OLD | NEW |
1 // Copyright 2011 the V8 project authors. All rights reserved. | 1 // Copyright 2011 the V8 project authors. All rights reserved. |
2 // Redistribution and use in source and binary forms, with or without | 2 // Redistribution and use in source and binary forms, with or without |
3 // modification, are permitted provided that the following conditions are | 3 // modification, are permitted provided that the following conditions are |
4 // met: | 4 // met: |
5 // | 5 // |
6 // * Redistributions of source code must retain the above copyright | 6 // * Redistributions of source code must retain the above copyright |
7 // notice, this list of conditions and the following disclaimer. | 7 // notice, this list of conditions and the following disclaimer. |
8 // * Redistributions in binary form must reproduce the above | 8 // * Redistributions in binary form must reproduce the above |
9 // copyright notice, this list of conditions and the following | 9 // copyright notice, this list of conditions and the following |
10 // disclaimer in the documentation and/or other materials provided | 10 // disclaimer in the documentation and/or other materials provided |
(...skipping 94 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
105 __ mov(dst, Operand(esp, stack_offset + kDestinationOffset)); | 105 __ mov(dst, Operand(esp, stack_offset + kDestinationOffset)); |
106 __ mov(src, Operand(esp, stack_offset + kSourceOffset)); | 106 __ mov(src, Operand(esp, stack_offset + kSourceOffset)); |
107 __ mov(count, Operand(esp, stack_offset + kSizeOffset)); | 107 __ mov(count, Operand(esp, stack_offset + kSizeOffset)); |
108 | 108 |
109 | 109 |
110 __ movdqu(xmm0, Operand(src, 0)); | 110 __ movdqu(xmm0, Operand(src, 0)); |
111 __ movdqu(Operand(dst, 0), xmm0); | 111 __ movdqu(Operand(dst, 0), xmm0); |
112 __ mov(edx, dst); | 112 __ mov(edx, dst); |
113 __ and_(edx, 0xF); | 113 __ and_(edx, 0xF); |
114 __ neg(edx); | 114 __ neg(edx); |
115 __ add(Operand(edx), Immediate(16)); | 115 __ add(edx, Immediate(16)); |
116 __ add(dst, Operand(edx)); | 116 __ add(dst, edx); |
117 __ add(src, Operand(edx)); | 117 __ add(src, edx); |
118 __ sub(Operand(count), edx); | 118 __ sub(count, edx); |
119 | 119 |
120 // edi is now aligned. Check if esi is also aligned. | 120 // edi is now aligned. Check if esi is also aligned. |
121 Label unaligned_source; | 121 Label unaligned_source; |
122 __ test(Operand(src), Immediate(0x0F)); | 122 __ test(src, Immediate(0x0F)); |
123 __ j(not_zero, &unaligned_source); | 123 __ j(not_zero, &unaligned_source); |
124 { | 124 { |
125 // Copy loop for aligned source and destination. | 125 // Copy loop for aligned source and destination. |
126 __ mov(edx, count); | 126 __ mov(edx, count); |
127 Register loop_count = ecx; | 127 Register loop_count = ecx; |
128 Register count = edx; | 128 Register count = edx; |
129 __ shr(loop_count, 5); | 129 __ shr(loop_count, 5); |
130 { | 130 { |
131 // Main copy loop. | 131 // Main copy loop. |
132 Label loop; | 132 Label loop; |
133 __ bind(&loop); | 133 __ bind(&loop); |
134 __ prefetch(Operand(src, 0x20), 1); | 134 __ prefetch(Operand(src, 0x20), 1); |
135 __ movdqa(xmm0, Operand(src, 0x00)); | 135 __ movdqa(xmm0, Operand(src, 0x00)); |
136 __ movdqa(xmm1, Operand(src, 0x10)); | 136 __ movdqa(xmm1, Operand(src, 0x10)); |
137 __ add(Operand(src), Immediate(0x20)); | 137 __ add(src, Immediate(0x20)); |
138 | 138 |
139 __ movdqa(Operand(dst, 0x00), xmm0); | 139 __ movdqa(Operand(dst, 0x00), xmm0); |
140 __ movdqa(Operand(dst, 0x10), xmm1); | 140 __ movdqa(Operand(dst, 0x10), xmm1); |
141 __ add(Operand(dst), Immediate(0x20)); | 141 __ add(dst, Immediate(0x20)); |
142 | 142 |
143 __ dec(loop_count); | 143 __ dec(loop_count); |
144 __ j(not_zero, &loop); | 144 __ j(not_zero, &loop); |
145 } | 145 } |
146 | 146 |
147 // At most 31 bytes to copy. | 147 // At most 31 bytes to copy. |
148 Label move_less_16; | 148 Label move_less_16; |
149 __ test(Operand(count), Immediate(0x10)); | 149 __ test(count, Immediate(0x10)); |
150 __ j(zero, &move_less_16); | 150 __ j(zero, &move_less_16); |
151 __ movdqa(xmm0, Operand(src, 0)); | 151 __ movdqa(xmm0, Operand(src, 0)); |
152 __ add(Operand(src), Immediate(0x10)); | 152 __ add(src, Immediate(0x10)); |
153 __ movdqa(Operand(dst, 0), xmm0); | 153 __ movdqa(Operand(dst, 0), xmm0); |
154 __ add(Operand(dst), Immediate(0x10)); | 154 __ add(dst, Immediate(0x10)); |
155 __ bind(&move_less_16); | 155 __ bind(&move_less_16); |
156 | 156 |
157 // At most 15 bytes to copy. Copy 16 bytes at end of string. | 157 // At most 15 bytes to copy. Copy 16 bytes at end of string. |
158 __ and_(count, 0xF); | 158 __ and_(count, 0xF); |
159 __ movdqu(xmm0, Operand(src, count, times_1, -0x10)); | 159 __ movdqu(xmm0, Operand(src, count, times_1, -0x10)); |
160 __ movdqu(Operand(dst, count, times_1, -0x10), xmm0); | 160 __ movdqu(Operand(dst, count, times_1, -0x10), xmm0); |
161 | 161 |
162 __ mov(eax, Operand(esp, stack_offset + kDestinationOffset)); | 162 __ mov(eax, Operand(esp, stack_offset + kDestinationOffset)); |
163 __ pop(esi); | 163 __ pop(esi); |
164 __ pop(edi); | 164 __ pop(edi); |
165 __ ret(0); | 165 __ ret(0); |
166 } | 166 } |
167 __ Align(16); | 167 __ Align(16); |
168 { | 168 { |
169 // Copy loop for unaligned source and aligned destination. | 169 // Copy loop for unaligned source and aligned destination. |
170 // If source is not aligned, we can't read it as efficiently. | 170 // If source is not aligned, we can't read it as efficiently. |
171 __ bind(&unaligned_source); | 171 __ bind(&unaligned_source); |
172 __ mov(edx, ecx); | 172 __ mov(edx, ecx); |
173 Register loop_count = ecx; | 173 Register loop_count = ecx; |
174 Register count = edx; | 174 Register count = edx; |
175 __ shr(loop_count, 5); | 175 __ shr(loop_count, 5); |
176 { | 176 { |
177 // Main copy loop | 177 // Main copy loop |
178 Label loop; | 178 Label loop; |
179 __ bind(&loop); | 179 __ bind(&loop); |
180 __ prefetch(Operand(src, 0x20), 1); | 180 __ prefetch(Operand(src, 0x20), 1); |
181 __ movdqu(xmm0, Operand(src, 0x00)); | 181 __ movdqu(xmm0, Operand(src, 0x00)); |
182 __ movdqu(xmm1, Operand(src, 0x10)); | 182 __ movdqu(xmm1, Operand(src, 0x10)); |
183 __ add(Operand(src), Immediate(0x20)); | 183 __ add(src, Immediate(0x20)); |
184 | 184 |
185 __ movdqa(Operand(dst, 0x00), xmm0); | 185 __ movdqa(Operand(dst, 0x00), xmm0); |
186 __ movdqa(Operand(dst, 0x10), xmm1); | 186 __ movdqa(Operand(dst, 0x10), xmm1); |
187 __ add(Operand(dst), Immediate(0x20)); | 187 __ add(dst, Immediate(0x20)); |
188 | 188 |
189 __ dec(loop_count); | 189 __ dec(loop_count); |
190 __ j(not_zero, &loop); | 190 __ j(not_zero, &loop); |
191 } | 191 } |
192 | 192 |
193 // At most 31 bytes to copy. | 193 // At most 31 bytes to copy. |
194 Label move_less_16; | 194 Label move_less_16; |
195 __ test(Operand(count), Immediate(0x10)); | 195 __ test(count, Immediate(0x10)); |
196 __ j(zero, &move_less_16); | 196 __ j(zero, &move_less_16); |
197 __ movdqu(xmm0, Operand(src, 0)); | 197 __ movdqu(xmm0, Operand(src, 0)); |
198 __ add(Operand(src), Immediate(0x10)); | 198 __ add(src, Immediate(0x10)); |
199 __ movdqa(Operand(dst, 0), xmm0); | 199 __ movdqa(Operand(dst, 0), xmm0); |
200 __ add(Operand(dst), Immediate(0x10)); | 200 __ add(dst, Immediate(0x10)); |
201 __ bind(&move_less_16); | 201 __ bind(&move_less_16); |
202 | 202 |
203 // At most 15 bytes to copy. Copy 16 bytes at end of string. | 203 // At most 15 bytes to copy. Copy 16 bytes at end of string. |
204 __ and_(count, 0x0F); | 204 __ and_(count, 0x0F); |
205 __ movdqu(xmm0, Operand(src, count, times_1, -0x10)); | 205 __ movdqu(xmm0, Operand(src, count, times_1, -0x10)); |
206 __ movdqu(Operand(dst, count, times_1, -0x10), xmm0); | 206 __ movdqu(Operand(dst, count, times_1, -0x10), xmm0); |
207 | 207 |
208 __ mov(eax, Operand(esp, stack_offset + kDestinationOffset)); | 208 __ mov(eax, Operand(esp, stack_offset + kDestinationOffset)); |
209 __ pop(esi); | 209 __ pop(esi); |
210 __ pop(edi); | 210 __ pop(edi); |
(...skipping 14 matching lines...) Expand all Loading... |
225 __ mov(count, Operand(esp, stack_offset + kSizeOffset)); | 225 __ mov(count, Operand(esp, stack_offset + kSizeOffset)); |
226 | 226 |
227 // Copy the first word. | 227 // Copy the first word. |
228 __ mov(eax, Operand(src, 0)); | 228 __ mov(eax, Operand(src, 0)); |
229 __ mov(Operand(dst, 0), eax); | 229 __ mov(Operand(dst, 0), eax); |
230 | 230 |
231 // Increment src,dstso that dst is aligned. | 231 // Increment src,dstso that dst is aligned. |
232 __ mov(edx, dst); | 232 __ mov(edx, dst); |
233 __ and_(edx, 0x03); | 233 __ and_(edx, 0x03); |
234 __ neg(edx); | 234 __ neg(edx); |
235 __ add(Operand(edx), Immediate(4)); // edx = 4 - (dst & 3) | 235 __ add(edx, Immediate(4)); // edx = 4 - (dst & 3) |
236 __ add(dst, Operand(edx)); | 236 __ add(dst, edx); |
237 __ add(src, Operand(edx)); | 237 __ add(src, edx); |
238 __ sub(Operand(count), edx); | 238 __ sub(count, edx); |
239 // edi is now aligned, ecx holds number of remaning bytes to copy. | 239 // edi is now aligned, ecx holds number of remaning bytes to copy. |
240 | 240 |
241 __ mov(edx, count); | 241 __ mov(edx, count); |
242 count = edx; | 242 count = edx; |
243 __ shr(ecx, 2); // Make word count instead of byte count. | 243 __ shr(ecx, 2); // Make word count instead of byte count. |
244 __ rep_movs(); | 244 __ rep_movs(); |
245 | 245 |
246 // At most 3 bytes left to copy. Copy 4 bytes at end of string. | 246 // At most 3 bytes left to copy. Copy 4 bytes at end of string. |
247 __ and_(count, 3); | 247 __ and_(count, 3); |
248 __ mov(eax, Operand(src, count, times_1, -4)); | 248 __ mov(eax, Operand(src, count, times_1, -4)); |
(...skipping 12 matching lines...) Expand all Loading... |
261 CPU::FlushICache(buffer, actual_size); | 261 CPU::FlushICache(buffer, actual_size); |
262 OS::ProtectCode(buffer, actual_size); | 262 OS::ProtectCode(buffer, actual_size); |
263 return FUNCTION_CAST<OS::MemCopyFunction>(buffer); | 263 return FUNCTION_CAST<OS::MemCopyFunction>(buffer); |
264 } | 264 } |
265 | 265 |
266 #undef __ | 266 #undef __ |
267 | 267 |
268 } } // namespace v8::internal | 268 } } // namespace v8::internal |
269 | 269 |
270 #endif // V8_TARGET_ARCH_IA32 | 270 #endif // V8_TARGET_ARCH_IA32 |
OLD | NEW |