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 optionally byte swap elements
60 *
61 * <swap> - true if elements should be byte swapped
62 *
63 * @param src address of source
64 * @param dst address of destination
65 * @param byte_count number of bytes to copy
66 * @param elem_size size of the elements to copy-swap
67 */
68 template<bool swap>
69 static void conjoint_swap_if_needed(const void* src, void* dst, size_t byte_count, size_t elem_size) {
70 assert(src != NULL, "address must not be NULL");
71 assert(dst != NULL, "address must not be NULL");
72 assert(elem_size == 2 || elem_size == 4 || elem_size == 8,
73 "incorrect element size: " SIZE_FORMAT, elem_size);
74 assert(is_aligned(byte_count, elem_size),
75 "byte_count " SIZE_FORMAT " must be multiple of element size " SIZE_FORMAT, byte_count, elem_size);
76
77 address src_end = (address)src + byte_count;
78
79 if (dst <= src || dst >= src_end) {
80 do_conjoint_swap<RIGHT,swap>(src, dst, byte_count, elem_size);
81 } else {
82 do_conjoint_swap<LEFT,swap>(src, dst, byte_count, elem_size);
83 }
84 }
85
86 private:
87 /**
88 * Byte swap a 16-bit value
89 */
90 static uint16_t byte_swap(uint16_t x) {
91 return (x << 8) | (x >> 8);
92 }
93
94 /**
95 * Byte swap a 32-bit value
96 */
97 static uint32_t byte_swap(uint32_t x) {
98 uint16_t lo = (uint16_t)x;
99 uint16_t hi = (uint16_t)(x >> 16);
100
101 return ((uint32_t)byte_swap(lo) << 16) | (uint32_t)byte_swap(hi);
102 }
103
104 /**
105 * Byte swap a 64-bit value
106 */
107 static uint64_t byte_swap(uint64_t x) {
108 uint32_t lo = (uint32_t)x;
109 uint32_t hi = (uint32_t)(x >> 32);
110
111 return ((uint64_t)byte_swap(lo) << 32) | (uint64_t)byte_swap(hi);
112 }
113
114 enum CopyDirection {
115 RIGHT, // lower -> higher address
116 LEFT // higher -> lower address
117 };
118
119 /**
120 * Copy and byte swap elements
121 *
122 * <T> - type of element to copy
123 * <D> - copy direction
124 * <is_src_aligned> - true if src argument is aligned to element size
125 * <is_dst_aligned> - true if dst argument is aligned to element size
126 *
127 * @param src address of source
128 * @param dst address of destination
129 * @param byte_count number of bytes to copy
130 */
131 template <typename T, CopyDirection D, bool swap, bool is_src_aligned, bool is_dst_aligned>
132 static void do_conjoint_swap(const void* src, void* dst, size_t byte_count) {
133 const char* cur_src;
134 char* cur_dst;
135
136 switch (D) {
137 case RIGHT:
138 cur_src = (const char*)src;
139 cur_dst = (char*)dst;
140 break;
141 case LEFT:
142 cur_src = (const char*)src + byte_count - sizeof(T);
143 cur_dst = (char*)dst + byte_count - sizeof(T);
144 break;
145 }
146
147 for (size_t i = 0; i < byte_count / sizeof(T); i++) {
148 T tmp;
149
150 if (is_src_aligned) {
151 tmp = *(T*)cur_src;
152 } else {
153 memcpy(&tmp, cur_src, sizeof(T));
154 }
155
156 if (swap) {
157 tmp = byte_swap(tmp);
158 }
159
160 if (is_dst_aligned) {
161 *(T*)cur_dst = tmp;
162 } else {
163 memcpy(cur_dst, &tmp, sizeof(T));
164 }
165
166 switch (D) {
167 case RIGHT:
168 cur_src += sizeof(T);
169 cur_dst += sizeof(T);
170 break;
171 case LEFT:
172 cur_src -= sizeof(T);
173 cur_dst -= sizeof(T);
174 break;
175 }
176 }
177 }
178
179 /**
180 * Copy and byte swap elements
181 *
182 * <T> - type of element to copy
183 * <D> - copy direction
184 * <swap> - true if elements should be byte swapped
185 *
186 * @param src address of source
187 * @param dst address of destination
188 * @param byte_count number of bytes to copy
189 */
190 template <typename T, CopyDirection direction, bool swap>
191 static void do_conjoint_swap(const void* src, void* dst, size_t byte_count) {
192 if (is_aligned(src, sizeof(T))) {
193 if (is_aligned(dst, sizeof(T))) {
194 do_conjoint_swap<T,direction,swap,true,true>(src, dst, byte_count);
195 } else {
196 do_conjoint_swap<T,direction,swap,true,false>(src, dst, byte_count);
197 }
198 } else {
199 if (is_aligned(dst, sizeof(T))) {
200 do_conjoint_swap<T,direction,swap,false,true>(src, dst, byte_count);
201 } else {
202 do_conjoint_swap<T,direction,swap,false,false>(src, dst, byte_count);
203 }
204 }
205 }
206
207
208 /**
209 * Copy and byte swap elements
210 *
211 * <D> - copy direction
212 * <swap> - true if elements should be byte swapped
213 *
214 * @param src address of source
215 * @param dst address of destination
216 * @param byte_count number of bytes to copy
217 * @param elem_size size of the elements to copy-swap
218 */
219 template <CopyDirection D, bool swap>
220 static void do_conjoint_swap(const void* src, void* dst, size_t byte_count, size_t elem_size) {
221 switch (elem_size) {
222 case 2: do_conjoint_swap<uint16_t,D,swap>(src, dst, byte_count); break;
223 case 4: do_conjoint_swap<uint32_t,D,swap>(src, dst, byte_count); break;
224 case 8: do_conjoint_swap<uint64_t,D,swap>(src, dst, byte_count); break;
225 default: guarantee(false, "do_conjoint_swap: Invalid elem_size " SIZE_FORMAT "\n", elem_size);
226 }
227 }
228 };
229
230 void Copy::conjoint_copy(const void* src, void* dst, size_t byte_count, size_t elem_size) {
231 CopySwap::conjoint_swap_if_needed<false>(src, dst, byte_count, elem_size);
232 }
233
234 void Copy::conjoint_swap(const void* src, void* dst, size_t byte_count, size_t elem_size) {
235 CopySwap::conjoint_swap_if_needed<true>(src, dst, byte_count, elem_size);
236 }
237
238 // Fill bytes; larger units are filled atomically if everything is aligned.
239 void Copy::fill_to_memory_atomic(void* to, size_t size, jubyte value) {
240 address dst = (address) to;
241 uintptr_t bits = (uintptr_t) to | (uintptr_t) size;
242 if (bits % sizeof(jlong) == 0) {
243 jlong fill = (julong)( (jubyte)value ); // zero-extend
244 if (fill != 0) {
245 fill += fill << 8;
246 fill += fill << 16;
247 fill += fill << 32;
248 }
249 //Copy::fill_to_jlongs_atomic((jlong*) dst, size / sizeof(jlong));
250 for (uintptr_t off = 0; off < size; off += sizeof(jlong)) {
251 *(jlong*)(dst + off) = fill;
252 }
253 } else if (bits % sizeof(jint) == 0) {
254 jint fill = (juint)( (jubyte)value ); // zero-extend
255 if (fill != 0) {
256 fill += fill << 8;
257 fill += fill << 16;
258 }
259 //Copy::fill_to_jints_atomic((jint*) dst, size / sizeof(jint));
260 for (uintptr_t off = 0; off < size; off += sizeof(jint)) {
261 *(jint*)(dst + off) = fill;
262 }
263 } else if (bits % sizeof(jshort) == 0) {
264 jshort fill = (jushort)( (jubyte)value ); // zero-extend
265 fill += fill << 8;
266 //Copy::fill_to_jshorts_atomic((jshort*) dst, size / sizeof(jshort));
267 for (uintptr_t off = 0; off < size; off += sizeof(jshort)) {
268 *(jshort*)(dst + off) = fill;
269 }
270 } else {
271 // Not aligned, so no need to be atomic.
272 Copy::fill_to_bytes(dst, size, value);
273 }
274 }