1 /*
  2  * Copyright (c) 2008, 2018, 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 "gc/shared/cardTableModRefBS.hpp"
 30 #include "gc/shared/collectedHeap.inline.hpp"
 31 #include "interpreter/interpreter.hpp"
 32 #include "interpreter/interpreterRuntime.hpp"
 33 #include "interpreter/templateInterpreterGenerator.hpp"
 34 #include "memory/resourceArea.hpp"
 35 #include "prims/jvm_misc.hpp"
 36 #include "prims/methodHandles.hpp"
 37 #include "runtime/biasedLocking.hpp"
 38 #include "runtime/interfaceSupport.hpp"
 39 #include "runtime/objectMonitor.hpp"
 40 #include "runtime/os.hpp"
 41 #include "runtime/sharedRuntime.hpp"
 42 #include "runtime/stubRoutines.hpp"
 43 #include "utilities/hashtable.hpp"
 44 #include "utilities/macros.hpp"
 45 #if INCLUDE_ALL_GCS
 46 #include "gc/g1/g1BarrierSet.hpp"
 47 #include "gc/g1/g1CollectedHeap.inline.hpp"
 48 #include "gc/g1/heapRegion.hpp"
 49 #endif // INCLUDE_ALL_GCS
 50 
 51 // Returns whether given imm has equal bit fields <0:size-1> and <size:2*size-1>.
 52 inline bool Assembler::LogicalImmediate::has_equal_subpatterns(uintx imm, int size) {
 53   uintx mask = right_n_bits(size);
 54   uintx subpattern1 = mask_bits(imm, mask);
 55   uintx subpattern2 = mask_bits(imm >> size, mask);
 56   return subpattern1 == subpattern2;
 57 }
 58 
 59 // Returns least size that is a power of two from 2 to 64 with the proviso that given
 60 // imm is composed of repeating patterns of this size.
 61 inline int Assembler::LogicalImmediate::least_pattern_size(uintx imm) {
 62   int size = BitsPerWord;
 63   while (size > 2 && has_equal_subpatterns(imm, size >> 1)) {
 64     size >>= 1;
 65   }
 66   return size;
 67 }
 68 
 69 // Returns count of set bits in given imm. Based on variable-precision SWAR algorithm.
 70 inline int Assembler::LogicalImmediate::population_count(uintx x) {
 71   x -= ((x >> 1) & 0x5555555555555555L);
 72   x = (((x >> 2) & 0x3333333333333333L) + (x & 0x3333333333333333L));
 73   x = (((x >> 4) + x) & 0x0f0f0f0f0f0f0f0fL);
 74   x += (x >> 8);
 75   x += (x >> 16);
 76   x += (x >> 32);
 77   return(x & 0x7f);
 78 }
 79 
 80 // 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.
 81 inline uintx Assembler::LogicalImmediate::set_least_zeroes(uintx x) {
 82   return x | (x - 1);
 83 }
 84 
 85 
 86 #ifdef ASSERT
 87 
 88 // Restores immediate by encoded bit masks.
 89 uintx Assembler::LogicalImmediate::decode() {
 90   assert (_encoded, "should be");
 91 
 92   int len_code = (_immN << 6) | ((~_imms) & 0x3f);
 93   assert (len_code != 0, "should be");
 94 
 95   int len = 6;
 96   while (!is_set_nth_bit(len_code, len)) len--;
 97   int esize = 1 << len;
 98   assert (len > 0, "should be");
 99   assert ((_is32bit ? 32 : 64) >= esize, "should be");
100 
101   int levels = right_n_bits(len);
102   int S = _imms & levels;
103   int R = _immr & levels;
104 
105   assert (S != levels, "should be");
106 
107   uintx welem = right_n_bits(S + 1);
108   uintx wmask = (R == 0) ? welem : ((welem >> R) | (welem << (esize - R)));
109 
110   for (int size = esize; size < 64; size <<= 1) {
111     wmask |= (wmask << size);
112   }
113 
114   return wmask;
115 }
116 
117 #endif
118 
119 
120 // Constructs LogicalImmediate by given imm. Figures out if given imm can be used in AArch64 logical
121 // instructions (AND, ANDS, EOR, ORR) and saves its encoding.
122 void Assembler::LogicalImmediate::construct(uintx imm, bool is32) {
123   _is32bit = is32;
124 
125   if (is32) {
126     assert(((imm >> 32) == 0) || (((intx)imm >> 31) == -1), "32-bit immediate is out of range");
127 
128     // Replicate low 32 bits.
129     imm &= 0xffffffff;
130     imm |= imm << 32;
131   }
132 
133   // All-zeroes and all-ones can not be encoded.
134   if (imm != 0 && (~imm != 0)) {
135 
136     // Let LPS (least pattern size) be the least size (power of two from 2 to 64) of repeating
137     // patterns in the immediate. If immediate value can be encoded, it is encoded by pattern
138     // of exactly LPS size (due to structure of valid patterns). In order to verify
139     // that immediate value can be encoded, LPS is calculated and <LPS-1:0> bits of immediate
140     // are verified to be valid pattern.
141     int lps = least_pattern_size(imm);
142     uintx lps_mask = right_n_bits(lps);
143 
144     // A valid pattern has one of the following forms:
145     //  | 0 x A | 1 x B | 0 x C |, where B > 0 and C > 0, or
146     //  | 1 x A | 0 x B | 1 x C |, where B > 0 and C > 0.
147     // For simplicity, the second form of the pattern is inverted into the first form.
148     bool inverted = imm & 0x1;
149     uintx pattern = (inverted ? ~imm : imm) & lps_mask;
150 
151     //  | 0 x A | 1 x (B + C)   |
152     uintx without_least_zeroes = set_least_zeroes(pattern);
153 
154     // Pattern is valid iff without least zeroes it is a power of two - 1.
155     if ((without_least_zeroes & (without_least_zeroes + 1)) == 0) {
156 
157       // Count B as population count of pattern.
158       int bits_count = population_count(pattern);
159 
160       // Count B+C as population count of pattern without least zeroes
161       int left_range = population_count(without_least_zeroes);
162 
163       // S-prefix is a part of imms field which encodes LPS.
164       //  LPS  |  S prefix
165       //   64  |     not defined
166       //   32  |     0b0
167       //   16  |     0b10
168       //    8  |     0b110
169       //    4  |     0b1110
170       //    2  |     0b11110
171       int s_prefix = (lps == 64) ? 0 : ~set_least_zeroes(lps) & 0x3f;
172 
173       // immN bit is set iff LPS == 64.
174       _immN = (lps == 64) ? 1 : 0;
175       assert (!is32 || (_immN == 0), "32-bit immediate should be encoded with zero N-bit");
176 
177       // immr is the rotation size.
178       _immr = lps + (inverted ? 0 : bits_count) - left_range;
179 
180       // imms is the field that encodes bits count and S-prefix.
181       _imms = ((inverted ? (lps - bits_count) : bits_count) - 1) | s_prefix;
182 
183       _encoded = true;
184       assert (decode() == imm, "illegal encoding");
185 
186       return;
187     }
188   }
189 
190   _encoded = false;
191 }