1 /*
   2  * Copyright (c) 1999, 2017, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #include "precompiled.hpp"
  26 #include "c1/c1_MacroAssembler.hpp"
  27 #include "c1/c1_Runtime1.hpp"
  28 #include "classfile/systemDictionary.hpp"
  29 #include "gc/shared/collectedHeap.hpp"
  30 #include "interpreter/interpreter.hpp"
  31 #include "oops/arrayOop.hpp"
  32 #include "oops/markOop.hpp"
  33 #include "runtime/basicLock.hpp"
  34 #include "runtime/biasedLocking.hpp"
  35 #include "runtime/os.hpp"
  36 #include "runtime/sharedRuntime.hpp"
  37 #include "runtime/stubRoutines.hpp"
  38 
  39 void C1_MacroAssembler::inline_cache_check(Register receiver, Register iCache) {
  40   Label L;
  41   const Register temp_reg = G3_scratch;
  42   // Note: needs more testing of out-of-line vs. inline slow case
  43   verify_oop(receiver);
  44   load_klass(receiver, temp_reg);
  45   cmp_and_brx_short(temp_reg, iCache, Assembler::equal, Assembler::pt, L);
  46   AddressLiteral ic_miss(SharedRuntime::get_ic_miss_stub());
  47   jump_to(ic_miss, temp_reg);
  48   delayed()->nop();
  49   align(CodeEntryAlignment);
  50   bind(L);
  51 }
  52 
  53 
  54 void C1_MacroAssembler::explicit_null_check(Register base) {
  55   Unimplemented();
  56 }
  57 
  58 
  59 void C1_MacroAssembler::build_frame(int frame_size_in_bytes, int bang_size_in_bytes) {
  60   assert(bang_size_in_bytes >= frame_size_in_bytes, "stack bang size incorrect");
  61   generate_stack_overflow_check(bang_size_in_bytes);
  62   // Create the frame.
  63   save_frame_c1(frame_size_in_bytes);
  64 }
  65 
  66 
  67 void C1_MacroAssembler::verified_entry() {
  68   if (C1Breakpoint) breakpoint_trap();
  69   // build frame
  70   verify_FPU(0, "method_entry");
  71 }
  72 
  73 
  74 void C1_MacroAssembler::lock_object(Register Rmark, Register Roop, Register Rbox, Register Rscratch, Label& slow_case) {
  75   assert_different_registers(Rmark, Roop, Rbox, Rscratch);
  76 
  77   Label done;
  78 
  79   Address mark_addr(Roop, oopDesc::mark_offset_in_bytes());
  80 
  81   // The following move must be the first instruction of emitted since debug
  82   // information may be generated for it.
  83   // Load object header
  84   ld_ptr(mark_addr, Rmark);
  85 
  86   verify_oop(Roop);
  87 
  88   // save object being locked into the BasicObjectLock
  89   st_ptr(Roop, Rbox, BasicObjectLock::obj_offset_in_bytes());
  90 
  91   if (UseBiasedLocking) {
  92     biased_locking_enter(Roop, Rmark, Rscratch, done, &slow_case);
  93   }
  94 
  95   // Save Rbox in Rscratch to be used for the cas operation
  96   mov(Rbox, Rscratch);
  97 
  98   // and mark it unlocked
  99   or3(Rmark, markOopDesc::unlocked_value, Rmark);
 100 
 101   // save unlocked object header into the displaced header location on the stack
 102   st_ptr(Rmark, Rbox, BasicLock::displaced_header_offset_in_bytes());
 103 
 104   // compare object markOop with Rmark and if equal exchange Rscratch with object markOop
 105   assert(mark_addr.disp() == 0, "cas must take a zero displacement");
 106   cas_ptr(mark_addr.base(), Rmark, Rscratch);
 107   // if compare/exchange succeeded we found an unlocked object and we now have locked it
 108   // hence we are done
 109   cmp(Rmark, Rscratch);
 110   brx(Assembler::equal, false, Assembler::pt, done);
 111   delayed()->sub(Rscratch, SP, Rscratch);  //pull next instruction into delay slot
 112   // we did not find an unlocked object so see if this is a recursive case
 113   // sub(Rscratch, SP, Rscratch);
 114   assert(os::vm_page_size() > 0xfff, "page size too small - change the constant");
 115   andcc(Rscratch, 0xfffff003, Rscratch);
 116   brx(Assembler::notZero, false, Assembler::pn, slow_case);
 117   delayed()->st_ptr(Rscratch, Rbox, BasicLock::displaced_header_offset_in_bytes());
 118   bind(done);
 119 }
 120 
 121 
 122 void C1_MacroAssembler::unlock_object(Register Rmark, Register Roop, Register Rbox, Label& slow_case) {
 123   assert_different_registers(Rmark, Roop, Rbox);
 124 
 125   Label done;
 126 
 127   Address mark_addr(Roop, oopDesc::mark_offset_in_bytes());
 128   assert(mark_addr.disp() == 0, "cas must take a zero displacement");
 129 
 130   if (UseBiasedLocking) {
 131     // load the object out of the BasicObjectLock
 132     ld_ptr(Rbox, BasicObjectLock::obj_offset_in_bytes(), Roop);
 133     verify_oop(Roop);
 134     biased_locking_exit(mark_addr, Rmark, done);
 135   }
 136   // Test first it it is a fast recursive unlock
 137   ld_ptr(Rbox, BasicLock::displaced_header_offset_in_bytes(), Rmark);
 138   br_null_short(Rmark, Assembler::pt, done);
 139   if (!UseBiasedLocking) {
 140     // load object
 141     ld_ptr(Rbox, BasicObjectLock::obj_offset_in_bytes(), Roop);
 142     verify_oop(Roop);
 143   }
 144 
 145   // Check if it is still a light weight lock, this is is true if we see
 146   // the stack address of the basicLock in the markOop of the object
 147   cas_ptr(mark_addr.base(), Rbox, Rmark);
 148   cmp(Rbox, Rmark);
 149 
 150   brx(Assembler::notEqual, false, Assembler::pn, slow_case);
 151   delayed()->nop();
 152   // Done
 153   bind(done);
 154 }
 155 
 156 
 157 void C1_MacroAssembler::try_allocate(
 158   Register obj,                        // result: pointer to object after successful allocation
 159   Register var_size_in_bytes,          // object size in bytes if unknown at compile time; invalid otherwise
 160   int      con_size_in_bytes,          // object size in bytes if   known at compile time
 161   Register t1,                         // temp register, must be global register for incr_allocated_bytes
 162   Register t2,                         // temp register
 163   Label&   slow_case                   // continuation point if fast allocation fails
 164 ) {
 165   RegisterOrConstant size_in_bytes = var_size_in_bytes->is_valid()
 166     ? RegisterOrConstant(var_size_in_bytes) : RegisterOrConstant(con_size_in_bytes);
 167   if (UseTLAB) {
 168     tlab_allocate(obj, var_size_in_bytes, con_size_in_bytes, t1, slow_case);
 169   } else {
 170     eden_allocate(obj, var_size_in_bytes, con_size_in_bytes, t1, t2, slow_case);
 171     incr_allocated_bytes(size_in_bytes, t1, t2);
 172   }
 173 }
 174 
 175 
 176 void C1_MacroAssembler::initialize_header(Register obj, Register klass, Register len, Register t1, Register t2) {
 177   assert_different_registers(obj, klass, len, t1, t2);
 178   if (UseBiasedLocking && !len->is_valid()) {
 179     ld_ptr(klass, in_bytes(Klass::prototype_header_offset()), t1);
 180   } else {
 181     set((intx)markOopDesc::prototype(), t1);
 182   }
 183   st_ptr(t1, obj, oopDesc::mark_offset_in_bytes());
 184   if (UseCompressedClassPointers) {
 185     // Save klass
 186     mov(klass, t1);
 187     encode_klass_not_null(t1);
 188     stw(t1, obj, oopDesc::klass_offset_in_bytes());
 189   } else {
 190     st_ptr(klass, obj, oopDesc::klass_offset_in_bytes());
 191   }
 192   if (len->is_valid()) {
 193     st(len, obj, arrayOopDesc::length_offset_in_bytes());
 194   } else if (UseCompressedClassPointers) {
 195     // otherwise length is in the class gap
 196     store_klass_gap(G0, obj);
 197   }
 198 }
 199 
 200 
 201 void C1_MacroAssembler::initialize_body(Register base, Register index) {
 202   zero_memory(base, index);
 203 }
 204 
 205 
 206 void C1_MacroAssembler::allocate_object(
 207   Register obj,                        // result: pointer to object after successful allocation
 208   Register t1,                         // temp register
 209   Register t2,                         // temp register, must be a global register for try_allocate
 210   Register t3,                         // temp register
 211   int      hdr_size,                   // object header size in words
 212   int      obj_size,                   // object size in words
 213   Register klass,                      // object klass
 214   Label&   slow_case                   // continuation point if fast allocation fails
 215 ) {
 216   assert_different_registers(obj, t1, t2, t3, klass);
 217   assert(klass == G5, "must be G5");
 218 
 219   // allocate space & initialize header
 220   if (!is_simm13(obj_size * wordSize)) {
 221     // would need to use extra register to load
 222     // object size => go the slow case for now
 223     ba(slow_case);
 224     delayed()->nop();
 225     return;
 226   }
 227   try_allocate(obj, noreg, obj_size * wordSize, t2, t3, slow_case);
 228 
 229   initialize_object(obj, klass, noreg, obj_size * HeapWordSize, t1, t2, /* is_tlab_allocated */ UseTLAB);
 230 }
 231 
 232 void C1_MacroAssembler::initialize_object(
 233   Register obj,                        // result: pointer to object after successful allocation
 234   Register klass,                      // object klass
 235   Register var_size_in_bytes,          // object size in bytes if unknown at compile time; invalid otherwise
 236   int      con_size_in_bytes,          // object size in bytes if   known at compile time
 237   Register t1,                         // temp register
 238   Register t2,                         // temp register
 239   bool     is_tlab_allocated           // the object was allocated in a TLAB; relevant for the implementation of ZeroTLAB
 240   ) {
 241   const int hdr_size_in_bytes = instanceOopDesc::header_size() * HeapWordSize;
 242 
 243   initialize_header(obj, klass, noreg, t1, t2);
 244 
 245 #ifdef ASSERT
 246   {
 247     Label ok;
 248     ld(klass, in_bytes(Klass::layout_helper_offset()), t1);
 249     if (var_size_in_bytes != noreg) {
 250       cmp_and_brx_short(t1, var_size_in_bytes, Assembler::equal, Assembler::pt, ok);
 251     } else {
 252       cmp_and_brx_short(t1, con_size_in_bytes, Assembler::equal, Assembler::pt, ok);
 253     }
 254     stop("bad size in initialize_object");
 255     should_not_reach_here();
 256 
 257     bind(ok);
 258   }
 259 
 260 #endif
 261 
 262   if (!(UseTLAB && ZeroTLAB && is_tlab_allocated)) {
 263     // initialize body
 264     const int threshold = 5 * HeapWordSize;              // approximate break even point for code size
 265     if (var_size_in_bytes != noreg) {
 266       // use a loop
 267       add(obj, hdr_size_in_bytes, t1);               // compute address of first element
 268       sub(var_size_in_bytes, hdr_size_in_bytes, t2); // compute size of body
 269       initialize_body(t1, t2);
 270     } else if (con_size_in_bytes <= threshold) {
 271       // use explicit NULL stores
 272       for (int i = hdr_size_in_bytes; i < con_size_in_bytes; i += HeapWordSize)     st_ptr(G0, obj, i);
 273     } else if (con_size_in_bytes > hdr_size_in_bytes) {
 274       // use a loop
 275       const Register base  = t1;
 276       const Register index = t2;
 277       add(obj, hdr_size_in_bytes, base);               // compute address of first element
 278       // compute index = number of words to clear
 279       set(con_size_in_bytes - hdr_size_in_bytes, index);
 280       initialize_body(base, index);
 281     }
 282   }
 283 
 284   if (CURRENT_ENV->dtrace_alloc_probes()) {
 285     assert(obj == O0, "must be");
 286     call(CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::dtrace_object_alloc_id)),
 287          relocInfo::runtime_call_type);
 288     delayed()->nop();
 289   }
 290 
 291   verify_oop(obj);
 292 }
 293 
 294 
 295 void C1_MacroAssembler::allocate_array(
 296   Register obj,                        // result: pointer to array after successful allocation
 297   Register len,                        // array length
 298   Register t1,                         // temp register
 299   Register t2,                         // temp register
 300   Register t3,                         // temp register
 301   int      hdr_size,                   // object header size in words
 302   int      elt_size,                   // element size in bytes
 303   Register klass,                      // object klass
 304   Label&   slow_case                   // continuation point if fast allocation fails
 305 ) {
 306   assert_different_registers(obj, len, t1, t2, t3, klass);
 307   assert(klass == G5, "must be G5");
 308   assert(t1 == G1, "must be G1");
 309 
 310   // determine alignment mask
 311   assert(!(BytesPerWord & 1), "must be a multiple of 2 for masking code to work");
 312 
 313   // check for negative or excessive length
 314   // note: the maximum length allowed is chosen so that arrays of any
 315   //       element size with this length are always smaller or equal
 316   //       to the largest integer (i.e., array size computation will
 317   //       not overflow)
 318   set(max_array_allocation_length, t1);
 319   cmp(len, t1);
 320   br(Assembler::greaterUnsigned, false, Assembler::pn, slow_case);
 321 
 322   // compute array size
 323   // note: if 0 <= len <= max_length, len*elt_size + header + alignment is
 324   //       smaller or equal to the largest integer; also, since top is always
 325   //       aligned, we can do the alignment here instead of at the end address
 326   //       computation
 327   const Register arr_size = t1;
 328   switch (elt_size) {
 329     case  1: delayed()->mov(len,    arr_size); break;
 330     case  2: delayed()->sll(len, 1, arr_size); break;
 331     case  4: delayed()->sll(len, 2, arr_size); break;
 332     case  8: delayed()->sll(len, 3, arr_size); break;
 333     default: ShouldNotReachHere();
 334   }
 335   add(arr_size, hdr_size * wordSize + MinObjAlignmentInBytesMask, arr_size); // add space for header & alignment
 336   and3(arr_size, ~MinObjAlignmentInBytesMask, arr_size);                     // align array size
 337 
 338   // allocate space & initialize header
 339   if (UseTLAB) {
 340     tlab_allocate(obj, arr_size, 0, t2, slow_case);
 341   } else {
 342     eden_allocate(obj, arr_size, 0, t2, t3, slow_case);
 343   }
 344   initialize_header(obj, klass, len, t2, t3);
 345 
 346   // initialize body
 347   const Register base  = t2;
 348   const Register index = t3;
 349   add(obj, hdr_size * wordSize, base);               // compute address of first element
 350   sub(arr_size, hdr_size * wordSize, index);         // compute index = number of words to clear
 351   initialize_body(base, index);
 352 
 353   if (CURRENT_ENV->dtrace_alloc_probes()) {
 354     assert(obj == O0, "must be");
 355     call(CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::dtrace_object_alloc_id)),
 356          relocInfo::runtime_call_type);
 357     delayed()->nop();
 358   }
 359 
 360   verify_oop(obj);
 361 }
 362 
 363 
 364 #ifndef PRODUCT
 365 
 366 void C1_MacroAssembler::verify_stack_oop(int stack_offset) {
 367   if (!VerifyOops) return;
 368   verify_oop_addr(Address(SP, stack_offset + STACK_BIAS));
 369 }
 370 
 371 void C1_MacroAssembler::verify_not_null_oop(Register r) {
 372   Label not_null;
 373   br_notnull_short(r, Assembler::pt, not_null);
 374   stop("non-null oop required");
 375   bind(not_null);
 376   if (!VerifyOops) return;
 377   verify_oop(r);
 378 }
 379 
 380 void C1_MacroAssembler::invalidate_registers(bool iregisters, bool lregisters, bool oregisters,
 381                                              Register preserve1, Register preserve2) {
 382   if (iregisters) {
 383     for (int i = 0; i < 6; i++) {
 384       Register r = as_iRegister(i);
 385       if (r != preserve1 && r != preserve2)  set(0xdead, r);
 386     }
 387   }
 388   if (oregisters) {
 389     for (int i = 0; i < 6; i++) {
 390       Register r = as_oRegister(i);
 391       if (r != preserve1 && r != preserve2)  set(0xdead, r);
 392     }
 393   }
 394   if (lregisters) {
 395     for (int i = 0; i < 8; i++) {
 396       Register r = as_lRegister(i);
 397       if (r != preserve1 && r != preserve2)  set(0xdead, r);
 398     }
 399   }
 400 }
 401 
 402 
 403 #endif