1 /*
   2  * Copyright (c) 2005, 2010, 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 #ifndef CPU_X86_VM_C1_LINEARSCAN_X86_HPP
  26 #define CPU_X86_VM_C1_LINEARSCAN_X86_HPP
  27 
  28 inline bool LinearScan::is_processed_reg_num(int reg_num) {
  29 #ifndef _LP64
  30   // rsp and rbp (numbers 6 ancd 7) are ignored
  31   assert(FrameMap::rsp_opr->cpu_regnr() == 6, "wrong assumption below");
  32   assert(FrameMap::rbp_opr->cpu_regnr() == 7, "wrong assumption below");
  33   assert(reg_num >= 0, "invalid reg_num");
  34 
  35   return reg_num < 6 || reg_num > 7;
  36 #else
  37   // rsp and rbp, r10, r15 (numbers 6 ancd 7) are ignored
  38   assert(FrameMap::r10_opr->cpu_regnr() == 12, "wrong assumption below");
  39   assert(FrameMap::r15_opr->cpu_regnr() == 13, "wrong assumption below");
  40   assert(FrameMap::rsp_opr->cpu_regnrLo() == 14, "wrong assumption below");
  41   assert(FrameMap::rbp_opr->cpu_regnrLo() == 15, "wrong assumption below");
  42   assert(reg_num >= 0, "invalid reg_num");
  43 
  44   return reg_num < 12 || reg_num > 15;
  45 #endif // _LP64
  46 }
  47 
  48 inline int LinearScan::num_physical_regs(BasicType type) {
  49   // Intel requires two cpu registers for long,
  50   // but requires only one fpu register for double
  51   if (LP64_ONLY(false &&) type == T_LONG) {
  52     return 2;
  53   }
  54   return 1;
  55 }
  56 
  57 
  58 inline bool LinearScan::requires_adjacent_regs(BasicType type) {
  59   return false;
  60 }
  61 
  62 inline bool LinearScan::is_caller_save(int assigned_reg) {
  63   assert(assigned_reg >= 0 && assigned_reg < nof_regs, "should call this only for registers");
  64   return true; // no callee-saved registers on Intel
  65 
  66 }
  67 
  68 
  69 inline void LinearScan::pd_add_temps(LIR_Op* op) {
  70   switch (op->code()) {
  71     case lir_tan:
  72     case lir_sin:
  73     case lir_cos: {
  74       // The slow path for these functions may need to save and
  75       // restore all live registers but we don't want to save and
  76       // restore everything all the time, so mark the xmms as being
  77       // killed.  If the slow path were explicit or we could propagate
  78       // live register masks down to the assembly we could do better
  79       // but we don't have any easy way to do that right now.  We
  80       // could also consider not killing all xmm registers if we
  81       // assume that slow paths are uncommon but it's not clear that
  82       // would be a good idea.
  83       if (UseSSE > 0) {
  84 #ifndef PRODUCT
  85         if (TraceLinearScanLevel >= 2) {
  86           tty->print_cr("killing XMMs for trig");
  87         }
  88 #endif
  89         int op_id = op->id();
  90         for (int xmm = 0; xmm < FrameMap::nof_caller_save_xmm_regs; xmm++) {
  91           LIR_Opr opr = FrameMap::caller_save_xmm_reg_at(xmm);
  92           add_temp(reg_num(opr), op_id, noUse, T_ILLEGAL);
  93         }
  94       }
  95       break;
  96     }
  97   }
  98 }
  99 
 100 
 101 // Implementation of LinearScanWalker
 102 
 103 inline bool LinearScanWalker::pd_init_regs_for_alloc(Interval* cur) {
 104   if (allocator()->gen()->is_vreg_flag_set(cur->reg_num(), LIRGenerator::byte_reg)) {
 105     assert(cur->type() != T_FLOAT && cur->type() != T_DOUBLE, "cpu regs only");
 106     _first_reg = pd_first_byte_reg;
 107     _last_reg = pd_last_byte_reg;
 108     return true;
 109   } else if ((UseSSE >= 1 && cur->type() == T_FLOAT) || (UseSSE >= 2 && cur->type() == T_DOUBLE)) {
 110     _first_reg = pd_first_xmm_reg;
 111     _last_reg = pd_last_xmm_reg;
 112     return true;
 113   }
 114 
 115   return false;
 116 }
 117 
 118 
 119 class FpuStackAllocator VALUE_OBJ_CLASS_SPEC {
 120  private:
 121   Compilation* _compilation;
 122   LinearScan* _allocator;
 123 
 124   LIR_OpVisitState visitor;
 125 
 126   LIR_List* _lir;
 127   int _pos;
 128   FpuStackSim _sim;
 129   FpuStackSim _temp_sim;
 130 
 131   bool _debug_information_computed;
 132 
 133   LinearScan*   allocator()                      { return _allocator; }
 134   Compilation*  compilation() const              { return _compilation; }
 135 
 136   // unified bailout support
 137   void          bailout(const char* msg) const   { compilation()->bailout(msg); }
 138   bool          bailed_out() const               { return compilation()->bailed_out(); }
 139 
 140   int pos() { return _pos; }
 141   void set_pos(int pos) { _pos = pos; }
 142   LIR_Op* cur_op() { return lir()->instructions_list()->at(pos()); }
 143   LIR_List* lir() { return _lir; }
 144   void set_lir(LIR_List* lir) { _lir = lir; }
 145   FpuStackSim* sim() { return &_sim; }
 146   FpuStackSim* temp_sim() { return &_temp_sim; }
 147 
 148   int fpu_num(LIR_Opr opr);
 149   int tos_offset(LIR_Opr opr);
 150   LIR_Opr to_fpu_stack_top(LIR_Opr opr, bool dont_check_offset = false);
 151 
 152   // Helper functions for handling operations
 153   void insert_op(LIR_Op* op);
 154   void insert_exchange(int offset);
 155   void insert_exchange(LIR_Opr opr);
 156   void insert_free(int offset);
 157   void insert_free_if_dead(LIR_Opr opr);
 158   void insert_free_if_dead(LIR_Opr opr, LIR_Opr ignore);
 159   void insert_copy(LIR_Opr from, LIR_Opr to);
 160   void do_rename(LIR_Opr from, LIR_Opr to);
 161   void do_push(LIR_Opr opr);
 162   void pop_if_last_use(LIR_Op* op, LIR_Opr opr);
 163   void pop_always(LIR_Op* op, LIR_Opr opr);
 164   void clear_fpu_stack(LIR_Opr preserve);
 165   void handle_op1(LIR_Op1* op1);
 166   void handle_op2(LIR_Op2* op2);
 167   void handle_opCall(LIR_OpCall* opCall);
 168   void compute_debug_information(LIR_Op* op);
 169   void allocate_exception_handler(XHandler* xhandler);
 170   void allocate_block(BlockBegin* block);
 171 
 172 #ifndef PRODUCT
 173   void check_invalid_lir_op(LIR_Op* op);
 174 #endif
 175 
 176   // Helper functions for merging of fpu stacks
 177   void merge_insert_add(LIR_List* instrs, FpuStackSim* cur_sim, int reg);
 178   void merge_insert_xchg(LIR_List* instrs, FpuStackSim* cur_sim, int slot);
 179   void merge_insert_pop(LIR_List* instrs, FpuStackSim* cur_sim);
 180   bool merge_rename(FpuStackSim* cur_sim, FpuStackSim* sux_sim, int start_slot, int change_slot);
 181   void merge_fpu_stack(LIR_List* instrs, FpuStackSim* cur_sim, FpuStackSim* sux_sim);
 182   void merge_cleanup_fpu_stack(LIR_List* instrs, FpuStackSim* cur_sim, BitMap& live_fpu_regs);
 183   bool merge_fpu_stack_with_successors(BlockBegin* block);
 184 
 185  public:
 186   LIR_Opr to_fpu_stack(LIR_Opr opr); // used by LinearScan for creation of debug information
 187 
 188   FpuStackAllocator(Compilation* compilation, LinearScan* allocator);
 189   void allocate();
 190 };
 191 
 192 #endif // CPU_X86_VM_C1_LINEARSCAN_X86_HPP