lxdream 0.9.1| filename | src/sh4/ia32abi.h |
| changeset | 596:dfc0c93d882e |
| prev | 590:4db6a084ca3c |
| next | 604:1024c3a9cb88 |
| author | nkeynes |
| date | Tue Jan 22 10:11:45 2008 +0000 (16 years ago) |
| permissions | -rw-r--r-- |
| last change | Invoke emulator single-step for untranslatable delay slots (and fix a few related bugs) |
| view | annotate | diff | log | raw |
1 /**
2 * $Id$
3 *
4 * Provides the implementation for the ia32 ABI (eg prologue, epilogue, and
5 * calling conventions)
6 *
7 * Copyright (c) 2007 Nathan Keynes.
8 *
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
18 */
20 #ifndef __lxdream_ia32abi_H
21 #define __lxdream_ia32abi_H 1
23 #define load_ptr( reg, ptr ) load_imm32( reg, (uint32_t)ptr );
25 /**
26 * Note: clobbers EAX to make the indirect call - this isn't usually
27 * a problem since the callee will usually clobber it anyway.
28 */
29 #define CALL_FUNC0_SIZE 7
30 static inline void call_func0( void *ptr )
31 {
32 load_imm32(R_EAX, (uint32_t)ptr);
33 CALL_r32(R_EAX);
34 }
36 #define CALL_FUNC1_SIZE 11
37 static inline void call_func1( void *ptr, int arg1 )
38 {
39 PUSH_r32(arg1);
40 call_func0(ptr);
41 ADD_imm8s_r32( 4, R_ESP );
42 }
44 #define CALL_FUNC2_SIZE 12
45 static inline void call_func2( void *ptr, int arg1, int arg2 )
46 {
47 PUSH_r32(arg2);
48 PUSH_r32(arg1);
49 call_func0(ptr);
50 ADD_imm8s_r32( 8, R_ESP );
51 }
53 /**
54 * Write a double (64-bit) value into memory, with the first word in arg2a, and
55 * the second in arg2b
56 * NB: 30 bytes
57 */
58 #define MEM_WRITE_DOUBLE_SIZE 30
59 static inline void MEM_WRITE_DOUBLE( int addr, int arg2a, int arg2b )
60 {
61 ADD_imm8s_r32( 4, addr );
62 PUSH_r32(arg2b);
63 PUSH_r32(addr);
64 ADD_imm8s_r32( -4, addr );
65 PUSH_r32(arg2a);
66 PUSH_r32(addr);
67 call_func0(sh4_write_long);
68 ADD_imm8s_r32( 8, R_ESP );
69 call_func0(sh4_write_long);
70 ADD_imm8s_r32( 8, R_ESP );
71 }
73 /**
74 * Read a double (64-bit) value from memory, writing the first word into arg2a
75 * and the second into arg2b. The addr must not be in EAX
76 * NB: 27 bytes
77 */
78 #define MEM_READ_DOUBLE_SIZE 27
79 static inline void MEM_READ_DOUBLE( int addr, int arg2a, int arg2b )
80 {
81 PUSH_r32(addr);
82 call_func0(sh4_read_long);
83 POP_r32(R_ECX);
84 PUSH_r32(R_EAX);
85 ADD_imm8s_r32( 4, R_ECX );
86 PUSH_r32(R_ECX);
87 call_func0(sh4_read_long);
88 ADD_imm8s_r32( 4, R_ESP );
89 MOV_r32_r32( R_EAX, arg2b );
90 POP_r32(arg2a);
91 }
93 /**
94 * Emit the 'start of block' assembly. Sets up the stack frame and save
95 * SI/DI as required
96 */
97 void sh4_translate_begin_block( sh4addr_t pc )
98 {
99 PUSH_r32(R_EBP);
100 /* mov &sh4r, ebp */
101 load_ptr( R_EBP, &sh4r );
103 sh4_x86.in_delay_slot = FALSE;
104 sh4_x86.priv_checked = FALSE;
105 sh4_x86.fpuen_checked = FALSE;
106 sh4_x86.branch_taken = FALSE;
107 sh4_x86.backpatch_posn = 0;
108 sh4_x86.block_start_pc = pc;
109 sh4_x86.tlb_on = IS_MMU_ENABLED();
110 sh4_x86.tstate = TSTATE_NONE;
111 #ifdef STACK_ALIGN
112 sh4_x86.stack_posn = 8;
113 #endif
114 }
116 /**
117 * Exit the block with sh4r.pc already written
118 */
119 void exit_block_pcset( sh4addr_t pc )
120 {
121 load_imm32( R_ECX, ((pc - sh4_x86.block_start_pc)>>1)*sh4_cpu_period ); // 5
122 ADD_r32_sh4r( R_ECX, REG_OFFSET(slice_cycle) ); // 6
123 load_spreg( R_EAX, R_PC );
124 if( sh4_x86.tlb_on ) {
125 call_func1(xlat_get_code_by_vma,R_EAX);
126 } else {
127 call_func1(xlat_get_code,R_EAX);
128 }
129 POP_r32(R_EBP);
130 RET();
131 }
133 /**
134 * Exit the block with sh4r.new_pc written with the target pc
135 */
136 void exit_block_newpcset( sh4addr_t pc )
137 {
138 load_imm32( R_ECX, ((pc - sh4_x86.block_start_pc)>>1)*sh4_cpu_period ); // 5
139 ADD_r32_sh4r( R_ECX, REG_OFFSET(slice_cycle) ); // 6
140 load_spreg( R_EAX, R_NEW_PC );
141 store_spreg( R_EAX, R_PC );
142 if( sh4_x86.tlb_on ) {
143 call_func1(xlat_get_code_by_vma,R_EAX);
144 } else {
145 call_func1(xlat_get_code,R_EAX);
146 }
147 POP_r32(R_EBP);
148 RET();
149 }
151 #define EXIT_BLOCK_SIZE(pc) (24 + (IS_IN_ICACHE(pc)?5:CALL_FUNC1_SIZE))
154 /**
155 * Exit the block to an absolute PC
156 */
157 void exit_block( sh4addr_t pc, sh4addr_t endpc )
158 {
159 load_imm32( R_ECX, pc ); // 5
160 store_spreg( R_ECX, REG_OFFSET(pc) ); // 3
161 if( IS_IN_ICACHE(pc) ) {
162 MOV_moff32_EAX( xlat_get_lut_entry(GET_ICACHE_PHYS(pc)) ); // 5
163 } else if( sh4_x86.tlb_on ) {
164 call_func1(xlat_get_code_by_vma,R_ECX);
165 } else {
166 call_func1(xlat_get_code,R_ECX);
167 }
168 AND_imm8s_r32( 0xFC, R_EAX ); // 3
169 load_imm32( R_ECX, ((endpc - sh4_x86.block_start_pc)>>1)*sh4_cpu_period ); // 5
170 ADD_r32_sh4r( R_ECX, REG_OFFSET(slice_cycle) ); // 6
171 POP_r32(R_EBP);
172 RET();
173 }
175 #define EXIT_BLOCK_REL_SIZE(pc) (27 + (IS_IN_ICACHE(pc)?5:CALL_FUNC1_SIZE))
177 /**
178 * Exit the block to a relative PC
179 */
180 void exit_block_rel( sh4addr_t pc, sh4addr_t endpc )
181 {
182 load_imm32( R_ECX, pc - sh4_x86.block_start_pc ); // 5
183 ADD_sh4r_r32( R_PC, R_ECX );
184 store_spreg( R_ECX, REG_OFFSET(pc) ); // 3
185 if( IS_IN_ICACHE(pc) ) {
186 MOV_moff32_EAX( xlat_get_lut_entry(GET_ICACHE_PHYS(pc)) ); // 5
187 } else if( sh4_x86.tlb_on ) {
188 call_func1(xlat_get_code_by_vma,R_ECX);
189 } else {
190 call_func1(xlat_get_code,R_ECX);
191 }
192 AND_imm8s_r32( 0xFC, R_EAX ); // 3
193 load_imm32( R_ECX, ((endpc - sh4_x86.block_start_pc)>>1)*sh4_cpu_period ); // 5
194 ADD_r32_sh4r( R_ECX, REG_OFFSET(slice_cycle) ); // 6
195 POP_r32(R_EBP);
196 RET();
197 }
199 /**
200 * Write the block trailer (exception handling block)
201 */
202 void sh4_translate_end_block( sh4addr_t pc ) {
203 if( sh4_x86.branch_taken == FALSE ) {
204 // Didn't exit unconditionally already, so write the termination here
205 exit_block_rel( pc, pc );
206 }
207 if( sh4_x86.backpatch_posn != 0 ) {
208 unsigned int i;
209 // Raise exception
210 uint8_t *end_ptr = xlat_output;
211 MOV_r32_r32( R_EDX, R_ECX );
212 ADD_r32_r32( R_EDX, R_ECX );
213 ADD_r32_sh4r( R_ECX, R_PC );
214 MOV_moff32_EAX( &sh4_cpu_period );
215 MUL_r32( R_EDX );
216 ADD_r32_sh4r( R_EAX, REG_OFFSET(slice_cycle) );
218 call_func0( sh4_raise_exception );
219 ADD_imm8s_r32( 4, R_ESP );
220 load_spreg( R_EAX, R_PC );
221 if( sh4_x86.tlb_on ) {
222 call_func1(xlat_get_code_by_vma,R_EAX);
223 } else {
224 call_func1(xlat_get_code,R_EAX);
225 }
226 POP_r32(R_EBP);
227 RET();
229 // Exception already raised - just cleanup
230 uint8_t *preexc_ptr = xlat_output;
231 MOV_r32_r32( R_EDX, R_ECX );
232 ADD_r32_r32( R_EDX, R_ECX );
233 ADD_r32_sh4r( R_ECX, R_SPC );
234 MOV_moff32_EAX( &sh4_cpu_period );
235 MUL_r32( R_EDX );
236 ADD_r32_sh4r( R_EAX, REG_OFFSET(slice_cycle) );
237 load_spreg( R_EAX, R_PC );
238 if( sh4_x86.tlb_on ) {
239 call_func1(xlat_get_code_by_vma,R_EAX);
240 } else {
241 call_func1(xlat_get_code,R_EAX);
242 }
243 POP_r32(R_EBP);
244 RET();
246 for( i=0; i< sh4_x86.backpatch_posn; i++ ) {
247 *sh4_x86.backpatch_list[i].fixup_addr =
248 xlat_output - ((uint8_t *)sh4_x86.backpatch_list[i].fixup_addr) - 4;
249 if( sh4_x86.backpatch_list[i].exc_code < 0 ) {
250 load_imm32( R_EDX, sh4_x86.backpatch_list[i].fixup_icount );
251 int stack_adj = -1 - sh4_x86.backpatch_list[i].exc_code;
252 if( stack_adj > 0 ) {
253 ADD_imm8s_r32( stack_adj, R_ESP );
254 }
255 int rel = preexc_ptr - xlat_output;
256 JMP_rel(rel);
257 } else {
258 PUSH_imm32( sh4_x86.backpatch_list[i].exc_code );
259 load_imm32( R_EDX, sh4_x86.backpatch_list[i].fixup_icount );
260 int rel = end_ptr - xlat_output;
261 JMP_rel(rel);
262 }
263 }
264 }
265 }
267 void *xlat_get_native_pc()
268 {
269 void *result = NULL;
270 asm(
271 "mov %%ebp, %%eax\n\t"
272 "mov $0x8, %%ecx\n\t"
273 "mov %1, %%edx\n"
274 "frame_loop: test %%eax, %%eax\n\t"
275 "je frame_not_found\n\t"
276 "cmp (%%eax), %%edx\n\t"
277 "je frame_found\n\t"
278 "sub $0x1, %%ecx\n\t"
279 "je frame_not_found\n\t"
280 "movl (%%eax), %%eax\n\t"
281 "jmp frame_loop\n"
282 "frame_found: movl 0x4(%%eax), %0\n"
283 "frame_not_found:"
284 : "=r" (result)
285 : "r" (&sh4r)
286 : "eax", "ecx", "edx" );
287 return result;
288 }
290 #endif
.