4 * SH4 translation core module. This part handles the non-target-specific
5 * section of the translation.
7 * Copyright (c) 2005 Nathan Keynes.
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.
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.
24 #include "dreamcast.h"
25 #include "sh4/sh4core.h"
26 #include "sh4/sh4trans.h"
27 #include "sh4/xltcache.h"
30 static jmp_buf xlat_jmp_buf;
31 static gboolean xlat_running = FALSE;
33 gboolean sh4_xlat_is_running()
39 * Execute a timeslice using translated code only (ie translate/execute loop)
41 uint32_t sh4_xlat_run_slice( uint32_t nanosecs )
45 if( sh4r.sh4_state != SH4_STATE_RUNNING ) {
46 sh4_sleep_run_slice(nanosecs);
49 switch( setjmp(xlat_jmp_buf) ) {
50 case XLAT_EXIT_BREAKPOINT:
51 sh4_clear_breakpoint( sh4r.pc, BREAK_ONESHOT );
54 if( sh4r.sh4_state != SH4_STATE_STANDBY ) {
55 TMU_run_slice( sh4r.slice_cycle );
56 SCIF_run_slice( sh4r.slice_cycle );
58 return sh4r.slice_cycle;
60 case XLAT_EXIT_SYSRESET:
64 sh4_sleep_run_slice(nanosecs);
69 void * (*code)() = NULL;
70 while( sh4r.slice_cycle < nanosecs ) {
71 if( sh4r.event_pending <= sh4r.slice_cycle ) {
72 if( sh4r.event_types & PENDING_EVENT ) {
75 /* Eventq execute may (quite likely) deliver an immediate IRQ */
76 if( sh4r.event_types & PENDING_IRQ ) {
77 sh4_accept_interrupt();
83 if( sh4r.pc > 0xFFFFFF00 ) {
84 syscall_invoke( sh4r.pc );
85 sh4r.in_delay_slot = 0;
89 code = xlat_get_code_by_vma( sh4r.pc );
91 code = sh4_translate_basic_block( sh4r.pc );
99 sh4r.slice_cycle = nanosecs;
100 if( sh4r.sh4_state != SH4_STATE_STANDBY ) {
101 TMU_run_slice( nanosecs );
102 SCIF_run_slice( nanosecs );
107 uint8_t *xlat_output;
108 xlat_cache_block_t xlat_current_block;
109 struct xlat_recovery_record xlat_recovery[MAX_RECOVERY_SIZE];
110 uint32_t xlat_recovery_posn;
112 void sh4_translate_add_recovery( uint32_t icount )
114 xlat_recovery[xlat_recovery_posn].xlat_offset =
115 ((uintptr_t)xlat_output) - ((uintptr_t)xlat_current_block->code);
116 xlat_recovery[xlat_recovery_posn].sh4_icount = icount;
117 xlat_recovery_posn++;
121 * Translate a linear basic block, ie all instructions from the start address
122 * (inclusive) until the next branch/jump instruction or the end of the page
124 * @return the address of the translated block
125 * eg due to lack of buffer space.
127 void * sh4_translate_basic_block( sh4addr_t start )
129 sh4addr_t pc = start;
130 sh4addr_t lastpc = (pc&0xFFFFF000)+0x1000;
132 xlat_current_block = xlat_start_block( start );
133 xlat_output = (uint8_t *)xlat_current_block->code;
134 xlat_recovery_posn = 0;
135 uint8_t *eob = xlat_output + xlat_current_block->size;
137 if( GET_ICACHE_END() < lastpc ) {
138 lastpc = GET_ICACHE_END();
141 sh4_translate_begin_block(pc);
144 /* check for breakpoints at this pc */
145 for( i=0; i<sh4_breakpoint_count; i++ ) {
146 if( sh4_breakpoints[i].address == pc ) {
147 sh4_translate_emit_breakpoint(pc);
151 if( eob - xlat_output < MAX_INSTRUCTION_SIZE ) {
152 uint8_t *oldstart = xlat_current_block->code;
153 xlat_current_block = xlat_extend_block( xlat_output - oldstart + MAX_INSTRUCTION_SIZE );
154 xlat_output = xlat_current_block->code + (xlat_output - oldstart);
155 eob = xlat_current_block->code + xlat_current_block->size;
157 done = sh4_translate_instruction( pc );
158 assert( xlat_output <= eob );
160 if ( pc >= lastpc ) {
166 // Add end-of-block recovery for post-instruction checks
167 sh4_translate_add_recovery( (pc - start)>>1 );
169 int epilogue_size = sh4_translate_end_block_size();
170 uint32_t recovery_size = sizeof(struct xlat_recovery_record)*xlat_recovery_posn;
171 uint32_t finalsize = (xlat_output - xlat_current_block->code) + epilogue_size + recovery_size;
172 if( xlat_current_block->size < finalsize ) {
173 uint8_t *oldstart = xlat_current_block->code;
174 xlat_current_block = xlat_extend_block( finalsize );
175 xlat_output = xlat_current_block->code + (xlat_output - oldstart);
177 sh4_translate_end_block(pc);
178 assert( xlat_output <= (xlat_current_block->code + xlat_current_block->size - recovery_size) );
180 /* Write the recovery records onto the end of the code block */
181 memcpy( xlat_output, xlat_recovery, recovery_size);
182 xlat_current_block->recover_table_offset = xlat_output - (uint8_t *)xlat_current_block->code;
183 xlat_current_block->recover_table_size = xlat_recovery_posn;
184 xlat_commit_block( finalsize, pc-start );
185 return xlat_current_block->code;
189 * "Execute" the supplied recovery record. Currently this only updates
190 * sh4r.pc and sh4r.slice_cycle according to the currently executing
191 * instruction. In future this may be more sophisticated (ie will
192 * call into generated code).
194 void sh4_translate_run_recovery( xlat_recovery_record_t recovery )
196 sh4r.slice_cycle += (recovery->sh4_icount * sh4_cpu_period);
197 sh4r.pc += (recovery->sh4_icount<<1);
200 void sh4_translate_unwind_stack( gboolean abort_after, unwind_thunk_t thunk )
202 void *pc = xlat_get_native_pc();
204 assert( pc != NULL );
205 void *code = xlat_get_code( sh4r.pc );
206 xlat_recovery_record_t recover = xlat_get_recovery(code, pc, TRUE);
207 if( recover != NULL ) {
208 // Can be null if there is no recovery necessary
209 sh4_translate_run_recovery(recover);
211 if( thunk != NULL ) {
214 // finally longjmp back into sh4_xlat_run_slice
215 xlat_running = FALSE;
216 longjmp(xlat_jmp_buf, XLAT_EXIT_CONTINUE);
219 void sh4_translate_exit( int exit_code )
221 void *pc = xlat_get_native_pc();
223 // could be null if we're not actually running inside the translator
224 void *code = xlat_get_code( sh4r.pc );
225 xlat_recovery_record_t recover = xlat_get_recovery(code, pc, TRUE);
226 if( recover != NULL ) {
227 // Can be null if there is no recovery necessary
228 sh4_translate_run_recovery(recover);
231 // finally longjmp back into sh4_xlat_run_slice
232 xlat_running = FALSE;
233 longjmp(xlat_jmp_buf, exit_code);
236 void sh4_translate_breakpoint_hit(uint32_t pc)
238 if( sh4_starting && sh4r.slice_cycle == 0 && pc == sh4r.pc ) {
241 sh4_translate_exit( XLAT_EXIT_BREAKPOINT );
245 * Exit the current block at the end of the current instruction, flush the
246 * translation cache (completely) and return control to sh4_xlat_run_slice.
248 * As a special case, if the current instruction is actually the last
249 * instruction in the block (ie it's in a delay slot), this function
250 * returns to allow normal completion of the translation block. Otherwise
251 * this function never returns.
253 * Must only be invoked (indirectly) from within translated code.
255 void sh4_translate_flush_cache()
257 void *pc = xlat_get_native_pc();
258 assert( pc != NULL );
260 void *code = xlat_get_code( sh4r.pc );
261 xlat_recovery_record_t recover = xlat_get_recovery(code, pc, TRUE);
262 if( recover != NULL ) {
263 // Can be null if there is no recovery necessary
264 sh4_translate_run_recovery(recover);
266 xlat_running = FALSE;
267 longjmp(xlat_jmp_buf, XLAT_EXIT_CONTINUE);
274 void *xlat_get_code_by_vma( sh4vma_t vma )
278 if( IS_IN_ICACHE(vma) ) {
279 return xlat_get_code( GET_ICACHE_PHYS(vma) );
282 if( vma > 0xFFFFFF00 ) {
287 if( !mmu_update_icache(vma) ) {
288 // fault - off to the fault handler
289 if( !mmu_update_icache(sh4r.pc) ) {
290 // double fault - halt
291 ERROR( "Double fault - halting" );
297 assert( IS_IN_ICACHE(sh4r.pc) );
298 result = xlat_get_code( GET_ICACHE_PHYS(sh4r.pc) );
.