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 address src_end = src + byte_count;
68
69 if (dst <= src || dst >= src_end) {
70 do_conjoint_swap<RIGHT>(src, dst, byte_count, elem_size);
71 } else {
72 do_conjoint_swap<LEFT>(src, dst, byte_count, elem_size);
73 }
74 }
75
76 private:
77 /**
78 * Byte swap a 16-bit value
79 */
80 static uint16_t byte_swap(uint16_t x) {
81 return (x << 8) | (x >> 8);
82 }
83
84 /**
85 * Byte swap a 32-bit value
86 */
87 static uint32_t byte_swap(uint32_t x) {
88 uint16_t lo = (uint16_t)x;
89 uint16_t hi = (uint16_t)(x >> 16);
90
91 return ((uint32_t)byte_swap(lo) << 16) | (uint32_t)byte_swap(hi);
92 }
93
94 /**
95 * Byte swap a 64-bit value
96 */
97 static uint64_t byte_swap(uint64_t x) {
98 uint32_t lo = (uint32_t)x;
99 uint32_t hi = (uint32_t)(x >> 32);
100
101 return ((uint64_t)byte_swap(lo) << 32) | (uint64_t)byte_swap(hi);
102 }
103
104 enum CopyDirection {
105 RIGHT, // lower -> higher address
106 LEFT // higher -> lower address
107 };
108
109 /**
110 * Copy and byte swap elements
111 *
112 * <T> - type of element to copy
113 * <D> - copy direction
114 * <is_src_aligned> - true if src argument is aligned to element size
115 * <is_dst_aligned> - true if dst argument is aligned to element size
116 *
117 * @param src address of source
118 * @param dst address of destination
119 * @param byte_count number of bytes to copy
120 */
121 template <typename T, CopyDirection D, bool is_src_aligned, bool is_dst_aligned>
122 static void do_conjoint_swap(address src, address dst, size_t byte_count) {
123 address cur_src, cur_dst;
124
125 switch (D) {
126 case RIGHT:
127 cur_src = src;
128 cur_dst = dst;
129 break;
130 case LEFT:
131 cur_src = src + byte_count - sizeof(T);
132 cur_dst = dst + byte_count - sizeof(T);
133 break;
134 }
135
136 for (size_t i = 0; i < byte_count / sizeof(T); i++) {
137 T tmp;
138
139 if (is_src_aligned) {
140 tmp = *(T*)cur_src;
141 } else {
142 memcpy(&tmp, cur_src, sizeof(T));
143 }
144
145 tmp = byte_swap(tmp);
146
147 if (is_dst_aligned) {
148 *(T*)cur_dst = tmp;
149 } else {
150 memcpy(cur_dst, &tmp, sizeof(T));
151 }
152
153 switch (D) {
154 case RIGHT:
155 cur_src += sizeof(T);
156 cur_dst += sizeof(T);
157 break;
158 case LEFT:
159 cur_src -= sizeof(T);
160 cur_dst -= sizeof(T);
161 break;
162 }
163 }
164 }
165
166 /**
167 * Copy and byte swap elements
168 *
169 * <T> - type of element to copy
170 * <D> - copy direction
171 *
172 * @param src address of source
173 * @param dst address of destination
174 * @param byte_count number of bytes to copy
175 */
176 template <typename T, CopyDirection direction>
177 static void do_conjoint_swap(address src, address dst, size_t byte_count) {
178 if (is_ptr_aligned(src, sizeof(T))) {
179 if (is_ptr_aligned(dst, sizeof(T))) {
180 do_conjoint_swap<T,direction,true,true>(src, dst, byte_count);
181 } else {
182 do_conjoint_swap<T,direction,true,false>(src, dst, byte_count);
183 }
184 } else {
185 if (is_ptr_aligned(dst, sizeof(T))) {
186 do_conjoint_swap<T,direction,false,true>(src, dst, byte_count);
187 } else {
188 do_conjoint_swap<T,direction,false,false>(src, dst, byte_count);
189 }
190 }
191 }
192
193
194 /**
195 * Copy and byte swap elements
196 *
197 * <D> - copy direction
198 *
199 * @param src address of source
200 * @param dst address of destination
201 * @param byte_count number of bytes to copy
202 * @param elem_size size of the elements to copy-swap
203 */
204 template <CopyDirection D>
205 static void do_conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
206 switch (elem_size) {
207 case 2: do_conjoint_swap<uint16_t,D>(src, dst, byte_count); break;
208 case 4: do_conjoint_swap<uint32_t,D>(src, dst, byte_count); break;
209 case 8: do_conjoint_swap<uint64_t,D>(src, dst, byte_count); break;
210 default: guarantee(false, "do_conjoint_swap: Invalid elem_size %zd\n", elem_size);
211 }
212 }
213 };
214
215 void Copy::conjoint_swap(address src, address dst, size_t byte_count, size_t elem_size) {
216 CopySwap::conjoint_swap(src, dst, byte_count, elem_size);
217 }
218
219 // Fill bytes; larger units are filled atomically if everything is aligned.
220 void Copy::fill_to_memory_atomic(void* to, size_t size, jubyte value) {
221 address dst = (address) to;
222 uintptr_t bits = (uintptr_t) to | (uintptr_t) size;
223 if (bits % sizeof(jlong) == 0) {
224 jlong fill = (julong)( (jubyte)value ); // zero-extend
225 if (fill != 0) {
226 fill += fill << 8;
227 fill += fill << 16;
228 fill += fill << 32;
229 }
230 //Copy::fill_to_jlongs_atomic((jlong*) dst, size / sizeof(jlong));
231 for (uintptr_t off = 0; off < size; off += sizeof(jlong)) {
232 *(jlong*)(dst + off) = fill;
233 }
234 } else if (bits % sizeof(jint) == 0) {
235 jint fill = (juint)( (jubyte)value ); // zero-extend
236 if (fill != 0) {
237 fill += fill << 8;
238 fill += fill << 16;
239 }
240 //Copy::fill_to_jints_atomic((jint*) dst, size / sizeof(jint));
241 for (uintptr_t off = 0; off < size; off += sizeof(jint)) {
242 *(jint*)(dst + off) = fill;
243 }
244 } else if (bits % sizeof(jshort) == 0) {
245 jshort fill = (jushort)( (jubyte)value ); // zero-extend
246 fill += fill << 8;
247 //Copy::fill_to_jshorts_atomic((jshort*) dst, size / sizeof(jshort));
248 for (uintptr_t off = 0; off < size; off += sizeof(jshort)) {
249 *(jshort*)(dst + off) = fill;
250 }
251 } else {
252 // Not aligned, so no need to be atomic.
253 Copy::fill_to_bytes(dst, size, value);
254 }
255 }