1 /*
   2  * Copyright (c) 2008, 2015, 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 "asm/assembler.hpp"
  27 #include "asm/assembler.inline.hpp"
  28 #include "ci/ciEnv.hpp"
  29 #include "code/codeCacheExtensions.hpp"
  30 #include "gc/shared/cardTableModRefBS.hpp"
  31 #include "gc/shared/collectedHeap.inline.hpp"
  32 #include "interpreter/interpreter.hpp"
  33 #include "interpreter/interpreterRuntime.hpp"
  34 #include "interpreter/templateInterpreterGenerator.hpp"
  35 #include "memory/resourceArea.hpp"
  36 #include "prims/jvm_misc.hpp"
  37 #include "prims/methodHandles.hpp"
  38 #include "runtime/biasedLocking.hpp"
  39 #include "runtime/interfaceSupport.hpp"
  40 #include "runtime/objectMonitor.hpp"
  41 #include "runtime/os.hpp"
  42 #include "runtime/sharedRuntime.hpp"
  43 #include "runtime/stubRoutines.hpp"
  44 #include "utilities/hashtable.hpp"
  45 #include "utilities/macros.hpp"
  46 #if INCLUDE_ALL_GCS
  47 #include "gc/g1/g1CollectedHeap.inline.hpp"
  48 #include "gc/g1/g1SATBCardTableModRefBS.hpp"
  49 #include "gc/g1/heapRegion.hpp"
  50 #endif // INCLUDE_ALL_GCS
  51 
  52 // Returns whether given imm has equal bit fields <0:size-1> and <size:2*size-1>.
  53 inline bool Assembler::LogicalImmediate::has_equal_subpatterns(uintx imm, int size) {
  54   uintx mask = right_n_bits(size);
  55   uintx subpattern1 = mask_bits(imm, mask);
  56   uintx subpattern2 = mask_bits(imm >> size, mask);
  57   return subpattern1 == subpattern2;
  58 }
  59 
  60 // Returns least size that is a power of two from 2 to 64 with the proviso that given
  61 // imm is composed of repeating patterns of this size.
  62 inline int Assembler::LogicalImmediate::least_pattern_size(uintx imm) {
  63   int size = BitsPerWord;
  64   while (size > 2 && has_equal_subpatterns(imm, size >> 1)) {
  65     size >>= 1;
  66   }
  67   return size;
  68 }
  69 
  70 // Returns count of set bits in given imm. Based on variable-precision SWAR algorithm.
  71 inline int Assembler::LogicalImmediate::population_count(uintx x) {
  72   x -= ((x >> 1) & 0x5555555555555555L);
  73   x = (((x >> 2) & 0x3333333333333333L) + (x & 0x3333333333333333L));
  74   x = (((x >> 4) + x) & 0x0f0f0f0f0f0f0f0fL);
  75   x += (x >> 8);
  76   x += (x >> 16);
  77   x += (x >> 32);
  78   return(x & 0x7f);
  79 }
  80 
  81 // Let given x be <A:B> where B = 0 and least bit of A = 1. Returns <A:C>, where C is B-size set bits.
  82 inline uintx Assembler::LogicalImmediate::set_least_zeroes(uintx x) {
  83   return x | (x - 1);
  84 }
  85 
  86 
  87 #ifdef ASSERT
  88 
  89 // Restores immediate by encoded bit masks.
  90 uintx Assembler::LogicalImmediate::decode() {
  91   assert (_encoded, "should be");
  92 
  93   int len_code = (_immN << 6) | ((~_imms) & 0x3f);
  94   assert (len_code != 0, "should be");
  95 
  96   int len = 6;
  97   while (!is_set_nth_bit(len_code, len)) len--;
  98   int esize = 1 << len;
  99   assert (len > 0, "should be");
 100   assert ((_is32bit ? 32 : 64) >= esize, "should be");
 101 
 102   int levels = right_n_bits(len);
 103   int S = _imms & levels;
 104   int R = _immr & levels;
 105 
 106   assert (S != levels, "should be");
 107 
 108   uintx welem = right_n_bits(S + 1);
 109   uintx wmask = (R == 0) ? welem : ((welem >> R) | (welem << (esize - R)));
 110 
 111   for (int size = esize; size < 64; size <<= 1) {
 112     wmask |= (wmask << size);
 113   }
 114 
 115   return wmask;
 116 }
 117 
 118 #endif
 119 
 120 
 121 // Constructs LogicalImmediate by given imm. Figures out if given imm can be used in AArch64 logical
 122 // instructions (AND, ANDS, EOR, ORR) and saves its encoding.
 123 void Assembler::LogicalImmediate::construct(uintx imm, bool is32) {
 124   _is32bit = is32;
 125 
 126   if (is32) {
 127     assert(((imm >> 32) == 0) || (((intx)imm >> 31) == -1), "32-bit immediate is out of range");
 128 
 129     // Replicate low 32 bits.
 130     imm &= 0xffffffff;
 131     imm |= imm << 32;
 132   }
 133 
 134   // All-zeroes and all-ones can not be encoded.
 135   if (imm != 0 && (~imm != 0)) {
 136 
 137     // Let LPS (least pattern size) be the least size (power of two from 2 to 64) of repeating
 138     // patterns in the immediate. If immediate value can be encoded, it is encoded by pattern
 139     // of exactly LPS size (due to structure of valid patterns). In order to verify
 140     // that immediate value can be encoded, LPS is calculated and <LPS-1:0> bits of immediate
 141     // are verified to be valid pattern.
 142     int lps = least_pattern_size(imm);
 143     uintx lps_mask = right_n_bits(lps);
 144 
 145     // A valid pattern has one of the following forms:
 146     //  | 0 x A | 1 x B | 0 x C |, where B > 0 and C > 0, or
 147     //  | 1 x A | 0 x B | 1 x C |, where B > 0 and C > 0.
 148     // For simplicity, the second form of the pattern is inverted into the first form.
 149     bool inverted = imm & 0x1;
 150     uintx pattern = (inverted ? ~imm : imm) & lps_mask;
 151 
 152     //  | 0 x A | 1 x (B + C)   |
 153     uintx without_least_zeroes = set_least_zeroes(pattern);
 154 
 155     // Pattern is valid iff without least zeroes it is a power of two - 1.
 156     if ((without_least_zeroes & (without_least_zeroes + 1)) == 0) {
 157 
 158       // Count B as population count of pattern.
 159       int bits_count = population_count(pattern);
 160 
 161       // Count B+C as population count of pattern without least zeroes
 162       int left_range = population_count(without_least_zeroes);
 163 
 164       // S-prefix is a part of imms field which encodes LPS.
 165       //  LPS  |  S prefix
 166       //   64  |     not defined
 167       //   32  |     0b0
 168       //   16  |     0b10
 169       //    8  |     0b110
 170       //    4  |     0b1110
 171       //    2  |     0b11110
 172       int s_prefix = (lps == 64) ? 0 : ~set_least_zeroes(lps) & 0x3f;
 173 
 174       // immN bit is set iff LPS == 64.
 175       _immN = (lps == 64) ? 1 : 0;
 176       assert (!is32 || (_immN == 0), "32-bit immediate should be encoded with zero N-bit");
 177 
 178       // immr is the rotation size.
 179       _immr = lps + (inverted ? 0 : bits_count) - left_range;
 180 
 181       // imms is the field that encodes bits count and S-prefix.
 182       _imms = ((inverted ? (lps - bits_count) : bits_count) - 1) | s_prefix;
 183 
 184       _encoded = true;
 185       assert (decode() == imm, "illegal encoding");
 186 
 187       return;
 188     }
 189   }
 190 
 191   _encoded = false;
 192 }