| 1 | diff --git a/libpcsxcore/new_dynarec/linkage_arm.S b/libpcsxcore/new_dynarec/linkage_arm.S |
| 2 | index fcb4e1a..e2547c4 100644 |
| 3 | --- a/libpcsxcore/new_dynarec/linkage_arm.S |
| 4 | +++ b/libpcsxcore/new_dynarec/linkage_arm.S |
| 5 | @@ -439,7 +439,7 @@ FUNCTION(cc_interrupt): |
| 6 | str r1, [fp, #LO_pending_exception] |
| 7 | and r2, r2, r10, lsr #17 |
| 8 | add r3, fp, #LO_restore_candidate |
| 9 | - str r10, [fp, #LO_cycle] /* PCSX cycles */ |
| 10 | +@@@ str r10, [fp, #LO_cycle] /* PCSX cycles */ |
| 11 | @@ str r10, [fp, #LO_reg_cop0+36] /* Count */ |
| 12 | ldr r4, [r2, r3] |
| 13 | mov r10, lr |
| 14 | @@ -519,7 +519,7 @@ FUNCTION(jump_syscall_hle): |
| 15 | mov r1, #0 /* in delay slot */ |
| 16 | add r2, r2, r10 |
| 17 | mov r0, #0x20 /* cause */ |
| 18 | - str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 19 | +@@@ str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 20 | bl psxException |
| 21 | |
| 22 | /* note: psxException might do recursive recompiler call from it's HLE code, |
| 23 | @@ -540,7 +540,7 @@ FUNCTION(jump_hlecall): |
| 24 | str r0, [fp, #LO_pcaddr] |
| 25 | add r2, r2, r10 |
| 26 | adr lr, pcsx_return |
| 27 | - str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 28 | +@@@ str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 29 | bx r1 |
| 30 | .size jump_hlecall, .-jump_hlecall |
| 31 | |
| 32 | @@ -550,7 +550,7 @@ FUNCTION(jump_intcall): |
| 33 | str r0, [fp, #LO_pcaddr] |
| 34 | add r2, r2, r10 |
| 35 | adr lr, pcsx_return |
| 36 | - str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 37 | +@@@ str r2, [fp, #LO_cycle] /* PCSX cycle counter */ |
| 38 | b execI |
| 39 | .size jump_hlecall, .-jump_hlecall |
| 40 | |
| 41 | @@ -559,7 +559,7 @@ FUNCTION(new_dyna_leave): |
| 42 | ldr r0, [fp, #LO_last_count] |
| 43 | add r12, fp, #28 |
| 44 | add r10, r0, r10 |
| 45 | - str r10, [fp, #LO_cycle] |
| 46 | +@@@ str r10, [fp, #LO_cycle] |
| 47 | ldmfd sp!, {r4, r5, r6, r7, r8, r9, sl, fp, ip, pc} |
| 48 | .size new_dyna_leave, .-new_dyna_leave |
| 49 | |
| 50 | @@ -676,7 +676,7 @@ FUNCTION(new_dyna_start): |
| 51 | \readop r0, [r1, r3, lsl #\tab_shift] |
| 52 | .endif |
| 53 | movcc pc, lr |
| 54 | - str r2, [fp, #LO_cycle] |
| 55 | +@@@ str r2, [fp, #LO_cycle] |
| 56 | bx r1 |
| 57 | .endm |
| 58 | |
| 59 | @@ -711,7 +711,7 @@ FUNCTION(jump_handler_read32): |
| 60 | mov r0, r1 |
| 61 | add r2, r2, r12 |
| 62 | push {r2, lr} |
| 63 | - str r2, [fp, #LO_cycle] |
| 64 | +@@@ str r2, [fp, #LO_cycle] |
| 65 | blx r3 |
| 66 | |
| 67 | ldr r0, [fp, #LO_next_interupt] |
| 68 | @@ -739,7 +739,7 @@ FUNCTION(jump_handler_write_h): |
| 69 | add r2, r2, r12 |
| 70 | mov r0, r1 |
| 71 | push {r2, lr} |
| 72 | - str r2, [fp, #LO_cycle] |
| 73 | +@@@ str r2, [fp, #LO_cycle] |
| 74 | blx r3 |
| 75 | |
| 76 | ldr r0, [fp, #LO_next_interupt] |
| 77 | diff --git a/libpcsxcore/new_dynarec/linkage_arm64.S b/libpcsxcore/new_dynarec/linkage_arm64.S |
| 78 | index 060ac48..e3007d3 100644 |
| 79 | --- a/libpcsxcore/new_dynarec/linkage_arm64.S |
| 80 | +++ b/libpcsxcore/new_dynarec/linkage_arm64.S |
| 81 | @@ -131,7 +131,7 @@ FUNCTION(cc_interrupt): |
| 82 | str wzr, [rFP, #LO_pending_exception] |
| 83 | and w2, w2, rCC, lsr #17 |
| 84 | add x3, rFP, #LO_restore_candidate |
| 85 | - str rCC, [rFP, #LO_cycle] /* PCSX cycles */ |
| 86 | +## str rCC, [rFP, #LO_cycle] /* PCSX cycles */ |
| 87 | # str rCC, [rFP, #LO_reg_cop0+36] /* Count */ |
| 88 | ldr w19, [x3, w2, uxtw] |
| 89 | mov x21, lr |
| 90 | @@ -254,7 +254,7 @@ FUNCTION(new_dyna_start): |
| 91 | FUNCTION(new_dyna_leave): |
| 92 | ldr w0, [rFP, #LO_last_count] |
| 93 | add rCC, rCC, w0 |
| 94 | - str rCC, [rFP, #LO_cycle] |
| 95 | +## str rCC, [rFP, #LO_cycle] |
| 96 | ldp x19, x20, [sp, #16*1] |
| 97 | ldp x21, x22, [sp, #16*2] |
| 98 | ldp x23, x24, [sp, #16*3] |
| 99 | @@ -272,7 +272,7 @@ FUNCTION(new_dyna_leave): |
| 100 | /* w0 = adddr/data, x1 = rhandler, w2 = cycles, x3 = whandler */ |
| 101 | ldr w4, [rFP, #LO_last_count] |
| 102 | add w4, w4, w2 |
| 103 | - str w4, [rFP, #LO_cycle] |
| 104 | +## str w4, [rFP, #LO_cycle] |
| 105 | .endm |
| 106 | |
| 107 | .macro memhandler_post |
| 108 | diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c |
| 109 | index e7b55b6..caa06d0 100644 |
| 110 | --- a/libpcsxcore/new_dynarec/new_dynarec.c |
| 111 | +++ b/libpcsxcore/new_dynarec/new_dynarec.c |
| 112 | @@ -43,10 +43,10 @@ static int sceBlock; |
| 113 | #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0])) |
| 114 | #endif |
| 115 | |
| 116 | -//#define DISASM |
| 117 | -//#define assem_debug printf |
| 118 | +#define DISASM |
| 119 | +#define assem_debug printf |
| 120 | //#define inv_debug printf |
| 121 | -#define assem_debug(...) |
| 122 | +//#define assem_debug(...) |
| 123 | #define inv_debug(...) |
| 124 | |
| 125 | #ifdef __i386__ |
| 126 | @@ -424,6 +424,9 @@ static int doesnt_expire_soon(void *tcaddr) |
| 127 | // This is called from the recompiled JR/JALR instructions |
| 128 | void noinline *get_addr(u_int vaddr) |
| 129 | { |
| 130 | +#ifdef DRC_DBG |
| 131 | +printf("get_addr %08x, pc=%08x\n", vaddr, psxRegs.pc); |
| 132 | +#endif |
| 133 | u_int page=get_page(vaddr); |
| 134 | u_int vpage=get_vpage(vaddr); |
| 135 | struct ll_entry *head; |
| 136 | @@ -4221,13 +4224,15 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert) |
| 137 | } |
| 138 | emit_addimm_and_set_flags(cycles,HOST_CCREG); |
| 139 | jaddr=out; |
| 140 | - emit_jns(0); |
| 141 | + //emit_jns(0); |
| 142 | + emit_jmp(0); |
| 143 | } |
| 144 | else |
| 145 | { |
| 146 | emit_cmpimm(HOST_CCREG,-CLOCK_ADJUST(count+2)); |
| 147 | jaddr=out; |
| 148 | - emit_jns(0); |
| 149 | + //emit_jns(0); |
| 150 | + emit_jmp(0); |
| 151 | } |
| 152 | add_stub(CC_STUB,jaddr,idle?idle:out,(*adj==0||invert||idle)?0:(count+2),i,addr,taken,0); |
| 153 | } |
| 154 | @@ -4635,7 +4640,8 @@ static void rjump_assemble(int i,struct regstat *i_regs) |
| 155 | // special case for RFE |
| 156 | emit_jmp(0); |
| 157 | else |
| 158 | - emit_jns(0); |
| 159 | + //emit_jns(0); |
| 160 | + emit_jmp(0); |
| 161 | //load_regs_bt(branch_regs[i].regmap,branch_regs[i].dirty,-1); |
| 162 | #ifdef USE_MINI_HT |
| 163 | if(rs1[i]==31) { |
| 164 | @@ -4737,7 +4743,8 @@ static void cjump_assemble(int i,struct regstat *i_regs) |
| 165 | else if(nop) { |
| 166 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc); |
| 167 | void *jaddr=out; |
| 168 | - emit_jns(0); |
| 169 | + //emit_jns(0); |
| 170 | + emit_jmp(0); |
| 171 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0); |
| 172 | } |
| 173 | else { |
| 174 | @@ -4924,7 +4931,8 @@ static void cjump_assemble(int i,struct regstat *i_regs) |
| 175 | emit_loadreg(CCREG,HOST_CCREG); |
| 176 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG); |
| 177 | void *jaddr=out; |
| 178 | - emit_jns(0); |
| 179 | + //emit_jns(0); |
| 180 | + emit_jmp(0); |
| 181 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0); |
| 182 | emit_storereg(CCREG,HOST_CCREG); |
| 183 | } |
| 184 | @@ -4933,7 +4941,8 @@ static void cjump_assemble(int i,struct regstat *i_regs) |
| 185 | assert(cc==HOST_CCREG); |
| 186 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc); |
| 187 | void *jaddr=out; |
| 188 | - emit_jns(0); |
| 189 | + //emit_jns(0); |
| 190 | + emit_jmp(0); |
| 191 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0); |
| 192 | } |
| 193 | } |
| 194 | @@ -5032,7 +5041,8 @@ static void sjump_assemble(int i,struct regstat *i_regs) |
| 195 | else if(nevertaken) { |
| 196 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc); |
| 197 | void *jaddr=out; |
| 198 | - emit_jns(0); |
| 199 | + //emit_jns(0); |
| 200 | + emit_jmp(0); |
| 201 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0); |
| 202 | } |
| 203 | else { |
| 204 | @@ -5188,7 +5198,8 @@ static void sjump_assemble(int i,struct regstat *i_regs) |
| 205 | emit_loadreg(CCREG,HOST_CCREG); |
| 206 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG); |
| 207 | void *jaddr=out; |
| 208 | - emit_jns(0); |
| 209 | + //emit_jns(0); |
| 210 | + emit_jmp(0); |
| 211 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0); |
| 212 | emit_storereg(CCREG,HOST_CCREG); |
| 213 | } |
| 214 | @@ -5197,7 +5208,8 @@ static void sjump_assemble(int i,struct regstat *i_regs) |
| 215 | assert(cc==HOST_CCREG); |
| 216 | emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc); |
| 217 | void *jaddr=out; |
| 218 | - emit_jns(0); |
| 219 | + //emit_jns(0); |
| 220 | + emit_jmp(0); |
| 221 | add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0); |
| 222 | } |
| 223 | } |
| 224 | @@ -5685,7 +5697,7 @@ void unneeded_registers(int istart,int iend,int r) |
| 225 | // R0 is always unneeded |
| 226 | u|=1; |
| 227 | // Save it |
| 228 | - unneeded_reg[i]=u; |
| 229 | + unneeded_reg[i]=1;//u; |
| 230 | gte_unneeded[i]=gte_u; |
| 231 | /* |
| 232 | printf("ur (%d,%d) %x: ",istart,iend,start+i*4); |
| 233 | @@ -8209,6 +8221,7 @@ int new_recompile_block(int addr) |
| 234 | |
| 235 | // This allocates registers (if possible) one instruction prior |
| 236 | // to use, which can avoid a load-use penalty on certain CPUs. |
| 237 | +#if 0 |
| 238 | for(i=0;i<slen-1;i++) |
| 239 | { |
| 240 | if(!i||(itype[i-1]!=UJUMP&&itype[i-1]!=CJUMP&&itype[i-1]!=SJUMP&&itype[i-1]!=RJUMP)) |
| 241 | @@ -8365,6 +8378,7 @@ int new_recompile_block(int addr) |
| 242 | } |
| 243 | } |
| 244 | } |
| 245 | +#endif |
| 246 | |
| 247 | /* Pass 6 - Optimize clean/dirty state */ |
| 248 | clean_registers(0,slen-1,1); |
| 249 | @@ -8659,6 +8673,11 @@ int new_recompile_block(int addr) |
| 250 | case SPAN: |
| 251 | pagespan_assemble(i,®s[i]);break; |
| 252 | } |
| 253 | + |
| 254 | +#ifdef DRC_DBG |
| 255 | + if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP) |
| 256 | + wb_dirtys(regs[i].regmap,regs[i].dirty); |
| 257 | +#endif |
| 258 | if(itype[i]==UJUMP||itype[i]==RJUMP||(source[i]>>16)==0x1000) |
| 259 | literal_pool(1024); |
| 260 | else |
| 261 | @@ -8767,7 +8786,7 @@ int new_recompile_block(int addr) |
| 262 | } |
| 263 | } |
| 264 | // External Branch Targets (jump_in) |
| 265 | - if(copy+slen*4>(void *)shadow+sizeof(shadow)) copy=shadow; |
| 266 | + if(copy+slen*4>(void *)shadow+sizeof(shadow)) {copy=shadow;printf("shadow overflow\n");} |
| 267 | for(i=0;i<slen;i++) |
| 268 | { |
| 269 | if(bt[i]||i==0) |
| 270 | @@ -8882,6 +8901,10 @@ int new_recompile_block(int addr) |
| 271 | } |
| 272 | expirep=(expirep+1)&65535; |
| 273 | } |
| 274 | +#ifdef DRC_DBG |
| 275 | +printf("new_recompile_block done\n"); |
| 276 | +fflush(stdout); |
| 277 | +#endif |
| 278 | return 0; |
| 279 | } |
| 280 | |