1 /*
2 * Copyright (c) 2006, 2016, 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 "runtime/sharedRuntime.hpp"
27 #include "utilities/copy.hpp"
28
29
30 // Copy bytes; larger units are filled atomically if everything is aligned.
31 void Copy::conjoint_memory_atomic(void* from, void* to, size_t size) {
32 address src = (address) from;
33 address dst = (address) to;
34 uintptr_t bits = (uintptr_t) src | (uintptr_t) dst | (uintptr_t) size;
35
36 // (Note: We could improve performance by ignoring the low bits of size,
37 // and putting a short cleanup loop after each bulk copy loop.
38 // There are plenty of other ways to make this faster also,
39 // and it's a slippery slope. For now, let's keep this code simple
40 // since the simplicity helps clarify the atomicity semantics of
41 // this operation. There are also CPU-specific assembly versions
42 // which may or may not want to include such optimizations.)
43
44 if (bits % sizeof(jlong) == 0) {
45 Copy::conjoint_jlongs_atomic((jlong*) src, (jlong*) dst, size / sizeof(jlong));
46 } else if (bits % sizeof(jint) == 0) {
47 Copy::conjoint_jints_atomic((jint*) src, (jint*) dst, size / sizeof(jint));
48 } else if (bits % sizeof(jshort) == 0) {
49 Copy::conjoint_jshorts_atomic((jshort*) src, (jshort*) dst, size / sizeof(jshort));
50 } else {
51 // Not aligned, so no need to be atomic.
52 Copy::conjoint_jbytes((void*) src, (void*) dst, size);
53 }
54 }
55
56 class CopySwap : AllStatic {
57 public:
58 /**
59 * Copy and byte swap elements
60 *
61 * @param src address of source
62 * @param dst address of destination
63 * @param byte_count number of bytes to copy
64 * @param elem_size size of the elements to copy-swap
65 */
66 static void conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
67 assert(src != NULL, "address must not be NULL");
68 assert(dst != NULL, "address must not be NULL");
69 assert(elem_size == 2 || elem_size == 4 || elem_size == 8,
70 "incorrect element size: " SIZE_FORMAT, elem_size);
71 assert(is_size_aligned(byte_count, elem_size),
72 "byte_count " SIZE_FORMAT " must be multiple of element size " SIZE_FORMAT, byte_count, elem_size);
73
74 address src_end = src + byte_count;
75
76 if (dst <= src || dst >= src_end) {
77 do_conjoint_swap<RIGHT>(src, dst, byte_count, elem_size);
78 } else {
79 do_conjoint_swap<LEFT>(src, dst, byte_count, elem_size);
80 }
81 }
82
83 private:
84 /**
85 * Byte swap a 16-bit value
86 */
87 static uint16_t byte_swap(uint16_t x) {
88 return (x << 8) | (x >> 8);
89 }
90
91 /**
92 * Byte swap a 32-bit value
93 */
94 static uint32_t byte_swap(uint32_t x) {
95 uint16_t lo = (uint16_t)x;
96 uint16_t hi = (uint16_t)(x >> 16);
97
98 return ((uint32_t)byte_swap(lo) << 16) | (uint32_t)byte_swap(hi);
99 }
100
101 /**
102 * Byte swap a 64-bit value
103 */
104 static uint64_t byte_swap(uint64_t x) {
105 uint32_t lo = (uint32_t)x;
106 uint32_t hi = (uint32_t)(x >> 32);
107
108 return ((uint64_t)byte_swap(lo) << 32) | (uint64_t)byte_swap(hi);
109 }
110
111 enum CopyDirection {
112 RIGHT, // lower -> higher address
113 LEFT // higher -> lower address
114 };
115
116 /**
117 * Copy and byte swap elements
118 *
119 * <T> - type of element to copy
120 * <D> - copy direction
121 * <is_src_aligned> - true if src argument is aligned to element size
122 * <is_dst_aligned> - true if dst argument is aligned to element size
123 *
124 * @param src address of source
125 * @param dst address of destination
126 * @param byte_count number of bytes to copy
127 */
128 template <typename T, CopyDirection D, bool is_src_aligned, bool is_dst_aligned>
129 static void do_conjoint_swap(address src, address dst, size_t byte_count) {
130 address cur_src, cur_dst;
131
132 switch (D) {
133 case RIGHT:
134 cur_src = src;
135 cur_dst = dst;
136 break;
137 case LEFT:
138 cur_src = src + byte_count - sizeof(T);
139 cur_dst = dst + byte_count - sizeof(T);
140 break;
141 }
142
143 for (size_t i = 0; i < byte_count / sizeof(T); i++) {
144 T tmp;
145
146 if (is_src_aligned) {
147 tmp = *(T*)cur_src;
148 } else {
149 memcpy(&tmp, cur_src, sizeof(T));
150 }
151
152 tmp = byte_swap(tmp);
153
154 if (is_dst_aligned) {
155 *(T*)cur_dst = tmp;
156 } else {
157 memcpy(cur_dst, &tmp, sizeof(T));
158 }
159
160 switch (D) {
161 case RIGHT:
162 cur_src += sizeof(T);
163 cur_dst += sizeof(T);
164 break;
165 case LEFT:
166 cur_src -= sizeof(T);
167 cur_dst -= sizeof(T);
168 break;
169 }
170 }
171 }
172
173 /**
174 * Copy and byte swap elements
175 *
176 * <T> - type of element to copy
177 * <D> - copy direction
178 *
179 * @param src address of source
180 * @param dst address of destination
181 * @param byte_count number of bytes to copy
182 */
183 template <typename T, CopyDirection direction>
184 static void do_conjoint_swap(address src, address dst, size_t byte_count) {
185 if (is_ptr_aligned(src, sizeof(T))) {
186 if (is_ptr_aligned(dst, sizeof(T))) {
187 do_conjoint_swap<T,direction,true,true>(src, dst, byte_count);
188 } else {
189 do_conjoint_swap<T,direction,true,false>(src, dst, byte_count);
190 }
191 } else {
192 if (is_ptr_aligned(dst, sizeof(T))) {
193 do_conjoint_swap<T,direction,false,true>(src, dst, byte_count);
194 } else {
195 do_conjoint_swap<T,direction,false,false>(src, dst, byte_count);
196 }
197 }
198 }
199
200
201 /**
202 * Copy and byte swap elements
203 *
204 * <D> - copy direction
205 *
206 * @param src address of source
207 * @param dst address of destination
208 * @param byte_count number of bytes to copy
209 * @param elem_size size of the elements to copy-swap
210 */
211 template <CopyDirection D>
212 static void do_conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
213 switch (elem_size) {
214 case 2: do_conjoint_swap<uint16_t,D>(src, dst, byte_count); break;
215 case 4: do_conjoint_swap<uint32_t,D>(src, dst, byte_count); break;
216 case 8: do_conjoint_swap<uint64_t,D>(src, dst, byte_count); break;
217 default: guarantee(false, "do_conjoint_swap: Invalid elem_size " SIZE_FORMAT "\n", elem_size);
218 }
219 }
220 };
221
222 void Copy::conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
223 CopySwap::conjoint_swap(src, dst, byte_count, elem_size);
224 }
225
226 // Fill bytes; larger units are filled atomically if everything is aligned.
227 void Copy::fill_to_memory_atomic(void* to, size_t size, jubyte value) {
228 address dst = (address) to;
229 uintptr_t bits = (uintptr_t) to | (uintptr_t) size;
230 if (bits % sizeof(jlong) == 0) {
231 jlong fill = (julong)( (jubyte)value ); // zero-extend
232 if (fill != 0) {
233 fill += fill << 8;
234 fill += fill << 16;
235 fill += fill << 32;
236 }
237 //Copy::fill_to_jlongs_atomic((jlong*) dst, size / sizeof(jlong));
238 for (uintptr_t off = 0; off < size; off += sizeof(jlong)) {
239 *(jlong*)(dst + off) = fill;
240 }
241 } else if (bits % sizeof(jint) == 0) {
242 jint fill = (juint)( (jubyte)value ); // zero-extend
243 if (fill != 0) {
244 fill += fill << 8;
245 fill += fill << 16;
246 }
247 //Copy::fill_to_jints_atomic((jint*) dst, size / sizeof(jint));
248 for (uintptr_t off = 0; off < size; off += sizeof(jint)) {
249 *(jint*)(dst + off) = fill;
250 }
251 } else if (bits % sizeof(jshort) == 0) {
252 jshort fill = (jushort)( (jubyte)value ); // zero-extend
253 fill += fill << 8;
254 //Copy::fill_to_jshorts_atomic((jshort*) dst, size / sizeof(jshort));
255 for (uintptr_t off = 0; off < size; off += sizeof(jshort)) {
256 *(jshort*)(dst + off) = fill;
257 }
258 } else {
259 // Not aligned, so no need to be atomic.
260 Copy::fill_to_bytes(dst, size, value);
261 }
262 }