X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=pcsx_rearmed.git;a=blobdiff_plain;f=libpcsxcore%2Fnew_dynarec%2Fnew_dynarec.c;h=013af9b839d3bbc1adb7b2cbb55e5fd3a63eaa13;hp=9b8f153c0fbcf82852cb17eb60102fe3ac72ef57;hb=252c20fc731d772878a62473c09f1688b3cac3dc;hpb=57871462a0b157066bbc4a763c59b61085436609 diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c index 9b8f153c..013af9b8 100644 --- a/libpcsxcore/new_dynarec/new_dynarec.c +++ b/libpcsxcore/new_dynarec/new_dynarec.c @@ -22,14 +22,7 @@ #include //include for uint64_t #include -#include "../recomp.h" -#include "../recomph.h" //include for function prototypes -#include "../macros.h" -#include "../r4300.h" -#include "../ops.h" -#include "../interupt.h" - -#include "../../memory/memory.h" +#include "emu_if.h" //emulator interface #include @@ -142,7 +135,7 @@ struct ll_entry #define CCREG 36 // Cycle count #define INVCP 37 // Pointer to invalid_code #define TEMPREG 38 -#define FTEMP 38 // FPU temporary register +#define FTEMP 38 // FPU/LDL/LDR temporary register #define PTEMP 39 // Prefetch temporary register #define TLREG 40 // TLB mapping offset #define RHASH 41 // Return address hash @@ -182,6 +175,10 @@ struct ll_entry #define OTHER 23 // Other #define SPAN 24 // Branch/delay slot spans 2 pages #define NI 25 // Not implemented +#define HLECALL 26// PCSX fake opcodes for HLE +#define COP2 27 // Coprocessor 2 move +#define C2LS 28 // Coprocessor 2 load/store +#define C2OP 29 // Coprocessor 2 operation /* stubs */ #define CC_STUB 1 @@ -220,7 +217,10 @@ void cc_interrupt(); void fp_exception(); void fp_exception_ds(); void jump_syscall(); +void jump_syscall_hle(); void jump_eret(); +void jump_hlecall(); +void new_dyna_leave(); // TLB void TLBWI_new(); @@ -258,8 +258,9 @@ void nullf() {} #define assem_debug nullf #define inv_debug nullf -void tlb_hacks() +static void tlb_hacks() { +#ifndef DISABLE_TLB // Goldeneye hack if (strncmp((char *) ROM_HEADER->nom, "GOLDENEYE",9) == 0) { @@ -301,18 +302,35 @@ void tlb_hacks() } } } +#endif } -// Get address from virtual address -// This is called from the recompiled JR/JALR instructions -void *get_addr(u_int vaddr) +static u_int get_page(u_int vaddr) { u_int page=(vaddr^0x80000000)>>12; - u_int vpage=page; +#ifndef DISABLE_TLB if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[vaddr>>12]^0x80000000)>>12; +#endif if(page>2048) page=2048+(page&2047); + return page; +} + +static u_int get_vpage(u_int vaddr) +{ + u_int vpage=(vaddr^0x80000000)>>12; +#ifndef DISABLE_TLB if(vpage>262143&&tlb_LUT_r[vaddr>>12]) vpage&=2047; // jump_dirty uses a hash of the virtual address instead +#endif if(vpage>2048) vpage=2048+(vpage&2047); + return vpage; +} + +// Get address from virtual address +// This is called from the recompiled JR/JALR instructions +void *get_addr(u_int vaddr) +{ + u_int page=get_page(vaddr); + u_int vpage=get_vpage(vaddr); struct ll_entry *head; //printf("TRACE: count=%d next=%d (get_addr %x,page %d)\n",Count,next_interupt,vaddr,page); head=jump_in[page]; @@ -339,10 +357,12 @@ void *get_addr(u_int vaddr) invalid_code[vaddr>>12]=0; memory_map[vaddr>>12]|=0x40000000; if(vpage<2048) { +#ifndef DISABLE_TLB if(tlb_LUT_r[vaddr>>12]) { invalid_code[tlb_LUT_r[vaddr>>12]>>12]=0; memory_map[tlb_LUT_r[vaddr>>12]>>12]|=0x40000000; } +#endif restore_candidate[vpage>>3]|=1<<(vpage&7); } else restore_candidate[page>>3]|=1<<(page&7); @@ -386,16 +406,15 @@ void *get_addr_ht(u_int vaddr) void *get_addr_32(u_int vaddr,u_int flags) { +#ifdef FORCE32 + return get_addr(vaddr); +#endif //printf("TRACE: count=%d next=%d (get_addr_32 %x,flags %x)\n",Count,next_interupt,vaddr,flags); int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; if(ht_bin[0]==vaddr) return (void *)ht_bin[1]; if(ht_bin[2]==vaddr) return (void *)ht_bin[3]; - u_int page=(vaddr^0x80000000)>>12; - u_int vpage=page; - if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[vaddr>>12]^0x80000000)>>12; - if(page>2048) page=2048+(page&2047); - if(vpage>262143&&tlb_LUT_r[vaddr>>12]) vpage&=2047; // jump_dirty uses a hash of the virtual address instead - if(vpage>2048) vpage=2048+(vpage&2047); + u_int page=get_page(vaddr); + u_int vpage=get_vpage(vaddr); struct ll_entry *head; head=jump_in[page]; while(head!=NULL) { @@ -430,10 +449,12 @@ void *get_addr_32(u_int vaddr,u_int flags) invalid_code[vaddr>>12]=0; memory_map[vaddr>>12]|=0x40000000; if(vpage<2048) { +#ifndef DISABLE_TLB if(tlb_LUT_r[vaddr>>12]) { invalid_code[tlb_LUT_r[vaddr>>12]>>12]=0; memory_map[tlb_LUT_r[vaddr>>12]>>12]|=0x40000000; } +#endif restore_candidate[vpage>>3]|=1<<(vpage&7); } else restore_candidate[page>>3]|=1<<(page&7); @@ -615,7 +636,7 @@ void lsn(u_char hsn[], int i, int *preferred_reg) } // On some architectures stores need invc_ptr #if defined(HOST_IMM8) - if(itype[i+j]==STORE || itype[i+j]==STORELR || (opcode[i+j]&0x3b)==0x39) { + if(itype[i+j]==STORE || itype[i+j]==STORELR || (opcode[i+j]&0x3b)==0x39 || (opcode[i+j]&0x3b)==0x3a) { hsn[INVCP]=j; } #endif @@ -652,7 +673,7 @@ void lsn(u_char hsn[], int i, int *preferred_reg) hsn[RHTBL]=1; } // Coprocessor load/store needs FTEMP, even if not declared - if(itype[i]==C1LS) { + if(itype[i]==C1LS||itype[i]==C2LS) { hsn[FTEMP]=0; } // Load L/R also uses FTEMP as a temporary register @@ -664,7 +685,7 @@ void lsn(u_char hsn[], int i, int *preferred_reg) hsn[FTEMP]=0; } // Don't remove the TLB registers either - if(itype[i]==LOAD || itype[i]==LOADLR || itype[i]==STORE || itype[i]==STORELR || itype[i]==C1LS ) { + if(itype[i]==LOAD || itype[i]==LOADLR || itype[i]==STORE || itype[i]==STORELR || itype[i]==C1LS || itype[i]==C2LS) { hsn[TLREG]=0; } // Don't remove the miniht registers @@ -705,7 +726,7 @@ int needed_again(int r, int i) j++; break; } - if(itype[i+j]==SYSCALL||((source[i+j]&0xfc00003f)==0x0d)) + if(itype[i+j]==SYSCALL||itype[i+j]==HLECALL||((source[i+j]&0xfc00003f)==0x0d)) { break; } @@ -951,14 +972,10 @@ void ll_add(struct ll_entry **head,int vaddr,void *addr) // Add virtual address mapping for 32-bit compiled block void ll_add_32(struct ll_entry **head,int vaddr,u_int reg32,void *addr) { - struct ll_entry *new_entry; - new_entry=malloc(sizeof(struct ll_entry)); - assert(new_entry!=NULL); - new_entry->vaddr=vaddr; - new_entry->reg32=reg32; - new_entry->addr=addr; - new_entry->next=*head; - *head=new_entry; + ll_add(head,vaddr,addr); +#ifndef FORCE32 + (*head)->reg32=reg32; +#endif } // Check if an address is already compiled @@ -974,9 +991,7 @@ void *check_addr(u_int vaddr) if(((ht_bin[3]-MAX_OUTPUT_BLOCK_SIZE-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) if(isclean(ht_bin[3])) return (void *)ht_bin[3]; } - u_int page=(vaddr^0x80000000)>>12; - if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[vaddr>>12]^0x80000000)>>12; - if(page>2048) page=2048+(page&2047); + u_int page=get_page(vaddr); struct ll_entry *head; head=jump_in[page]; while(head!=NULL) { @@ -1104,12 +1119,8 @@ int invalidate_page(u_int page) void invalidate_block(u_int block) { int modified; - u_int page,vpage; - page=vpage=block^0x80000; - if(page>262143&&tlb_LUT_r[block]) page=(tlb_LUT_r[block]^0x80000000)>>12; - if(page>2048) page=2048+(page&2047); - if(vpage>262143&&tlb_LUT_r[block]) vpage&=2047; // jump_dirty uses a hash of the virtual address instead - if(vpage>2048) vpage=2048+(vpage&2047); + u_int page=get_page(block<<12); + u_int vpage=get_vpage(block<<12); inv_debug("INVALIDATE: %x (%d)\n",block<<12,page); //inv_debug("invalid_code[block]=%d\n",invalid_code[block]); u_int first,last; @@ -1128,12 +1139,14 @@ void invalidate_block(u_int block) if((((end-1-(u_int)rdram)>>12)&2047)>last) last=((end-1-(u_int)rdram)>>12)&2047; } } +#ifndef DISABLE_TLB if(page<2048&&(signed int)start>=(signed int)0xC0000000&&(signed int)end>=(signed int)0xC0000000) { if(((start+memory_map[start>>12]-(u_int)rdram)>>12)<=page&&((end-1+memory_map[(end-1)>>12]-(u_int)rdram)>>12)>=page) { if((((start+memory_map[start>>12]-(u_int)rdram)>>12)&2047)>12]-(u_int)rdram)>>12)&2047; if((((end-1+memory_map[(end-1)>>12]-(u_int)rdram)>>12)&2047)>last) last=((end-1+memory_map[(end-1)>>12]-(u_int)rdram)>>12)&2047; } } +#endif } head=head->next; } @@ -1152,6 +1165,7 @@ void invalidate_block(u_int block) // Don't trap writes invalid_code[block]=1; +#ifndef DISABLE_TLB // If there is a valid TLB entry for this page, remove write protect if(tlb_LUT_w[block]) { assert(tlb_LUT_r[block]==tlb_LUT_w[block]); @@ -1162,6 +1176,7 @@ void invalidate_block(u_int block) if(real_block>=0x80000&&real_block<0x80800) memory_map[real_block]=((u_int)rdram-0x80000000)>>2; } else if(block>=0x80000&&block<0x80800) memory_map[block]=((u_int)rdram-0x80000000)>>2; +#endif #ifdef __arm__ if(modified) __clear_cache((void *)BASE_ADDR,(void *)BASE_ADDR+(1<>12; - if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[vaddr>>12]^0x80000000)>>12; - if(page>4095) page=2048+(page&2047); + u_int page=get_page(vaddr); inv_debug("add_link: %x -> %x (%d)\n",(int)src,vaddr,page); ll_add(jump_out+page,vaddr,src); //int ptr=get_pointer(src); @@ -1251,7 +1266,9 @@ void clean_blocks(u_int page) void * clean_addr=(void *)get_clean_addr((int)head->addr); if((((u_int)clean_addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) { u_int ppage=page; +#ifndef DISABLE_TLB if(page<2048&&tlb_LUT_r[head->vaddr>>12]) ppage=(tlb_LUT_r[head->vaddr>>12]^0x80000000)>>12; +#endif inv_debug("INV: Restored %x (%x/%x)\n",head->vaddr, (int)head->addr, (int)clean_addr); //printf("page=%x, addr=%x\n",page,head->vaddr); //assert(head->vaddr>>12==(page|0x80000)); @@ -1632,6 +1649,22 @@ void c1ls_alloc(struct regstat *current,int i) alloc_reg_temp(current,i,-1); } +void c2ls_alloc(struct regstat *current,int i) +{ + clear_const(current,rt1[i]); + if(needed_again(rs1[i],i)) alloc_reg(current,i,rs1[i]); + alloc_reg(current,i,FTEMP); + // If using TLB, need a register for pointer to the mapping table + if(using_tlb) alloc_reg(current,i,TLREG); + #if defined(HOST_IMM8) + // On CPUs without 32-bit immediates we need a pointer to invalid_code + else if((opcode[i]&0x3b)==0x3a) // SWC2/SDC2 + alloc_reg(current,i,INVCP); + #endif + // We need a temporary register for address generation + alloc_reg_temp(current,i,-1); +} + #ifndef multdiv_alloc void multdiv_alloc(struct regstat *current,int i) { @@ -1769,6 +1802,10 @@ void float_alloc(struct regstat *current,int i) alloc_reg(current,i,CSREG); // Load status alloc_reg_temp(current,i,-1); } +void c2op_alloc(struct regstat *current,int i) +{ + alloc_reg_temp(current,i,-1); +} void fcomp_alloc(struct regstat *current,int i) { alloc_reg(current,i,CSREG); // Load status @@ -1794,6 +1831,7 @@ void delayslot_alloc(struct regstat *current,int i) case RJUMP: case FJUMP: case SYSCALL: + case HLECALL: case SPAN: assem_debug("jump in the delay slot. this shouldn't happen.\n");//exit(1); printf("Disabled speculative precompilation\n"); @@ -1829,11 +1867,15 @@ void delayslot_alloc(struct regstat *current,int i) cop0_alloc(current,i); break; case COP1: + case COP2: cop1_alloc(current,i); break; case C1LS: c1ls_alloc(current,i); break; + case C2LS: + c2ls_alloc(current,i); + break; case FCONV: fconv_alloc(current,i); break; @@ -1843,6 +1885,9 @@ void delayslot_alloc(struct regstat *current,int i) case FCOMP: fcomp_alloc(current,i); break; + case C2OP: + c2op_alloc(current,i); + break; } } @@ -1863,9 +1908,9 @@ static void pagespan_alloc(struct regstat *current,int i) if(opcode[i]==0&&(opcode2[i]&0x3E)==8) // JR/JALR { alloc_reg(current,i,rs1[i]); - if (rt1[i]==31) { - alloc_reg(current,i,31); - dirty_reg(current,31); + if (rt1[i]!=0) { + alloc_reg(current,i,rt1[i]); + dirty_reg(current,rt1[i]); } } if((opcode[i]&0x2E)==4) // BEQ/BNE/BEQL/BNEL @@ -1919,10 +1964,12 @@ void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32 if((dirty>>hr)&1) { if(regmap[hr]<64) { emit_storereg(r,hr); +#ifndef FORCE32 if((is32>>regmap[hr])&1) { emit_sarimm(hr,31,hr); emit_storereg(r|64,hr); } +#endif }else{ emit_storereg(r|64,hr); } @@ -1953,14 +2000,6 @@ int rchecksum() sum^=((u_int *)reg)[i]; return sum; } -int fchecksum() -{ - int i; - int sum=0; - for(i=0;i<64;i++) - sum^=((u_int *)reg_cop1_fgr_64)[i]; - return sum; -} void rlist() { int i; @@ -1968,10 +2007,12 @@ void rlist() for(i=0;i<32;i++) printf("r%d:%8x%8x ",i,((int *)(reg+i))[1],((int *)(reg+i))[0]); printf("\n"); +#ifndef DISABLE_COP1 printf("TRACE: "); for(i=0;i<32;i++) printf("f%d:%8x%8x ",i,((int*)reg_cop1_simple[i])[1],*((int*)reg_cop1_simple[i])); printf("\n"); +#endif } void enabletrace() @@ -2741,8 +2782,13 @@ void load_assemble(int i,struct regstat *i_regs) //gen_tlb_addr_r(tl,map); //emit_movsbl_indexed((int)rdram-0x80000000,tl,tl); int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,3,tl); else x=((constmap[i][s]+offset)^3)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (tl!=addr) emit_mov(addr,tl); +#endif emit_movsbl_indexed_tlb(x,tl,map,tl); } if(jaddr) @@ -2760,8 +2806,13 @@ void load_assemble(int i,struct regstat *i_regs) #endif { int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,2,tl); else x=((constmap[i][s]+offset)^2)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (tl!=addr) emit_mov(addr,tl); +#endif //#ifdef //emit_movswl_indexed_tlb(x,tl,map,tl); //else @@ -2804,8 +2855,13 @@ void load_assemble(int i,struct regstat *i_regs) //gen_tlb_addr_r(tl,map); //emit_movzbl_indexed((int)rdram-0x80000000,tl,tl); int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,3,tl); else x=((constmap[i][s]+offset)^3)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (tl!=addr) emit_mov(addr,tl); +#endif emit_movzbl_indexed_tlb(x,tl,map,tl); } if(jaddr) @@ -2823,8 +2879,13 @@ void load_assemble(int i,struct regstat *i_regs) #endif { int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,2,tl); else x=((constmap[i][s]+offset)^2)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (tl!=addr) emit_mov(addr,tl); +#endif //#ifdef //emit_movzwl_indexed_tlb(x,tl,map,tl); //#else @@ -2921,7 +2982,7 @@ void store_assemble(int i,struct regstat *i_regs) int addr,temp; int offset; int jaddr=0,jaddr2,type; - int memtarget,c=0; + int memtarget=0,c=0; int agr=AGEN1+(i&1); u_int hr,reglist=0; th=get_reg(i_regs->regmap,rs2[i]|64); @@ -2981,8 +3042,13 @@ void store_assemble(int i,struct regstat *i_regs) if (opcode[i]==0x28) { // SB if(!c||memtarget) { int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,3,temp); else x=((constmap[i][s]+offset)^3)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (addr!=temp) emit_mov(addr,temp); +#endif //gen_tlb_addr_w(temp,map); //emit_writebyte_indexed(tl,(int)rdram-0x80000000,temp); emit_writebyte_indexed_tlb(tl,x,temp,map,temp); @@ -2992,8 +3058,13 @@ void store_assemble(int i,struct regstat *i_regs) if (opcode[i]==0x29) { // SH if(!c||memtarget) { int x=0; +#ifdef BIG_ENDIAN_MIPS if(!c) emit_xorimm(addr,2,temp); else x=((constmap[i][s]+offset)^2)-(constmap[i][s]+offset); +#else + if(c) x=(constmap[i][s]+offset)-(constmap[i][s]+offset); + else if (addr!=temp) emit_mov(addr,temp); +#endif //#ifdef //emit_writehword_indexed_tlb(tl,x,temp,map,temp); //#else @@ -3027,6 +3098,9 @@ void store_assemble(int i,struct regstat *i_regs) } type=STORED_STUB; } + if(!using_tlb&&(!c||memtarget)) + // addr could be a temp, make sure it survives STORE*_STUB + reglist|=1<>16)&0x1f; + s=get_reg(i_regs->regmap,rs1[i]); + tl=get_reg(i_regs->regmap,FTEMP); + offset=imm[i]; + assert(rs1[i]>0); + assert(tl>=0); + assert(!using_tlb); + + for(hr=0;hrregmap[hr]>=0) reglist|=1<regmap[HOST_CCREG]==CCREG) + reglist&=~(1<regmap,agr); + if(ar<0) ar=get_reg(i_regs->regmap,-1); + reglist|=1<=0) ar=s; + assert(ar>=0); + + if (opcode[i]==0x3a) { // SWC2 + cop2_get_dreg(copr,tl,HOST_TEMPREG); + } + if(s>=0) c=(i_regs->wasconst>>s)&1; + if(!c) { + emit_cmpimm(offset||c||s<0?ar:s,0x800000); + jaddr2=(int)out; + emit_jno(0); + } + else if(((signed int)(constmap[i][s]+offset))>=(signed int)0x80800000) { + jaddr2=(int)out; + emit_jmp(0); // inline_readstub/inline_writestub? Very rare case + } + if (opcode[i]==0x32) { // LWC2 + #ifdef HOST_IMM_ADDR32 + if(c) emit_readword_tlb(constmap[i][s]+offset,-1,tl); + else + #endif + emit_readword_indexed(0,ar,tl); + type=LOADW_STUB; + } + if (opcode[i]==0x3a) { // SWC2 +#ifdef DESTRUCTIVE_SHIFT + if(!offset&&!c&&s>=0) emit_mov(s,ar); +#endif + emit_writeword_indexed(tl,0,ar); + type=STOREW_STUB; + } + if(jaddr2) + add_stub(type,jaddr2,(int)out,i,ar,(int)i_regs,ccadj[i],reglist); + if (opcode[i]==0x3a) { // SWC2 +#if defined(HOST_IMM8) + int ir=get_reg(i_regs->regmap,INVCP); + assert(ir>=0); + emit_cmpmem_indexedsr12_reg(ir,ar,1); +#else + emit_cmpmem_indexedsr12_imm((int)invalid_code,ar,1); +#endif + jaddr3=(int)out; + emit_jne(0); + add_stub(INVCODE_STUB,jaddr3,(int)out,reglist|(1<regmap,CCREG); + assert(ccreg==HOST_CCREG); + assert(!is_delayslot); + emit_movimm(start+i*4+4,0); // Get PC + emit_movimm((int)psxHLEt[source[i]&7],1); + emit_addimm(HOST_CCREG,CLOCK_DIVIDER*ccadj[i],HOST_CCREG); // XXX + emit_jmp((int)jump_hlecall); } void ds_assemble(int i,struct regstat *i_regs) @@ -3566,6 +3739,12 @@ void ds_assemble(int i,struct regstat *i_regs) cop1_assemble(i,i_regs);break; case C1LS: c1ls_assemble(i,i_regs);break; + case COP2: + cop2_assemble(i,i_regs);break; + case C2LS: + c2ls_assemble(i,i_regs);break; + case C2OP: + c2op_assemble(i,i_regs);break; case FCONV: fconv_assemble(i,i_regs);break; case FLOAT: @@ -3577,6 +3756,7 @@ void ds_assemble(int i,struct regstat *i_regs) case MOV: mov_assemble(i,i_regs);break; case SYSCALL: + case HLECALL: case SPAN: case UJUMP: case RJUMP: @@ -3734,9 +3914,10 @@ static void loop_preload(signed char pre[],signed char entry[]) } // Generate address for load/store instruction +// goes to AGEN for writes, FTEMP for LOADLR and cop1/2 loads void address_generation(int i,struct regstat *i_regs,signed char entry[]) { - if(itype[i]==LOAD||itype[i]==LOADLR||itype[i]==STORE||itype[i]==STORELR||itype[i]==C1LS) { + if(itype[i]==LOAD||itype[i]==LOADLR||itype[i]==STORE||itype[i]==STORELR||itype[i]==C1LS||itype[i]==C2LS) { int ra; int agr=AGEN1+(i&1); int mgr=MGEN1+(i&1); @@ -3751,8 +3932,8 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) ra=get_reg(i_regs->regmap,agr); if(ra<0) ra=get_reg(i_regs->regmap,-1); } - if(itype[i]==C1LS) { - if (opcode[i]==0x31||opcode[i]==0x35) // LWC1/LDC1 + if(itype[i]==C1LS||itype[i]==C2LS) { + if ((opcode[i]&0x3b)==0x31||(opcode[i]&0x3b)==0x32) // LWC1/LDC1/LWC2/LDC2 ra=get_reg(i_regs->regmap,FTEMP); else { // SWC1/SDC1 ra=get_reg(i_regs->regmap,agr); @@ -3790,7 +3971,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) else if(c) { if(rm>=0) { if(!entry||entry[rm]!=mgr) { - if(itype[i]==STORE||itype[i]==STORELR||opcode[i]==0x39||opcode[i]==0x3D) { + if(itype[i]==STORE||itype[i]==STORELR||(opcode[i]&0x3b)==0x39||(opcode[i]&0x3b)==0x3a) { // Stores to memory go thru the mapper to detect self-modifying // code, loads don't. if((unsigned int)(constmap[i][rs]+offset)>=0xC0000000 || @@ -3810,7 +3991,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) emit_movimm((constmap[i][rs]+offset)&0xFFFFFFF8,ra); // LDL/LDR }else{ #ifdef HOST_IMM_ADDR32 - if((itype[i]!=LOAD&&opcode[i]!=0x31&&opcode[i]!=0x35) || + if((itype[i]!=LOAD&&(opcode[i]&0x3b)!=0x31&&(opcode[i]&0x3b)!=0x32) || // LWC1/LDC1/LWC2/LDC2 (using_tlb&&((signed int)constmap[i][rs]+offset)>=(signed int)0xC0000000)) #endif emit_movimm(constmap[i][rs]+offset,ra); @@ -3828,7 +4009,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) } } // Preload constants for next instruction - if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS) { + if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS||itype[i+1]==C2LS) { int agr,ra; #ifndef HOST_IMM_ADDR32 // Mapper entry @@ -3839,7 +4020,8 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) int offset=imm[i+1]; int c=(regs[i+1].wasconst>>rs)&1; if(c) { - if(itype[i+1]==STORE||itype[i+1]==STORELR||opcode[i+1]==0x39||opcode[i+1]==0x3D) { + if(itype[i+1]==STORE||itype[i+1]==STORELR + ||(opcode[i+1]&0x3b)==0x39||(opcode[i+1]&0x3b)==0x3a) { // SWC1/SDC1, SWC2/SDC2 // Stores to memory go thru the mapper to detect self-modifying // code, loads don't. if((unsigned int)(constmap[i+1][rs]+offset)>=0xC0000000 || @@ -3869,7 +4051,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) emit_movimm((constmap[i+1][rs]+offset)&0xFFFFFFF8,ra); // LDL/LDR }else{ #ifdef HOST_IMM_ADDR32 - if((itype[i+1]!=LOAD&&opcode[i+1]!=0x31&&opcode[i+1]!=0x35) || + if((itype[i+1]!=LOAD&&(opcode[i+1]&0x3b)!=0x31&&(opcode[i+1]&0x3b)!=0x32) || // LWC1/LDC1/LWC2/LDC2 (using_tlb&&((signed int)constmap[i+1][rs]+offset)>=(signed int)0xC0000000)) #endif emit_movimm(constmap[i+1][rs]+offset,ra); @@ -4041,6 +4223,7 @@ void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty) if((i_dirty>>hr)&1) { if(i_regmap[hr]<64) { emit_storereg(i_regmap[hr],hr); +#ifndef FORCE32 if( ((i_is32>>i_regmap[hr])&1) ) { #ifdef DESTRUCTIVE_WRITEBACK emit_sarimm(hr,31,hr); @@ -4050,6 +4233,7 @@ void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty) emit_storereg(i_regmap[hr]|64,HOST_TEMPREG); #endif } +#endif }else{ if( !((i_is32>>(i_regmap[hr]&63))&1) ) { emit_storereg(i_regmap[hr],hr); @@ -4075,6 +4259,7 @@ void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,in if((i_dirty>>hr)&1) { if(i_regmap[hr]<64) { emit_storereg(i_regmap[hr],hr); +#ifndef FORCE32 if( ((i_is32>>i_regmap[hr])&1) ) { #ifdef DESTRUCTIVE_WRITEBACK emit_sarimm(hr,31,hr); @@ -4084,6 +4269,7 @@ void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,in emit_storereg(i_regmap[hr]|64,HOST_TEMPREG); #endif } +#endif }else{ if( !((i_is32>>(i_regmap[hr]&63))&1) ) { emit_storereg(i_regmap[hr],hr); @@ -4379,7 +4565,7 @@ void ds_assemble_entry(int i) wb_register(CCREG,regs[t].regmap_entry,regs[t].wasdirty,regs[t].was32); load_regs(regs[t].regmap_entry,regs[t].regmap,regs[t].was32,rs1[t],rs2[t]); address_generation(t,®s[t],regs[t].regmap_entry); - if(itype[t]==STORE||itype[t]==STORELR||(opcode[t]&0x3b)==0x39) + if(itype[t]==STORE||itype[t]==STORELR||(opcode[t]&0x3b)==0x39||(opcode[t]&0x3b)==0x3a) load_regs(regs[t].regmap_entry,regs[t].regmap,regs[t].was32,INVCP,INVCP); cop1_usable=0; is_delayslot=0; @@ -4406,6 +4592,12 @@ void ds_assemble_entry(int i) cop1_assemble(t,®s[t]);break; case C1LS: c1ls_assemble(t,®s[t]);break; + case COP2: + cop2_assemble(t,®s[t]);break; + case C2LS: + c2ls_assemble(t,®s[t]);break; + case C2OP: + c2op_assemble(t,®s[t]);break; case FCONV: fconv_assemble(t,®s[t]);break; case FLOAT: @@ -4417,6 +4609,7 @@ void ds_assemble_entry(int i) case MOV: mov_assemble(t,®s[t]);break; case SYSCALL: + case HLECALL: case SPAN: case UJUMP: case RJUMP: @@ -4883,11 +5076,11 @@ void rjump_assemble(int i,struct regstat *i_regs) wb_invalidate(regs[i].regmap,branch_regs[i].regmap,regs[i].dirty,regs[i].is32, bc_unneeded,bc_unneeded_upper); load_regs(regs[i].regmap,branch_regs[i].regmap,regs[i].was32,rs1[i],CCREG); - if(rt1[i]==31) { + if(rt1[i]!=0) { int rt,return_address; - assert(rt1[i+1]!=31); - assert(rt2[i+1]!=31); - rt=get_reg(branch_regs[i].regmap,31); + assert(rt1[i+1]!=rt1[i]); + assert(rt2[i+1]!=rt1[i]); + rt=get_reg(branch_regs[i].regmap,rt1[i]); assem_debug("branch(%d): eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",i,branch_regs[i].regmap[0],branch_regs[i].regmap[1],branch_regs[i].regmap[2],branch_regs[i].regmap[3],branch_regs[i].regmap[5],branch_regs[i].regmap[6],branch_regs[i].regmap[7]); assert(rt>=0); return_address=start+i*4+8; @@ -6003,7 +6196,7 @@ static void pagespan_assemble(int i,struct regstat *i_regs) emit_mov(s1l,addr); if(opcode2[i]==9) // JALR { - int rt=get_reg(i_regs->regmap,31); + int rt=get_reg(i_regs->regmap,rt1[i]); emit_movimm(start+i*4+8,rt); } } @@ -6192,12 +6385,8 @@ static void pagespan_ds() { assem_debug("initial delay slot:\n"); u_int vaddr=start+1; - u_int page=(0x80000000^vaddr)>>12; - u_int vpage=page; - if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[page^0x80000]^0x80000000)>>12; - if(page>2048) page=2048+(page&2047); - if(vpage>262143&&tlb_LUT_r[vaddr>>12]) vpage&=2047; // jump_dirty uses a hash of the virtual address instead - if(vpage>2048) vpage=2048+(vpage&2047); + u_int page=get_page(vaddr); + u_int vpage=get_vpage(vaddr); ll_add(jump_dirty+vpage,vaddr,(void *)out); do_dirty_stub_ds(); ll_add(jump_in+page,vaddr,(void *)out); @@ -6208,7 +6397,7 @@ static void pagespan_ds() emit_writeword(HOST_BTREG,(int)&branch_target); load_regs(regs[0].regmap_entry,regs[0].regmap,regs[0].was32,rs1[0],rs2[0]); address_generation(0,®s[0],regs[0].regmap_entry); - if(itype[0]==STORE||itype[0]==STORELR||(opcode[0]&0x3b)==0x39) + if(itype[0]==STORE||itype[0]==STORELR||(opcode[0]&0x3b)==0x39||(opcode[0]&0x3b)==0x3a) load_regs(regs[0].regmap_entry,regs[0].regmap,regs[0].was32,INVCP,INVCP); cop1_usable=0; is_delayslot=0; @@ -6235,6 +6424,12 @@ static void pagespan_ds() cop1_assemble(0,®s[0]);break; case C1LS: c1ls_assemble(0,®s[0]);break; + case COP2: + cop2_assemble(0,®s[0]);break; + case C2LS: + c2ls_assemble(0,®s[0]);break; + case C2OP: + c2op_assemble(0,®s[0]);break; case FCONV: fconv_assemble(0,®s[0]);break; case FLOAT: @@ -6246,6 +6441,7 @@ static void pagespan_ds() case MOV: mov_assemble(0,®s[0]);break; case SYSCALL: + case HLECALL: case SPAN: case UJUMP: case RJUMP: @@ -6469,7 +6665,7 @@ void unneeded_registers(int istart,int iend,int r) } } } - else if(itype[i]==SYSCALL) + else if(itype[i]==SYSCALL||itype[i]==HLECALL) { // SYSCALL instruction (software interrupt) u=1; @@ -6522,6 +6718,12 @@ void unneeded_registers(int istart,int iend,int r) } printf("\n");*/ } +#ifdef FORCE32 + for (i=iend;i>=istart;i--) + { + unneeded_reg_upper[i]=branch_unneeded_reg_upper[i]=-1LL; + } +#endif } // Identify registers which are likely to contain 32-bit values @@ -6712,18 +6914,22 @@ static void provisional_32bit() if(op2==0) is32|=1LL<>dep2[i+1])&1) r32|=1LL<>11)&0x1f); // MTC1 else printf (" %x: %s\n",start+i*4,insn[i]); break; + case COP2: + if(opcode2[i]<3) + printf (" %x: %s r%d,cpr2[%d]\n",start+i*4,insn[i],rt1[i],(source[i]>>11)&0x1f); // MFC2 + else if(opcode2[i]>3) + printf (" %x: %s r%d,cpr2[%d]\n",start+i*4,insn[i],rs1[i],(source[i]>>11)&0x1f); // MTC2 + else printf (" %x: %s\n",start+i*4,insn[i]); + break; case C1LS: printf (" %x: %s cpr1[%d],r%d+%x\n",start+i*4,insn[i],(source[i]>>16)&0x1f,rs1[i],imm[i]); break; + case C2LS: + printf (" %x: %s cpr2[%d],r%d+%x\n",start+i*4,insn[i],(source[i]>>16)&0x1f,rs1[i],imm[i]); + break; default: //printf (" %s %8x\n",insn[i],source[i]); printf (" %x: %s\n",start+i*4,insn[i]); @@ -7381,10 +7601,12 @@ void new_dynarec_init() PROT_READ | PROT_WRITE | PROT_EXEC, MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS, -1, 0) <= 0) {printf("mmap() failed\n");} +#ifdef MUPEN64 rdword=&readmem_dword; fake_pc.f.r.rs=&readmem_dword; fake_pc.f.r.rt=&readmem_dword; fake_pc.f.r.rd=&readmem_dword; +#endif int n; for(n=0x80000;n<0x80800;n++) invalid_code[n]=1; @@ -7409,32 +7631,44 @@ void new_dynarec_init() memory_map[n]=((u_int)rdram-0x80000000)>>2; for(n=526336;n<1048576;n++) // 0x80800000 .. 0xFFFFFFFF memory_map[n]=-1; +#ifdef MUPEN64 for(n=0;n<0x8000;n++) { // 0 .. 0x7FFFFFFF writemem[n] = write_nomem_new; writememb[n] = write_nomemb_new; writememh[n] = write_nomemh_new; +#ifndef FORCE32 writememd[n] = write_nomemd_new; +#endif readmem[n] = read_nomem_new; readmemb[n] = read_nomemb_new; readmemh[n] = read_nomemh_new; +#ifndef FORCE32 readmemd[n] = read_nomemd_new; +#endif } for(n=0x8000;n<0x8080;n++) { // 0x80000000 .. 0x807FFFFF writemem[n] = write_rdram_new; writememb[n] = write_rdramb_new; writememh[n] = write_rdramh_new; +#ifndef FORCE32 writememd[n] = write_rdramd_new; +#endif } for(n=0xC000;n<0x10000;n++) { // 0xC0000000 .. 0xFFFFFFFF writemem[n] = write_nomem_new; writememb[n] = write_nomemb_new; writememh[n] = write_nomemh_new; +#ifndef FORCE32 writememd[n] = write_nomemd_new; +#endif readmem[n] = read_nomem_new; readmemb[n] = read_nomemb_new; readmemh[n] = read_nomemh_new; +#ifndef FORCE32 readmemd[n] = read_nomemd_new; +#endif } +#endif tlb_hacks(); arch_init(); } @@ -7474,14 +7708,36 @@ int new_recompile_block(int addr) //rlist(); start = (u_int)addr&~3; //assert(((u_int)addr&1)==0); +#ifdef PCSX + if (Config.HLE && start == 0x80001000) { + // XXX: is this enough? Maybe check hleSoftCall? + u_int page=get_page(start); + ll_add(jump_in+page,start,out); + invalid_code[start>>12]=0; + emit_movimm(start,0); + emit_writeword(0,(int)&pcaddr); + emit_jmp((int)new_dyna_leave); // enough?? + return 0; + } + else if ((u_int)addr < 0x00200000) { + // used for BIOS calls mostly? + source = (u_int *)((u_int)rdram+start-0); + pagelimit = 0x00200000; + } + else +#endif +#ifdef MUPEN64 if ((int)addr >= 0xa4000000 && (int)addr < 0xa4001000) { source = (u_int *)((u_int)SP_DMEM+start-0xa4000000); pagelimit = 0xa4001000; } - else if ((int)addr >= 0x80000000 && (int)addr < 0x80800000) { + else +#endif + if ((int)addr >= 0x80000000 && (int)addr < 0x80800000) { source = (u_int *)((u_int)rdram+start-0x80000000); pagelimit = 0x80800000; } +#ifndef DISABLE_TLB else if ((signed int)addr >= (signed int)0xC0000000) { //printf("addr=%x mm=%x\n",(u_int)addr,(memory_map[start>>12]<<2)); //if(tlb_LUT_r[start>>12]) @@ -7505,6 +7761,7 @@ int new_recompile_block(int addr) } //printf("source= %x\n",(int)source); } +#endif else { printf("Compile at bogus memory address: %x \n", (int)addr); exit(1); @@ -7792,7 +8049,30 @@ int new_recompile_block(int addr) case 0x3C: strcpy(insn[i],"SCD"); type=NI; break; case 0x3D: strcpy(insn[i],"SDC1"); type=C1LS; break; case 0x3F: strcpy(insn[i],"SD"); type=STORE; break; - default: strcpy(insn[i],"???"); type=NI; break; +#ifdef PCSX + case 0x12: strcpy(insn[i],"COP2"); type=NI; + op2=(source[i]>>21)&0x1f; + switch(op2) + { + case 0x00: strcpy(insn[i],"MFC2"); type=COP2; break; + case 0x02: strcpy(insn[i],"CFC2"); type=COP2; break; + case 0x04: strcpy(insn[i],"MTC2"); type=COP2; break; + case 0x06: strcpy(insn[i],"CTC2"); type=COP2; break; + default: + if (gte_handlers[source[i]&0x3f]!=NULL) { + snprintf(insn[i], sizeof(insn[i]), "COP2 %x", source[i]&0x3f); + type=C2OP; + } + break; + } + break; + case 0x32: strcpy(insn[i],"LWC2"); type=C2LS; break; + case 0x3A: strcpy(insn[i],"SWC2"); type=C2LS; break; + case 0x3B: strcpy(insn[i],"HLECALL"); type=HLECALL; break; +#endif + default: strcpy(insn[i],"???"); type=NI; + printf("NI %08x @%08x\n", source[i], addr + i*4); + break; } itype[i]=type; opcode2[i]=op2; @@ -7861,9 +8141,9 @@ int new_recompile_block(int addr) rs2[i]=0; rt1[i]=0; rt2[i]=0; - // The JALR instruction writes to r31. + // The JALR instruction writes to rd. if (op2&1) { - rt1[i]=31; + rt1[i]=(source[i]>>11)&0x1f; } rs2[i]=CCREG; break; @@ -7965,6 +8245,7 @@ int new_recompile_block(int addr) if(op2==16) if((source[i]&0x3f)==0x18) rs2[i]=CCREG; // ERET break; case COP1: + case COP2: rs1[i]=0; rs2[i]=0; rt1[i]=0; @@ -7981,6 +8262,13 @@ int new_recompile_block(int addr) rt2[i]=0; imm[i]=(short)source[i]; break; + case C2LS: + rs1[i]=(source[i]>>21)&0x1F; + rs2[i]=0; + rt1[i]=0; + rt2[i]=0; + imm[i]=(short)source[i]; + break; case FLOAT: case FCONV: rs1[i]=0; @@ -7995,6 +8283,7 @@ int new_recompile_block(int addr) rt2[i]=0; break; case SYSCALL: + case HLECALL: rs1[i]=CCREG; rs2[i]=0; rt1[i]=0; @@ -8018,7 +8307,7 @@ int new_recompile_block(int addr) else ba[i]=-1; /* Is this the end of the block? */ if(i>0&&(itype[i-1]==UJUMP||itype[i-1]==RJUMP||(source[i-1]>>16)==0x1000)) { - if(rt1[i-1]!=31) { // Continue past subroutine call (JAL) + if(rt1[i-1]==0) { // Continue past subroutine call (JAL) done=1; // Does the block continue due to a branch? for(j=i-1;j>=0;j--) @@ -8038,6 +8327,7 @@ int new_recompile_block(int addr) if(i>MAXBLOCK/2) done=1; } if(i>0&&itype[i-1]==SYSCALL&&stop_after_jal) done=1; + if(itype[i-1]==HLECALL) done=1; assert(i0&&(itype[i-1]==RJUMP||itype[i-1]==UJUMP||itype[i-1]==CJUMP||itype[i-1]==SJUMP||itype[i-1]==FJUMP||itype[i]==SYSCALL)) + if(i>0&&(itype[i-1]==RJUMP||itype[i-1]==UJUMP||itype[i-1]==CJUMP||itype[i-1]==SJUMP||itype[i-1]==FJUMP||itype[i]==SYSCALL||itype[i]==HLECALL)) { cc=0; } @@ -9089,13 +9392,13 @@ int new_recompile_block(int addr) if(dep1[i+1]==(regs[i].regmap_entry[hr]&63)) nr|=1<>2; if(t>0&&(itype[t-1]!=UJUMP&&itype[t-1]!=RJUMP&&itype[t-1]!=CJUMP&&itype[t-1]!=SJUMP&&itype[t-1]!=FJUMP)) // loop_preload can't handle jumps into delay slots @@ -9328,7 +9634,8 @@ int new_recompile_block(int addr) else if(branch_regs[i].regmap[hr]>=0) f_regmap[hr]=branch_regs[i].regmap[hr]; if(itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS ||itype[i+1]==SHIFT||itype[i+1]==COP1||itype[i+1]==FLOAT - ||itype[i+1]==FCOMP||itype[i+1]==FCONV) + ||itype[i+1]==FCOMP||itype[i+1]==FCONV + ||itype[i+1]==COP2||itype[i+1]==C2LS||itype[i+1]==C2OP) { // Test both in case the delay slot is ooo, // could be done better... @@ -9367,8 +9674,8 @@ int new_recompile_block(int addr) while(k>1&®s[k-1].regmap[hr]==-1) { if(itype[k-1]==STORE||itype[k-1]==STORELR ||itype[k-1]==C1LS||itype[k-1]==SHIFT||itype[k-1]==COP1 - ||itype[k-1]==FLOAT||itype[k-1]==FCONV - ||itype[k-1]==FCOMP) { + ||itype[k-1]==FLOAT||itype[k-1]==FCONV||itype[k-1]==FCOMP + ||itype[k-1]==COP2||itype[k-1]==C2LS||itype[k-1]==C2OP) { if(count_free_regs(regs[k-1].regmap)<2) { //printf("no free regs for store %x\n",start+(k-1)*4); break; @@ -9477,7 +9784,8 @@ int new_recompile_block(int addr) } if(itype[j]==STORE||itype[j]==STORELR||itype[j]==C1LS ||itype[j]==SHIFT||itype[j]==COP1||itype[j]==FLOAT - ||itype[j]==FCOMP||itype[j]==FCONV) { + ||itype[j]==FCOMP||itype[j]==FCONV + ||itype[j]==COP2||itype[j]==C2LS||itype[j]==C2OP) { if(count_free_regs(regs[j].regmap)<2) { //printf("No free regs for store %x\n",start+j*4); break; @@ -9519,7 +9827,8 @@ int new_recompile_block(int addr) if(regs[j].regmap[HOST_CCREG]!=-1) break; if(itype[j]==STORE||itype[j]==STORELR||itype[j]==C1LS ||itype[j]==SHIFT||itype[j]==COP1||itype[j]==FLOAT - ||itype[j]==FCOMP||itype[j]==FCONV) { + ||itype[j]==FCOMP||itype[j]==FCONV + ||itype[j]==COP2||itype[j]==C2LS||itype[j]==C2OP) { if(count_free_regs(regs[j].regmap)<2) { //printf("no free regs for store %x\n",start+j*4); break; @@ -9552,7 +9861,8 @@ int new_recompile_block(int addr) while(regs[k-1].regmap[HOST_CCREG]==-1) { if(itype[k-1]==STORE||itype[k-1]==STORELR||itype[k-1]==C1LS ||itype[k-1]==SHIFT||itype[k-1]==COP1||itype[k-1]==FLOAT - ||itype[k-1]==FCONV||itype[k-1]==FCOMP) { + ||itype[k-1]==FCONV||itype[k-1]==FCOMP + ||itype[k-1]==COP2||itype[k-1]==C2LS||itype[k-1]==C2OP) { if(count_free_regs(regs[k-1].regmap)<2) { //printf("no free regs for store %x\n",start+(k-1)*4); break; @@ -9583,7 +9893,8 @@ int new_recompile_block(int addr) if(itype[i]!=STORE&&itype[i]!=STORELR&&itype[i]!=C1LS&&itype[i]!=SHIFT&& itype[i]!=NOP&&itype[i]!=MOV&&itype[i]!=ALU&&itype[i]!=SHIFTIMM&& itype[i]!=IMM16&&itype[i]!=LOAD&&itype[i]!=COP1&&itype[i]!=FLOAT&& - itype[i]!=FCONV&&itype[i]!=FCOMP) + itype[i]!=FCONV&&itype[i]!=FCOMP&& + itype[i]!=COP2&&itype[i]!=C2LS&&itype[i]!=C2OP) { memcpy(f_regmap,regs[i].regmap,sizeof(f_regmap)); } @@ -9598,7 +9909,8 @@ int new_recompile_block(int addr) { if(!bt[i+1]) { - if(itype[i]==ALU||itype[i]==MOV||itype[i]==LOAD||itype[i]==SHIFTIMM||itype[i]==IMM16||(itype[i]==COP1&&opcode2[i]<3)) + if(itype[i]==ALU||itype[i]==MOV||itype[i]==LOAD||itype[i]==SHIFTIMM||itype[i]==IMM16 + ||((itype[i]==COP1||itype[i]==COP2)&&opcode2[i]<3)) { if(rs1[i+1]) { if((hr=get_reg(regs[i+1].regmap,rs1[i+1]))>=0) @@ -9665,7 +9977,7 @@ int new_recompile_block(int addr) } } #ifndef HOST_IMM_ADDR32 - if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS) { + if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS||itype[i+1]==C2LS) { hr=get_reg(regs[i+1].regmap,TLREG); if(hr>=0) { int sr=get_reg(regs[i+1].regmap,rs1[i+1]); @@ -9703,7 +10015,8 @@ int new_recompile_block(int addr) } } #endif - if(itype[i+1]==STORE||itype[i+1]==STORELR||opcode[i+1]==0x39||opcode[i+1]==0x3D) { // SB/SH/SW/SD/SWC1/SDC1 + if(itype[i+1]==STORE||itype[i+1]==STORELR + ||(opcode[i+1]&0x3b)==0x39||(opcode[i+1]&0x3b)==0x3a) { // SB/SH/SW/SD/SWC1/SDC1/SWC2/SDC2 if(get_reg(regs[i+1].regmap,rs1[i+1])<0) { hr=get_reg2(regs[i].regmap,regs[i+1].regmap,-1); if(hr<0) hr=get_reg(regs[i+1].regmap,-1); @@ -9722,7 +10035,7 @@ int new_recompile_block(int addr) } } } - if(itype[i+1]==LOADLR||opcode[i+1]==0x31||opcode[i+1]==0x35) { // LWC1/LDC1 + if(itype[i+1]==LOADLR||(opcode[i+1]&0x3b)==0x31||(opcode[i+1]&0x3b)==0x32) { // LWC1/LDC1, LWC2/LDC2 if(get_reg(regs[i+1].regmap,rs1[i+1])<0) { int nr; hr=get_reg(regs[i+1].regmap,FTEMP); @@ -9757,12 +10070,12 @@ int new_recompile_block(int addr) } } } - if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR/*||itype[i+1]==C1LS*/) { + if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR/*||itype[i+1]==C1LS||||itype[i+1]==C2LS*/) { if(itype[i+1]==LOAD) hr=get_reg(regs[i+1].regmap,rt1[i+1]); - if(itype[i+1]==LOADLR||opcode[i+1]==0x31||opcode[i+1]==0x35) // LWC1/LDC1 + if(itype[i+1]==LOADLR||(opcode[i+1]&0x3b)==0x31||(opcode[i+1]&0x3b)==0x32) // LWC1/LDC1, LWC2/LDC2 hr=get_reg(regs[i+1].regmap,FTEMP); - if(itype[i+1]==STORE||itype[i+1]==STORELR||opcode[i+1]==0x39||opcode[i+1]==0x3D) { // SWC1/SDC1 + if(itype[i+1]==STORE||itype[i+1]==STORELR||(opcode[i+1]&0x3b)==0x39||(opcode[i+1]&0x3b)==0x3a) { // SWC1/SDC1/SWC2/SDC2 hr=get_reg(regs[i+1].regmap,AGEN1+((i+1)&1)); if(hr<0) hr=get_reg(regs[i+1].regmap,-1); } @@ -9857,7 +10170,7 @@ int new_recompile_block(int addr) if((regs[i].was32>>dep2[i+1])&1) r32|=1LL<>r)&1) { @@ -9937,6 +10251,7 @@ int new_recompile_block(int addr) else printf(" r%d",r); } } +#endif printf("\n"); #if defined(__i386__) || defined(__x86_64__) printf("pre: eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",regmap_pre[i][0],regmap_pre[i][1],regmap_pre[i][2],regmap_pre[i][3],regmap_pre[i][5],regmap_pre[i][6],regmap_pre[i][7]); @@ -10058,6 +10373,7 @@ int new_recompile_block(int addr) #endif printf("\n"); } +#ifndef FORCE32 printf(" 32:"); for(r=0;r<=CCREG;r++) { if((regs[i].is32>>r)&1) { @@ -10068,6 +10384,7 @@ int new_recompile_block(int addr) } } printf("\n"); +#endif /*printf(" p32:"); for(r=0;r<=CCREG;r++) { if((p32[i]>>r)&1) { @@ -10105,6 +10422,7 @@ int new_recompile_block(int addr) if((branch_regs[i].dirty>>10)&1) printf("r10 "); if((branch_regs[i].dirty>>12)&1) printf("r12 "); #endif +#ifndef FORCE32 printf(" 32:"); for(r=0;r<=CCREG;r++) { if((branch_regs[i].is32>>r)&1) { @@ -10115,6 +10433,7 @@ int new_recompile_block(int addr) } } printf("\n"); +#endif } } @@ -10172,7 +10491,7 @@ int new_recompile_block(int addr) load_regs(regs[i].regmap_entry,regs[i].regmap,regs[i].was32,rs1[i+1],rs1[i+1]); if(rs2[i+1]!=rs1[i+1]&&rs2[i+1]!=rs1[i]&&rs2[i+1]!=rs2[i]) load_regs(regs[i].regmap_entry,regs[i].regmap,regs[i].was32,rs2[i+1],rs2[i+1]); - if(itype[i+1]==STORE||itype[i+1]==STORELR||(opcode[i+1]&0x3b)==0x39) + if(itype[i+1]==STORE||itype[i+1]==STORELR||(opcode[i+1]&0x3b)==0x39||(opcode[i+1]&0x3b)==0x3a) load_regs(regs[i].regmap_entry,regs[i].regmap,regs[i].was32,INVCP,INVCP); } else if(i+1>12; - u_int vpage=page; - if(page>262143&&tlb_LUT_r[vaddr>>12]) page=(tlb_LUT_r[page^0x80000]^0x80000000)>>12; - if(page>2048) page=2048+(page&2047); - if(vpage>262143&&tlb_LUT_r[vaddr>>12]) vpage&=2047; // jump_dirty uses a hash of the virtual address instead - if(vpage>2048) vpage=2048+(vpage&2047); + u_int page=get_page(vaddr); + u_int vpage=get_vpage(vaddr); literal_pool(256); //if(!(is32[i]&(~unneeded_reg_upper[i])&~(1LL<>12;i<=(start+slen*4)>>12;i++) { invalid_code[i]=0; +#ifndef DISABLE_TLB memory_map[i]|=0x40000000; if((signed int)start>=(signed int)0xC0000000) { assert(using_tlb); @@ -10429,6 +10753,7 @@ int new_recompile_block(int addr) memory_map[j]|=0x40000000; //printf("write protect physical page: %x (virtual %x)\n",j<<12,start); } +#endif } /* Pass 10 - Free memory by expiring oldest blocks */ @@ -10485,3 +10810,5 @@ int new_recompile_block(int addr) } return 0; } + +// vim:shiftwidth=2:expandtab