@@ -116,7 +116,7 @@ void MacroAssembler::get_thread(Register thread) {
   ThreadLocalStorage::pd_tlsAccessMode tlsMode = ThreadLocalStorage::pd_getTlsAccessMode ();
   if (tlsMode == ThreadLocalStorage::pd_tlsAccessIndirect) {            // T1
      // Use thread as a temporary: mov r, gs:[0]; mov r, [r+tlsOffset]
-     emit_byte (segment);
+     emit_int8 (segment);
      // ExternalAddress doesn't work because it can't take NULL
      AddressLiteral null(0, relocInfo::none);
      movptr (thread, null);
@@ -125,7 +125,7 @@ void MacroAssembler::get_thread(Register thread) {
   } else
   if (tlsMode == ThreadLocalStorage::pd_tlsAccessDirect) {              // T2
      // mov r, gs:[tlsOffset]
-     emit_byte (segment);
+     emit_int8 (segment);
      AddressLiteral tls_off((address)ThreadLocalStorage::pd_getTlsOffset(), relocInfo::none);
      movptr (thread, tls_off);
      return ;

hotspot/src/os_cpu/windows_x86/vm/assembler_windows_x86.cpp

@@ -30,7 +30,7 @@
 
 
 void MacroAssembler::int3() {
-  emit_byte(0xCC);
+  emit_int8((unsigned char)0xCC);
 }
 
 #ifndef _LP64

hotspot/src/share/vm/asm/assembler.cpp

@@ -109,37 +109,6 @@ void AbstractAssembler::flush() {
   ICache::invalidate_range(addr_at(0), offset());
 }
 
-
-void AbstractAssembler::a_byte(int x) {
-  emit_byte(x);
-}
-
-
-void AbstractAssembler::a_long(jint x) {
-  emit_long(x);
-}
-
-// Labels refer to positions in the (to be) generated code.  There are bound
-// and unbound
-//
-// Bound labels refer to known positions in the already generated code.
-// offset() is the position the label refers to.
-//
-// Unbound labels refer to unknown positions in the code to be generated; it
-// may contain a list of unresolved displacements that refer to it
-#ifndef PRODUCT
-void AbstractAssembler::print(Label& L) {
-  if (L.is_bound()) {
-    tty->print_cr("bound label to %d|%d", L.loc_pos(), L.loc_sect());
-  } else if (L.is_unbound()) {
-    L.print_instructions((MacroAssembler*)this);
-  } else {
-    tty->print_cr("label in inconsistent state (loc = %d)", L.loc());
-  }
-}
-#endif // PRODUCT
-
-
 void AbstractAssembler::bind(Label& L) {
   if (L.is_bound()) {
     // Assembler can bind a label more than once to the same place.
@@ -342,28 +311,3 @@ bool MacroAssembler::needs_explicit_null_check(intptr_t offset) {
 #endif
   return offset < 0 || os::vm_page_size() <= offset;
 }
-
-#ifndef PRODUCT
-void Label::print_instructions(MacroAssembler* masm) const {
-  CodeBuffer* cb = masm->code();
-  for (int i = 0; i < _patch_index; ++i) {
-    int branch_loc;
-    if (i >= PatchCacheSize) {
-      branch_loc = _patch_overflow->at(i - PatchCacheSize);
-    } else {
-      branch_loc = _patches[i];
-    }
-    int branch_pos  = CodeBuffer::locator_pos(branch_loc);
-    int branch_sect = CodeBuffer::locator_sect(branch_loc);
-    address branch = cb->locator_address(branch_loc);
-    tty->print_cr("unbound label");
-    tty->print("@ %d|%d ", branch_pos, branch_sect);
-    if (branch_sect == CodeBuffer::SECT_CONSTS) {
-      tty->print_cr(PTR_FORMAT, *(address*)branch);
-      continue;
-    }
-    masm->pd_print_patched_instruction(branch);
-    tty->cr();
-  }
-}
-#endif // ndef PRODUCT

hotspot/src/share/vm/asm/assembler.hpp

@@ -216,17 +216,6 @@ class AbstractAssembler : public ResourceObj  {
   bool isByte(int x) const             { return 0 <= x && x < 0x100; }
   bool isShiftCount(int x) const       { return 0 <= x && x < 32; }
 
-  void emit_int8(   int8_t  x) { code_section()->emit_int8(   x); }
-  void emit_int16(  int16_t x) { code_section()->emit_int16(  x); }
-  void emit_int32(  int32_t x) { code_section()->emit_int32(  x); }
-  void emit_int64(  int64_t x) { code_section()->emit_int64(  x); }
-
-  void emit_float(  jfloat  x) { code_section()->emit_float(  x); }
-  void emit_double( jdouble x) { code_section()->emit_double( x); }
-  void emit_address(address x) { code_section()->emit_address(x); }
-
-  void emit_byte(int x)  { emit_int8 (x); }  // deprecated
-  void emit_word(int x)  { emit_int16(x); }  // deprecated
   void emit_long(jint x) { emit_int32(x); }  // deprecated
 
   // Instruction boundaries (required when emitting relocatable values).
@@ -277,9 +266,6 @@ class AbstractAssembler : public ResourceObj  {
   };
 #endif
 
-  // Label functions
-  void print(Label& L);
-
  public:
 
   // Creation
@@ -288,6 +274,15 @@ class AbstractAssembler : public ResourceObj  {
   // ensure buf contains all code (call this before using/copying the code)
   void flush();
 
+  void emit_int8(   int8_t  x) { code_section()->emit_int8(   x); }
+  void emit_int16(  int16_t x) { code_section()->emit_int16(  x); }
+  void emit_int32(  int32_t x) { code_section()->emit_int32(  x); }
+  void emit_int64(  int64_t x) { code_section()->emit_int64(  x); }
+
+  void emit_float(  jfloat  x) { code_section()->emit_float(  x); }
+  void emit_double( jdouble x) { code_section()->emit_double( x); }
+  void emit_address(address x) { code_section()->emit_address(x); }
+
   // min and max values for signed immediate ranges
   static int min_simm(int nbits) { return -(intptr_t(1) << (nbits - 1))    ; }
   static int max_simm(int nbits) { return  (intptr_t(1) << (nbits - 1)) - 1; }
@@ -327,8 +322,6 @@ class AbstractAssembler : public ResourceObj  {
   void    clear_inst_mark()       {        code_section()->clear_mark(); }
 
   // Constants in code
-  void a_byte(int x);
-  void a_long(jint x);
   void relocate(RelocationHolder const& rspec, int format = 0) {
     assert(!pd_check_instruction_mark()
         || inst_mark() == NULL || inst_mark() == code_section()->end(),
@@ -441,15 +434,6 @@ class AbstractAssembler : public ResourceObj  {
    */
   void pd_patch_instruction(address branch, address target);
 
-#ifndef PRODUCT
-  /**
-   * Platform-dependent method of printing an instruction that needs to be
-   * patched.
-   *
-   * @param branch the instruction to be patched in the buffer.
-   */
-  static void pd_print_patched_instruction(address branch);
-#endif // PRODUCT
 };
 
 #ifdef TARGET_ARCH_x86

hotspot/src/share/vm/c1/c1_GraphBuilder.cpp

@@ -3442,6 +3442,11 @@ bool GraphBuilder::try_inline_intrinsics(ciMethod* callee) {
       preserves_state = true;
       break;
 
+    case vmIntrinsics::_loadFence :
+    case vmIntrinsics::_storeFence:
+    case vmIntrinsics::_fullFence :
+      break;
+
     default                       : return false; // do not inline
   }
   // create intrinsic node

hotspot/src/share/vm/c1/c1_LIRGenerator.cpp

@@ -2977,6 +2977,16 @@ void LIRGenerator::do_Intrinsic(Intrinsic* x) {
     do_CompareAndSwap(x, longType);
     break;
 
+  case vmIntrinsics::_loadFence :
+    if (os::is_MP()) __ membar_acquire();
+    break;
+  case vmIntrinsics::_storeFence:
+    if (os::is_MP()) __ membar_release();
+    break;
+  case vmIntrinsics::_fullFence :
+    if (os::is_MP()) __ membar();
+    break;
+
   case vmIntrinsics::_Reference_get:
     do_Reference_get(x);
     break;

hotspot/src/share/vm/ci/ciField.cpp

@@ -366,10 +366,12 @@ bool ciField::will_link(ciInstanceKlass* accessing_klass,
 // ------------------------------------------------------------------
 // ciField::print
 void ciField::print() {
-  tty->print("<ciField ");
+  tty->print("<ciField name=");
   _holder->print_name();
   tty->print(".");
   _name->print_symbol();
+  tty->print(" signature=");
+  _signature->print_symbol();
   tty->print(" offset=%d type=", _offset);
   if (_type != NULL) _type->print_name();
   else               tty->print("(reference)");

hotspot/src/share/vm/classfile/classLoaderData.cpp

@@ -169,16 +169,18 @@ void ClassLoaderData::add_dependency(Handle dependency, TRAPS) {
     ok = (objArrayOop)ok->obj_at(1);
   }
 
+  // Must handle over GC points
+  assert (last != NULL, "dependencies should be initialized");
+  objArrayHandle last_handle(THREAD, last);
+
   // Create a new dependency node with fields for (class_loader or mirror, next)
   objArrayOop deps = oopFactory::new_objectArray(2, CHECK);
   deps->obj_at_put(0, dependency());
 
-  // Must handle over more GC points
+  // Must handle over GC points
   objArrayHandle new_dependency(THREAD, deps);
 
   // Add the dependency under lock
-  assert (last != NULL, "dependencies should be initialized");
-  objArrayHandle last_handle(THREAD, last);
   locked_add_dependency(last_handle, new_dependency);
 }
 

hotspot/src/share/vm/classfile/vmSymbols.hpp

@@ -756,6 +756,15 @@
   do_intrinsic(_unpark,                   sun_misc_Unsafe,        unpark_name, unpark_signature,                 F_RN)  \
    do_name(     unpark_name,                                     "unpark")                                              \
    do_alias(    unpark_signature,                               /*(LObject;)V*/ object_void_signature)                  \
+  do_intrinsic(_loadFence,                sun_misc_Unsafe,        loadFence_name, loadFence_signature,           F_RN)  \
+   do_name(     loadFence_name,                                  "loadFence")                                           \
+   do_alias(    loadFence_signature,                              void_method_signature)                                \
+  do_intrinsic(_storeFence,               sun_misc_Unsafe,        storeFence_name, storeFence_signature,         F_RN)  \
+   do_name(     storeFence_name,                                 "storeFence")                                          \
+   do_alias(    storeFence_signature,                             void_method_signature)                                \
+  do_intrinsic(_fullFence,                sun_misc_Unsafe,        fullFence_name, fullFence_signature,           F_RN)  \
+   do_name(     fullFence_name,                                  "fullFence")                                           \
+   do_alias(    fullFence_signature,                              void_method_signature)                                \
                                                                                                                         \
   /* unsafe memory references (there are a lot of them...) */                                                           \
   do_signature(getObject_signature,       "(Ljava/lang/Object;J)Ljava/lang/Object;")                                    \
@@ -897,12 +906,14 @@
   do_intrinsic(_getAndAddLong,            sun_misc_Unsafe,        getAndAddLong_name, getAndAddLong_signature, F_R)     \
    do_name(     getAndAddLong_name,                               "getAndAddLong")                                      \
    do_signature(getAndAddLong_signature,                          "(Ljava/lang/Object;JJ)J" )                           \
-  do_intrinsic(_getAndSetInt,             sun_misc_Unsafe,        getAndSet_name, getAndSetInt_signature, F_R)          \
-   do_name(     getAndSet_name,                                   "getAndSet")                                          \
+  do_intrinsic(_getAndSetInt,             sun_misc_Unsafe,        getAndSetInt_name, getAndSetInt_signature, F_R)       \
+   do_name(     getAndSetInt_name,                                "getAndSetInt")                                       \
    do_alias(    getAndSetInt_signature,                         /*"(Ljava/lang/Object;JI)I"*/ getAndAddInt_signature)   \
-  do_intrinsic(_getAndSetLong,            sun_misc_Unsafe,        getAndSet_name, getAndSetLong_signature, F_R)         \
+  do_intrinsic(_getAndSetLong,            sun_misc_Unsafe,        getAndSetLong_name, getAndSetLong_signature, F_R)     \
+   do_name(     getAndSetLong_name,                               "getAndSetLong")                                      \
    do_alias(    getAndSetLong_signature,                        /*"(Ljava/lang/Object;JJ)J"*/ getAndAddLong_signature)  \
-  do_intrinsic(_getAndSetObject,          sun_misc_Unsafe,        getAndSet_name, getAndSetObject_signature,  F_R)      \
+  do_intrinsic(_getAndSetObject,          sun_misc_Unsafe,        getAndSetObject_name, getAndSetObject_signature,  F_R)\
+   do_name(     getAndSetObject_name,                             "getAndSetObject")                                    \
    do_signature(getAndSetObject_signature,                        "(Ljava/lang/Object;JLjava/lang/Object;)Ljava/lang/Object;" ) \
                                                                                                                         \
   /* prefetch_signature is shared by all prefetch variants */                                                           \

hotspot/src/share/vm/compiler/compilerOracle.cpp

@@ -538,6 +538,7 @@ void CompilerOracle::parse_from_line(char* line) {
 
   if (match != NULL) {
     if (!_quiet) {
+      ResourceMark rm;
       tty->print("CompilerOracle: %s ", command_names[command]);
       match->print();
     }

hotspot/src/share/vm/opto/addnode.cpp

@@ -189,6 +189,11 @@ Node *AddNode::Ideal(PhaseGVN *phase, bool can_reshape) {
       set_req(1, addx);
       set_req(2, a22);
       progress = this;
+      PhaseIterGVN *igvn = phase->is_IterGVN();
+      if (add2->outcnt() == 0 && igvn) {
+        // add disconnected.
+        igvn->_worklist.push(add2);
+      }
     }
   }
 
@@ -624,6 +629,11 @@ Node *AddPNode::Ideal(PhaseGVN *phase, bool can_reshape) {
     if( t22->singleton() && (t22 != Type::TOP) ) {  // Right input is an add of a constant?
       set_req(Address, phase->transform(new (phase->C) AddPNode(in(Base),in(Address),add->in(1))));
       set_req(Offset, add->in(2));
+      PhaseIterGVN *igvn = phase->is_IterGVN();
+      if (add->outcnt() == 0 && igvn) {
+        // add disconnected.
+        igvn->_worklist.push((Node*)add);
+      }
       return this;              // Made progress
     }
   }

hotspot/src/share/vm/opto/bytecodeInfo.cpp

@@ -403,7 +403,7 @@ const char* InlineTree::check_can_parse(ciMethod* callee) {
 //------------------------------print_inlining---------------------------------
 // Really, the failure_msg can be a success message also.
 void InlineTree::print_inlining(ciMethod* callee_method, int caller_bci, const char* failure_msg) const {
-  CompileTask::print_inlining(callee_method, inline_level(), caller_bci, failure_msg ? failure_msg : "inline");
+  C->print_inlining(callee_method, inline_level(), caller_bci, failure_msg ? failure_msg : "inline");
   if (callee_method == NULL)  tty->print(" callee not monotonic or profiled");
   if (Verbose && callee_method) {
     const InlineTree *top = this;

hotspot/src/share/vm/opto/callGenerator.cpp

@@ -274,6 +274,9 @@ class LateInlineCallGenerator : public DirectCallGenerator {
   virtual void do_late_inline();
 
   virtual JVMState* generate(JVMState* jvms) {
+    Compile *C = Compile::current();
+    C->print_inlining_skip(this);
+
     // Record that this call site should be revisited once the main
     // parse is finished.
     Compile::current()->add_late_inline(this);
@@ -284,7 +287,6 @@ class LateInlineCallGenerator : public DirectCallGenerator {
     // as is done for allocations and macro expansion.
     return DirectCallGenerator::generate(jvms);
   }
-
 };
 
 
@@ -307,7 +309,9 @@ void LateInlineCallGenerator::do_late_inline() {
 
   // Make sure the state is a MergeMem for parsing.
   if (!map->in(TypeFunc::Memory)->is_MergeMem()) {
-    map->set_req(TypeFunc::Memory, MergeMemNode::make(C, map->in(TypeFunc::Memory)));
+    Node* mem = MergeMemNode::make(C, map->in(TypeFunc::Memory));
+    C->initial_gvn()->set_type_bottom(mem);
+    map->set_req(TypeFunc::Memory, mem);
   }
 
   // Make enough space for the expression stack and transfer the incoming arguments
@@ -320,6 +324,8 @@ void LateInlineCallGenerator::do_late_inline() {
     }
   }
 
+  C->print_inlining_insert(this);
+
   CompileLog* log = C->log();
   if (log != NULL) {
     log->head("late_inline method='%d'", log->identify(method()));
@@ -608,7 +614,7 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
         if (cg != NULL && cg->is_inline())
           return cg;
       } else {
-        if (PrintInlining)  CompileTask::print_inlining(callee, jvms->depth() - 1, jvms->bci(), "receiver not constant");
+        if (PrintInlining)  C->print_inlining(callee, jvms->depth() - 1, jvms->bci(), "receiver not constant");
       }
     }
     break;

hotspot/src/share/vm/opto/callGenerator.hpp

@@ -147,9 +147,9 @@ class CallGenerator : public ResourceObj {
                                                 CallGenerator* cg);
   virtual Node* generate_predicate(JVMState* jvms) { return NULL; };
 
-  static void print_inlining(ciMethod* callee, int inline_level, int bci, const char* msg) {
+  static void print_inlining(Compile* C, ciMethod* callee, int inline_level, int bci, const char* msg) {
     if (PrintInlining)
-      CompileTask::print_inlining(callee, inline_level, bci, msg);
+      C->print_inlining(callee, inline_level, bci, msg);
   }
 };