From: notaz Date: Tue, 20 Sep 2016 23:24:37 +0000 (+0300) Subject: Merge branch 'mainline' into libretro X-Git-Tag: r24l~831 X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=9766e77ddec3157cbf8b7da407b99f47467a8f72;hp=38b1da12aade33bf94bdbe71a3988db6f57fb012;p=pcsx_rearmed.git Merge branch 'mainline' into libretro --- diff --git a/Makefile b/Makefile index dc4e07da..29f02fdc 100644 --- a/Makefile +++ b/Makefile @@ -67,7 +67,6 @@ endif OBJS += libpcsxcore/new_dynarec/emu_if.o libpcsxcore/new_dynarec/new_dynarec.o: libpcsxcore/new_dynarec/assem_arm.c \ libpcsxcore/new_dynarec/pcsxmem_inline.c -libpcsxcore/new_dynarec/new_dynarec.o: CFLAGS += -Wno-all -Wno-pointer-sign ifdef DRC_DBG libpcsxcore/new_dynarec/emu_if.o: CFLAGS += -D_FILE_OFFSET_BITS=64 CFLAGS += -DDRC_DBG diff --git a/Makefile.libretro b/Makefile.libretro index 01c1fcfa..0f6608d9 100644 --- a/Makefile.libretro +++ b/Makefile.libretro @@ -185,7 +185,7 @@ else ifeq ($(platform), qnx) DRC_CACHE_BASE = 0 BUILTIN_GPU = neon ARCH = arm - CFLAGS += -DBASE_ADDR_FIXED=0 -D__BLACKBERRY_QNX__ -marm -mcpu=cortex-a9 -mtune=cortex-a9 -mfpu=neon -mfloat-abi=softfp + CFLAGS += -D__BLACKBERRY_QNX__ -marm -mcpu=cortex-a9 -mtune=cortex-a9 -mfpu=neon -mfloat-abi=softfp ASFLAGS += -mcpu=cortex-a9 -mfpu=neon -mfloat-abi=softfp MAIN_LDLIBS += -lsocket LIBPTHREAD := @@ -268,6 +268,17 @@ ifeq ($(NO_UNDEF_CHECK)$(shell ld -v 2> /dev/null | awk '{print $$1}'),GNU) MAIN_LDFLAGS += -Wl,--no-undefined endif +# try to autodetect stuff for the lazy +ifndef ARCH +ARCH = $(shell $(CC) -dumpmachine | awk -F- '{print $$1}') +endif +ifndef HAVE_NEON +HAVE_NEON = $(shell $(CC) -E -dD - < /dev/null 2> /dev/null | grep -q __ARM_NEON__ && echo 1 || echo 0) +endif +ifeq ($(shell ld -v 2> /dev/null | awk '{print $$1}'),GNU) +MAIN_LDFLAGS += -Wl,--no-undefined +endif + TARGET ?= libretro.so PLATFORM = libretro BUILTIN_GPU ?= peops diff --git a/frontend/libretro.c b/frontend/libretro.c index 444719a5..1e865091 100644 --- a/frontend/libretro.c +++ b/frontend/libretro.c @@ -10,6 +10,10 @@ #include #include #include +#ifdef __MACH__ +#include +#include +#endif #include "../libpcsxcore/misc.h" #include "../libpcsxcore/psxcounters.h" @@ -1550,6 +1554,11 @@ void retro_init(void) int i, ret; bool found_bios = false; +#ifdef __MACH__ + // magic sauce to make the dynarec work on iOS + syscall(SYS_ptrace, 0 /*PTRACE_TRACEME*/, 0, 0, 0); +#endif + #ifdef _3DS psxMapHook = pl_3ds_mmap; psxUnmapHook = pl_3ds_munmap; diff --git a/libpcsxcore/new_dynarec/assem_arm.c b/libpcsxcore/new_dynarec/assem_arm.c index 708c8aea..21640f84 100644 --- a/libpcsxcore/new_dynarec/assem_arm.c +++ b/libpcsxcore/new_dynarec/assem_arm.c @@ -19,7 +19,6 @@ * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */ -#ifdef PCSX #include "../gte.h" #define FLAGLESS #include "../gte.h" @@ -27,10 +26,12 @@ #include "../gte_arm.h" #include "../gte_neon.h" #include "pcnt.h" -#endif #include "arm_features.h" -#if !BASE_ADDR_FIXED +#if defined(BASE_ADDR_FIXED) +#elif defined(BASE_ADDR_DYNAMIC) +char *translation_cache; +#else char translation_cache[1 << TARGET_SIZE_2] __attribute__((aligned(4096))); #endif @@ -40,19 +41,16 @@ char translation_cache[1 << TARGET_SIZE_2] __attribute__((aligned(4096))); #define CALLER_SAVE_REGS 0x120f #endif +#define unused __attribute__((unused)) + extern int cycle_count; extern int last_count; extern int pcaddr; extern int pending_exception; extern int branch_target; extern uint64_t readmem_dword; -#ifdef MUPEN64 -extern precomp_instr fake_pc; -#endif extern void *dynarec_local; -extern u_int memory_map[1048576]; extern u_int mini_ht[32][2]; -extern u_int rounding_modes[4]; void indirect_jump_indexed(); void indirect_jump(); @@ -119,13 +117,11 @@ const u_int invalidate_addr_reg[16] = { 0, 0}; -#include "fpu.h" - -unsigned int needs_clear_cache[1<<(TARGET_SIZE_2-17)]; +static u_int needs_clear_cache[1<<(TARGET_SIZE_2-17)]; /* Linker */ -void set_jump_target(int addr,u_int target) +static void set_jump_target(int addr,u_int target) { u_char *ptr=(u_char *)addr; u_int *ptr2=(u_int *)ptr; @@ -159,7 +155,8 @@ void set_jump_target(int addr,u_int target) // This optionally copies the instruction from the target of the branch into // the space before the branch. Works, but the difference in speed is // usually insignificant. -void set_jump_target_fillslot(int addr,u_int target,int copy) +#if 0 +static void set_jump_target_fillslot(int addr,u_int target,int copy) { u_char *ptr=(u_char *)addr; u_int *ptr2=(u_int *)ptr; @@ -188,46 +185,43 @@ void set_jump_target_fillslot(int addr,u_int target,int copy) *ptr2=(*ptr2&0xFF000000)|(((target-(u_int)ptr2-8)<<6)>>8); } } +#endif /* Literal pool */ -add_literal(int addr,int val) +static void add_literal(int addr,int val) { assert(literalcount>6)+8; } // Find the "clean" entry point from a "dirty" entry point // by skipping past the call to verify_code -u_int get_clean_addr(int addr) +static u_int get_clean_addr(int addr) { int *ptr=(int *)addr; #ifndef HAVE_ARMV7 @@ -244,9 +238,8 @@ u_int get_clean_addr(int addr) return (u_int)ptr; } -int verify_dirty(int addr) +static int verify_dirty(u_int *ptr) { - u_int *ptr=(u_int *)addr; #ifndef HAVE_ARMV7 // get from literal pool assert((*ptr&0xFFFF0000)==0xe59f0000); @@ -266,30 +259,18 @@ int verify_dirty(int addr) #endif if((*ptr&0xFF000000)!=0xeb000000) ptr++; assert((*ptr&0xFF000000)==0xeb000000); // bl instruction -#ifndef DISABLE_TLB - u_int verifier=(int)ptr+((signed int)(*ptr<<8)>>6)+8; // get target of bl - if(verifier==(u_int)verify_code_vm||verifier==(u_int)verify_code_ds) { - unsigned int page=source>>12; - unsigned int map_value=memory_map[page]; - if(map_value>=0x80000000) return 0; - while(page<((source+len-1)>>12)) { - if((memory_map[++page]<<2)!=(map_value<<2)) return 0; - } - source = source+(map_value<<2); - } -#endif //printf("verify_dirty: %x %x %x\n",source,copy,len); return !memcmp((void *)source,(void *)copy,len); } // This doesn't necessarily find all clean entry points, just // guarantees that it's not dirty -int isclean(int addr) +static int isclean(int addr) { #ifndef HAVE_ARMV7 - int *ptr=((u_int *)addr)+4; + u_int *ptr=((u_int *)addr)+4; #else - int *ptr=((u_int *)addr)+6; + u_int *ptr=((u_int *)addr)+6; #endif if((*ptr&0xFF000000)!=0xeb000000) ptr++; if((*ptr&0xFF000000)!=0xeb000000) return 1; // bl instruction @@ -300,7 +281,7 @@ int isclean(int addr) } // get source that block at addr was compiled from (host pointers) -void get_bounds(int addr,u_int *start,u_int *end) +static void get_bounds(int addr,u_int *start,u_int *end) { u_int *ptr=(u_int *)addr; #ifndef HAVE_ARMV7 @@ -322,13 +303,6 @@ void get_bounds(int addr,u_int *start,u_int *end) #endif if((*ptr&0xFF000000)!=0xeb000000) ptr++; assert((*ptr&0xFF000000)==0xeb000000); // bl instruction -#ifndef DISABLE_TLB - u_int verifier=(int)ptr+((signed int)(*ptr<<8)>>6)+8; // get target of bl - if(verifier==(u_int)verify_code_vm||verifier==(u_int)verify_code_ds) { - if(memory_map[source>>12]>=0x80000000) source = 0; - else source = source+(memory_map[source>>12]<<2); - } -#endif *start=source; *end=source+len; } @@ -337,25 +311,25 @@ void get_bounds(int addr,u_int *start,u_int *end) // Note: registers are allocated clean (unmodified state) // if you intend to modify the register, you must call dirty_reg(). -void alloc_reg(struct regstat *cur,int i,signed char reg) +static void alloc_reg(struct regstat *cur,int i,signed char reg) { int r,hr; int preferred_reg = (reg&7); if(reg==CCREG) preferred_reg=HOST_CCREG; if(reg==PTEMP||reg==FTEMP) preferred_reg=12; - + // Don't allocate unused registers if((cur->u>>reg)&1) return; - + // see if it's already allocated for(hr=0;hrregmap[hr]==reg) return; } - + // Keep the same mapping if the register was already allocated in a loop preferred_reg = loop_reg(i,reg,preferred_reg); - + // Try to allocate the preferred register if(cur->regmap[preferred_reg]==-1) { cur->regmap[preferred_reg]=reg; @@ -376,7 +350,7 @@ void alloc_reg(struct regstat *cur,int i,signed char reg) cur->isconst&=~(1<uu>>reg)&1) return; - + // see if the upper half is already allocated for(hr=0;hrregmap[hr]==reg+64) return; } - + // Keep the same mapping if the register was already allocated in a loop preferred_reg = loop_reg(i,reg,preferred_reg); - + // Try to allocate the preferred register if(cur->regmap[preferred_reg]==-1) { cur->regmap[preferred_reg]=reg|64; @@ -542,7 +516,7 @@ void alloc_reg64(struct regstat *cur,int i,signed char reg) cur->isconst&=~(1<regmap[hr]==reg) return; } - + // Try to allocate any available register for(hr=HOST_REGS-1;hr>=0;hr--) { if(hr!=EXCLUDE_REG&&cur->regmap[hr]==-1) { @@ -691,7 +665,7 @@ void alloc_reg_temp(struct regstat *cur,int i,signed char reg) return; } } - + // Find an unneeded register for(hr=HOST_REGS-1;hr>=0;hr--) { @@ -720,7 +694,7 @@ void alloc_reg_temp(struct regstat *cur,int i,signed char reg) } } } - + // Ok, now we have to evict someone // Pick a register we hopefully won't need soon // TODO: we might want to follow unconditional jumps here @@ -790,12 +764,13 @@ void alloc_reg_temp(struct regstat *cur,int i,signed char reg) } SysPrintf("This shouldn't happen");exit(1); } + // Allocate a specific ARM register. -void alloc_arm_reg(struct regstat *cur,int i,signed char reg,char hr) +static void alloc_arm_reg(struct regstat *cur,int i,signed char reg,int hr) { int n; int dirty=0; - + // see if it's already allocated (and dealloc it) for(n=0;nregmap[n]=-1; } } - + cur->regmap[hr]=reg; cur->dirty&=~(1<dirty|=dirty<=33554432) { @@ -913,96 +874,99 @@ u_int genjmp(u_int addr) return ((u_int)offset>>2)&0xffffff; } -void emit_mov(int rs,int rt) +static void emit_mov(int rs,int rt) { assem_debug("mov %s,%s\n",regname[rt],regname[rs]); output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)); } -void emit_movs(int rs,int rt) +static void emit_movs(int rs,int rt) { assem_debug("movs %s,%s\n",regname[rt],regname[rs]); output_w32(0xe1b00000|rd_rn_rm(rt,0,rs)); } -void emit_add(int rs1,int rs2,int rt) +static void emit_add(int rs1,int rs2,int rt) { assem_debug("add %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0800000|rd_rn_rm(rt,rs1,rs2)); } -void emit_adds(int rs1,int rs2,int rt) +static void emit_adds(int rs1,int rs2,int rt) { assem_debug("adds %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0900000|rd_rn_rm(rt,rs1,rs2)); } -void emit_adcs(int rs1,int rs2,int rt) +static void emit_adcs(int rs1,int rs2,int rt) { assem_debug("adcs %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0b00000|rd_rn_rm(rt,rs1,rs2)); } -void emit_sbc(int rs1,int rs2,int rt) +static void emit_sbc(int rs1,int rs2,int rt) { assem_debug("sbc %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0c00000|rd_rn_rm(rt,rs1,rs2)); } -void emit_sbcs(int rs1,int rs2,int rt) +static void emit_sbcs(int rs1,int rs2,int rt) { assem_debug("sbcs %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0d00000|rd_rn_rm(rt,rs1,rs2)); } -void emit_neg(int rs, int rt) +static void emit_neg(int rs, int rt) { assem_debug("rsb %s,%s,#0\n",regname[rt],regname[rs]); output_w32(0xe2600000|rd_rn_rm(rt,rs,0)); } -void emit_negs(int rs, int rt) +static void emit_negs(int rs, int rt) { assem_debug("rsbs %s,%s,#0\n",regname[rt],regname[rs]); output_w32(0xe2700000|rd_rn_rm(rt,rs,0)); } -void emit_sub(int rs1,int rs2,int rt) +static void emit_sub(int rs1,int rs2,int rt) { assem_debug("sub %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0400000|rd_rn_rm(rt,rs1,rs2)); } -void emit_subs(int rs1,int rs2,int rt) +static void emit_subs(int rs1,int rs2,int rt) { assem_debug("subs %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0500000|rd_rn_rm(rt,rs1,rs2)); } -void emit_zeroreg(int rt) +static void emit_zeroreg(int rt) { assem_debug("mov %s,#0\n",regname[rt]); output_w32(0xe3a00000|rd_rn_rm(rt,0,0)); } -void emit_loadlp(u_int imm,u_int rt) +static void emit_loadlp(u_int imm,u_int rt) { add_literal((int)out,imm); assem_debug("ldr %s,pc+? [=%x]\n",regname[rt],imm); output_w32(0xe5900000|rd_rn_rm(rt,15,0)); } -void emit_movw(u_int imm,u_int rt) + +static void emit_movw(u_int imm,u_int rt) { assert(imm<65536); assem_debug("movw %s,#%d (0x%x)\n",regname[rt],imm,imm); output_w32(0xe3000000|rd_rn_rm(rt,0,0)|(imm&0xfff)|((imm<<4)&0xf0000)); } -void emit_movt(u_int imm,u_int rt) + +static void emit_movt(u_int imm,u_int rt) { assem_debug("movt %s,#%d (0x%x)\n",regname[rt],imm&0xffff0000,imm&0xffff0000); output_w32(0xe3400000|rd_rn_rm(rt,0,0)|((imm>>16)&0xfff)|((imm>>12)&0xf0000)); } -void emit_movimm(u_int imm,u_int rt) + +static void emit_movimm(u_int imm,u_int rt) { u_int armval; if(genimm(imm,&armval)) { @@ -1029,21 +993,20 @@ void emit_movimm(u_int imm,u_int rt) #endif } } -void emit_pcreladdr(u_int rt) + +static void emit_pcreladdr(u_int rt) { assem_debug("add %s,pc,#?\n",regname[rt]); output_w32(0xe2800000|rd_rn_rm(rt,15,0)); } -void emit_loadreg(int r, int hr) +static void emit_loadreg(int r, int hr) { -#ifdef FORCE32 if(r&64) { SysPrintf("64bit load in 32bit mode!\n"); assert(0); return; } -#endif if((r&63)==0) emit_zeroreg(hr); else { @@ -1060,15 +1023,14 @@ void emit_loadreg(int r, int hr) output_w32(0xe5900000|rd_rn_rm(hr,FP,0)|offset); } } -void emit_storereg(int r, int hr) + +static void emit_storereg(int r, int hr) { -#ifdef FORCE32 if(r&64) { SysPrintf("64bit store in 32bit mode!\n"); assert(0); return; } -#endif int addr=((int)reg)+((r&63)<>4); if((r&63)==HIREG) addr=(int)&hi+((r&64)>>4); if((r&63)==LOREG) addr=(int)&lo+((r&64)>>4); @@ -1080,13 +1042,13 @@ void emit_storereg(int r, int hr) output_w32(0xe5800000|rd_rn_rm(hr,FP,0)|offset); } -void emit_test(int rs, int rt) +static void emit_test(int rs, int rt) { assem_debug("tst %s,%s\n",regname[rs],regname[rt]); output_w32(0xe1100000|rd_rn_rm(0,rs,rt)); } -void emit_testimm(int rs,int imm) +static void emit_testimm(int rs,int imm) { u_int armval; assem_debug("tst %s,#%d\n",regname[rs],imm); @@ -1094,7 +1056,7 @@ void emit_testimm(int rs,int imm) output_w32(0xe3100000|rd_rn_rm(0,rs,0)|armval); } -void emit_testeqimm(int rs,int imm) +static void emit_testeqimm(int rs,int imm) { u_int armval; assem_debug("tsteq %s,$%d\n",regname[rs],imm); @@ -1102,36 +1064,37 @@ void emit_testeqimm(int rs,int imm) output_w32(0x03100000|rd_rn_rm(0,rs,0)|armval); } -void emit_not(int rs,int rt) +static void emit_not(int rs,int rt) { assem_debug("mvn %s,%s\n",regname[rt],regname[rs]); output_w32(0xe1e00000|rd_rn_rm(rt,0,rs)); } -void emit_mvnmi(int rs,int rt) +static void emit_mvnmi(int rs,int rt) { assem_debug("mvnmi %s,%s\n",regname[rt],regname[rs]); output_w32(0x41e00000|rd_rn_rm(rt,0,rs)); } -void emit_and(u_int rs1,u_int rs2,u_int rt) +static void emit_and(u_int rs1,u_int rs2,u_int rt) { assem_debug("and %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0000000|rd_rn_rm(rt,rs1,rs2)); } -void emit_or(u_int rs1,u_int rs2,u_int rt) +static void emit_or(u_int rs1,u_int rs2,u_int rt) { assem_debug("orr %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe1800000|rd_rn_rm(rt,rs1,rs2)); } -void emit_or_and_set_flags(int rs1,int rs2,int rt) + +static void emit_or_and_set_flags(int rs1,int rs2,int rt) { assem_debug("orrs %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe1900000|rd_rn_rm(rt,rs1,rs2)); } -void emit_orrshl_imm(u_int rs,u_int imm,u_int rt) +static void emit_orrshl_imm(u_int rs,u_int imm,u_int rt) { assert(rs<16); assert(rt<16); @@ -1140,7 +1103,7 @@ void emit_orrshl_imm(u_int rs,u_int imm,u_int rt) output_w32(0xe1800000|rd_rn_rm(rt,rt,rs)|(imm<<7)); } -void emit_orrshr_imm(u_int rs,u_int imm,u_int rt) +static void emit_orrshr_imm(u_int rs,u_int imm,u_int rt) { assert(rs<16); assert(rt<16); @@ -1149,13 +1112,13 @@ void emit_orrshr_imm(u_int rs,u_int imm,u_int rt) output_w32(0xe1800020|rd_rn_rm(rt,rt,rs)|(imm<<7)); } -void emit_xor(u_int rs1,u_int rs2,u_int rt) +static void emit_xor(u_int rs1,u_int rs2,u_int rt) { assem_debug("eor %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe0200000|rd_rn_rm(rt,rs1,rs2)); } -void emit_addimm(u_int rs,int imm,u_int rt) +static void emit_addimm(u_int rs,int imm,u_int rt) { assert(rs<16); assert(rt<16); @@ -1196,7 +1159,7 @@ void emit_addimm(u_int rs,int imm,u_int rt) else if(rs!=rt) emit_mov(rs,rt); } -void emit_addimm_and_set_flags(int imm,int rt) +static void emit_addimm_and_set_flags(int imm,int rt) { assert(imm>-65536&&imm<65536); u_int armval; @@ -1218,49 +1181,28 @@ void emit_addimm_and_set_flags(int imm,int rt) output_w32(0xe2900000|rd_rn_imm_shift(rt,rt,imm&0xff,0)); } } -void emit_addimm_no_flags(u_int imm,u_int rt) + +static void emit_addimm_no_flags(u_int imm,u_int rt) { emit_addimm(rt,imm,rt); } -void emit_addnop(u_int r) +static void emit_addnop(u_int r) { assert(r<16); assem_debug("add %s,%s,#0 (nop)\n",regname[r],regname[r]); output_w32(0xe2800000|rd_rn_rm(r,r,0)); } -void emit_adcimm(u_int rs,int imm,u_int rt) +static void emit_adcimm(u_int rs,int imm,u_int rt) { u_int armval; genimm_checked(imm,&armval); assem_debug("adc %s,%s,#%d\n",regname[rt],regname[rs],imm); output_w32(0xe2a00000|rd_rn_rm(rt,rs,0)|armval); } -/*void emit_sbcimm(int imm,u_int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("sbc %s,%s,#%d\n",regname[rt],regname[rt],imm); - output_w32(0xe2c00000|rd_rn_rm(rt,rt,0)|armval); -}*/ -void emit_sbbimm(int imm,u_int rt) -{ - assem_debug("sbb $%d,%%%s\n",imm,regname[rt]); - assert(rt<8); - if(imm<128&&imm>=-128) { - output_byte(0x83); - output_modrm(3,rt,3); - output_byte(imm); - } - else - { - output_byte(0x81); - output_modrm(3,rt,3); - output_w32(imm); - } -} -void emit_rscimm(int rs,int imm,u_int rt) + +static void emit_rscimm(int rs,int imm,u_int rt) { assert(0); u_int armval; @@ -1269,7 +1211,7 @@ void emit_rscimm(int rs,int imm,u_int rt) output_w32(0xe2e00000|rd_rn_rm(rt,rs,0)|armval); } -void emit_addimm64_32(int rsh,int rsl,int imm,int rth,int rtl) +static void emit_addimm64_32(int rsh,int rsl,int imm,int rth,int rtl) { // TODO: if(genimm(imm,&armval)) ... // else @@ -1278,14 +1220,7 @@ void emit_addimm64_32(int rsh,int rsl,int imm,int rth,int rtl) emit_adcimm(rsh,0,rth); } -void emit_sbb(int rs1,int rs2) -{ - assem_debug("sbb %%%s,%%%s\n",regname[rs2],regname[rs1]); - output_byte(0x19); - output_modrm(3,rs1,rs2); -} - -void emit_andimm(int rs,int imm,int rt) +static void emit_andimm(int rs,int imm,int rt) { u_int armval; if(imm==0) { @@ -1321,7 +1256,7 @@ void emit_andimm(int rs,int imm,int rt) } } -void emit_orimm(int rs,int imm,int rt) +static void emit_orimm(int rs,int imm,int rt) { u_int armval; if(imm==0) { @@ -1338,7 +1273,7 @@ void emit_orimm(int rs,int imm,int rt) } } -void emit_xorimm(int rs,int imm,int rt) +static void emit_xorimm(int rs,int imm,int rt) { u_int armval; if(imm==0) { @@ -1355,7 +1290,7 @@ void emit_xorimm(int rs,int imm,int rt) } } -void emit_shlimm(int rs,u_int imm,int rt) +static void emit_shlimm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1364,7 +1299,7 @@ void emit_shlimm(int rs,u_int imm,int rt) output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|(imm<<7)); } -void emit_lsls_imm(int rs,int imm,int rt) +static void emit_lsls_imm(int rs,int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1372,7 +1307,7 @@ void emit_lsls_imm(int rs,int imm,int rt) output_w32(0xe1b00000|rd_rn_rm(rt,0,rs)|(imm<<7)); } -void emit_lslpls_imm(int rs,int imm,int rt) +static unused void emit_lslpls_imm(int rs,int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1380,7 +1315,7 @@ void emit_lslpls_imm(int rs,int imm,int rt) output_w32(0x51b00000|rd_rn_rm(rt,0,rs)|(imm<<7)); } -void emit_shrimm(int rs,u_int imm,int rt) +static void emit_shrimm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1388,7 +1323,7 @@ void emit_shrimm(int rs,u_int imm,int rt) output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x20|(imm<<7)); } -void emit_sarimm(int rs,u_int imm,int rt) +static void emit_sarimm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1396,7 +1331,7 @@ void emit_sarimm(int rs,u_int imm,int rt) output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x40|(imm<<7)); } -void emit_rorimm(int rs,u_int imm,int rt) +static void emit_rorimm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -1404,7 +1339,7 @@ void emit_rorimm(int rs,u_int imm,int rt) output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x60|(imm<<7)); } -void emit_shldimm(int rs,int rs2,u_int imm,int rt) +static void emit_shldimm(int rs,int rs2,u_int imm,int rt) { assem_debug("shld %%%s,%%%s,%d\n",regname[rt],regname[rs2],imm); assert(imm>0); @@ -1416,7 +1351,7 @@ void emit_shldimm(int rs,int rs2,u_int imm,int rt) output_w32(0xe1800020|rd_rn_rm(rt,rt,rs2)|((32-imm)<<7)); } -void emit_shrdimm(int rs,int rs2,u_int imm,int rt) +static void emit_shrdimm(int rs,int rs2,u_int imm,int rt) { assem_debug("shrd %%%s,%%%s,%d\n",regname[rt],regname[rs2],imm); assert(imm>0); @@ -1428,7 +1363,7 @@ void emit_shrdimm(int rs,int rs2,u_int imm,int rt) output_w32(0xe1800000|rd_rn_rm(rt,rt,rs2)|((32-imm)<<7)); } -void emit_signextend16(int rs,int rt) +static void emit_signextend16(int rs,int rt) { #ifndef HAVE_ARMV6 emit_shlimm(rs,16,rt); @@ -1439,7 +1374,7 @@ void emit_signextend16(int rs,int rt) #endif } -void emit_signextend8(int rs,int rt) +static void emit_signextend8(int rs,int rt) { #ifndef HAVE_ARMV6 emit_shlimm(rs,24,rt); @@ -1450,7 +1385,7 @@ void emit_signextend8(int rs,int rt) #endif } -void emit_shl(u_int rs,u_int shift,u_int rt) +static void emit_shl(u_int rs,u_int shift,u_int rt) { assert(rs<16); assert(rt<16); @@ -1459,7 +1394,8 @@ void emit_shl(u_int rs,u_int shift,u_int rt) assem_debug("lsl %s,%s,%s\n",regname[rt],regname[rs],regname[shift]); output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x10|(shift<<8)); } -void emit_shr(u_int rs,u_int shift,u_int rt) + +static void emit_shr(u_int rs,u_int shift,u_int rt) { assert(rs<16); assert(rt<16); @@ -1467,7 +1403,8 @@ void emit_shr(u_int rs,u_int shift,u_int rt) assem_debug("lsr %s,%s,%s\n",regname[rt],regname[rs],regname[shift]); output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x30|(shift<<8)); } -void emit_sar(u_int rs,u_int shift,u_int rt) + +static void emit_sar(u_int rs,u_int shift,u_int rt) { assert(rs<16); assert(rt<16); @@ -1475,33 +1412,8 @@ void emit_sar(u_int rs,u_int shift,u_int rt) assem_debug("asr %s,%s,%s\n",regname[rt],regname[rs],regname[shift]); output_w32(0xe1a00000|rd_rn_rm(rt,0,rs)|0x50|(shift<<8)); } -void emit_shlcl(int r) -{ - assem_debug("shl %%%s,%%cl\n",regname[r]); - assert(0); -} -void emit_shrcl(int r) -{ - assem_debug("shr %%%s,%%cl\n",regname[r]); - assert(0); -} -void emit_sarcl(int r) -{ - assem_debug("sar %%%s,%%cl\n",regname[r]); - assert(0); -} -void emit_shldcl(int r1,int r2) -{ - assem_debug("shld %%%s,%%%s,%%cl\n",regname[r1],regname[r2]); - assert(0); -} -void emit_shrdcl(int r1,int r2) -{ - assem_debug("shrd %%%s,%%%s,%%cl\n",regname[r1],regname[r2]); - assert(0); -} -void emit_orrshl(u_int rs,u_int shift,u_int rt) +static void emit_orrshl(u_int rs,u_int shift,u_int rt) { assert(rs<16); assert(rt<16); @@ -1509,7 +1421,8 @@ void emit_orrshl(u_int rs,u_int shift,u_int rt) assem_debug("orr %s,%s,%s,lsl %s\n",regname[rt],regname[rt],regname[rs],regname[shift]); output_w32(0xe1800000|rd_rn_rm(rt,rt,rs)|0x10|(shift<<8)); } -void emit_orrshr(u_int rs,u_int shift,u_int rt) + +static void emit_orrshr(u_int rs,u_int shift,u_int rt) { assert(rs<16); assert(rt<16); @@ -1518,7 +1431,7 @@ void emit_orrshr(u_int rs,u_int shift,u_int rt) output_w32(0xe1800000|rd_rn_rm(rt,rt,rs)|0x30|(shift<<8)); } -void emit_cmpimm(int rs,int imm) +static void emit_cmpimm(int rs,int imm) { u_int armval; if(genimm(imm,&armval)) { @@ -1540,85 +1453,79 @@ void emit_cmpimm(int rs,int imm) } } -void emit_cmovne(u_int *addr,int rt) -{ - assem_debug("cmovne %x,%%%s",(int)addr,regname[rt]); - assert(0); -} -void emit_cmovl(u_int *addr,int rt) -{ - assem_debug("cmovl %x,%%%s",(int)addr,regname[rt]); - assert(0); -} -void emit_cmovs(u_int *addr,int rt) -{ - assem_debug("cmovs %x,%%%s",(int)addr,regname[rt]); - assert(0); -} -void emit_cmovne_imm(int imm,int rt) +static void emit_cmovne_imm(int imm,int rt) { assem_debug("movne %s,#%d\n",regname[rt],imm); u_int armval; genimm_checked(imm,&armval); output_w32(0x13a00000|rd_rn_rm(rt,0,0)|armval); } -void emit_cmovl_imm(int imm,int rt) + +static void emit_cmovl_imm(int imm,int rt) { assem_debug("movlt %s,#%d\n",regname[rt],imm); u_int armval; genimm_checked(imm,&armval); output_w32(0xb3a00000|rd_rn_rm(rt,0,0)|armval); } -void emit_cmovb_imm(int imm,int rt) + +static void emit_cmovb_imm(int imm,int rt) { assem_debug("movcc %s,#%d\n",regname[rt],imm); u_int armval; genimm_checked(imm,&armval); output_w32(0x33a00000|rd_rn_rm(rt,0,0)|armval); } -void emit_cmovs_imm(int imm,int rt) + +static void emit_cmovs_imm(int imm,int rt) { assem_debug("movmi %s,#%d\n",regname[rt],imm); u_int armval; genimm_checked(imm,&armval); output_w32(0x43a00000|rd_rn_rm(rt,0,0)|armval); } -void emit_cmove_reg(int rs,int rt) + +static void emit_cmove_reg(int rs,int rt) { assem_debug("moveq %s,%s\n",regname[rt],regname[rs]); output_w32(0x01a00000|rd_rn_rm(rt,0,rs)); } -void emit_cmovne_reg(int rs,int rt) + +static void emit_cmovne_reg(int rs,int rt) { assem_debug("movne %s,%s\n",regname[rt],regname[rs]); output_w32(0x11a00000|rd_rn_rm(rt,0,rs)); } -void emit_cmovl_reg(int rs,int rt) + +static void emit_cmovl_reg(int rs,int rt) { assem_debug("movlt %s,%s\n",regname[rt],regname[rs]); output_w32(0xb1a00000|rd_rn_rm(rt,0,rs)); } -void emit_cmovs_reg(int rs,int rt) + +static void emit_cmovs_reg(int rs,int rt) { assem_debug("movmi %s,%s\n",regname[rt],regname[rs]); output_w32(0x41a00000|rd_rn_rm(rt,0,rs)); } -void emit_slti32(int rs,int imm,int rt) +static void emit_slti32(int rs,int imm,int rt) { if(rs!=rt) emit_zeroreg(rt); emit_cmpimm(rs,imm); if(rs==rt) emit_movimm(0,rt); emit_cmovl_imm(1,rt); } -void emit_sltiu32(int rs,int imm,int rt) + +static void emit_sltiu32(int rs,int imm,int rt) { if(rs!=rt) emit_zeroreg(rt); emit_cmpimm(rs,imm); if(rs==rt) emit_movimm(0,rt); emit_cmovb_imm(1,rt); } -void emit_slti64_32(int rsh,int rsl,int imm,int rt) + +static void emit_slti64_32(int rsh,int rsl,int imm,int rt) { assert(rsh!=rt); emit_slti32(rsl,imm,rt); @@ -1635,7 +1542,8 @@ void emit_slti64_32(int rsh,int rsl,int imm,int rt) emit_cmovl_imm(1,rt); } } -void emit_sltiu64_32(int rsh,int rsl,int imm,int rt) + +static void emit_sltiu64_32(int rsh,int rsl,int imm,int rt) { assert(rsh!=rt); emit_sltiu32(rsl,imm,rt); @@ -1651,26 +1559,29 @@ void emit_sltiu64_32(int rsh,int rsl,int imm,int rt) } } -void emit_cmp(int rs,int rt) +static void emit_cmp(int rs,int rt) { assem_debug("cmp %s,%s\n",regname[rs],regname[rt]); output_w32(0xe1500000|rd_rn_rm(0,rs,rt)); } -void emit_set_gz32(int rs, int rt) + +static void emit_set_gz32(int rs, int rt) { //assem_debug("set_gz32\n"); emit_cmpimm(rs,1); emit_movimm(1,rt); emit_cmovl_imm(0,rt); } -void emit_set_nz32(int rs, int rt) + +static void emit_set_nz32(int rs, int rt) { //assem_debug("set_nz32\n"); if(rs!=rt) emit_movs(rs,rt); else emit_test(rs,rs); emit_cmovne_imm(1,rt); } -void emit_set_gz64_32(int rsh, int rsl, int rt) + +static void emit_set_gz64_32(int rsh, int rsl, int rt) { //assem_debug("set_gz64\n"); emit_set_gz32(rsl,rt); @@ -1678,13 +1589,15 @@ void emit_set_gz64_32(int rsh, int rsl, int rt) emit_cmovne_imm(1,rt); emit_cmovs_imm(0,rt); } -void emit_set_nz64_32(int rsh, int rsl, int rt) + +static void emit_set_nz64_32(int rsh, int rsl, int rt) { //assem_debug("set_nz64\n"); emit_or_and_set_flags(rsh,rsl,rt); emit_cmovne_imm(1,rt); } -void emit_set_if_less32(int rs1, int rs2, int rt) + +static void emit_set_if_less32(int rs1, int rs2, int rt) { //assem_debug("set if less (%%%s,%%%s),%%%s\n",regname[rs1],regname[rs2],regname[rt]); if(rs1!=rt&&rs2!=rt) emit_zeroreg(rt); @@ -1692,7 +1605,8 @@ void emit_set_if_less32(int rs1, int rs2, int rt) if(rs1==rt||rs2==rt) emit_movimm(0,rt); emit_cmovl_imm(1,rt); } -void emit_set_if_carry32(int rs1, int rs2, int rt) + +static void emit_set_if_carry32(int rs1, int rs2, int rt) { //assem_debug("set if carry (%%%s,%%%s),%%%s\n",regname[rs1],regname[rs2],regname[rt]); if(rs1!=rt&&rs2!=rt) emit_zeroreg(rt); @@ -1700,7 +1614,8 @@ void emit_set_if_carry32(int rs1, int rs2, int rt) if(rs1==rt||rs2==rt) emit_movimm(0,rt); emit_cmovb_imm(1,rt); } -void emit_set_if_less64_32(int u1, int l1, int u2, int l2, int rt) + +static void emit_set_if_less64_32(int u1, int l1, int u2, int l2, int rt) { //assem_debug("set if less64 (%%%s,%%%s,%%%s,%%%s),%%%s\n",regname[u1],regname[l1],regname[u2],regname[l2],regname[rt]); assert(u1!=rt); @@ -1710,7 +1625,8 @@ void emit_set_if_less64_32(int u1, int l1, int u2, int l2, int rt) emit_sbcs(u1,u2,HOST_TEMPREG); emit_cmovl_imm(1,rt); } -void emit_set_if_carry64_32(int u1, int l1, int u2, int l2, int rt) + +static void emit_set_if_carry64_32(int u1, int l1, int u2, int l2, int rt) { //assem_debug("set if carry64 (%%%s,%%%s,%%%s,%%%s),%%%s\n",regname[u1],regname[l1],regname[u2],regname[l2],regname[rt]); assert(u1!=rt); @@ -1721,111 +1637,97 @@ void emit_set_if_carry64_32(int u1, int l1, int u2, int l2, int rt) emit_cmovb_imm(1,rt); } -void emit_call(int a) +static void emit_call(int a) { assem_debug("bl %x (%x+%x)\n",a,(int)out,a-(int)out-8); u_int offset=genjmp(a); output_w32(0xeb000000|offset); } -void emit_jmp(int a) + +static void emit_jmp(int a) { assem_debug("b %x (%x+%x)\n",a,(int)out,a-(int)out-8); u_int offset=genjmp(a); output_w32(0xea000000|offset); } -void emit_jne(int a) + +static void emit_jne(int a) { assem_debug("bne %x\n",a); u_int offset=genjmp(a); output_w32(0x1a000000|offset); } -void emit_jeq(int a) + +static void emit_jeq(int a) { assem_debug("beq %x\n",a); u_int offset=genjmp(a); output_w32(0x0a000000|offset); } -void emit_js(int a) + +static void emit_js(int a) { assem_debug("bmi %x\n",a); u_int offset=genjmp(a); output_w32(0x4a000000|offset); } -void emit_jns(int a) + +static void emit_jns(int a) { assem_debug("bpl %x\n",a); u_int offset=genjmp(a); output_w32(0x5a000000|offset); } -void emit_jl(int a) + +static void emit_jl(int a) { assem_debug("blt %x\n",a); u_int offset=genjmp(a); output_w32(0xba000000|offset); } -void emit_jge(int a) + +static void emit_jge(int a) { assem_debug("bge %x\n",a); u_int offset=genjmp(a); output_w32(0xaa000000|offset); } -void emit_jno(int a) + +static void emit_jno(int a) { assem_debug("bvc %x\n",a); u_int offset=genjmp(a); output_w32(0x7a000000|offset); } -void emit_jc(int a) + +static void emit_jc(int a) { assem_debug("bcs %x\n",a); u_int offset=genjmp(a); output_w32(0x2a000000|offset); } -void emit_jcc(int a) + +static void emit_jcc(int a) { assem_debug("bcc %x\n",a); u_int offset=genjmp(a); output_w32(0x3a000000|offset); } -void emit_pushimm(int imm) -{ - assem_debug("push $%x\n",imm); - assert(0); -} -void emit_pusha() -{ - assem_debug("pusha\n"); - assert(0); -} -void emit_popa() -{ - assem_debug("popa\n"); - assert(0); -} -void emit_pushreg(u_int r) -{ - assem_debug("push %%%s\n",regname[r]); - assert(0); -} -void emit_popreg(u_int r) -{ - assem_debug("pop %%%s\n",regname[r]); - assert(0); -} -void emit_callreg(u_int r) +static void emit_callreg(u_int r) { assert(r<15); assem_debug("blx %s\n",regname[r]); output_w32(0xe12fff30|r); } -void emit_jmpreg(u_int r) + +static void emit_jmpreg(u_int r) { assem_debug("mov pc,%s\n",regname[r]); output_w32(0xe1a00000|rd_rn_rm(15,0,r)); } -void emit_readword_indexed(int offset, int rs, int rt) +static void emit_readword_indexed(int offset, int rs, int rt) { assert(offset>-4096&&offset<4096); assem_debug("ldr %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -1835,37 +1737,44 @@ void emit_readword_indexed(int offset, int rs, int rt) output_w32(0xe5100000|rd_rn_rm(rt,rs,0)|(-offset)); } } -void emit_readword_dualindexedx4(int rs1, int rs2, int rt) + +static void emit_readword_dualindexedx4(int rs1, int rs2, int rt) { assem_debug("ldr %s,%s,%s lsl #2\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe7900000|rd_rn_rm(rt,rs1,rs2)|0x100); } -void emit_ldrcc_dualindexed(int rs1, int rs2, int rt) + +static void emit_ldrcc_dualindexed(int rs1, int rs2, int rt) { assem_debug("ldrcc %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x37900000|rd_rn_rm(rt,rs1,rs2)); } -void emit_ldrccb_dualindexed(int rs1, int rs2, int rt) + +static void emit_ldrccb_dualindexed(int rs1, int rs2, int rt) { assem_debug("ldrccb %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x37d00000|rd_rn_rm(rt,rs1,rs2)); } -void emit_ldrccsb_dualindexed(int rs1, int rs2, int rt) + +static void emit_ldrccsb_dualindexed(int rs1, int rs2, int rt) { assem_debug("ldrccsb %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x319000d0|rd_rn_rm(rt,rs1,rs2)); } -void emit_ldrcch_dualindexed(int rs1, int rs2, int rt) + +static void emit_ldrcch_dualindexed(int rs1, int rs2, int rt) { assem_debug("ldrcch %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x319000b0|rd_rn_rm(rt,rs1,rs2)); } -void emit_ldrccsh_dualindexed(int rs1, int rs2, int rt) + +static void emit_ldrccsh_dualindexed(int rs1, int rs2, int rt) { assem_debug("ldrccsh %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x319000f0|rd_rn_rm(rt,rs1,rs2)); } -void emit_readword_indexed_tlb(int addr, int rs, int map, int rt) + +static void emit_readword_indexed_tlb(int addr, int rs, int map, int rt) { if(map<0) emit_readword_indexed(addr, rs, rt); else { @@ -1873,7 +1782,8 @@ void emit_readword_indexed_tlb(int addr, int rs, int map, int rt) emit_readword_dualindexedx4(rs, map, rt); } } -void emit_readdword_indexed_tlb(int addr, int rs, int map, int rh, int rl) + +static void emit_readdword_indexed_tlb(int addr, int rs, int map, int rh, int rl) { if(map<0) { if(rh>=0) emit_readword_indexed(addr, rs, rh); @@ -1885,7 +1795,8 @@ void emit_readdword_indexed_tlb(int addr, int rs, int map, int rh, int rl) emit_readword_indexed_tlb(addr, rs, map, rl); } } -void emit_movsbl_indexed(int offset, int rs, int rt) + +static void emit_movsbl_indexed(int offset, int rs, int rt) { assert(offset>-256&&offset<256); assem_debug("ldrsb %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -1895,7 +1806,8 @@ void emit_movsbl_indexed(int offset, int rs, int rt) output_w32(0xe15000d0|rd_rn_rm(rt,rs,0)|(((-offset)<<4)&0xf00)|((-offset)&0xf)); } } -void emit_movsbl_indexed_tlb(int addr, int rs, int map, int rt) + +static void emit_movsbl_indexed_tlb(int addr, int rs, int map, int rt) { if(map<0) emit_movsbl_indexed(addr, rs, rt); else { @@ -1911,7 +1823,8 @@ void emit_movsbl_indexed_tlb(int addr, int rs, int map, int rt) } } } -void emit_movswl_indexed(int offset, int rs, int rt) + +static void emit_movswl_indexed(int offset, int rs, int rt) { assert(offset>-256&&offset<256); assem_debug("ldrsh %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -1921,7 +1834,8 @@ void emit_movswl_indexed(int offset, int rs, int rt) output_w32(0xe15000f0|rd_rn_rm(rt,rs,0)|(((-offset)<<4)&0xf00)|((-offset)&0xf)); } } -void emit_movzbl_indexed(int offset, int rs, int rt) + +static void emit_movzbl_indexed(int offset, int rs, int rt) { assert(offset>-4096&&offset<4096); assem_debug("ldrb %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -1931,12 +1845,14 @@ void emit_movzbl_indexed(int offset, int rs, int rt) output_w32(0xe5500000|rd_rn_rm(rt,rs,0)|(-offset)); } } -void emit_movzbl_dualindexedx4(int rs1, int rs2, int rt) + +static void emit_movzbl_dualindexedx4(int rs1, int rs2, int rt) { assem_debug("ldrb %s,%s,%s lsl #2\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe7d00000|rd_rn_rm(rt,rs1,rs2)|0x100); } -void emit_movzbl_indexed_tlb(int addr, int rs, int map, int rt) + +static void emit_movzbl_indexed_tlb(int addr, int rs, int map, int rt) { if(map<0) emit_movzbl_indexed(addr, rs, rt); else { @@ -1948,7 +1864,8 @@ void emit_movzbl_indexed_tlb(int addr, int rs, int map, int rt) } } } -void emit_movzwl_indexed(int offset, int rs, int rt) + +static void emit_movzwl_indexed(int offset, int rs, int rt) { assert(offset>-256&&offset<256); assem_debug("ldrh %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -1958,6 +1875,7 @@ void emit_movzwl_indexed(int offset, int rs, int rt) output_w32(0xe15000b0|rd_rn_rm(rt,rs,0)|(((-offset)<<4)&0xf00)|((-offset)&0xf)); } } + static void emit_ldrd(int offset, int rs, int rt) { assert(offset>-256&&offset<256); @@ -1968,53 +1886,48 @@ static void emit_ldrd(int offset, int rs, int rt) output_w32(0xe14000d0|rd_rn_rm(rt,rs,0)|(((-offset)<<4)&0xf00)|((-offset)&0xf)); } } -void emit_readword(int addr, int rt) + +static void emit_readword(int addr, int rt) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<4096); assem_debug("ldr %s,fp+%d\n",regname[rt],offset); output_w32(0xe5900000|rd_rn_rm(rt,FP,0)|offset); } -void emit_movsbl(int addr, int rt) + +static unused void emit_movsbl(int addr, int rt) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<256); assem_debug("ldrsb %s,fp+%d\n",regname[rt],offset); output_w32(0xe1d000d0|rd_rn_rm(rt,FP,0)|((offset<<4)&0xf00)|(offset&0xf)); } -void emit_movswl(int addr, int rt) + +static unused void emit_movswl(int addr, int rt) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<256); assem_debug("ldrsh %s,fp+%d\n",regname[rt],offset); output_w32(0xe1d000f0|rd_rn_rm(rt,FP,0)|((offset<<4)&0xf00)|(offset&0xf)); } -void emit_movzbl(int addr, int rt) + +static unused void emit_movzbl(int addr, int rt) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<4096); assem_debug("ldrb %s,fp+%d\n",regname[rt],offset); output_w32(0xe5d00000|rd_rn_rm(rt,FP,0)|offset); } -void emit_movzwl(int addr, int rt) + +static unused void emit_movzwl(int addr, int rt) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<256); assem_debug("ldrh %s,fp+%d\n",regname[rt],offset); output_w32(0xe1d000b0|rd_rn_rm(rt,FP,0)|((offset<<4)&0xf00)|(offset&0xf)); } -void emit_movzwl_reg(int rs, int rt) -{ - assem_debug("movzwl %%%s,%%%s\n",regname[rs]+1,regname[rt]); - assert(0); -} -void emit_xchg(int rs, int rt) -{ - assem_debug("xchg %%%s,%%%s\n",regname[rs],regname[rt]); - assert(0); -} -void emit_writeword_indexed(int rt, int offset, int rs) +static void emit_writeword_indexed(int rt, int offset, int rs) { assert(offset>-4096&&offset<4096); assem_debug("str %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -2024,12 +1937,14 @@ void emit_writeword_indexed(int rt, int offset, int rs) output_w32(0xe5000000|rd_rn_rm(rt,rs,0)|(-offset)); } } -void emit_writeword_dualindexedx4(int rt, int rs1, int rs2) + +static void emit_writeword_dualindexedx4(int rt, int rs1, int rs2) { assem_debug("str %s,%s,%s lsl #2\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe7800000|rd_rn_rm(rt,rs1,rs2)|0x100); } -void emit_writeword_indexed_tlb(int rt, int addr, int rs, int map, int temp) + +static void emit_writeword_indexed_tlb(int rt, int addr, int rs, int map, int temp) { if(map<0) emit_writeword_indexed(rt, addr, rs); else { @@ -2037,7 +1952,8 @@ void emit_writeword_indexed_tlb(int rt, int addr, int rs, int map, int temp) emit_writeword_dualindexedx4(rt, rs, map); } } -void emit_writedword_indexed_tlb(int rh, int rl, int addr, int rs, int map, int temp) + +static void emit_writedword_indexed_tlb(int rh, int rl, int addr, int rs, int map, int temp) { if(map<0) { if(rh>=0) emit_writeword_indexed(rh, addr, rs); @@ -2053,7 +1969,8 @@ void emit_writedword_indexed_tlb(int rh, int rl, int addr, int rs, int map, int } } } -void emit_writehword_indexed(int rt, int offset, int rs) + +static void emit_writehword_indexed(int rt, int offset, int rs) { assert(offset>-256&&offset<256); assem_debug("strh %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -2063,7 +1980,8 @@ void emit_writehword_indexed(int rt, int offset, int rs) output_w32(0xe14000b0|rd_rn_rm(rt,rs,0)|(((-offset)<<4)&0xf00)|((-offset)&0xf)); } } -void emit_writebyte_indexed(int rt, int offset, int rs) + +static void emit_writebyte_indexed(int rt, int offset, int rs) { assert(offset>-4096&&offset<4096); assem_debug("strb %s,%s+%d\n",regname[rt],regname[rs],offset); @@ -2073,12 +1991,14 @@ void emit_writebyte_indexed(int rt, int offset, int rs) output_w32(0xe5400000|rd_rn_rm(rt,rs,0)|(-offset)); } } -void emit_writebyte_dualindexedx4(int rt, int rs1, int rs2) + +static void emit_writebyte_dualindexedx4(int rt, int rs1, int rs2) { assem_debug("strb %s,%s,%s lsl #2\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0xe7c00000|rd_rn_rm(rt,rs1,rs2)|0x100); } -void emit_writebyte_indexed_tlb(int rt, int addr, int rs, int map, int temp) + +static void emit_writebyte_indexed_tlb(int rt, int addr, int rs, int map, int temp) { if(map<0) emit_writebyte_indexed(rt, addr, rs); else { @@ -2090,64 +2010,50 @@ void emit_writebyte_indexed_tlb(int rt, int addr, int rs, int map, int temp) } } } -void emit_strcc_dualindexed(int rs1, int rs2, int rt) + +static void emit_strcc_dualindexed(int rs1, int rs2, int rt) { assem_debug("strcc %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x37800000|rd_rn_rm(rt,rs1,rs2)); } -void emit_strccb_dualindexed(int rs1, int rs2, int rt) + +static void emit_strccb_dualindexed(int rs1, int rs2, int rt) { assem_debug("strccb %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x37c00000|rd_rn_rm(rt,rs1,rs2)); } -void emit_strcch_dualindexed(int rs1, int rs2, int rt) + +static void emit_strcch_dualindexed(int rs1, int rs2, int rt) { assem_debug("strcch %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x318000b0|rd_rn_rm(rt,rs1,rs2)); } -void emit_writeword(int rt, int addr) + +static void emit_writeword(int rt, int addr) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<4096); assem_debug("str %s,fp+%d\n",regname[rt],offset); output_w32(0xe5800000|rd_rn_rm(rt,FP,0)|offset); } -void emit_writehword(int rt, int addr) + +static unused void emit_writehword(int rt, int addr) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<256); assem_debug("strh %s,fp+%d\n",regname[rt],offset); output_w32(0xe1c000b0|rd_rn_rm(rt,FP,0)|((offset<<4)&0xf00)|(offset&0xf)); } -void emit_writebyte(int rt, int addr) + +static unused void emit_writebyte(int rt, int addr) { u_int offset = addr-(u_int)&dynarec_local; assert(offset<4096); assem_debug("strb %s,fp+%d\n",regname[rt],offset); output_w32(0xe5c00000|rd_rn_rm(rt,FP,0)|offset); } -void emit_writeword_imm(int imm, int addr) -{ - assem_debug("movl $%x,%x\n",imm,addr); - assert(0); -} -void emit_writebyte_imm(int imm, int addr) -{ - assem_debug("movb $%x,%x\n",imm,addr); - assert(0); -} -void emit_mul(int rs) -{ - assem_debug("mul %%%s\n",regname[rs]); - assert(0); -} -void emit_imul(int rs) -{ - assem_debug("imul %%%s\n",regname[rs]); - assert(0); -} -void emit_umull(u_int rs1, u_int rs2, u_int hi, u_int lo) +static void emit_umull(u_int rs1, u_int rs2, u_int hi, u_int lo) { assem_debug("umull %s, %s, %s, %s\n",regname[lo],regname[hi],regname[rs1],regname[rs2]); assert(rs1<16); @@ -2156,7 +2062,8 @@ void emit_umull(u_int rs1, u_int rs2, u_int hi, u_int lo) assert(lo<16); output_w32(0xe0800090|(hi<<16)|(lo<<12)|(rs2<<8)|rs1); } -void emit_smull(u_int rs1, u_int rs2, u_int hi, u_int lo) + +static void emit_smull(u_int rs1, u_int rs2, u_int hi, u_int lo) { assem_debug("smull %s, %s, %s, %s\n",regname[lo],regname[hi],regname[rs1],regname[rs2]); assert(rs1<16); @@ -2166,35 +2073,19 @@ void emit_smull(u_int rs1, u_int rs2, u_int hi, u_int lo) output_w32(0xe0c00090|(hi<<16)|(lo<<12)|(rs2<<8)|rs1); } -void emit_div(int rs) -{ - assem_debug("div %%%s\n",regname[rs]); - assert(0); -} -void emit_idiv(int rs) -{ - assem_debug("idiv %%%s\n",regname[rs]); - assert(0); -} -void emit_cdq() -{ - assem_debug("cdq\n"); - assert(0); -} - -void emit_clz(int rs,int rt) +static void emit_clz(int rs,int rt) { assem_debug("clz %s,%s\n",regname[rt],regname[rs]); output_w32(0xe16f0f10|rd_rn_rm(rt,0,rs)); } -void emit_subcs(int rs1,int rs2,int rt) +static void emit_subcs(int rs1,int rs2,int rt) { assem_debug("subcs %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x20400000|rd_rn_rm(rt,rs1,rs2)); } -void emit_shrcc_imm(int rs,u_int imm,int rt) +static void emit_shrcc_imm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -2202,7 +2093,7 @@ void emit_shrcc_imm(int rs,u_int imm,int rt) output_w32(0x31a00000|rd_rn_rm(rt,0,rs)|0x20|(imm<<7)); } -void emit_shrne_imm(int rs,u_int imm,int rt) +static void emit_shrne_imm(int rs,u_int imm,int rt) { assert(imm>0); assert(imm<32); @@ -2210,73 +2101,73 @@ void emit_shrne_imm(int rs,u_int imm,int rt) output_w32(0x11a00000|rd_rn_rm(rt,0,rs)|0x20|(imm<<7)); } -void emit_negmi(int rs, int rt) +static void emit_negmi(int rs, int rt) { assem_debug("rsbmi %s,%s,#0\n",regname[rt],regname[rs]); output_w32(0x42600000|rd_rn_rm(rt,rs,0)); } -void emit_negsmi(int rs, int rt) +static void emit_negsmi(int rs, int rt) { assem_debug("rsbsmi %s,%s,#0\n",regname[rt],regname[rs]); output_w32(0x42700000|rd_rn_rm(rt,rs,0)); } -void emit_orreq(u_int rs1,u_int rs2,u_int rt) +static void emit_orreq(u_int rs1,u_int rs2,u_int rt) { assem_debug("orreq %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x01800000|rd_rn_rm(rt,rs1,rs2)); } -void emit_orrne(u_int rs1,u_int rs2,u_int rt) +static void emit_orrne(u_int rs1,u_int rs2,u_int rt) { assem_debug("orrne %s,%s,%s\n",regname[rt],regname[rs1],regname[rs2]); output_w32(0x11800000|rd_rn_rm(rt,rs1,rs2)); } -void emit_bic_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_bic_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("bic %s,%s,%s lsl %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0xe1C00000|rd_rn_rm(rt,rs1,rs2)|0x10|(shift<<8)); } -void emit_biceq_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_biceq_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("biceq %s,%s,%s lsl %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0x01C00000|rd_rn_rm(rt,rs1,rs2)|0x10|(shift<<8)); } -void emit_bicne_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_bicne_lsl(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("bicne %s,%s,%s lsl %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0x11C00000|rd_rn_rm(rt,rs1,rs2)|0x10|(shift<<8)); } -void emit_bic_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_bic_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("bic %s,%s,%s lsr %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0xe1C00000|rd_rn_rm(rt,rs1,rs2)|0x30|(shift<<8)); } -void emit_biceq_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_biceq_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("biceq %s,%s,%s lsr %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0x01C00000|rd_rn_rm(rt,rs1,rs2)|0x30|(shift<<8)); } -void emit_bicne_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) +static void emit_bicne_lsr(u_int rs1,u_int rs2,u_int shift,u_int rt) { assem_debug("bicne %s,%s,%s lsr %s\n",regname[rt],regname[rs1],regname[rs2],regname[shift]); output_w32(0x11C00000|rd_rn_rm(rt,rs1,rs2)|0x30|(shift<<8)); } -void emit_teq(int rs, int rt) +static void emit_teq(int rs, int rt) { assem_debug("teq %s,%s\n",regname[rs],regname[rt]); output_w32(0xe1300000|rd_rn_rm(0,rs,rt)); } -void emit_rsbimm(int rs, int imm, int rt) +static void emit_rsbimm(int rs, int imm, int rt) { u_int armval; genimm_checked(imm,&armval); @@ -2285,7 +2176,7 @@ void emit_rsbimm(int rs, int imm, int rt) } // Load 2 immediates optimizing for small code size -void emit_mov2imm_compact(int imm1,u_int rt1,int imm2,u_int rt2) +static void emit_mov2imm_compact(int imm1,u_int rt1,int imm2,u_int rt2) { emit_movimm(imm1,rt1); u_int armval; @@ -2301,7 +2192,7 @@ void emit_mov2imm_compact(int imm1,u_int rt1,int imm2,u_int rt2) // Conditionally select one of two immediates, optimizing for small code size // This will only be called if HAVE_CMOV_IMM is defined -void emit_cmov2imm_e_ne_compact(int imm1,int imm2,u_int rt) +static void emit_cmov2imm_e_ne_compact(int imm1,int imm2,u_int rt) { u_int armval; if(genimm(imm2-imm1,&armval)) { @@ -2335,13 +2226,7 @@ void emit_cmov2imm_e_ne_compact(int imm1,int imm2,u_int rt) } // special case for checking invalid_code -void emit_cmpmem_indexedsr12_imm(int addr,int r,int imm) -{ - assert(0); -} - -// special case for checking invalid_code -void emit_cmpmem_indexedsr12_reg(int base,int r,int imm) +static void emit_cmpmem_indexedsr12_reg(int base,int r,int imm) { assert(imm<128&&imm>=0); assert(r>=0&&r<16); @@ -2350,14 +2235,7 @@ void emit_cmpmem_indexedsr12_reg(int base,int r,int imm) emit_cmpimm(HOST_TEMPREG,imm); } -// special case for tlb mapping -void emit_addsr12(int rs1,int rs2,int rt) -{ - assem_debug("add %s,%s,%s lsr #12\n",regname[rt],regname[rs1],regname[rs2]); - output_w32(0xe0800620|rd_rn_rm(rt,rs1,rs2)); -} - -void emit_callne(int a) +static void emit_callne(int a) { assem_debug("blne %x\n",a); u_int offset=genjmp(a); @@ -2365,260 +2243,90 @@ void emit_callne(int a) } // Used to preload hash table entries -void emit_prefetch(void *addr) -{ - assem_debug("prefetch %x\n",(int)addr); - output_byte(0x0F); - output_byte(0x18); - output_modrm(0,5,1); - output_w32((int)addr); -} -void emit_prefetchreg(int r) +static unused void emit_prefetchreg(int r) { assem_debug("pld %s\n",regname[r]); output_w32(0xf5d0f000|rd_rn_rm(0,r,0)); } // Special case for mini_ht -void emit_ldreq_indexed(int rs, u_int offset, int rt) +static void emit_ldreq_indexed(int rs, u_int offset, int rt) { assert(offset<4096); assem_debug("ldreq %s,[%s, #%d]\n",regname[rt],regname[rs],offset); output_w32(0x05900000|rd_rn_rm(rt,rs,0)|offset); } -void emit_flds(int r,int sr) +static unused void emit_bicne_imm(int rs,int imm,int rt) { - assem_debug("flds s%d,[%s]\n",sr,regname[r]); - output_w32(0xed900a00|((sr&14)<<11)|((sr&1)<<22)|(r<<16)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("bicne %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x13c00000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_vldr(int r,int vr) +static unused void emit_biccs_imm(int rs,int imm,int rt) { - assem_debug("vldr d%d,[%s]\n",vr,regname[r]); - output_w32(0xed900b00|(vr<<12)|(r<<16)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("biccs %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x23c00000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_fsts(int sr,int r) +static unused void emit_bicvc_imm(int rs,int imm,int rt) { - assem_debug("fsts s%d,[%s]\n",sr,regname[r]); - output_w32(0xed800a00|((sr&14)<<11)|((sr&1)<<22)|(r<<16)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("bicvc %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x73c00000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_vstr(int vr,int r) +static unused void emit_bichi_imm(int rs,int imm,int rt) { - assem_debug("vstr d%d,[%s]\n",vr,regname[r]); - output_w32(0xed800b00|(vr<<12)|(r<<16)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("bichi %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x83c00000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_ftosizs(int s,int d) +static unused void emit_orrvs_imm(int rs,int imm,int rt) { - assem_debug("ftosizs s%d,s%d\n",d,s); - output_w32(0xeebd0ac0|((d&14)<<11)|((d&1)<<22)|((s&14)>>1)|((s&1)<<5)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("orrvs %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x63800000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_ftosizd(int s,int d) +static void emit_orrne_imm(int rs,int imm,int rt) { - assem_debug("ftosizd s%d,d%d\n",d,s); - output_w32(0xeebd0bc0|((d&14)<<11)|((d&1)<<22)|(s&7)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("orrne %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x13800000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_fsitos(int s,int d) +static void emit_andne_imm(int rs,int imm,int rt) { - assem_debug("fsitos s%d,s%d\n",d,s); - output_w32(0xeeb80ac0|((d&14)<<11)|((d&1)<<22)|((s&14)>>1)|((s&1)<<5)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("andne %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x12000000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_fsitod(int s,int d) +static unused void emit_addpl_imm(int rs,int imm,int rt) { - assem_debug("fsitod d%d,s%d\n",d,s); - output_w32(0xeeb80bc0|((d&7)<<12)|((s&14)>>1)|((s&1)<<5)); -} + u_int armval; + genimm_checked(imm,&armval); + assem_debug("addpl %s,%s,#%d\n",regname[rt],regname[rs],imm); + output_w32(0x52800000|rd_rn_rm(rt,rs,0)|armval); +} -void emit_fcvtds(int s,int d) +static void emit_jno_unlikely(int a) { - assem_debug("fcvtds d%d,s%d\n",d,s); - output_w32(0xeeb70ac0|((d&7)<<12)|((s&14)>>1)|((s&1)<<5)); -} - -void emit_fcvtsd(int s,int d) -{ - assem_debug("fcvtsd s%d,d%d\n",d,s); - output_w32(0xeeb70bc0|((d&14)<<11)|((d&1)<<22)|(s&7)); -} - -void emit_fsqrts(int s,int d) -{ - assem_debug("fsqrts d%d,s%d\n",d,s); - output_w32(0xeeb10ac0|((d&14)<<11)|((d&1)<<22)|((s&14)>>1)|((s&1)<<5)); -} - -void emit_fsqrtd(int s,int d) -{ - assem_debug("fsqrtd s%d,d%d\n",d,s); - output_w32(0xeeb10bc0|((d&7)<<12)|(s&7)); -} - -void emit_fabss(int s,int d) -{ - assem_debug("fabss d%d,s%d\n",d,s); - output_w32(0xeeb00ac0|((d&14)<<11)|((d&1)<<22)|((s&14)>>1)|((s&1)<<5)); -} - -void emit_fabsd(int s,int d) -{ - assem_debug("fabsd s%d,d%d\n",d,s); - output_w32(0xeeb00bc0|((d&7)<<12)|(s&7)); -} - -void emit_fnegs(int s,int d) -{ - assem_debug("fnegs d%d,s%d\n",d,s); - output_w32(0xeeb10a40|((d&14)<<11)|((d&1)<<22)|((s&14)>>1)|((s&1)<<5)); -} - -void emit_fnegd(int s,int d) -{ - assem_debug("fnegd s%d,d%d\n",d,s); - output_w32(0xeeb10b40|((d&7)<<12)|(s&7)); -} - -void emit_fadds(int s1,int s2,int d) -{ - assem_debug("fadds s%d,s%d,s%d\n",d,s1,s2); - output_w32(0xee300a00|((d&14)<<11)|((d&1)<<22)|((s1&14)<<15)|((s1&1)<<7)|((s2&14)>>1)|((s2&1)<<5)); -} - -void emit_faddd(int s1,int s2,int d) -{ - assem_debug("faddd d%d,d%d,d%d\n",d,s1,s2); - output_w32(0xee300b00|((d&7)<<12)|((s1&7)<<16)|(s2&7)); -} - -void emit_fsubs(int s1,int s2,int d) -{ - assem_debug("fsubs s%d,s%d,s%d\n",d,s1,s2); - output_w32(0xee300a40|((d&14)<<11)|((d&1)<<22)|((s1&14)<<15)|((s1&1)<<7)|((s2&14)>>1)|((s2&1)<<5)); -} - -void emit_fsubd(int s1,int s2,int d) -{ - assem_debug("fsubd d%d,d%d,d%d\n",d,s1,s2); - output_w32(0xee300b40|((d&7)<<12)|((s1&7)<<16)|(s2&7)); -} - -void emit_fmuls(int s1,int s2,int d) -{ - assem_debug("fmuls s%d,s%d,s%d\n",d,s1,s2); - output_w32(0xee200a00|((d&14)<<11)|((d&1)<<22)|((s1&14)<<15)|((s1&1)<<7)|((s2&14)>>1)|((s2&1)<<5)); -} - -void emit_fmuld(int s1,int s2,int d) -{ - assem_debug("fmuld d%d,d%d,d%d\n",d,s1,s2); - output_w32(0xee200b00|((d&7)<<12)|((s1&7)<<16)|(s2&7)); -} - -void emit_fdivs(int s1,int s2,int d) -{ - assem_debug("fdivs s%d,s%d,s%d\n",d,s1,s2); - output_w32(0xee800a00|((d&14)<<11)|((d&1)<<22)|((s1&14)<<15)|((s1&1)<<7)|((s2&14)>>1)|((s2&1)<<5)); -} - -void emit_fdivd(int s1,int s2,int d) -{ - assem_debug("fdivd d%d,d%d,d%d\n",d,s1,s2); - output_w32(0xee800b00|((d&7)<<12)|((s1&7)<<16)|(s2&7)); -} - -void emit_fcmps(int x,int y) -{ - assem_debug("fcmps s14, s15\n"); - output_w32(0xeeb47a67); -} - -void emit_fcmpd(int x,int y) -{ - assem_debug("fcmpd d6, d7\n"); - output_w32(0xeeb46b47); -} - -void emit_fmstat() -{ - assem_debug("fmstat\n"); - output_w32(0xeef1fa10); -} - -void emit_bicne_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("bicne %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x13c00000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_biccs_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("biccs %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x23c00000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_bicvc_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("bicvc %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x73c00000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_bichi_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("bichi %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x83c00000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_orrvs_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("orrvs %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x63800000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_orrne_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("orrne %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x13800000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_andne_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("andne %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x12000000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_addpl_imm(int rs,int imm,int rt) -{ - u_int armval; - genimm_checked(imm,&armval); - assem_debug("addpl %s,%s,#%d\n",regname[rt],regname[rs],imm); - output_w32(0x52800000|rd_rn_rm(rt,rs,0)|armval); -} - -void emit_jno_unlikely(int a) -{ - //emit_jno(a); - assem_debug("addvc pc,pc,#? (%x)\n",/*a-(int)out-8,*/a); - output_w32(0x72800000|rd_rn_rm(15,15,0)); -} + //emit_jno(a); + assem_debug("addvc pc,pc,#? (%x)\n",/*a-(int)out-8,*/a); + output_w32(0x72800000|rd_rn_rm(15,15,0)); +} static void save_regs_all(u_int reglist) { @@ -2631,6 +2339,7 @@ static void save_regs_all(u_int reglist) assem_debug("}\n"); output_w32(0xe88b0000|reglist); } + static void restore_regs_all(u_int reglist) { int i; @@ -2642,12 +2351,14 @@ static void restore_regs_all(u_int reglist) assem_debug("}\n"); output_w32(0xe89b0000|reglist); } + // Save registers before function call static void save_regs(u_int reglist) { reglist&=CALLER_SAVE_REGS; // only save the caller-save registers, r0-r3, r12 save_regs_all(reglist); } + // Restore registers after function call static void restore_regs(u_int reglist) { @@ -2655,37 +2366,9 @@ static void restore_regs(u_int reglist) restore_regs_all(reglist); } -// Write back consts using r14 so we don't disturb the other registers -void wb_consts(signed char i_regmap[],uint64_t i_is32,u_int i_dirty,int i) -{ - int hr; - for(hr=0;hr=0&&((i_dirty>>hr)&1)) { - if(((regs[i].isconst>>hr)&1)&&i_regmap[hr]>0) { - if(i_regmap[hr]<64 || !((i_is32>>(i_regmap[hr]&63))&1) ) { - int value=constmap[i][hr]; - if(value==0) { - emit_zeroreg(HOST_TEMPREG); - } - else { - emit_movimm(value,HOST_TEMPREG); - } - emit_storereg(i_regmap[hr],HOST_TEMPREG); -#ifndef FORCE32 - if((i_is32>>i_regmap[hr])&1) { - if(value!=-1&&value!=0) emit_sarimm(HOST_TEMPREG,31,HOST_TEMPREG); - emit_storereg(i_regmap[hr]|64,HOST_TEMPREG); - } -#endif - } - } - } - } -} - /* Stubs/epilogue */ -void literal_pool(int n) +static void literal_pool(int n) { if(!literalcount) return; if(n) { @@ -2717,7 +2400,7 @@ void literal_pool(int n) literalcount=0; } -void literal_pool_jumpover(int n) +static void literal_pool_jumpover(int n) { if(!literalcount) return; if(n) { @@ -2729,10 +2412,12 @@ void literal_pool_jumpover(int n) set_jump_target(jaddr,(int)out); } -emit_extjump2(u_int addr, int target, int linker) +static void emit_extjump2(u_int addr, int target, int linker) { u_char *ptr=(u_char *)addr; assert((ptr[3]&0x0e)==0xa); + (void)ptr; + emit_loadlp(target,0); emit_loadlp(addr,1); assert(addr>=BASE_ADDR&&addr<(BASE_ADDR+(1<regmap; - int addr=get_reg(i_regmap,AGEN1+(i&1)); - int rth,rt; - int ds; + int rt; if(itype[i]==C1LS||itype[i]==C2LS||itype[i]==LOADLR) { - rth=get_reg(i_regmap,FTEMP|64); rt=get_reg(i_regmap,FTEMP); }else{ - rth=get_reg(i_regmap,rt1[i]|64); rt=get_reg(i_regmap,rt1[i]); } assert(rs>=0); -#ifdef PCSX int r,temp=-1,temp2=HOST_TEMPREG,regs_saved=0,restore_jump=0; reglist|=(1<=0); - int ftable=0; - if(type==LOADB_STUB||type==LOADBU_STUB) - ftable=(int)readmemb; - if(type==LOADH_STUB||type==LOADHU_STUB) - ftable=(int)readmemh; - if(type==LOADW_STUB) - ftable=(int)readmem; -#ifndef FORCE32 - if(type==LOADD_STUB) - ftable=(int)readmemd; -#endif - assert(ftable!=0); - emit_writeword(rs,(int)&address); - //emit_pusha(); - save_regs(reglist); -#ifndef PCSX - ds=i_regs!=®s[i]; - int real_rs=(itype[i]==LOADLR)?-1:get_reg(i_regmap,rs1[i]); - u_int cmask=ds?-1:(0x100f|~i_regs->wasconst); - if(!ds) load_all_consts(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty&~(1<regmap_entry,i_regs->was32,i_regs->wasdirty&cmask&~(1<>rs1[i])&1)<<1)+ds,3); -#endif - //emit_readword((int)&last_count,temp); - //emit_add(cc,temp,cc); - //emit_writeword(cc,(int)&Count); - //emit_mov(15,14); - emit_call((int)&indirect_jump_indexed); - //emit_callreg(rs); - //emit_readword_dualindexedx4(rs,HOST_TEMPREG,15); -#ifndef PCSX - // We really shouldn't need to update the count here, - // but not doing so causes random crashes... - emit_readword((int)&Count,HOST_TEMPREG); - emit_readword((int)&next_interupt,2); - emit_addimm(HOST_TEMPREG,-2*stubs[n][6]-2,HOST_TEMPREG); - emit_writeword(2,(int)&last_count); - emit_sub(HOST_TEMPREG,2,cc<0?HOST_TEMPREG:cc); - if(cc<0) { - emit_storereg(CCREG,HOST_TEMPREG); - } -#endif - //emit_popa(); - restore_regs(reglist); - //if((cc=get_reg(regmap,CCREG))>=0) { - // emit_loadreg(CCREG,cc); - //} - if(itype[i]==C1LS||itype[i]==C2LS||(rt>=0&&rt1[i]!=0)) { - assert(rt>=0); - if(type==LOADB_STUB) - emit_movsbl((int)&readmem_dword,rt); - if(type==LOADBU_STUB) - emit_movzbl((int)&readmem_dword,rt); - if(type==LOADH_STUB) - emit_movswl((int)&readmem_dword,rt); - if(type==LOADHU_STUB) - emit_movzwl((int)&readmem_dword,rt); - if(type==LOADW_STUB) - emit_readword((int)&readmem_dword,rt); - if(type==LOADD_STUB) { - emit_readword((int)&readmem_dword,rt); - if(rth>=0) emit_readword(((int)&readmem_dword)+4,rth); - } - } - emit_jmp(stubs[n][2]); // return address -#endif // !PCSX } -#ifdef PCSX // return memhandler, or get directly accessable address and return 0 -u_int get_direct_memhandler(void *table,u_int addr,int type,u_int *addr_host) +static u_int get_direct_memhandler(void *table,u_int addr,int type,u_int *addr_host) { u_int l1,l2=0; l1=((u_int *)table)[addr>>12]; @@ -3028,16 +2626,13 @@ u_int get_direct_memhandler(void *table,u_int addr,int type,u_int *addr_host) return l2<<1; } } -#endif -inline_readstub(int type, int i, u_int addr, signed char regmap[], int target, int adj, u_int reglist) +static void inline_readstub(int type, int i, u_int addr, signed char regmap[], int target, int adj, u_int reglist) { int rs=get_reg(regmap,target); - int rth=get_reg(regmap,target|64); int rt=get_reg(regmap,target); if(rs<0) rs=get_reg(regmap,-1); assert(rs>=0); -#ifdef PCSX u_int handler,host_addr=0,is_dynamic,far_call=0; int cc=get_reg(regmap,CCREG); if(pcsx_direct_read(type,addr,CLOCK_ADJUST(adj+1),cc,target?rs:-1,rt)) @@ -3111,90 +2706,9 @@ inline_readstub(int type, int i, u_int addr, signed char regmap[], int target, i } } restore_regs(reglist); -#else // if !PCSX - int ftable=0; - if(type==LOADB_STUB||type==LOADBU_STUB) - ftable=(int)readmemb; - if(type==LOADH_STUB||type==LOADHU_STUB) - ftable=(int)readmemh; - if(type==LOADW_STUB) - ftable=(int)readmem; -#ifndef FORCE32 - if(type==LOADD_STUB) - ftable=(int)readmemd; -#endif - assert(ftable!=0); - if(target==0) - emit_movimm(addr,rs); - emit_writeword(rs,(int)&address); - //emit_pusha(); - save_regs(reglist); -#ifndef PCSX - if((signed int)addr>=(signed int)0xC0000000) { - // Theoretically we can have a pagefault here, if the TLB has never - // been enabled and the address is outside the range 80000000..BFFFFFFF - // Write out the registers so the pagefault can be handled. This is - // a very rare case and likely represents a bug. - int ds=regmap!=regs[i].regmap; - if(!ds) load_all_consts(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty,i); - if(!ds) wb_dirtys(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty); - else wb_dirtys(branch_regs[i-1].regmap_entry,branch_regs[i-1].was32,branch_regs[i-1].wasdirty); - } -#endif - //emit_shrimm(rs,16,1); - int cc=get_reg(regmap,CCREG); - if(cc<0) { - emit_loadreg(CCREG,2); - } - //emit_movimm(ftable,0); - emit_movimm(((u_int *)ftable)[addr>>16],0); - //emit_readword((int)&last_count,12); - emit_addimm(cc<0?2:cc,CLOCK_ADJUST(adj+1),2); -#ifndef PCSX - if((signed int)addr>=(signed int)0xC0000000) { - // Pagefault address - int ds=regmap!=regs[i].regmap; - emit_movimm(start+i*4+(((regs[i].was32>>rs1[i])&1)<<1)+ds,3); - } -#endif - //emit_add(12,2,2); - //emit_writeword(2,(int)&Count); - //emit_call(((u_int *)ftable)[addr>>16]); - emit_call((int)&indirect_jump); -#ifndef PCSX - // We really shouldn't need to update the count here, - // but not doing so causes random crashes... - emit_readword((int)&Count,HOST_TEMPREG); - emit_readword((int)&next_interupt,2); - emit_addimm(HOST_TEMPREG,-CLOCK_ADJUST(adj+1),HOST_TEMPREG); - emit_writeword(2,(int)&last_count); - emit_sub(HOST_TEMPREG,2,cc<0?HOST_TEMPREG:cc); - if(cc<0) { - emit_storereg(CCREG,HOST_TEMPREG); - } -#endif - //emit_popa(); - restore_regs(reglist); - if(rt>=0) { - if(type==LOADB_STUB) - emit_movsbl((int)&readmem_dword,rt); - if(type==LOADBU_STUB) - emit_movzbl((int)&readmem_dword,rt); - if(type==LOADH_STUB) - emit_movswl((int)&readmem_dword,rt); - if(type==LOADHU_STUB) - emit_movzwl((int)&readmem_dword,rt); - if(type==LOADW_STUB) - emit_readword((int)&readmem_dword,rt); - if(type==LOADD_STUB) { - emit_readword((int)&readmem_dword,rt); - if(rth>=0) emit_readword(((int)&readmem_dword)+4,rth); - } - } -#endif // !PCSX } -do_writestub(int n) +static void do_writestub(int n) { assem_debug("do_writestub %x\n",start+stubs[n][3]*4); literal_pool(256); @@ -3205,19 +2719,14 @@ do_writestub(int n) struct regstat *i_regs=(struct regstat *)stubs[n][5]; u_int reglist=stubs[n][7]; signed char *i_regmap=i_regs->regmap; - int addr=get_reg(i_regmap,AGEN1+(i&1)); - int rth,rt,r; - int ds; + int rt,r; if(itype[i]==C1LS||itype[i]==C2LS) { - rth=get_reg(i_regmap,FTEMP|64); rt=get_reg(i_regmap,r=FTEMP); }else{ - rth=get_reg(i_regmap,rs2[i]|64); rt=get_reg(i_regmap,r=rs2[i]); } assert(rs>=0); assert(rt>=0); -#ifdef PCSX int rtmp,temp=-1,temp2=HOST_TEMPREG,regs_saved=0,restore_jump=0,ra; int reglist2=reglist|(1<=0); - int ftable=0; - if(type==STOREB_STUB) - ftable=(int)writememb; - if(type==STOREH_STUB) - ftable=(int)writememh; - if(type==STOREW_STUB) - ftable=(int)writemem; -#ifndef FORCE32 - if(type==STORED_STUB) - ftable=(int)writememd; -#endif - assert(ftable!=0); - emit_writeword(rs,(int)&address); - //emit_shrimm(rs,16,rs); - //emit_movmem_indexedx4(ftable,rs,rs); - if(type==STOREB_STUB) - emit_writebyte(rt,(int)&byte); - if(type==STOREH_STUB) - emit_writehword(rt,(int)&hword); - if(type==STOREW_STUB) - emit_writeword(rt,(int)&word); - if(type==STORED_STUB) { -#ifndef FORCE32 - emit_writeword(rt,(int)&dword); - emit_writeword(r?rth:rt,(int)&dword+4); -#else - SysPrintf("STORED_STUB\n"); -#endif - } - //emit_pusha(); - save_regs(reglist); -#ifndef PCSX - ds=i_regs!=®s[i]; - int real_rs=get_reg(i_regmap,rs1[i]); - u_int cmask=ds?-1:(0x100f|~i_regs->wasconst); - if(!ds) load_all_consts(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty&~(1<regmap_entry,i_regs->was32,i_regs->wasdirty&cmask&~(1<>rs1[i])&1)<<1)+ds,3); -#endif - //emit_readword((int)&last_count,temp); - //emit_addimm(cc,2*stubs[n][5]+2,cc); - //emit_add(cc,temp,cc); - //emit_writeword(cc,(int)&Count); - emit_call((int)&indirect_jump_indexed); - //emit_callreg(rs); - emit_readword((int)&Count,HOST_TEMPREG); - emit_readword((int)&next_interupt,2); - emit_addimm(HOST_TEMPREG,-2*stubs[n][6]-2,HOST_TEMPREG); - emit_writeword(2,(int)&last_count); - emit_sub(HOST_TEMPREG,2,cc<0?HOST_TEMPREG:cc); - if(cc<0) { - emit_storereg(CCREG,HOST_TEMPREG); - } - //emit_popa(); - restore_regs(reglist); - //if((cc=get_reg(regmap,CCREG))>=0) { - // emit_loadreg(CCREG,cc); - //} - emit_jmp(stubs[n][2]); // return address -#endif // !PCSX } -inline_writestub(int type, int i, u_int addr, signed char regmap[], int target, int adj, u_int reglist) +static void inline_writestub(int type, int i, u_int addr, signed char regmap[], int target, int adj, u_int reglist) { int rs=get_reg(regmap,-1); - int rth=get_reg(regmap,target|64); int rt=get_reg(regmap,target); assert(rs>=0); assert(rt>=0); -#ifdef PCSX u_int handler,host_addr=0; handler=get_direct_memhandler(mem_wtab,addr,type,&host_addr); if (handler==0) { @@ -3388,85 +2822,9 @@ inline_writestub(int type, int i, u_int addr, signed char regmap[], int target, if(cc<0) emit_storereg(CCREG,2); restore_regs(reglist); -#else // if !pcsx - int ftable=0; - if(type==STOREB_STUB) - ftable=(int)writememb; - if(type==STOREH_STUB) - ftable=(int)writememh; - if(type==STOREW_STUB) - ftable=(int)writemem; -#ifndef FORCE32 - if(type==STORED_STUB) - ftable=(int)writememd; -#endif - assert(ftable!=0); - emit_writeword(rs,(int)&address); - //emit_shrimm(rs,16,rs); - //emit_movmem_indexedx4(ftable,rs,rs); - if(type==STOREB_STUB) - emit_writebyte(rt,(int)&byte); - if(type==STOREH_STUB) - emit_writehword(rt,(int)&hword); - if(type==STOREW_STUB) - emit_writeword(rt,(int)&word); - if(type==STORED_STUB) { -#ifndef FORCE32 - emit_writeword(rt,(int)&dword); - emit_writeword(target?rth:rt,(int)&dword+4); -#else - SysPrintf("STORED_STUB\n"); -#endif - } - //emit_pusha(); - save_regs(reglist); -#ifndef PCSX - // rearmed note: load_all_consts prevents BIOS boot, some bug? - if((signed int)addr>=(signed int)0xC0000000) { - // Theoretically we can have a pagefault here, if the TLB has never - // been enabled and the address is outside the range 80000000..BFFFFFFF - // Write out the registers so the pagefault can be handled. This is - // a very rare case and likely represents a bug. - int ds=regmap!=regs[i].regmap; - if(!ds) load_all_consts(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty,i); - if(!ds) wb_dirtys(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty); - else wb_dirtys(branch_regs[i-1].regmap_entry,branch_regs[i-1].was32,branch_regs[i-1].wasdirty); - } -#endif - //emit_shrimm(rs,16,1); - int cc=get_reg(regmap,CCREG); - if(cc<0) { - emit_loadreg(CCREG,2); - } - //emit_movimm(ftable,0); - emit_movimm(((u_int *)ftable)[addr>>16],0); - //emit_readword((int)&last_count,12); - emit_addimm(cc<0?2:cc,CLOCK_ADJUST(adj+1),2); -#ifndef PCSX - if((signed int)addr>=(signed int)0xC0000000) { - // Pagefault address - int ds=regmap!=regs[i].regmap; - emit_movimm(start+i*4+(((regs[i].was32>>rs1[i])&1)<<1)+ds,3); - } -#endif - //emit_add(12,2,2); - //emit_writeword(2,(int)&Count); - //emit_call(((u_int *)ftable)[addr>>16]); - emit_call((int)&indirect_jump); - emit_readword((int)&Count,HOST_TEMPREG); - emit_readword((int)&next_interupt,2); - emit_addimm(HOST_TEMPREG,-CLOCK_ADJUST(adj+1),HOST_TEMPREG); - emit_writeword(2,(int)&last_count); - emit_sub(HOST_TEMPREG,2,cc<0?HOST_TEMPREG:cc); - if(cc<0) { - emit_storereg(CCREG,HOST_TEMPREG); - } - //emit_popa(); - restore_regs(reglist); -#endif } -do_unalignedwritestub(int n) +static void do_unalignedwritestub(int n) { assem_debug("do_unalignedwritestub %x\n",start+stubs[n][3]*4); literal_pool(256); @@ -3479,7 +2837,6 @@ do_unalignedwritestub(int n) signed char *i_regmap=i_regs->regmap; int temp2=get_reg(i_regmap,FTEMP); int rt; - int ds, real_rs; rt=get_reg(i_regmap,rs2[i]); assert(rt>=0); assert(addr>=0); @@ -3506,14 +2863,6 @@ do_unalignedwritestub(int n) emit_writeword(temp2,(int)&address); save_regs(reglist); -#ifndef PCSX - ds=i_regs!=®s[i]; - real_rs=get_reg(i_regmap,rs1[i]); - u_int cmask=ds?-1:(0x100f|~i_regs->wasconst); - if(!ds) load_all_consts(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty&~(1<regmap_entry,i_regs->was32,i_regs->wasdirty&cmask&~(1<>rs1[i])&1)<<1)+ds,3); -#endif emit_call((int)&indirect_jump_indexed); restore_regs(reglist); @@ -3567,12 +2912,7 @@ do_unalignedwritestub(int n) #endif } -void printregs(int edi,int esi,int ebp,int esp,int b,int d,int c,int a) -{ - printf("regs: %x %x %x %x %x %x %x (%x)\n",a,b,c,d,ebp,esi,edi,(&edi)[-1]); -} - -do_invstub(int n) +static void do_invstub(int n) { literal_pool(20); u_int reglist=stubs[n][3]; @@ -3587,10 +2927,7 @@ do_invstub(int n) int do_dirty_stub(int i) { assem_debug("do_dirty_stub %x\n",start+i*4); - u_int addr=(int)start<(int)0xC0000000?(u_int)source:(u_int)start; - #ifdef PCSX - addr=(u_int)source; - #endif + u_int addr=(u_int)source; // Careful about the code output here, verify_dirty needs to parse it. #ifndef HAVE_ARMV7 emit_loadlp(addr,1); @@ -3612,7 +2949,7 @@ int do_dirty_stub(int i) return entry; } -void do_dirty_stub_ds() +static void do_dirty_stub_ds() { // Careful about the code output here, verify_dirty needs to parse it. #ifndef HAVE_ARMV7 @@ -3630,7 +2967,7 @@ void do_dirty_stub_ds() emit_call((int)&verify_code_ds); } -do_cop1stub(int n) +static void do_cop1stub(int n) { literal_pool(256); assem_debug("do_cop1stub %x\n",start+stubs[n][3]*4); @@ -3651,105 +2988,9 @@ do_cop1stub(int n) emit_jmp(ds?(int)fp_exception_ds:(int)fp_exception); } -#ifndef DISABLE_TLB - -/* TLB */ - -int do_tlb_r(int s,int ar,int map,int x,int a,int shift,int c,u_int addr) -{ - if(c) { - if((signed int)addr>=(signed int)0xC0000000) { - // address_generation already loaded the const - emit_readword_dualindexedx4(FP,map,map); - } - else - return -1; // No mapping - } - else { - assert(s!=map); - emit_movimm(((int)memory_map-(int)&dynarec_local)>>2,map); - emit_addsr12(map,s,map); - // Schedule this while we wait on the load - //if(x) emit_xorimm(s,x,ar); - if(shift>=0) emit_shlimm(s,3,shift); - if(~a) emit_andimm(s,a,ar); - emit_readword_dualindexedx4(FP,map,map); - } - return map; -} -int do_tlb_r_branch(int map, int c, u_int addr, int *jaddr) -{ - if(!c||(signed int)addr>=(signed int)0xC0000000) { - emit_test(map,map); - *jaddr=(int)out; - emit_js(0); - } - return map; -} - -int gen_tlb_addr_r(int ar, int map) { - if(map>=0) { - assem_debug("add %s,%s,%s lsl #2\n",regname[ar],regname[ar],regname[map]); - output_w32(0xe0800100|rd_rn_rm(ar,ar,map)); - } -} - -int do_tlb_w(int s,int ar,int map,int x,int c,u_int addr) -{ - if(c) { - if(addr<0x80800000||addr>=0xC0000000) { - // address_generation already loaded the const - emit_readword_dualindexedx4(FP,map,map); - } - else - return -1; // No mapping - } - else { - assert(s!=map); - emit_movimm(((int)memory_map-(int)&dynarec_local)>>2,map); - emit_addsr12(map,s,map); - // Schedule this while we wait on the load - //if(x) emit_xorimm(s,x,ar); - emit_readword_dualindexedx4(FP,map,map); - } - return map; -} -int do_tlb_w_branch(int map, int c, u_int addr, int *jaddr) -{ - if(!c||addr<0x80800000||addr>=0xC0000000) { - emit_testimm(map,0x40000000); - *jaddr=(int)out; - emit_jne(0); - } -} - -int gen_tlb_addr_w(int ar, int map) { - if(map>=0) { - assem_debug("add %s,%s,%s lsl #2\n",regname[ar],regname[ar],regname[map]); - output_w32(0xe0800100|rd_rn_rm(ar,ar,map)); - } -} - -// Generate the address of the memory_map entry, relative to dynarec_local -generate_map_const(u_int addr,int reg) { - //printf("generate_map_const(%x,%s)\n",addr,regname[reg]); - emit_movimm((addr>>12)+(((u_int)memory_map-(u_int)&dynarec_local)>>2),reg); -} - -#else - -static int do_tlb_r(int a, ...) { return 0; } -static int do_tlb_r_branch(int a, ...) { return 0; } -static int gen_tlb_addr_r(int a, ...) { return 0; } -static int do_tlb_w(int a, ...) { return 0; } -static int do_tlb_w_branch(int a, ...) { return 0; } -static int gen_tlb_addr_w(int a, ...) { return 0; } - -#endif // DISABLE_TLB - /* Special assem */ -void shift_assemble_arm(int i,struct regstat *i_regs) +static void shift_assemble_arm(int i,struct regstat *i_regs) { if(rt1[i]) { if(opcode2[i]<=0x07) // SLLV/SRLV/SRAV @@ -3859,7 +3100,6 @@ void shift_assemble_arm(int i,struct regstat *i_regs) } } -#ifdef PCSX static void speculate_mov(int rs,int rt) { if(rt!=0) { @@ -3977,13 +3217,10 @@ static int get_ptr_mem_type(u_int a) return MTYPE_A000; return MTYPE_8000; } -#endif static int emit_fastpath_cmp_jump(int i,int addr,int *addr_reg_override) { - int jaddr,type=0; - -#ifdef PCSX + int jaddr=0,type=0; int mr=rs1[i]; if(((smrv_strong|smrv_weak)>>mr)&1) { type=get_ptr_mem_type(smrv[mr]); @@ -4022,7 +3259,6 @@ static int emit_fastpath_cmp_jump(int i,int addr,int *addr_reg_override) type=0; } } -#endif if(type==0) { @@ -4046,7 +3282,7 @@ static int emit_fastpath_cmp_jump(int i,int addr,int *addr_reg_override) #define shift_assemble shift_assemble_arm -void loadlr_assemble_arm(int i,struct regstat *i_regs) +static void loadlr_assemble_arm(int i,struct regstat *i_regs) { int s,th,tl,temp,temp2,addr,map=-1; int offset; @@ -4072,55 +3308,31 @@ void loadlr_assemble_arm(int i,struct regstat *i_regs) c=(i_regs->wasconst>>s)&1; if(c) { memtarget=((signed int)(constmap[i][s]+offset))<(signed int)0x80000000+RAM_SIZE; - if(using_tlb&&((signed int)(constmap[i][s]+offset))>=(signed int)0xC0000000) memtarget=1; } } - if(!using_tlb) { - if(!c) { - #ifdef RAM_OFFSET - map=get_reg(i_regs->regmap,ROREG); - if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); - #endif - emit_shlimm(addr,3,temp); - if (opcode[i]==0x22||opcode[i]==0x26) { - emit_andimm(addr,0xFFFFFFFC,temp2); // LWL/LWR - }else{ - emit_andimm(addr,0xFFFFFFF8,temp2); // LDL/LDR - } - jaddr=emit_fastpath_cmp_jump(i,temp2,&fastload_reg_override); + if(!c) { + #ifdef RAM_OFFSET + map=get_reg(i_regs->regmap,ROREG); + if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); + #endif + emit_shlimm(addr,3,temp); + if (opcode[i]==0x22||opcode[i]==0x26) { + emit_andimm(addr,0xFFFFFFFC,temp2); // LWL/LWR + }else{ + emit_andimm(addr,0xFFFFFFF8,temp2); // LDL/LDR } - else { - if(ram_offset&&memtarget) { - emit_addimm(temp2,ram_offset,HOST_TEMPREG); - fastload_reg_override=HOST_TEMPREG; - } - if (opcode[i]==0x22||opcode[i]==0x26) { - emit_movimm(((constmap[i][s]+offset)<<3)&24,temp); // LWL/LWR - }else{ - emit_movimm(((constmap[i][s]+offset)<<3)&56,temp); // LDL/LDR - } + jaddr=emit_fastpath_cmp_jump(i,temp2,&fastload_reg_override); + } + else { + if(ram_offset&&memtarget) { + emit_addimm(temp2,ram_offset,HOST_TEMPREG); + fastload_reg_override=HOST_TEMPREG; } - }else{ // using tlb - int a; - if(c) { - a=-1; - }else if (opcode[i]==0x22||opcode[i]==0x26) { - a=0xFFFFFFFC; // LWL/LWR + if (opcode[i]==0x22||opcode[i]==0x26) { + emit_movimm(((constmap[i][s]+offset)<<3)&24,temp); // LWL/LWR }else{ - a=0xFFFFFFF8; // LDL/LDR + emit_movimm(((constmap[i][s]+offset)<<3)&56,temp); // LDL/LDR } - map=get_reg(i_regs->regmap,TLREG); - assert(map>=0); - reglist&=~(1<>11)&0x1f; //assert(t>=0); // Why does this happen? OOT is weird if(t>=0&&rt1[i]!=0) { -#ifdef MUPEN64 - emit_addimm(FP,(int)&fake_pc-(int)&dynarec_local,0); - emit_movimm((source[i]>>11)&0x1f,1); - emit_writeword(0,(int)&PC); - emit_writebyte(1,(int)&(fake_pc.f.r.nrd)); - if(copr==9) { - emit_readword((int)&last_count,ECX); - emit_loadreg(CCREG,HOST_CCREG); // TODO: do proper reg alloc - emit_add(HOST_CCREG,ECX,HOST_CCREG); - emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); - emit_writeword(HOST_CCREG,(int)&Count); - } - emit_call((int)MFC0); - emit_readword((int)&readmem_dword,t); -#else emit_readword((int)®_cop0+copr*4,t); -#endif } } else if(opcode2[i]==4) // MTC0 @@ -4231,16 +3427,7 @@ void cop0_assemble(int i,struct regstat *i_regs) signed char s=get_reg(i_regs->regmap,rs1[i]); char copr=(source[i]>>11)&0x1f; assert(s>=0); -#ifdef MUPEN64 - emit_writeword(s,(int)&readmem_dword); - wb_register(rs1[i],i_regs->regmap,i_regs->dirty,i_regs->is32); - emit_addimm(FP,(int)&fake_pc-(int)&dynarec_local,0); - emit_movimm((source[i]>>11)&0x1f,1); - emit_writeword(0,(int)&PC); - emit_writebyte(1,(int)&(fake_pc.f.r.nrd)); -#else wb_register(rs1[i],i_regs->regmap,i_regs->dirty,i_regs->is32); -#endif if(copr==9||copr==11||copr==12||copr==13) { emit_readword((int)&last_count,HOST_TEMPREG); emit_loadreg(CCREG,HOST_CCREG); // TODO: do proper reg alloc @@ -4253,7 +3440,6 @@ void cop0_assemble(int i,struct regstat *i_regs) // The interrupt must be taken immediately, because a subsequent // instruction might disable interrupts again. if(copr==12||copr==13) { -#ifdef PCSX if (is_delayslot) { // burn cycles to cause cc_interrupt, which will // reschedule next_interupt. Relies on CCREG from above. @@ -4267,7 +3453,6 @@ void cop0_assemble(int i,struct regstat *i_regs) emit_loadreg(rs1[i],s); return; } -#endif emit_movimm(start+i*4+4,HOST_TEMPREG); emit_writeword(HOST_TEMPREG,(int)&pcaddr); emit_movimm(0,HOST_TEMPREG); @@ -4275,16 +3460,12 @@ void cop0_assemble(int i,struct regstat *i_regs) } //else if(copr==12&&is_delayslot) emit_call((int)MTC0_R12); //else -#ifdef PCSX if(s==HOST_CCREG) emit_loadreg(rs1[i],1); else if(s!=1) emit_mov(s,1); emit_movimm(copr,0); emit_call((int)pcsx_mtc0); -#else - emit_call((int)MTC0); -#endif if(copr==9||copr==11||copr==12||copr==13) { emit_readword((int)&Count,HOST_CCREG); emit_readword((int)&next_interupt,HOST_TEMPREG); @@ -4307,25 +3488,6 @@ void cop0_assemble(int i,struct regstat *i_regs) else { assert(opcode2[i]==0x10); -#ifndef DISABLE_TLB - if((source[i]&0x3f)==0x01) // TLBR - emit_call((int)TLBR); - if((source[i]&0x3f)==0x02) // TLBWI - emit_call((int)TLBWI_new); - if((source[i]&0x3f)==0x06) { // TLBWR - // The TLB entry written by TLBWR is dependent on the count, - // so update the cycle count - emit_readword((int)&last_count,ECX); - if(i_regs->regmap[HOST_CCREG]!=CCREG) emit_loadreg(CCREG,HOST_CCREG); - emit_add(HOST_CCREG,ECX,HOST_CCREG); - emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); - emit_writeword(HOST_CCREG,(int)&Count); - emit_call((int)TLBWR_new); - } - if((source[i]&0x3f)==0x08) // TLBP - emit_call((int)TLBP); -#endif -#ifdef PCSX if((source[i]&0x3f)==0x10) // RFE { emit_readword((int)&Status,0); @@ -4334,15 +3496,6 @@ void cop0_assemble(int i,struct regstat *i_regs) emit_orrshr_imm(1,2,0); emit_writeword(0,(int)&Status); } -#else - if((source[i]&0x3f)==0x18) // ERET - { - int count=ccadj[i]; - if(i_regs->regmap[HOST_CCREG]!=CCREG) emit_loadreg(CCREG,HOST_CCREG); - emit_addimm(HOST_CCREG,CLOCK_ADJUST(count),HOST_CCREG); // TODO: Should there be an extra cycle here? - emit_jmp((int)jump_eret); - } -#endif } } @@ -4445,7 +3598,7 @@ static void cop2_put_dreg(u_int copr,signed char sl,signed char temp) } } -void cop2_assemble(int i,struct regstat *i_regs) +static void cop2_assemble(int i,struct regstat *i_regs) { u_int copr=(source[i]>>11)&0x1f; signed char temp=get_reg(i_regs->regmap,-1); @@ -4535,7 +3688,6 @@ static void c2op_call_rgb_func(void *func,int lm,int need_ir,int need_flags) static void c2op_assemble(int i,struct regstat *i_regs) { - signed char temp=get_reg(i_regs->regmap,-1); u_int c2op=source[i]&0x3f; u_int hr,reglist_full=0,reglist; int need_flags,need_ir; @@ -4654,7 +3806,7 @@ static void c2op_assemble(int i,struct regstat *i_regs) } } -void cop1_unusable(int i,struct regstat *i_regs) +static void cop1_unusable(int i,struct regstat *i_regs) { // XXX: should just just do the exception instead if(!cop1_usable) { @@ -4665,629 +3817,28 @@ void cop1_unusable(int i,struct regstat *i_regs) } } -void cop1_assemble(int i,struct regstat *i_regs) +static void cop1_assemble(int i,struct regstat *i_regs) { -#ifndef DISABLE_COP1 - // Check cop1 unusable - if(!cop1_usable) { - signed char rs=get_reg(i_regs->regmap,CSREG); - assert(rs>=0); - emit_testimm(rs,0x20000000); - int jaddr=(int)out; - emit_jeq(0); - add_stub(FP_STUB,jaddr,(int)out,i,rs,(int)i_regs,is_delayslot,0); - cop1_usable=1; - } - if (opcode2[i]==0) { // MFC1 - signed char tl=get_reg(i_regs->regmap,rt1[i]); - if(tl>=0) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],tl); - emit_readword_indexed(0,tl,tl); - } - } - else if (opcode2[i]==1) { // DMFC1 - signed char tl=get_reg(i_regs->regmap,rt1[i]); - signed char th=get_reg(i_regs->regmap,rt1[i]|64); - if(tl>=0) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],tl); - if(th>=0) emit_readword_indexed(4,tl,th); - emit_readword_indexed(0,tl,tl); - } - } - else if (opcode2[i]==4) { // MTC1 - signed char sl=get_reg(i_regs->regmap,rs1[i]); - signed char temp=get_reg(i_regs->regmap,-1); - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_writeword_indexed(sl,0,temp); - } - else if (opcode2[i]==5) { // DMTC1 - signed char sl=get_reg(i_regs->regmap,rs1[i]); - signed char sh=rs1[i]>0?get_reg(i_regs->regmap,rs1[i]|64):sl; - signed char temp=get_reg(i_regs->regmap,-1); - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_writeword_indexed(sh,4,temp); - emit_writeword_indexed(sl,0,temp); - } - else if (opcode2[i]==2) // CFC1 - { - signed char tl=get_reg(i_regs->regmap,rt1[i]); - if(tl>=0) { - u_int copr=(source[i]>>11)&0x1f; - if(copr==0) emit_readword((int)&FCR0,tl); - if(copr==31) emit_readword((int)&FCR31,tl); - } - } - else if (opcode2[i]==6) // CTC1 - { - signed char sl=get_reg(i_regs->regmap,rs1[i]); - u_int copr=(source[i]>>11)&0x1f; - assert(sl>=0); - if(copr==31) - { - emit_writeword(sl,(int)&FCR31); - // Set the rounding mode - //FIXME - //char temp=get_reg(i_regs->regmap,-1); - //emit_andimm(sl,3,temp); - //emit_fldcw_indexed((int)&rounding_modes,temp); - } - } -#else cop1_unusable(i, i_regs); -#endif } -void fconv_assemble_arm(int i,struct regstat *i_regs) +static void fconv_assemble_arm(int i,struct regstat *i_regs) { -#ifndef DISABLE_COP1 - signed char temp=get_reg(i_regs->regmap,-1); - assert(temp>=0); - // Check cop1 unusable - if(!cop1_usable) { - signed char rs=get_reg(i_regs->regmap,CSREG); - assert(rs>=0); - emit_testimm(rs,0x20000000); - int jaddr=(int)out; - emit_jeq(0); - add_stub(FP_STUB,jaddr,(int)out,i,rs,(int)i_regs,is_delayslot,0); - cop1_usable=1; - } - - #if(defined(__VFP_FP__) && !defined(__SOFTFP__)) - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0d) { // trunc_w_s - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_flds(temp,15); - emit_ftosizs(15,15); // float->int, truncate - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - emit_fsts(15,temp); - return; - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0d) { // trunc_w_d - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_vldr(temp,7); - emit_ftosizd(7,13); // double->int, truncate - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - emit_fsts(13,temp); - return; - } - - if(opcode2[i]==0x14&&(source[i]&0x3f)==0x20) { // cvt_s_w - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_flds(temp,13); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - emit_fsitos(13,15); - emit_fsts(15,temp); - return; - } - if(opcode2[i]==0x14&&(source[i]&0x3f)==0x21) { // cvt_d_w - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_flds(temp,13); - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],temp); - emit_fsitod(13,7); - emit_vstr(7,temp); - return; - } - - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x21) { // cvt_d_s - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_flds(temp,13); - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],temp); - emit_fcvtds(13,7); - emit_vstr(7,temp); - return; - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x20) { // cvt_s_d - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_vldr(temp,7); - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - emit_fcvtsd(7,13); - emit_fsts(13,temp); - return; - } - #endif - - // C emulation code - - u_int hr,reglist=0; - for(hr=0;hrregmap[hr]>=0) reglist|=1<>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_s_w); - } - if(opcode2[i]==0x14&&(source[i]&0x3f)==0x21) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_d_w); - } - if(opcode2[i]==0x15&&(source[i]&0x3f)==0x20) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_s_l); - } - if(opcode2[i]==0x15&&(source[i]&0x3f)==0x21) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_d_l); - } - - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x21) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_d_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x24) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_w_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x25) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_l_s); - } - - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x20) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_s_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x24) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_w_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x25) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)cvt_l_d); - } - - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x08) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)round_l_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x09) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)trunc_l_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0a) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)ceil_l_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0b) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)floor_l_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0c) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)round_w_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0d) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)trunc_w_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0e) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)ceil_w_s); - } - if(opcode2[i]==0x10&&(source[i]&0x3f)==0x0f) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)floor_w_s); - } - - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x08) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)round_l_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x09) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)trunc_l_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0a) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)ceil_l_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0b) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)floor_l_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0c) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)round_w_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0d) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)trunc_w_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0e) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)ceil_w_d); - } - if(opcode2[i]==0x11&&(source[i]&0x3f)==0x0f) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - emit_call((int)floor_w_d); - } - - restore_regs(reglist); -#else cop1_unusable(i, i_regs); -#endif } #define fconv_assemble fconv_assemble_arm -void fcomp_assemble(int i,struct regstat *i_regs) +static void fcomp_assemble(int i,struct regstat *i_regs) { -#ifndef DISABLE_COP1 - signed char fs=get_reg(i_regs->regmap,FSREG); - signed char temp=get_reg(i_regs->regmap,-1); - assert(temp>=0); - // Check cop1 unusable - if(!cop1_usable) { - signed char cs=get_reg(i_regs->regmap,CSREG); - assert(cs>=0); - emit_testimm(cs,0x20000000); - int jaddr=(int)out; - emit_jeq(0); - add_stub(FP_STUB,jaddr,(int)out,i,cs,(int)i_regs,is_delayslot,0); - cop1_usable=1; - } - - if((source[i]&0x3f)==0x30) { - emit_andimm(fs,~0x800000,fs); - return; - } - - if((source[i]&0x3e)==0x38) { - // sf/ngle - these should throw exceptions for NaNs - emit_andimm(fs,~0x800000,fs); - return; - } - - #if(defined(__VFP_FP__) && !defined(__SOFTFP__)) - if(opcode2[i]==0x10) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_readword((int)®_cop1_simple[(source[i]>>16)&0x1f],HOST_TEMPREG); - emit_orimm(fs,0x800000,fs); - emit_flds(temp,14); - emit_flds(HOST_TEMPREG,15); - emit_fcmps(14,15); - emit_fmstat(); - if((source[i]&0x3f)==0x31) emit_bicvc_imm(fs,0x800000,fs); // c_un_s - if((source[i]&0x3f)==0x32) emit_bicne_imm(fs,0x800000,fs); // c_eq_s - if((source[i]&0x3f)==0x33) {emit_bicne_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ueq_s - if((source[i]&0x3f)==0x34) emit_biccs_imm(fs,0x800000,fs); // c_olt_s - if((source[i]&0x3f)==0x35) {emit_biccs_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ult_s - if((source[i]&0x3f)==0x36) emit_bichi_imm(fs,0x800000,fs); // c_ole_s - if((source[i]&0x3f)==0x37) {emit_bichi_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ule_s - if((source[i]&0x3f)==0x3a) emit_bicne_imm(fs,0x800000,fs); // c_seq_s - if((source[i]&0x3f)==0x3b) emit_bicne_imm(fs,0x800000,fs); // c_ngl_s - if((source[i]&0x3f)==0x3c) emit_biccs_imm(fs,0x800000,fs); // c_lt_s - if((source[i]&0x3f)==0x3d) emit_biccs_imm(fs,0x800000,fs); // c_nge_s - if((source[i]&0x3f)==0x3e) emit_bichi_imm(fs,0x800000,fs); // c_le_s - if((source[i]&0x3f)==0x3f) emit_bichi_imm(fs,0x800000,fs); // c_ngt_s - return; - } - if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],HOST_TEMPREG); - emit_orimm(fs,0x800000,fs); - emit_vldr(temp,6); - emit_vldr(HOST_TEMPREG,7); - emit_fcmpd(6,7); - emit_fmstat(); - if((source[i]&0x3f)==0x31) emit_bicvc_imm(fs,0x800000,fs); // c_un_d - if((source[i]&0x3f)==0x32) emit_bicne_imm(fs,0x800000,fs); // c_eq_d - if((source[i]&0x3f)==0x33) {emit_bicne_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ueq_d - if((source[i]&0x3f)==0x34) emit_biccs_imm(fs,0x800000,fs); // c_olt_d - if((source[i]&0x3f)==0x35) {emit_biccs_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ult_d - if((source[i]&0x3f)==0x36) emit_bichi_imm(fs,0x800000,fs); // c_ole_d - if((source[i]&0x3f)==0x37) {emit_bichi_imm(fs,0x800000,fs);emit_orrvs_imm(fs,0x800000,fs);} // c_ule_d - if((source[i]&0x3f)==0x3a) emit_bicne_imm(fs,0x800000,fs); // c_seq_d - if((source[i]&0x3f)==0x3b) emit_bicne_imm(fs,0x800000,fs); // c_ngl_d - if((source[i]&0x3f)==0x3c) emit_biccs_imm(fs,0x800000,fs); // c_lt_d - if((source[i]&0x3f)==0x3d) emit_biccs_imm(fs,0x800000,fs); // c_nge_d - if((source[i]&0x3f)==0x3e) emit_bichi_imm(fs,0x800000,fs); // c_le_d - if((source[i]&0x3f)==0x3f) emit_bichi_imm(fs,0x800000,fs); // c_ngt_d - return; - } - #endif - - // C only - - u_int hr,reglist=0; - for(hr=0;hrregmap[hr]>=0) reglist|=1<>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_simple[(source[i]>>16)&0x1f],ARG2_REG); - if((source[i]&0x3f)==0x30) emit_call((int)c_f_s); - if((source[i]&0x3f)==0x31) emit_call((int)c_un_s); - if((source[i]&0x3f)==0x32) emit_call((int)c_eq_s); - if((source[i]&0x3f)==0x33) emit_call((int)c_ueq_s); - if((source[i]&0x3f)==0x34) emit_call((int)c_olt_s); - if((source[i]&0x3f)==0x35) emit_call((int)c_ult_s); - if((source[i]&0x3f)==0x36) emit_call((int)c_ole_s); - if((source[i]&0x3f)==0x37) emit_call((int)c_ule_s); - if((source[i]&0x3f)==0x38) emit_call((int)c_sf_s); - if((source[i]&0x3f)==0x39) emit_call((int)c_ngle_s); - if((source[i]&0x3f)==0x3a) emit_call((int)c_seq_s); - if((source[i]&0x3f)==0x3b) emit_call((int)c_ngl_s); - if((source[i]&0x3f)==0x3c) emit_call((int)c_lt_s); - if((source[i]&0x3f)==0x3d) emit_call((int)c_nge_s); - if((source[i]&0x3f)==0x3e) emit_call((int)c_le_s); - if((source[i]&0x3f)==0x3f) emit_call((int)c_ngt_s); - } - if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],ARG2_REG); - if((source[i]&0x3f)==0x30) emit_call((int)c_f_d); - if((source[i]&0x3f)==0x31) emit_call((int)c_un_d); - if((source[i]&0x3f)==0x32) emit_call((int)c_eq_d); - if((source[i]&0x3f)==0x33) emit_call((int)c_ueq_d); - if((source[i]&0x3f)==0x34) emit_call((int)c_olt_d); - if((source[i]&0x3f)==0x35) emit_call((int)c_ult_d); - if((source[i]&0x3f)==0x36) emit_call((int)c_ole_d); - if((source[i]&0x3f)==0x37) emit_call((int)c_ule_d); - if((source[i]&0x3f)==0x38) emit_call((int)c_sf_d); - if((source[i]&0x3f)==0x39) emit_call((int)c_ngle_d); - if((source[i]&0x3f)==0x3a) emit_call((int)c_seq_d); - if((source[i]&0x3f)==0x3b) emit_call((int)c_ngl_d); - if((source[i]&0x3f)==0x3c) emit_call((int)c_lt_d); - if((source[i]&0x3f)==0x3d) emit_call((int)c_nge_d); - if((source[i]&0x3f)==0x3e) emit_call((int)c_le_d); - if((source[i]&0x3f)==0x3f) emit_call((int)c_ngt_d); - } - restore_regs(reglist); - emit_loadreg(FSREG,fs); -#else cop1_unusable(i, i_regs); -#endif } -void float_assemble(int i,struct regstat *i_regs) +static void float_assemble(int i,struct regstat *i_regs) { -#ifndef DISABLE_COP1 - signed char temp=get_reg(i_regs->regmap,-1); - assert(temp>=0); - // Check cop1 unusable - if(!cop1_usable) { - signed char cs=get_reg(i_regs->regmap,CSREG); - assert(cs>=0); - emit_testimm(cs,0x20000000); - int jaddr=(int)out; - emit_jeq(0); - add_stub(FP_STUB,jaddr,(int)out,i,cs,(int)i_regs,is_delayslot,0); - cop1_usable=1; - } - - #if(defined(__VFP_FP__) && !defined(__SOFTFP__)) - if((source[i]&0x3f)==6) // mov - { - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - if(opcode2[i]==0x10) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],HOST_TEMPREG); - emit_readword_indexed(0,temp,temp); - emit_writeword_indexed(temp,0,HOST_TEMPREG); - } - if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],HOST_TEMPREG); - emit_vldr(temp,7); - emit_vstr(7,HOST_TEMPREG); - } - } - return; - } - - if((source[i]&0x3f)>3) - { - if(opcode2[i]==0x10) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - emit_flds(temp,15); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - } - if((source[i]&0x3f)==4) // sqrt - emit_fsqrts(15,15); - if((source[i]&0x3f)==5) // abs - emit_fabss(15,15); - if((source[i]&0x3f)==7) // neg - emit_fnegs(15,15); - emit_fsts(15,temp); - } - if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - emit_vldr(temp,7); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],temp); - } - if((source[i]&0x3f)==4) // sqrt - emit_fsqrtd(7,7); - if((source[i]&0x3f)==5) // abs - emit_fabsd(7,7); - if((source[i]&0x3f)==7) // neg - emit_fnegd(7,7); - emit_vstr(7,temp); - } - return; - } - if((source[i]&0x3f)<4) - { - if(opcode2[i]==0x10) { - emit_readword((int)®_cop1_simple[(source[i]>>11)&0x1f],temp); - } - if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],temp); - } - if(((source[i]>>11)&0x1f)!=((source[i]>>16)&0x1f)) { - if(opcode2[i]==0x10) { - emit_readword((int)®_cop1_simple[(source[i]>>16)&0x1f],HOST_TEMPREG); - emit_flds(temp,15); - emit_flds(HOST_TEMPREG,13); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - if(((source[i]>>16)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - } - } - if((source[i]&0x3f)==0) emit_fadds(15,13,15); - if((source[i]&0x3f)==1) emit_fsubs(15,13,15); - if((source[i]&0x3f)==2) emit_fmuls(15,13,15); - if((source[i]&0x3f)==3) emit_fdivs(15,13,15); - if(((source[i]>>16)&0x1f)==((source[i]>>6)&0x1f)) { - emit_fsts(15,HOST_TEMPREG); - }else{ - emit_fsts(15,temp); - } - } - else if(opcode2[i]==0x11) { - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],HOST_TEMPREG); - emit_vldr(temp,7); - emit_vldr(HOST_TEMPREG,6); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - if(((source[i]>>16)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],temp); - } - } - if((source[i]&0x3f)==0) emit_faddd(7,6,7); - if((source[i]&0x3f)==1) emit_fsubd(7,6,7); - if((source[i]&0x3f)==2) emit_fmuld(7,6,7); - if((source[i]&0x3f)==3) emit_fdivd(7,6,7); - if(((source[i]>>16)&0x1f)==((source[i]>>6)&0x1f)) { - emit_vstr(7,HOST_TEMPREG); - }else{ - emit_vstr(7,temp); - } - } - } - else { - if(opcode2[i]==0x10) { - emit_flds(temp,15); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_simple[(source[i]>>6)&0x1f],temp); - } - if((source[i]&0x3f)==0) emit_fadds(15,15,15); - if((source[i]&0x3f)==1) emit_fsubs(15,15,15); - if((source[i]&0x3f)==2) emit_fmuls(15,15,15); - if((source[i]&0x3f)==3) emit_fdivs(15,15,15); - emit_fsts(15,temp); - } - else if(opcode2[i]==0x11) { - emit_vldr(temp,7); - if(((source[i]>>11)&0x1f)!=((source[i]>>6)&0x1f)) { - emit_readword((int)®_cop1_double[(source[i]>>6)&0x1f],temp); - } - if((source[i]&0x3f)==0) emit_faddd(7,7,7); - if((source[i]&0x3f)==1) emit_fsubd(7,7,7); - if((source[i]&0x3f)==2) emit_fmuld(7,7,7); - if((source[i]&0x3f)==3) emit_fdivd(7,7,7); - emit_vstr(7,temp); - } - } - return; - } - #endif - - u_int hr,reglist=0; - for(hr=0;hrregmap[hr]>=0) reglist|=1<>11)&0x1f],ARG1_REG); - if((source[i]&0x3f)<4) { - emit_readword((int)®_cop1_simple[(source[i]>>16)&0x1f],ARG2_REG); - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG3_REG); - }else{ - emit_readword((int)®_cop1_simple[(source[i]>> 6)&0x1f],ARG2_REG); - } - switch(source[i]&0x3f) - { - case 0x00: emit_call((int)add_s);break; - case 0x01: emit_call((int)sub_s);break; - case 0x02: emit_call((int)mul_s);break; - case 0x03: emit_call((int)div_s);break; - case 0x04: emit_call((int)sqrt_s);break; - case 0x05: emit_call((int)abs_s);break; - case 0x06: emit_call((int)mov_s);break; - case 0x07: emit_call((int)neg_s);break; - } - restore_regs(reglist); - } - if(opcode2[i]==0x11) { // Double precision - save_regs(reglist); - emit_readword((int)®_cop1_double[(source[i]>>11)&0x1f],ARG1_REG); - if((source[i]&0x3f)<4) { - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],ARG2_REG); - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG3_REG); - }else{ - emit_readword((int)®_cop1_double[(source[i]>> 6)&0x1f],ARG2_REG); - } - switch(source[i]&0x3f) - { - case 0x00: emit_call((int)add_d);break; - case 0x01: emit_call((int)sub_d);break; - case 0x02: emit_call((int)mul_d);break; - case 0x03: emit_call((int)div_d);break; - case 0x04: emit_call((int)sqrt_d);break; - case 0x05: emit_call((int)abs_d);break; - case 0x06: emit_call((int)mov_d);break; - case 0x07: emit_call((int)neg_d);break; - } - restore_regs(reglist); - } -#else cop1_unusable(i, i_regs); -#endif } -void multdiv_assemble_arm(int i,struct regstat *i_regs) +static void multdiv_assemble_arm(int i,struct regstat *i_regs) { // case 0x18: MULT // case 0x19: MULTU @@ -5397,183 +3948,7 @@ void multdiv_assemble_arm(int i,struct regstat *i_regs) } } else // 64-bit -#ifndef FORCE32 - { - if(opcode2[i]==0x1C) // DMULT - { - assert(opcode2[i]!=0x1C); - signed char m1h=get_reg(i_regs->regmap,rs1[i]|64); - signed char m1l=get_reg(i_regs->regmap,rs1[i]); - signed char m2h=get_reg(i_regs->regmap,rs2[i]|64); - signed char m2l=get_reg(i_regs->regmap,rs2[i]); - assert(m1h>=0); - assert(m2h>=0); - assert(m1l>=0); - assert(m2l>=0); - emit_pushreg(m2h); - emit_pushreg(m2l); - emit_pushreg(m1h); - emit_pushreg(m1l); - emit_call((int)&mult64); - emit_popreg(m1l); - emit_popreg(m1h); - emit_popreg(m2l); - emit_popreg(m2h); - signed char hih=get_reg(i_regs->regmap,HIREG|64); - signed char hil=get_reg(i_regs->regmap,HIREG); - if(hih>=0) emit_loadreg(HIREG|64,hih); - if(hil>=0) emit_loadreg(HIREG,hil); - signed char loh=get_reg(i_regs->regmap,LOREG|64); - signed char lol=get_reg(i_regs->regmap,LOREG); - if(loh>=0) emit_loadreg(LOREG|64,loh); - if(lol>=0) emit_loadreg(LOREG,lol); - } - if(opcode2[i]==0x1D) // DMULTU - { - signed char m1h=get_reg(i_regs->regmap,rs1[i]|64); - signed char m1l=get_reg(i_regs->regmap,rs1[i]); - signed char m2h=get_reg(i_regs->regmap,rs2[i]|64); - signed char m2l=get_reg(i_regs->regmap,rs2[i]); - assert(m1h>=0); - assert(m2h>=0); - assert(m1l>=0); - assert(m2l>=0); - save_regs(CALLER_SAVE_REGS); - if(m1l!=0) emit_mov(m1l,0); - if(m1h==0) emit_readword((int)&dynarec_local,1); - else if(m1h>1) emit_mov(m1h,1); - if(m2l<2) emit_readword((int)&dynarec_local+m2l*4,2); - else if(m2l>2) emit_mov(m2l,2); - if(m2h<3) emit_readword((int)&dynarec_local+m2h*4,3); - else if(m2h>3) emit_mov(m2h,3); - emit_call((int)&multu64); - restore_regs(CALLER_SAVE_REGS); - signed char hih=get_reg(i_regs->regmap,HIREG|64); - signed char hil=get_reg(i_regs->regmap,HIREG); - signed char loh=get_reg(i_regs->regmap,LOREG|64); - signed char lol=get_reg(i_regs->regmap,LOREG); - /*signed char temp=get_reg(i_regs->regmap,-1); - signed char rh=get_reg(i_regs->regmap,HIREG|64); - signed char rl=get_reg(i_regs->regmap,HIREG); - assert(m1h>=0); - assert(m2h>=0); - assert(m1l>=0); - assert(m2l>=0); - assert(temp>=0); - //emit_mov(m1l,EAX); - //emit_mul(m2l); - emit_umull(rl,rh,m1l,m2l); - emit_storereg(LOREG,rl); - emit_mov(rh,temp); - //emit_mov(m1h,EAX); - //emit_mul(m2l); - emit_umull(rl,rh,m1h,m2l); - emit_adds(rl,temp,temp); - emit_adcimm(rh,0,rh); - emit_storereg(HIREG,rh); - //emit_mov(m2h,EAX); - //emit_mul(m1l); - emit_umull(rl,rh,m1l,m2h); - emit_adds(rl,temp,temp); - emit_adcimm(rh,0,rh); - emit_storereg(LOREG|64,temp); - emit_mov(rh,temp); - //emit_mov(m2h,EAX); - //emit_mul(m1h); - emit_umull(rl,rh,m1h,m2h); - emit_adds(rl,temp,rl); - emit_loadreg(HIREG,temp); - emit_adcimm(rh,0,rh); - emit_adds(rl,temp,rl); - emit_adcimm(rh,0,rh); - // DEBUG - /* - emit_pushreg(m2h); - emit_pushreg(m2l); - emit_pushreg(m1h); - emit_pushreg(m1l); - emit_call((int)&multu64); - emit_popreg(m1l); - emit_popreg(m1h); - emit_popreg(m2l); - emit_popreg(m2h); - signed char hih=get_reg(i_regs->regmap,HIREG|64); - signed char hil=get_reg(i_regs->regmap,HIREG); - if(hih>=0) emit_loadreg(HIREG|64,hih); // DEBUG - if(hil>=0) emit_loadreg(HIREG,hil); // DEBUG - */ - // Shouldn't be necessary - //char loh=get_reg(i_regs->regmap,LOREG|64); - //char lol=get_reg(i_regs->regmap,LOREG); - //if(loh>=0) emit_loadreg(LOREG|64,loh); - //if(lol>=0) emit_loadreg(LOREG,lol); - } - if(opcode2[i]==0x1E) // DDIV - { - signed char d1h=get_reg(i_regs->regmap,rs1[i]|64); - signed char d1l=get_reg(i_regs->regmap,rs1[i]); - signed char d2h=get_reg(i_regs->regmap,rs2[i]|64); - signed char d2l=get_reg(i_regs->regmap,rs2[i]); - assert(d1h>=0); - assert(d2h>=0); - assert(d1l>=0); - assert(d2l>=0); - save_regs(CALLER_SAVE_REGS); - if(d1l!=0) emit_mov(d1l,0); - if(d1h==0) emit_readword((int)&dynarec_local,1); - else if(d1h>1) emit_mov(d1h,1); - if(d2l<2) emit_readword((int)&dynarec_local+d2l*4,2); - else if(d2l>2) emit_mov(d2l,2); - if(d2h<3) emit_readword((int)&dynarec_local+d2h*4,3); - else if(d2h>3) emit_mov(d2h,3); - emit_call((int)&div64); - restore_regs(CALLER_SAVE_REGS); - signed char hih=get_reg(i_regs->regmap,HIREG|64); - signed char hil=get_reg(i_regs->regmap,HIREG); - signed char loh=get_reg(i_regs->regmap,LOREG|64); - signed char lol=get_reg(i_regs->regmap,LOREG); - if(hih>=0) emit_loadreg(HIREG|64,hih); - if(hil>=0) emit_loadreg(HIREG,hil); - if(loh>=0) emit_loadreg(LOREG|64,loh); - if(lol>=0) emit_loadreg(LOREG,lol); - } - if(opcode2[i]==0x1F) // DDIVU - { - //u_int hr,reglist=0; - //for(hr=0;hrregmap[hr]>=0 && (i_regs->regmap[hr]&62)!=HIREG) reglist|=1<regmap,rs1[i]|64); - signed char d1l=get_reg(i_regs->regmap,rs1[i]); - signed char d2h=get_reg(i_regs->regmap,rs2[i]|64); - signed char d2l=get_reg(i_regs->regmap,rs2[i]); - assert(d1h>=0); - assert(d2h>=0); - assert(d1l>=0); - assert(d2l>=0); - save_regs(CALLER_SAVE_REGS); - if(d1l!=0) emit_mov(d1l,0); - if(d1h==0) emit_readword((int)&dynarec_local,1); - else if(d1h>1) emit_mov(d1h,1); - if(d2l<2) emit_readword((int)&dynarec_local+d2l*4,2); - else if(d2l>2) emit_mov(d2l,2); - if(d2h<3) emit_readword((int)&dynarec_local+d2h*4,3); - else if(d2h>3) emit_mov(d2h,3); - emit_call((int)&divu64); - restore_regs(CALLER_SAVE_REGS); - signed char hih=get_reg(i_regs->regmap,HIREG|64); - signed char hil=get_reg(i_regs->regmap,HIREG); - signed char loh=get_reg(i_regs->regmap,LOREG|64); - signed char lol=get_reg(i_regs->regmap,LOREG); - if(hih>=0) emit_loadreg(HIREG|64,hih); - if(hil>=0) emit_loadreg(HIREG,hil); - if(loh>=0) emit_loadreg(LOREG|64,loh); - if(lol>=0) emit_loadreg(LOREG,lol); - } - } -#else - assert(0); -#endif + assert(0); } else { @@ -5588,25 +3963,25 @@ void multdiv_assemble_arm(int i,struct regstat *i_regs) } #define multdiv_assemble multdiv_assemble_arm -void do_preload_rhash(int r) { +static void do_preload_rhash(int r) { // Don't need this for ARM. On x86, this puts the value 0xf8 into the // register. On ARM the hash can be done with a single instruction (below) } -void do_preload_rhtbl(int ht) { +static void do_preload_rhtbl(int ht) { emit_addimm(FP,(int)&mini_ht-(int)&dynarec_local,ht); } -void do_rhash(int rs,int rh) { +static void do_rhash(int rs,int rh) { emit_andimm(rs,0xf8,rh); } -void do_miniht_load(int ht,int rh) { +static void do_miniht_load(int ht,int rh) { assem_debug("ldr %s,[%s,%s]!\n",regname[rh],regname[ht],regname[rh]); output_w32(0xe7b00000|rd_rn_rm(rh,ht,rh)); } -void do_miniht_jump(int rs,int rh,int ht) { +static void do_miniht_jump(int rs,int rh,int ht) { emit_cmp(rh,rs); emit_ldreq_indexed(ht,4,15); #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK @@ -5617,7 +3992,7 @@ void do_miniht_jump(int rs,int rh,int ht) { #endif } -void do_miniht_insert(u_int return_address,int rt,int temp) { +static void do_miniht_insert(u_int return_address,int rt,int temp) { #ifndef HAVE_ARMV7 emit_movimm(return_address,rt); // PC into link register add_to_linker((int)out,return_address,1); @@ -5634,36 +4009,10 @@ void do_miniht_insert(u_int return_address,int rt,int temp) { #endif } -// Sign-extend to 64 bits and write out upper half of a register -// This is useful where we have a 32-bit value in a register, and want to -// keep it in a 32-bit register, but can't guarantee that it won't be read -// as a 64-bit value later. -void wb_sx(signed char pre[],signed char entry[],uint64_t dirty,uint64_t is32_pre,uint64_t is32,uint64_t u,uint64_t uu) -{ -#ifndef FORCE32 - if(is32_pre==is32) return; - int hr,reg; - for(hr=0;hr=0) { - if((dirty>>hr)&1) { - if( ((is32_pre&~is32&~uu)>>reg)&1 ) { - emit_sarimm(hr,31,HOST_TEMPREG); - emit_storereg(reg|64,HOST_TEMPREG); - } - } - } - //} - } - } -#endif -} - -void wb_valid(signed char pre[],signed char entry[],u_int dirty_pre,u_int dirty,uint64_t is32_pre,uint64_t u,uint64_t uu) +static void wb_valid(signed char pre[],signed char entry[],u_int dirty_pre,u_int dirty,uint64_t is32_pre,uint64_t u,uint64_t uu) { //if(dirty_pre==dirty) return; - int hr,reg,new_hr; + int hr,reg; for(hr=0;hr> 12) & 31); + if (!(needs_clear_cache[offset >> 17] & mask)) { + char *start = (char *)((u_long)target & ~4095ul); + start_tcache_write(start, start + 4096); + needs_clear_cache[offset >> 17] |= mask; + } +} + // Clearing the cache is rather slow on ARM Linux, so mark the areas // that need to be cleared, and then only clear these areas once. -void do_clear_cache() +static void do_clear_cache() { int i,j; for (i=0;i<(1<<(TARGET_SIZE_2-17));i++) @@ -5754,7 +4114,7 @@ void do_clear_cache() u_int bitmap=needs_clear_cache[i]; if(bitmap) { u_int start,end; - for(j=0;j<32;j++) + for(j=0;j<32;j++) { if(bitmap&(1< - -extern int FCR0, FCR31; - -void cvt_s_w(int *source,float *dest) -{ - *dest = *source; -} -void cvt_d_w(int *source,double *dest) -{ - *dest = *source; -} -void cvt_s_l(long long *source,float *dest) -{ - *dest = *source; -} -void cvt_d_l(long long *source,double *dest) -{ - *dest = *source; -} -void cvt_d_s(float *source,double *dest) -{ - *dest = *source; -} -void cvt_s_d(double *source,float *dest) -{ - *dest = *source; -} - -void round_l_s(float *source,long long *dest) -{ - *dest = roundf(*source); -} -void round_w_s(float *source,int *dest) -{ - *dest = roundf(*source); -} -void trunc_l_s(float *source,long long *dest) -{ - *dest = truncf(*source); -} -void trunc_w_s(float *source,int *dest) -{ - *dest = truncf(*source); -} -void ceil_l_s(float *source,long long *dest) -{ - *dest = ceilf(*source); -} -void ceil_w_s(float *source,int *dest) -{ - *dest = ceilf(*source); -} -void floor_l_s(float *source,long long *dest) -{ - *dest = floorf(*source); -} -void floor_w_s(float *source,int *dest) -{ - *dest = floorf(*source); -} - -void round_l_d(double *source,long long *dest) -{ - *dest = round(*source); -} -void round_w_d(double *source,int *dest) -{ - *dest = round(*source); -} -void trunc_l_d(double *source,long long *dest) -{ - *dest = trunc(*source); -} -void trunc_w_d(double *source,int *dest) -{ - *dest = trunc(*source); -} -void ceil_l_d(double *source,long long *dest) -{ - *dest = ceil(*source); -} -void ceil_w_d(double *source,int *dest) -{ - *dest = ceil(*source); -} -void floor_l_d(double *source,long long *dest) -{ - *dest = floor(*source); -} -void floor_w_d(double *source,int *dest) -{ - *dest = floor(*source); -} - -void cvt_w_s(float *source,int *dest) -{ - switch(FCR31&3) - { - case 0: round_w_s(source,dest);return; - case 1: trunc_w_s(source,dest);return; - case 2: ceil_w_s(source,dest);return; - case 3: floor_w_s(source,dest);return; - } -} -void cvt_w_d(double *source,int *dest) -{ - switch(FCR31&3) - { - case 0: round_w_d(source,dest);return; - case 1: trunc_w_d(source,dest);return; - case 2: ceil_w_d(source,dest);return; - case 3: floor_w_d(source,dest);return; - } -} -void cvt_l_s(float *source,long long *dest) -{ - switch(FCR31&3) - { - case 0: round_l_s(source,dest);return; - case 1: trunc_l_s(source,dest);return; - case 2: ceil_l_s(source,dest);return; - case 3: floor_l_s(source,dest);return; - } -} -void cvt_l_d(double *source,long long *dest) -{ - switch(FCR31&3) - { - case 0: round_l_d(source,dest);return; - case 1: trunc_l_d(source,dest);return; - case 2: ceil_l_d(source,dest);return; - case 3: floor_l_d(source,dest);return; - } -} - -void c_f_s() -{ - FCR31 &= ~0x800000; -} -void c_un_s(float *source,float *target) -{ - FCR31=(isnan(*source) || isnan(*target)) ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_eq_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ueq_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_olt_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ult_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_ole_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ule_s(float *source,float *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_sf_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31&=~0x800000; -} -void c_ngle_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31&=~0x800000; -} - -void c_seq_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ngl_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_lt_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_nge_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_le_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ngt_s(float *source,float *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_f_d() -{ - FCR31 &= ~0x800000; -} -void c_un_d(double *source,double *target) -{ - FCR31=(isnan(*source) || isnan(*target)) ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_eq_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ueq_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_olt_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ult_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_ole_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31&=~0x800000;return;} - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ule_d(double *source,double *target) -{ - if (isnan(*source) || isnan(*target)) {FCR31|=0x800000;return;} - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_sf_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31&=~0x800000; -} -void c_ngle_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31&=~0x800000; -} - -void c_seq_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ngl_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source==*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_lt_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_nge_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<*target ? FCR31|0x800000 : FCR31&~0x800000; -} - -void c_le_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} -void c_ngt_d(double *source,double *target) -{ - //if (isnan(*source) || isnan(*target)) // FIXME - exception - FCR31 = *source<=*target ? FCR31|0x800000 : FCR31&~0x800000; -} - - -void add_s(float *source1,float *source2,float *target) -{ - *target=(*source1)+(*source2); -} -void sub_s(float *source1,float *source2,float *target) -{ - *target=(*source1)-(*source2); -} -void mul_s(float *source1,float *source2,float *target) -{ - *target=(*source1)*(*source2); -} -void div_s(float *source1,float *source2,float *target) -{ - *target=(*source1)/(*source2); -} -void sqrt_s(float *source,float *target) -{ - *target=sqrtf(*source); -} -void abs_s(float *source,float *target) -{ - *target=fabsf(*source); -} -void mov_s(float *source,float *target) -{ - *target=*source; -} -void neg_s(float *source,float *target) -{ - *target=-(*source); -} -void add_d(double *source1,double *source2,double *target) -{ - *target=(*source1)+(*source2); -} -void sub_d(double *source1,double *source2,double *target) -{ - *target=(*source1)-(*source2); -} -void mul_d(double *source1,double *source2,double *target) -{ - *target=(*source1)*(*source2); -} -void div_d(double *source1,double *source2,double *target) -{ - *target=(*source1)/(*source2); -} -void sqrt_d(double *source,double *target) -{ - *target=sqrt(*source); -} -void abs_d(double *source,double *target) -{ - *target=fabs(*source); -} -void mov_d(double *source,double *target) -{ - *target=*source; -} -void neg_d(double *source,double *target) -{ - *target=-(*source); -} - diff --git a/libpcsxcore/new_dynarec/fpu.h b/libpcsxcore/new_dynarec/fpu.h deleted file mode 100644 index 881ddbe0..00000000 --- a/libpcsxcore/new_dynarec/fpu.h +++ /dev/null @@ -1,74 +0,0 @@ -void cvt_s_w(int *source,float *dest); -void cvt_d_w(int *source,double *dest); -void cvt_s_l(long long *source,float *dest); -void cvt_d_l(long long *source,double *dest); -void cvt_w_s(float *source,int *dest); -void cvt_w_d(double *source,int *dest); -void cvt_l_s(float *source,long long *dest); -void cvt_l_d(double *source,long long *dest); -void cvt_d_s(float *source,double *dest); -void cvt_s_d(double *source,float *dest); -void round_l_s(float *source,long long *dest); -void round_w_s(float *source,int *dest); -void trunc_l_s(float *source,long long *dest); -void trunc_w_s(float *source,int *dest); -void ceil_l_s(float *source,long long *dest); -void ceil_w_s(float *source,int *dest); -void floor_l_s(float *source,long long *dest); -void floor_w_s(float *source,int *dest); -void round_l_d(double *source,long long *dest); -void round_w_d(double *source,int *dest); -void trunc_l_d(double *source,long long *dest); -void trunc_w_d(double *source,int *dest); -void ceil_l_d(double *source,long long *dest); -void ceil_w_d(double *source,int *dest); -void floor_l_d(double *source,long long *dest); -void floor_w_d(double *source,int *dest); -void c_f_s(); -void c_un_s(float *source,float *target); -void c_eq_s(float *source,float *target); -void c_ueq_s(float *source,float *target); -void c_olt_s(float *source,float *target); -void c_ult_s(float *source,float *target); -void c_ole_s(float *source,float *target); -void c_ule_s(float *source,float *target); -void c_sf_s(float *source,float *target); -void c_ngle_s(float *source,float *target); -void c_seq_s(float *source,float *target); -void c_ngl_s(float *source,float *target); -void c_lt_s(float *source,float *target); -void c_nge_s(float *source,float *target); -void c_le_s(float *source,float *target); -void c_ngt_s(float *source,float *target); -void c_f_d(); -void c_un_d(double *source,double *target); -void c_eq_d(double *source,double *target); -void c_ueq_d(double *source,double *target); -void c_olt_d(double *source,double *target); -void c_ult_d(double *source,double *target); -void c_ole_d(double *source,double *target); -void c_ule_d(double *source,double *target); -void c_sf_d(double *source,double *target); -void c_ngle_d(double *source,double *target); -void c_seq_d(double *source,double *target); -void c_ngl_d(double *source,double *target); -void c_lt_d(double *source,double *target); -void c_nge_d(double *source,double *target); -void c_le_d(double *source,double *target); -void c_ngt_d(double *source,double *target); -void add_s(float *source1,float *source2,float *target); -void sub_s(float *source1,float *source2,float *target); -void mul_s(float *source1,float *source2,float *target); -void div_s(float *source1,float *source2,float *target); -void sqrt_s(float *source,float *target); -void abs_s(float *source,float *target); -void mov_s(float *source,float *target); -void neg_s(float *source,float *target); -void add_d(double *source1,double *source2,double *target); -void sub_d(double *source1,double *source2,double *target); -void mul_d(double *source1,double *source2,double *target); -void div_d(double *source1,double *source2,double *target); -void sqrt_d(double *source,double *target); -void abs_d(double *source,double *target); -void mov_d(double *source,double *target); -void neg_d(double *source,double *target); diff --git a/libpcsxcore/new_dynarec/linkage_arm.S b/libpcsxcore/new_dynarec/linkage_arm.S index 95af8b4e..147b0dfc 100644 --- a/libpcsxcore/new_dynarec/linkage_arm.S +++ b/libpcsxcore/new_dynarec/linkage_arm.S @@ -20,6 +20,7 @@ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */ #include "arm_features.h" +#include "new_dynarec_config.h" #include "linkage_offsets.h" @@ -160,9 +161,10 @@ ptr_hash_table: #endif .endm +/* r0 = virtual target address */ +/* r1 = instruction to patch */ .macro dyna_linker_main - /* r0 = virtual target address */ - /* r1 = instruction to patch */ +#ifndef NO_WRITE_EXEC load_varadr_ext r3, jump_in /* get_page */ lsr r2, r0, #12 @@ -242,6 +244,11 @@ ptr_hash_table: str r3, [r6, #12] mov pc, r1 8: +#else + /* XXX: should be able to do better than this... */ + bl get_addr_ht + mov pc, r0 +#endif .endm diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c index 65c8f03e..1618b0fa 100644 --- a/libpcsxcore/new_dynarec/new_dynarec.c +++ b/libpcsxcore/new_dynarec/new_dynarec.c @@ -23,7 +23,18 @@ #include #include #include +#ifdef __MACH__ +#include +#endif +#ifdef _3DS +#include <3ds_utils.h> +#endif +#ifdef VITA +#include +static int sceBlock; +#endif +#include "new_dynarec_config.h" #include "emu_if.h" //emulator interface //#define DISASM @@ -42,22 +53,6 @@ #include "assem_arm.h" #endif -#ifdef __BLACKBERRY_QNX__ -#undef __clear_cache -#define __clear_cache(start,end) msync(start, (size_t)((void*)end - (void*)start), MS_SYNC | MS_CACHE_ONLY | MS_INVALIDATE_ICACHE); -#elif defined(__MACH__) -#include -#define __clear_cache mach_clear_cache -static void __clear_cache(void *start, void *end) { - size_t len = (char *)end - (char *)start; - sys_dcache_flush(start, len); - sys_icache_invalidate(start, len); -} -#elif defined(_3DS) -#include "3ds_utils.h" -#define __clear_cache(start,end) svcFlushProcessDataCache(0xFFFF8001, start, (u32)(end)-(u32)(start)) -#endif - #define MAXBLOCK 4096 #define MAX_OUTPUT_BLOCK_SIZE 262144 @@ -86,22 +81,29 @@ struct ll_entry struct ll_entry *next; }; - u_int start; - u_int *source; - char insn[MAXBLOCK][10]; - u_char itype[MAXBLOCK]; - u_char opcode[MAXBLOCK]; - u_char opcode2[MAXBLOCK]; - u_char bt[MAXBLOCK]; - u_char rs1[MAXBLOCK]; - u_char rs2[MAXBLOCK]; - u_char rt1[MAXBLOCK]; - u_char rt2[MAXBLOCK]; - u_char us1[MAXBLOCK]; - u_char us2[MAXBLOCK]; - u_char dep1[MAXBLOCK]; - u_char dep2[MAXBLOCK]; - u_char lt1[MAXBLOCK]; + // used by asm: + u_char *out; + u_int hash_table[65536][4] __attribute__((aligned(16))); + struct ll_entry *jump_in[4096] __attribute__((aligned(16))); + struct ll_entry *jump_dirty[4096]; + + static struct ll_entry *jump_out[4096]; + static u_int start; + static u_int *source; + static char insn[MAXBLOCK][10]; + static u_char itype[MAXBLOCK]; + static u_char opcode[MAXBLOCK]; + static u_char opcode2[MAXBLOCK]; + static u_char bt[MAXBLOCK]; + static u_char rs1[MAXBLOCK]; + static u_char rs2[MAXBLOCK]; + static u_char rt1[MAXBLOCK]; + static u_char rt2[MAXBLOCK]; + static u_char us1[MAXBLOCK]; + static u_char us2[MAXBLOCK]; + static u_char dep1[MAXBLOCK]; + static u_char dep2[MAXBLOCK]; + static u_char lt1[MAXBLOCK]; static uint64_t gte_rs[MAXBLOCK]; // gte: 32 data and 32 ctl regs static uint64_t gte_rt[MAXBLOCK]; static uint64_t gte_unneeded[MAXBLOCK]; @@ -110,59 +112,47 @@ struct ll_entry static u_int smrv_weak; // same, but somewhat less likely static u_int smrv_strong_next; // same, but after current insn executes static u_int smrv_weak_next; - int imm[MAXBLOCK]; - u_int ba[MAXBLOCK]; - char likely[MAXBLOCK]; - char is_ds[MAXBLOCK]; - char ooo[MAXBLOCK]; - uint64_t unneeded_reg[MAXBLOCK]; - uint64_t unneeded_reg_upper[MAXBLOCK]; - uint64_t branch_unneeded_reg[MAXBLOCK]; - uint64_t branch_unneeded_reg_upper[MAXBLOCK]; - uint64_t p32[MAXBLOCK]; - uint64_t pr32[MAXBLOCK]; - signed char regmap_pre[MAXBLOCK][HOST_REGS]; + static int imm[MAXBLOCK]; + static u_int ba[MAXBLOCK]; + static char likely[MAXBLOCK]; + static char is_ds[MAXBLOCK]; + static char ooo[MAXBLOCK]; + static uint64_t unneeded_reg[MAXBLOCK]; + static uint64_t unneeded_reg_upper[MAXBLOCK]; + static uint64_t branch_unneeded_reg[MAXBLOCK]; + static uint64_t branch_unneeded_reg_upper[MAXBLOCK]; + static signed char regmap_pre[MAXBLOCK][HOST_REGS]; static uint64_t current_constmap[HOST_REGS]; static uint64_t constmap[MAXBLOCK][HOST_REGS]; static struct regstat regs[MAXBLOCK]; static struct regstat branch_regs[MAXBLOCK]; - signed char minimum_free_regs[MAXBLOCK]; - u_int needed_reg[MAXBLOCK]; - uint64_t requires_32bit[MAXBLOCK]; - u_int wont_dirty[MAXBLOCK]; - u_int will_dirty[MAXBLOCK]; - int ccadj[MAXBLOCK]; - int slen; - u_int instr_addr[MAXBLOCK]; + static signed char minimum_free_regs[MAXBLOCK]; + static u_int needed_reg[MAXBLOCK]; + static u_int wont_dirty[MAXBLOCK]; + static u_int will_dirty[MAXBLOCK]; + static int ccadj[MAXBLOCK]; + static int slen; + static u_int instr_addr[MAXBLOCK]; static u_int link_addr[MAXBLOCK][3]; - int linkcount; - u_int stubs[MAXBLOCK*3][8]; - int stubcount; - u_int literals[1024][2]; - int literalcount; - int is_delayslot; - int cop1_usable; - u_char *out; - struct ll_entry *jump_in[4096] __attribute__((aligned(16))); - struct ll_entry *jump_out[4096]; - struct ll_entry *jump_dirty[4096]; - u_int hash_table[65536][4] __attribute__((aligned(16))); - char shadow[1048576] __attribute__((aligned(16))); - void *copy; - int expirep; -#ifndef PCSX - u_int using_tlb; -#else - static const u_int using_tlb=0; -#endif - int new_dynarec_did_compile; - int new_dynarec_hacks; - u_int stop_after_jal; + static int linkcount; + static u_int stubs[MAXBLOCK*3][8]; + static int stubcount; + static u_int literals[1024][2]; + static int literalcount; + static int is_delayslot; + static int cop1_usable; + static char shadow[1048576] __attribute__((aligned(16))); + static void *copy; + static int expirep; + static u_int stop_after_jal; #ifndef RAM_FIXED static u_int ram_offset; #else static const u_int ram_offset=0; #endif + + int new_dynarec_hacks; + int new_dynarec_did_compile; extern u_char restore_candidate[512]; extern int cycle_count; @@ -174,20 +164,20 @@ struct ll_entry #define CSREG 35 // Coprocessor status #define CCREG 36 // Cycle count #define INVCP 37 // Pointer to invalid_code -#define MMREG 38 // Pointer to memory_map +//#define MMREG 38 // Pointer to memory_map #define ROREG 39 // ram offset (if rdram!=0x80000000) #define TEMPREG 40 #define FTEMP 40 // FPU temporary register #define PTEMP 41 // Prefetch temporary register -#define TLREG 42 // TLB mapping offset +//#define TLREG 42 // TLB mapping offset #define RHASH 43 // Return address hash #define RHTBL 44 // Return address hash table address #define RTEMP 45 // JR/JALR address register #define MAXREG 45 #define AGEN1 46 // Address generation temporary register -#define AGEN2 47 // Address generation temporary register -#define MGEN1 48 // Maptable address generation temporary register -#define MGEN2 49 // Maptable address generation temporary register +//#define AGEN2 47 // Address generation temporary register +//#define MGEN1 48 // Maptable address generation temporary register +//#define MGEN2 49 // Maptable address generation temporary register #define BTREG 50 // Branch target temporary register /* instruction types */ @@ -196,7 +186,7 @@ struct ll_entry #define STORE 2 // Store #define LOADLR 3 // Unaligned load #define STORELR 4 // Unaligned store -#define MOV 5 // Move +#define MOV 5 // Move #define ALU 6 // Arithmetic/logic #define MULTDIV 7 // Multiply/divide #define SHIFT 8 // Shift by register @@ -250,7 +240,6 @@ void *get_addr_ht(u_int vaddr); void invalidate_block(u_int block); void invalidate_addr(u_int addr); void remove_hash(int vaddr); -void jump_vaddr(); void dyna_linker(); void dyna_linker_ds(); void verify_code(); @@ -259,40 +248,86 @@ void verify_code_ds(); void cc_interrupt(); void fp_exception(); void fp_exception_ds(); -void jump_syscall(); void jump_syscall_hle(); -void jump_eret(); void jump_hlecall(); void jump_intcall(); void new_dyna_leave(); -// TLB -void TLBWI_new(); -void TLBWR_new(); -void read_nomem_new(); -void read_nomemb_new(); -void read_nomemh_new(); -void read_nomemd_new(); -void write_nomem_new(); -void write_nomemb_new(); -void write_nomemh_new(); -void write_nomemd_new(); -void write_rdram_new(); -void write_rdramb_new(); -void write_rdramh_new(); -void write_rdramd_new(); -extern u_int memory_map[1048576]; - // Needed by assembler -void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32); -void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty); -void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr); -void load_all_regs(signed char i_regmap[]); -void load_needed_regs(signed char i_regmap[],signed char next_regmap[]); -void load_regs_entry(int t); -void load_all_consts(signed char regmap[],int is32,u_int dirty,int i); +static void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32); +static void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty); +static void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr); +static void load_all_regs(signed char i_regmap[]); +static void load_needed_regs(signed char i_regmap[],signed char next_regmap[]); +static void load_regs_entry(int t); +static void load_all_consts(signed char regmap[],int is32,u_int dirty,int i); + +static int verify_dirty(u_int *ptr); +static int get_final_value(int hr, int i, int *value); +static void add_stub(int type,int addr,int retaddr,int a,int b,int c,int d,int e); +static void add_to_linker(int addr,int target,int ext); + +static int tracedebug=0; + +static void mprotect_w_x(void *start, void *end, int is_x) +{ +#ifdef NO_WRITE_EXEC + #if defined(VITA) + // *Open* enables write on all memory that was + // allocated by sceKernelAllocMemBlockForVM()? + if (is_x) + sceKernelCloseVMDomain(); + else + sceKernelOpenVMDomain(); + #else + u_long mstart = (u_long)start & ~4095ul; + u_long mend = (u_long)end; + if (mprotect((void *)mstart, mend - mstart, + PROT_READ | (is_x ? PROT_EXEC : PROT_WRITE)) != 0) + SysPrintf("mprotect(%c) failed: %s\n", is_x ? 'x' : 'w', strerror(errno)); + #endif +#endif +} + +static void start_tcache_write(void *start, void *end) +{ + mprotect_w_x(start, end, 0); +} + +static void end_tcache_write(void *start, void *end) +{ +#ifdef __arm__ + size_t len = (char *)end - (char *)start; + #if defined(__BLACKBERRY_QNX__) + msync(start, len, MS_SYNC | MS_CACHE_ONLY | MS_INVALIDATE_ICACHE); + #elif defined(__MACH__) + sys_cache_control(kCacheFunctionPrepareForExecution, start, len); + #elif defined(VITA) + sceKernelSyncVMDomain(sceBlock, start, len); + #elif defined(_3DS) + ctr_flush_invalidate_cache(); + #else + __clear_cache(start, end); + #endif + (void)len; +#endif + + mprotect_w_x(start, end, 1); +} + +static void *start_block(void) +{ + u_char *end = out + MAX_OUTPUT_BLOCK_SIZE; + if (end > (u_char *)BASE_ADDR + (1<nom, "GOLDENEYE",9) == 0) - { - u_int addr; - int n; - switch (ROM_HEADER->Country_code&0xFF) - { - case 0x45: // U - addr=0x34b30; - break; - case 0x4A: // J - addr=0x34b70; - break; - case 0x50: // E - addr=0x329f0; - break; - default: - // Unknown country code - addr=0; - break; - } - u_int rom_addr=(u_int)rom; - #ifdef ROM_COPY - // Since memory_map is 32-bit, on 64-bit systems the rom needs to be - // in the lower 4G of memory to use this hack. Copy it if necessary. - if((void *)rom>(void *)0xffffffff) { - munmap(ROM_COPY, 67108864); - if(mmap(ROM_COPY, 12582912, - PROT_READ | PROT_WRITE, - MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS, - -1, 0) <= 0) {printf("mmap() failed\n");} - memcpy(ROM_COPY,rom,12582912); - rom_addr=(u_int)ROM_COPY; - } - #endif - if(addr) { - for(n=0x7F000;n<0x80000;n++) { - memory_map[n]=(((u_int)(rom_addr+addr-0x7F000000))>>2)|0x40000000; - } - } - } -#endif -} - static u_int get_page(u_int vaddr) { -#ifndef PCSX - u_int page=(vaddr^0x80000000)>>12; -#else u_int page=vaddr&~0xe0000000; if (page < 0x1000000) page &= ~0x0e00000; // RAM mirrors page>>=12; -#endif -#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; } -#ifndef PCSX -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; -} -#else // no virtual mem in PCSX static u_int get_vpage(u_int vaddr) { return get_page(vaddr); } -#endif // Get address from virtual address // This is called from the recompiled JR/JALR instructions @@ -400,10 +369,10 @@ void *get_addr(u_int vaddr) while(head!=NULL) { if(head->vaddr==vaddr) { //printf("TRACE: count=%d next=%d (get_addr match %x: %x)\n",Count,next_interupt,vaddr,(int)head->addr); - int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; + u_int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; ht_bin[3]=ht_bin[1]; ht_bin[2]=ht_bin[0]; - ht_bin[1]=(int)head->addr; + ht_bin[1]=(u_int)head->addr; ht_bin[0]=vaddr; return head->addr; } @@ -419,22 +388,13 @@ void *get_addr(u_int vaddr) //printf("restore candidate: %x (%d) d=%d\n",vaddr,page,invalid_code[vaddr>>12]); invalid_code[vaddr>>12]=0; inv_code_start=inv_code_end=~0; -#ifndef DISABLE_TLB - memory_map[vaddr>>12]|=0x40000000; -#endif 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); - int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; + u_int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; if(ht_bin[0]==vaddr) { - ht_bin[1]=(int)head->addr; // Replace existing entry + ht_bin[1]=(u_int)head->addr; // Replace existing entry } else { @@ -464,7 +424,7 @@ void *get_addr(u_int vaddr) void *get_addr_ht(u_int vaddr) { //printf("TRACE: count=%d next=%d (get_addr_ht %x)\n",Count,next_interupt,vaddr); - int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; + u_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]; return get_addr(vaddr); @@ -526,7 +486,7 @@ static void flush_dirty_uppers(struct regstat *cur) for (hr=0;hrdirty>>hr)&1) { reg=cur->regmap[hr]; - if(reg>=64) + if(reg>=64) if((cur->is32>>(reg&63))&1) cur->regmap[hr]=-1; } } @@ -665,10 +625,6 @@ void lsn(u_char hsn[], int i, int *preferred_reg) if(opcode[i]==0x2a||opcode[i]==0x2e||opcode[i]==0x2c||opcode[i]==0x2d) { 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 || itype[i]==C2LS) { - hsn[TLREG]=0; - } // Don't remove the miniht registers if(itype[i]==UJUMP||itype[i]==RJUMP) { @@ -683,7 +639,7 @@ int needed_again(int r, int i) int j; int b=-1; int rn=10; - + if(i>0&&(itype[i-1]==UJUMP||itype[i-1]==RJUMP||(source[i-1]>>16)==0x1000)) { if(ba[i-1]start+slen*4-4) @@ -736,6 +692,7 @@ int needed_again(int r, int i) } }*/ if(rn<10) return 1; + (void)b; return 0; } @@ -786,7 +743,7 @@ int loop_reg(int i, int r, int hr) void alloc_all(struct regstat *cur,int i) { int hr; - + for(hr=0;hrregmap[hr]&63)!=rs1[i])&&((cur->regmap[hr]&63)!=rs2[i])&& @@ -805,119 +762,6 @@ void alloc_all(struct regstat *cur,int i) } } -#ifndef FORCE32 -void div64(int64_t dividend,int64_t divisor) -{ - lo=dividend/divisor; - hi=dividend%divisor; - //printf("TRACE: ddiv %8x%8x %8x%8x\n" ,(int)reg[HIREG],(int)(reg[HIREG]>>32) - // ,(int)reg[LOREG],(int)(reg[LOREG]>>32)); -} -void divu64(uint64_t dividend,uint64_t divisor) -{ - lo=dividend/divisor; - hi=dividend%divisor; - //printf("TRACE: ddivu %8x%8x %8x%8x\n",(int)reg[HIREG],(int)(reg[HIREG]>>32) - // ,(int)reg[LOREG],(int)(reg[LOREG]>>32)); -} - -void mult64(uint64_t m1,uint64_t m2) -{ - unsigned long long int op1, op2, op3, op4; - unsigned long long int result1, result2, result3, result4; - unsigned long long int temp1, temp2, temp3, temp4; - int sign = 0; - - if (m1 < 0) - { - op2 = -m1; - sign = 1 - sign; - } - else op2 = m1; - if (m2 < 0) - { - op4 = -m2; - sign = 1 - sign; - } - else op4 = m2; - - op1 = op2 & 0xFFFFFFFF; - op2 = (op2 >> 32) & 0xFFFFFFFF; - op3 = op4 & 0xFFFFFFFF; - op4 = (op4 >> 32) & 0xFFFFFFFF; - - temp1 = op1 * op3; - temp2 = (temp1 >> 32) + op1 * op4; - temp3 = op2 * op3; - temp4 = (temp3 >> 32) + op2 * op4; - - result1 = temp1 & 0xFFFFFFFF; - result2 = temp2 + (temp3 & 0xFFFFFFFF); - result3 = (result2 >> 32) + temp4; - result4 = (result3 >> 32); - - lo = result1 | (result2 << 32); - hi = (result3 & 0xFFFFFFFF) | (result4 << 32); - if (sign) - { - hi = ~hi; - if (!lo) hi++; - else lo = ~lo + 1; - } -} - -void multu64(uint64_t m1,uint64_t m2) -{ - unsigned long long int op1, op2, op3, op4; - unsigned long long int result1, result2, result3, result4; - unsigned long long int temp1, temp2, temp3, temp4; - - op1 = m1 & 0xFFFFFFFF; - op2 = (m1 >> 32) & 0xFFFFFFFF; - op3 = m2 & 0xFFFFFFFF; - op4 = (m2 >> 32) & 0xFFFFFFFF; - - temp1 = op1 * op3; - temp2 = (temp1 >> 32) + op1 * op4; - temp3 = op2 * op3; - temp4 = (temp3 >> 32) + op2 * op4; - - result1 = temp1 & 0xFFFFFFFF; - result2 = temp2 + (temp3 & 0xFFFFFFFF); - result3 = (result2 >> 32) + temp4; - result4 = (result3 >> 32); - - lo = result1 | (result2 << 32); - hi = (result3 & 0xFFFFFFFF) | (result4 << 32); - - //printf("TRACE: dmultu %8x%8x %8x%8x\n",(int)reg[HIREG],(int)(reg[HIREG]>>32) - // ,(int)reg[LOREG],(int)(reg[LOREG]>>32)); -} - -uint64_t ldl_merge(uint64_t original,uint64_t loaded,u_int bits) -{ - if(bits) { - original<<=64-bits; - original>>=64-bits; - loaded<<=bits; - original|=loaded; - } - else original=loaded; - return original; -} -uint64_t ldr_merge(uint64_t original,uint64_t loaded,u_int bits) -{ - if(bits^56) { - original>>=64-(bits^56); - original<<=64-(bits^56); - loaded>>=bits^56; - original|=loaded; - } - else original=loaded; - return original; -} -#endif - #ifdef __i386__ #include "assem_x86.c" #endif @@ -996,7 +840,7 @@ void *check_addr(u_int vaddr) void remove_hash(int vaddr) { //printf("remove hash: %x\n",vaddr); - int *ht_bin=hash_table[(((vaddr)>>16)^vaddr)&0xFFFF]; + u_int *ht_bin=hash_table[(((vaddr)>>16)^vaddr)&0xFFFF]; if(ht_bin[2]==vaddr) { ht_bin[2]=ht_bin[3]=-1; } @@ -1011,7 +855,7 @@ void ll_remove_matching_addrs(struct ll_entry **head,int addr,int shift) { struct ll_entry *next; while(*head) { - if(((u_int)((*head)->addr)>>shift)==(addr>>shift) || + if(((u_int)((*head)->addr)>>shift)==(addr>>shift) || ((u_int)((*head)->addr-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(addr>>shift)) { inv_debug("EXP: Remove pointer to %x (%x)\n",(int)(*head)->addr,(*head)->vaddr); @@ -1032,7 +876,7 @@ void ll_clear(struct ll_entry **head) { struct ll_entry *cur; struct ll_entry *next; - if(cur=*head) { + if((cur=*head)) { *head=0; while(cur) { next=cur->next; @@ -1043,7 +887,7 @@ void ll_clear(struct ll_entry **head) } // Dereference the pointers and remove if it matches -void ll_kill_pointers(struct ll_entry *head,int addr,int shift) +static void ll_kill_pointers(struct ll_entry *head,int addr,int shift) { while(head) { int ptr=get_pointer(head->addr); @@ -1052,10 +896,11 @@ void ll_kill_pointers(struct ll_entry *head,int addr,int shift) (((ptr-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(addr>>shift))) { inv_debug("EXP: Kill pointer at %x (%x)\n",(int)head->addr,head->vaddr); - u_int host_addr=(u_int)kill_pointer(head->addr); + void *host_addr=find_extjump_insn(head->addr); #ifdef __arm__ - needs_clear_cache[(host_addr-(u_int)BASE_ADDR)>>17]|=1<<(((host_addr-(u_int)BASE_ADDR)>>12)&31); + mark_clear_cache(host_addr); #endif + set_jump_target((int)host_addr,(int)head->addr); } head=head->next; } @@ -1079,10 +924,11 @@ void invalidate_page(u_int page) jump_out[page]=0; while(head!=NULL) { inv_debug("INVALIDATE: kill pointer to %x (%x)\n",head->vaddr,(int)head->addr); - u_int host_addr=(u_int)kill_pointer(head->addr); + void *host_addr=find_extjump_insn(head->addr); #ifdef __arm__ - needs_clear_cache[(host_addr-(u_int)BASE_ADDR)>>17]|=1<<(((host_addr-(u_int)BASE_ADDR)>>12)&31); + mark_clear_cache(host_addr); #endif + set_jump_target((int)host_addr,(int)head->addr); next=head->next; free(head); head=next; @@ -1107,21 +953,9 @@ static void invalidate_block_range(u_int block, u_int first, u_int last) #ifdef __arm__ do_clear_cache(); #endif - + // 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]); - // CHECK: Is this right? - memory_map[block]=((tlb_LUT_w[block]&0xFFFFF000)-(block<<12)+(unsigned int)rdram-0x80000000)>>2; - u_int real_block=tlb_LUT_w[block]>>12; - invalid_code[real_block]=1; - 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 USE_MINI_HT memset(mini_ht,-1,sizeof(mini_ht)); @@ -1150,14 +984,6 @@ 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; } @@ -1166,7 +992,6 @@ void invalidate_block(u_int block) void invalidate_addr(u_int addr) { -#ifdef PCSX //static int rhits; // this check is done by the caller //if (inv_code_start<=addr&&addr<=inv_code_end) { rhits++; return; } @@ -1220,7 +1045,6 @@ void invalidate_addr(u_int addr) return; } } -#endif invalidate_block(addr>>12); } @@ -1228,7 +1052,7 @@ void invalidate_addr(u_int addr) // Anything could have changed, so invalidate everything. void invalidate_all_pages() { - u_int page,n; + u_int page; for(page=0;page<4096;page++) invalidate_page(page); for(page=0;page<1048576;page++) @@ -1236,25 +1060,9 @@ void invalidate_all_pages() restore_candidate[(page&2047)>>3]|=1<<(page&7); restore_candidate[((page&2047)>>3)+256]|=1<<(page&7); } - #ifdef __arm__ - __clear_cache((void *)BASE_ADDR,(void *)BASE_ADDR+(1<>2; - if(!tlb_LUT_w[page]||!invalid_code[page]) - memory_map[page]|=0x40000000; // Write protect - } - else memory_map[page]=-1; - if(page==0x80000) page=0xC0000; - } - tlb_hacks(); - #endif } // Add an entry to jump_out after making a link @@ -1264,6 +1072,7 @@ void add_link(u_int vaddr,void *src) inv_debug("add_link: %x -> %x (%d)\n",(int)src,vaddr,page); int *ptr=(int *)(src+4); assert((*ptr&0x0fff0000)==0x059f0000); + (void)ptr; ll_add(jump_out+page,vaddr,src); //int ptr=get_pointer(src); //inv_debug("add_link: Pointer is to %x\n",(int)ptr); @@ -1283,7 +1092,7 @@ void clean_blocks(u_int page) // Don't restore blocks which are about to expire from the cache if((((u_int)head->addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) { u_int start,end; - if(verify_dirty((int)head->addr)) { + if(verify_dirty(head->addr)) { //printf("Possibly Restore %x (%x)\n",head->vaddr, (int)head->addr); u_int i; u_int inv=0; @@ -1293,13 +1102,6 @@ void clean_blocks(u_int page) inv|=invalid_code[i]; } } -#ifndef DISABLE_TLB - if((signed int)head->vaddr>=(signed int)0xC0000000) { - u_int addr = (head->vaddr+(memory_map[head->vaddr>>12]<<2)); - //printf("addr=%x start=%x end=%x\n",addr,start,end); - if(addr=end) inv=1; - } -#endif else if((signed int)head->vaddr>=(signed int)0x80000000+RAM_SIZE) { inv=1; } @@ -1307,19 +1109,16 @@ 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)); ll_add_flags(jump_in+ppage,head->vaddr,head->reg_sv_flags,clean_addr); - int *ht_bin=hash_table[((head->vaddr>>16)^head->vaddr)&0xFFFF]; + u_int *ht_bin=hash_table[((head->vaddr>>16)^head->vaddr)&0xFFFF]; if(ht_bin[0]==head->vaddr) { - ht_bin[1]=(int)clean_addr; // Replace existing entry + ht_bin[1]=(u_int)clean_addr; // Replace existing entry } if(ht_bin[2]==head->vaddr) { - ht_bin[3]=(int)clean_addr; // Replace existing entry + ht_bin[3]=(u_int)clean_addr; // Replace existing entry } } } @@ -1646,8 +1445,6 @@ void load_alloc(struct regstat *current,int i) } else current->is32|=1LL<>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); } @@ -2090,12 +1873,6 @@ 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() @@ -2129,11 +1906,6 @@ void memdebug(int i) //printf("TRACE: %x\n",(&i)[-1]); } -void tlb_debug(u_int cause, u_int addr, u_int iaddr) -{ - printf("TLB Exception: instruction=%x addr=%x cause=%x\n",iaddr, addr, cause); -} - void alu_assemble(int i,struct regstat *i_regs) { if(opcode2[i]>=0x20&&opcode2[i]<=0x23) { // ADD/ADDU/SUB/SUBU @@ -2620,23 +2392,25 @@ void imm16_assemble(int i,struct regstat *i_regs) emit_mov(sh,th); } } - if(opcode[i]==0x0d) //ORI - if(sl<0) { - emit_orimm(tl,imm[i],tl); - }else{ - if(!((i_regs->wasconst>>sl)&1)) - emit_orimm(sl,imm[i],tl); - else - emit_movimm(constmap[i][sl]|imm[i],tl); + if(opcode[i]==0x0d) { // ORI + if(sl<0) { + emit_orimm(tl,imm[i],tl); + }else{ + if(!((i_regs->wasconst>>sl)&1)) + emit_orimm(sl,imm[i],tl); + else + emit_movimm(constmap[i][sl]|imm[i],tl); + } } - if(opcode[i]==0x0e) //XORI - if(sl<0) { - emit_xorimm(tl,imm[i],tl); - }else{ - if(!((i_regs->wasconst>>sl)&1)) - emit_xorimm(sl,imm[i],tl); - else - emit_movimm(constmap[i][sl]^imm[i],tl); + if(opcode[i]==0x0e) { // XORI + if(sl<0) { + emit_xorimm(tl,imm[i],tl); + }else{ + if(!((i_regs->wasconst>>sl)&1)) + emit_xorimm(sl,imm[i],tl); + else + emit_movimm(constmap[i][sl]^imm[i],tl); + } } } else { @@ -2815,14 +2589,12 @@ void load_assemble(int i,struct regstat *i_regs) c=(i_regs->wasconst>>s)&1; if (c) { memtarget=((signed int)(constmap[i][s]+offset))<(signed int)0x80000000+RAM_SIZE; - if(using_tlb&&((signed int)(constmap[i][s]+offset))>=(signed int)0xC0000000) memtarget=1; } } //printf("load_assemble: c=%d\n",c); //if(c) printf("load_assemble: const=%x\n",(int)constmap[i][s]+offset); // FIXME: Even if the load is a NOP, we should check for pagefaults... -#ifdef PCSX - if(tl<0&&(!c||(((u_int)constmap[i][s]+offset)>>16)==0x1f80) + if((tl<0&&(!c||(((u_int)constmap[i][s]+offset)>>16)==0x1f80)) ||rt1[i]==0) { // could be FIFO, must perform the read // ||dummy read @@ -2830,7 +2602,6 @@ void load_assemble(int i,struct regstat *i_regs) tl=get_reg(i_regs->regmap,-1); assert(tl>=0); } -#endif if(offset||s<0||c) addr=tl; else addr=s; //if(tl<0) tl=get_reg(i_regs->regmap,-1); @@ -2840,34 +2611,22 @@ void load_assemble(int i,struct regstat *i_regs) assert(tl>=0); // Even if the load is a NOP, we must check for pagefaults and I/O reglist&=~(1<=0) reglist&=~(1<regmap,ROREG); - if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); - #endif -//#define R29_HACK 1 - #ifdef R29_HACK - // Strmnnrmn's speed hack - if(rs1[i]!=29||start<0x80001000||start>=0x80000000+RAM_SIZE) - #endif - { - jaddr=emit_fastpath_cmp_jump(i,addr,&fastload_reg_override); - } + if(!c) { + #ifdef RAM_OFFSET + map=get_reg(i_regs->regmap,ROREG); + if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); + #endif + #ifdef R29_HACK + // Strmnnrmn's speed hack + if(rs1[i]!=29||start<0x80001000||start>=0x80000000+RAM_SIZE) + #endif + { + jaddr=emit_fastpath_cmp_jump(i,addr,&fastload_reg_override); } - else if(ram_offset&&memtarget) { - emit_addimm(addr,ram_offset,HOST_TEMPREG); - fastload_reg_override=HOST_TEMPREG; - } - }else{ // using tlb - int x=0; - if (opcode[i]==0x20||opcode[i]==0x24) x=3; // LB/LBU - if (opcode[i]==0x21||opcode[i]==0x25) x=2; // LH/LHU - map=get_reg(i_regs->regmap,TLREG); - assert(map>=0); - reglist&=~(1<regmap,rt1[i])); // ignore loads to r0 and unneeded reg if (opcode[i]==0x20) { // LB @@ -2880,7 +2639,6 @@ void load_assemble(int i,struct regstat *i_regs) #endif { //emit_xorimm(addr,3,tl); - //gen_tlb_addr_r(tl,map); //emit_movsbl_indexed((int)rdram-0x80000000,tl,tl); int x=0,a=tl; #ifdef BIG_ENDIAN_MIPS @@ -2921,7 +2679,6 @@ void load_assemble(int i,struct regstat *i_regs) //emit_movswl_indexed_tlb(x,tl,map,tl); //else if(map>=0) { - gen_tlb_addr_r(a,map); emit_movswl_indexed(x,a,tl); }else{ #if 1 //def RAM_OFFSET @@ -2967,7 +2724,6 @@ void load_assemble(int i,struct regstat *i_regs) #endif { //emit_xorimm(addr,3,tl); - //gen_tlb_addr_r(tl,map); //emit_movzbl_indexed((int)rdram-0x80000000,tl,tl); int x=0,a=tl; #ifdef BIG_ENDIAN_MIPS @@ -3008,7 +2764,6 @@ void load_assemble(int i,struct regstat *i_regs) //emit_movzwl_indexed_tlb(x,tl,map,tl); //#else if(map>=0) { - gen_tlb_addr_r(a,map); emit_movzwl_indexed(x,a,tl); }else{ #if 1 //def RAM_OFFSET @@ -3052,7 +2807,6 @@ void load_assemble(int i,struct regstat *i_regs) if(!dummy) { int a=addr; if(fastload_reg_override) a=fastload_reg_override; - //gen_tlb_addr_r(tl,map); //if(th>=0) emit_readword_indexed((int)rdram-0x80000000,addr,th); //emit_readword_indexed((int)rdram-0x7FFFFFFC,addr,tl); #ifdef HOST_IMM_ADDR32 @@ -3097,7 +2851,7 @@ void load_assemble(int i,struct regstat *i_regs) emit_call((int)memdebug); //emit_popa(); restore_regs(0x100f); - }/**/ + }*/ } #ifndef loadlr_assemble @@ -3113,7 +2867,7 @@ void store_assemble(int i,struct regstat *i_regs) int s,th,tl,map=-1; int addr,temp; int offset; - int jaddr=0,jaddr2,type; + int jaddr=0,type; int memtarget=0,c=0; int agr=AGEN1+(i&1); int faststore_reg_override=0; @@ -3128,7 +2882,6 @@ void store_assemble(int i,struct regstat *i_regs) c=(i_regs->wasconst>>s)&1; if(c) { memtarget=((signed int)(constmap[i][s]+offset))<(signed int)0x80000000+RAM_SIZE; - if(using_tlb&&((signed int)(constmap[i][s]+offset))>=(signed int)0xC0000000) memtarget=1; } } assert(tl>=0); @@ -3139,48 +2892,12 @@ void store_assemble(int i,struct regstat *i_regs) if(i_regs->regmap[HOST_CCREG]==CCREG) reglist&=~(1<=0x80000000+RAM_SIZE) - #endif - emit_cmpimm(addr,RAM_SIZE); - #ifdef DESTRUCTIVE_SHIFT - if(s==addr) emit_mov(s,temp); - #endif - #ifdef R29_HACK - memtarget=1; - if(rs1[i]!=29||start<0x80001000||start>=0x80000000+RAM_SIZE) - #endif - { - jaddr=(int)out; - #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK - // Hint to branch predictor that the branch is unlikely to be taken - if(rs1[i]>=28) - emit_jno_unlikely(0); - else - #endif - emit_jno(0); - } - #else - jaddr=emit_fastpath_cmp_jump(i,addr,&faststore_reg_override); - #endif - } - else if(ram_offset&&memtarget) { - emit_addimm(addr,ram_offset,HOST_TEMPREG); - faststore_reg_override=HOST_TEMPREG; - } - }else{ // using tlb - int x=0; - if (opcode[i]==0x28) x=3; // SB - if (opcode[i]==0x29) x=2; // SH - map=get_reg(i_regs->regmap,TLREG); - assert(map>=0); - reglist&=~(1<=0) { - gen_tlb_addr_w(a,map); emit_writehword_indexed(tl,x,a); }else //emit_writehword_indexed(tl,(int)rdram-0x80000000+x,a); @@ -3248,15 +2963,13 @@ void store_assemble(int i,struct regstat *i_regs) } type=STORED_STUB; } -#ifdef PCSX if(jaddr) { // PCSX store handlers don't check invcode again reglist|=1<waswritten&(1<waswritten&(1<isconst>>s)&1; if(c) { memtarget=((signed int)(constmap[i][s]+offset))<(signed int)0x80000000+RAM_SIZE; - if(using_tlb&&((signed int)(constmap[i][s]+offset))>=(signed int)0xC0000000) memtarget=1; } } assert(tl>=0); @@ -3367,41 +3079,26 @@ void storelr_assemble(int i,struct regstat *i_regs) if(i_regs->regmap[hr]>=0) reglist|=1<=0); - if(!using_tlb) { - if(!c) { - emit_cmpimm(s<0||offset?temp:s,RAM_SIZE); - if(!offset&&s!=temp) emit_mov(s,temp); - jaddr=(int)out; - emit_jno(0); - } - else - { - if(!memtarget||!rs1[i]) { - jaddr=(int)out; - emit_jmp(0); - } - } - #ifdef RAM_OFFSET - int map=get_reg(i_regs->regmap,ROREG); - if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); - gen_tlb_addr_w(temp,map); - #else - if((u_int)rdram!=0x80000000) - emit_addimm_no_flags((u_int)rdram-(u_int)0x80000000,temp); - #endif - }else{ // using tlb - int map=get_reg(i_regs->regmap,TLREG); - assert(map>=0); - reglist&=~(1<=0) emit_mov(s,temp); - do_tlb_w_branch(map,c,constmap[i][s]+offset,&jaddr); - if(!jaddr&&!memtarget) { + if(!c) { + emit_cmpimm(s<0||offset?temp:s,RAM_SIZE); + if(!offset&&s!=temp) emit_mov(s,temp); + jaddr=(int)out; + emit_jno(0); + } + else + { + if(!memtarget||!rs1[i]) { jaddr=(int)out; emit_jmp(0); } - gen_tlb_addr_w(temp,map); } + #ifdef RAM_OFFSET + int map=get_reg(i_regs->regmap,ROREG); + if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG); + #else + if((u_int)rdram!=0x80000000) + emit_addimm_no_flags((u_int)rdram-(u_int)0x80000000,temp); + #endif if (opcode[i]==0x2C||opcode[i]==0x2D) { // SDL/SDR temp2=get_reg(i_regs->regmap,FTEMP); @@ -3544,7 +3241,7 @@ void storelr_assemble(int i,struct regstat *i_regs) } if(!c||!memtarget) add_stub(STORELR_STUB,jaddr,(int)out,i,(int)i_regs,temp,ccadj[i],reglist); - if(!using_tlb&&!(i_regs->waswritten&(1<waswritten&(1<regmap,ROREG); if(map<0) map=HOST_TEMPREG; @@ -3562,7 +3259,7 @@ void storelr_assemble(int i,struct regstat *i_regs) #if defined(HAVE_CONDITIONAL_CALL) && !defined(DESTRUCTIVE_SHIFT) emit_callne(invalidate_addr_reg[temp]); #else - jaddr2=(int)out; + int jaddr2=(int)out; emit_jne(0); add_stub(INVCODE_STUB,jaddr2,(int)out,reglist|(1<regmap,FTEMP|64); - tl=get_reg(i_regs->regmap,FTEMP); - s=get_reg(i_regs->regmap,rs1[i]); - temp=get_reg(i_regs->regmap,agr); - if(temp<0) temp=get_reg(i_regs->regmap,-1); - offset=imm[i]; - assert(tl>=0); - assert(rs1[i]>0); - assert(temp>=0); - for(hr=0;hrregmap[hr]>=0) reglist|=1<regmap[HOST_CCREG]==CCREG) reglist&=~(1<wasconst>>s)&1; - if(s>=0) c=(i_regs->wasconst>>s)&1; - // Check cop1 unusable - if(!cop1_usable) { - signed char rs=get_reg(i_regs->regmap,CSREG); - assert(rs>=0); - emit_testimm(rs,0x20000000); - jaddr=(int)out; - emit_jeq(0); - add_stub(FP_STUB,jaddr,(int)out,i,rs,(int)i_regs,is_delayslot,0); - cop1_usable=1; - } - if (opcode[i]==0x39) { // SWC1 (get float address) - emit_readword((int)®_cop1_simple[(source[i]>>16)&0x1f],tl); - } - if (opcode[i]==0x3D) { // SDC1 (get double address) - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],tl); - } - // Generate address + offset - if(!using_tlb) { - if(!c) - emit_cmpimm(offset||c||s<0?ar:s,RAM_SIZE); - } - else - { - map=get_reg(i_regs->regmap,TLREG); - assert(map>=0); - reglist&=~(1<>16)&0x1f],temp); - } - if (opcode[i]==0x35) { // LDC1 (get target address) - emit_readword((int)®_cop1_double[(source[i]>>16)&0x1f],temp); - } - if(!using_tlb) { - if(!c) { - jaddr2=(int)out; - emit_jno(0); - } - else if(((signed int)(constmap[i][s]+offset))>=(signed int)0x80000000+RAM_SIZE) { - jaddr2=(int)out; - emit_jmp(0); // inline_readstub/inline_writestub? Very rare case - } - #ifdef DESTRUCTIVE_SHIFT - if (opcode[i]==0x39||opcode[i]==0x3D) { // SWC1/SDC1 - if(!offset&&!c&&s>=0) emit_mov(s,ar); - } - #endif - }else{ - if (opcode[i]==0x31||opcode[i]==0x35) { // LWC1/LDC1 - do_tlb_r_branch(map,c,constmap[i][s]+offset,&jaddr2); - } - if (opcode[i]==0x39||opcode[i]==0x3D) { // SWC1/SDC1 - do_tlb_w_branch(map,c,constmap[i][s]+offset,&jaddr2); - } - } - if (opcode[i]==0x31) { // LWC1 - //if(s>=0&&!c&&!offset) emit_mov(s,tl); - //gen_tlb_addr_r(ar,map); - //emit_readword_indexed((int)rdram-0x80000000,tl,tl); - #ifdef HOST_IMM_ADDR32 - if(c) emit_readword_tlb(constmap[i][s]+offset,map,tl); - else - #endif - emit_readword_indexed_tlb(0,offset||c||s<0?tl:s,map,tl); - type=LOADW_STUB; - } - if (opcode[i]==0x35) { // LDC1 - assert(th>=0); - //if(s>=0&&!c&&!offset) emit_mov(s,tl); - //gen_tlb_addr_r(ar,map); - //emit_readword_indexed((int)rdram-0x80000000,tl,th); - //emit_readword_indexed((int)rdram-0x7FFFFFFC,tl,tl); - #ifdef HOST_IMM_ADDR32 - if(c) emit_readdword_tlb(constmap[i][s]+offset,map,th,tl); - else - #endif - emit_readdword_indexed_tlb(0,offset||c||s<0?tl:s,map,th,tl); - type=LOADD_STUB; - } - if (opcode[i]==0x39) { // SWC1 - //emit_writeword_indexed(tl,(int)rdram-0x80000000,temp); - emit_writeword_indexed_tlb(tl,0,offset||c||s<0?temp:s,map,temp); - type=STOREW_STUB; - } - if (opcode[i]==0x3D) { // SDC1 - assert(th>=0); - //emit_writeword_indexed(th,(int)rdram-0x80000000,temp); - //emit_writeword_indexed(tl,(int)rdram-0x7FFFFFFC,temp); - emit_writedword_indexed_tlb(th,tl,0,offset||c||s<0?temp:s,map,temp); - type=STORED_STUB; - } - if(!using_tlb&&!(i_regs->waswritten&(1<regmap,INVCP); - assert(ir>=0); - emit_cmpmem_indexedsr12_reg(ir,temp,1); - #else - emit_cmpmem_indexedsr12_imm((int)invalid_code,temp,1); - #endif - #if defined(HAVE_CONDITIONAL_CALL) && !defined(DESTRUCTIVE_SHIFT) - emit_callne(invalidate_addr_reg[temp]); - #else - jaddr3=(int)out; - emit_jne(0); - add_stub(INVCODE_STUB,jaddr3,(int)out,reglist|(1<regmap,CCREG)<0) - emit_loadreg(CCREG,HOST_CCREG); - emit_add(HOST_CCREG,ECX,HOST_CCREG); - emit_addimm(HOST_CCREG,2*ccadj[i],HOST_CCREG); - emit_writeword(HOST_CCREG,(int)&Count); - emit_call((int)memdebug); - emit_popa(); - }/**/ -#else + +void c1ls_assemble(int i,struct regstat *i_regs) +{ cop1_unusable(i, i_regs); -#endif } void c2ls_assemble(int i,struct regstat *i_regs) @@ -3775,7 +3290,7 @@ void c2ls_assemble(int i,struct regstat *i_regs) int ar; int offset; int memtarget=0,c=0; - int jaddr2=0,jaddr3,type; + int jaddr2=0,type; int agr=AGEN1+(i&1); int fastio_reg_override=0; u_int hr,reglist=0; @@ -3785,7 +3300,6 @@ void c2ls_assemble(int i,struct regstat *i_regs) offset=imm[i]; assert(rs1[i]>0); assert(tl>=0); - assert(!using_tlb); for(hr=0;hrregmap[hr]>=0) reglist|=1<regmap,CCREG); assert(ccreg==HOST_CCREG); assert(!is_delayslot); + (void)ccreg; emit_movimm(start+i*4,EAX); // Get PC emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); // CHECK: is this right? There should probably be an extra cycle... emit_jmp((int)jump_syscall_hle); // XXX @@ -3928,6 +3443,7 @@ void hlecall_assemble(int i,struct regstat *i_regs) signed char ccreg=get_reg(i_regs->regmap,CCREG); assert(ccreg==HOST_CCREG); assert(!is_delayslot); + (void)ccreg; emit_movimm(start+i*4+4,0); // Get PC emit_movimm((int)psxHLEt[source[i]&7],1); emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); // XXX @@ -3939,6 +3455,7 @@ void intcall_assemble(int i,struct regstat *i_regs) signed char ccreg=get_reg(i_regs->regmap,CCREG); assert(ccreg==HOST_CCREG); assert(!is_delayslot); + (void)ccreg; emit_movimm(start+i*4,0); // Get PC emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); emit_jmp((int)jump_intcall); @@ -4007,7 +3524,7 @@ int internal_branch(uint64_t i_is32,int addr) if(addr&1) return 0; // Indirect (register) jump if(addr>=start && addr>2; + //int t=(addr-start)>>2; // Delay slots are not valid branch targets //if(t>0&&(itype[t-1]==RJUMP||itype[t-1]==UJUMP||itype[t-1]==CJUMP||itype[t-1]==SJUMP||itype[t-1]==FJUMP)) return 0; // 64 -> 32 bit transition requires a recompile @@ -4017,11 +3534,7 @@ int internal_branch(uint64_t i_is32,int addr) else printf("optimizable: yes\n"); }*/ //if(is32[t]&~unneeded_reg_upper[t]&~i_is32) return 0; -#ifndef FORCE32 - if(requires_32bit[t]&~i_is32) return 0; - else -#endif - return 1; + return 1; } return 0; } @@ -4156,10 +3669,9 @@ 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||itype[i]==C2LS) { int ra=-1; int agr=AGEN1+(i&1); - int mgr=MGEN1+(i&1); if(itype[i]==LOAD) { ra=get_reg(i_regs->regmap,rt1[i]); - if(ra<0) ra=get_reg(i_regs->regmap,-1); + if(ra<0) ra=get_reg(i_regs->regmap,-1); assert(ra>=0); } if(itype[i]==LOADLR) { @@ -4178,17 +3690,11 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) } } int rs=get_reg(i_regs->regmap,rs1[i]); - int rm=get_reg(i_regs->regmap,TLREG); if(ra>=0) { int offset=imm[i]; int c=(i_regs->wasconst>>rs)&1; if(rs1[i]==0) { // Using r0 as a base address - /*if(rm>=0) { - if(!entry||entry[rm]!=mgr) { - generate_map_const(offset,rm); - } // else did it in the previous cycle - }*/ if(!entry||entry[ra]!=agr) { if (opcode[i]==0x22||opcode[i]==0x26) { emit_movimm(offset&0xFFFFFFFC,ra); // LWL/LWR @@ -4206,22 +3712,6 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[]) // printf("poor load scheduling!\n"); } else if(c) { -#ifndef DISABLE_TLB - if(rm>=0) { - if(!entry||entry[rm]!=mgr) { - 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 || - (unsigned int)(constmap[i][rs]+offset)<0x80000000+RAM_SIZE ) - generate_map_const(constmap[i][rs]+offset,rm); - }else{ - if((signed int)(constmap[i][rs]+offset)>=(signed int)0xC0000000) - generate_map_const(constmap[i][rs]+offset,rm); - } - } - } -#endif if(rs1[i]!=rt1[i]||itype[i]!=LOAD) { if(!entry||entry[ra]!=agr) { if (opcode[i]==0x22||opcode[i]==0x26) { @@ -4230,8 +3720,7 @@ 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]&0x3b)!=0x31&&(opcode[i]&0x3b)!=0x32) || // LWC1/LDC1/LWC2/LDC2 - (using_tlb&&((signed int)constmap[i][rs]+offset)>=(signed int)0xC0000000)) + if((itype[i]!=LOAD&&(opcode[i]&0x3b)!=0x31&&(opcode[i]&0x3b)!=0x32)) // LWC1/LDC1/LWC2/LDC2 #endif emit_movimm(constmap[i][rs]+offset,ra); regs[i].loadedconst|=1<regmap,agr); - if(ra>=0) { - int rs=get_reg(regs[i+1].regmap,rs1[i+1]); - 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]&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 || - (unsigned int)(constmap[i+1][rs]+offset)<0x80000000+RAM_SIZE ) - generate_map_const(constmap[i+1][rs]+offset,ra); - }else{ - if((signed int)(constmap[i+1][rs]+offset)>=(signed int)0xC0000000) - generate_map_const(constmap[i+1][rs]+offset,ra); - } - } - /*else if(rs1[i]==0) { - generate_map_const(offset,ra); - }*/ - } - #endif // Actual address agr=AGEN1+((i+1)&1); ra=get_reg(i_regs->regmap,agr); @@ -4291,8 +3754,7 @@ 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]&0x3b)!=0x31&&(opcode[i+1]&0x3b)!=0x32) || // LWC1/LDC1/LWC2/LDC2 - (using_tlb&&((signed int)constmap[i+1][rs]+offset)>=(signed int)0xC0000000)) + if((itype[i+1]!=LOAD&&(opcode[i+1]&0x3b)!=0x31&&(opcode[i+1]&0x3b)!=0x32)) // LWC1/LDC1/LWC2/LDC2 #endif emit_movimm(constmap[i+1][rs]+offset,ra); regs[i+1].loadedconst|=1<>hr)&1)) { - #ifdef HOST_IMM_ADDR32 - if(!using_tlb||((signed int)constmap[i][hr]+imm[i+2])<(signed int)0xC0000000) return 0; - #endif // Precompute load address *value=constmap[i][hr]+imm[i+2]; return 1; @@ -4341,9 +3800,6 @@ int get_final_value(int hr, int i, int *value) } if(itype[i+1]==LOAD&&rs1[i+1]==reg&&rt1[i+1]==reg) { - #ifdef HOST_IMM_ADDR32 - if(!using_tlb||((signed int)constmap[i][hr]+imm[i+1])<(signed int)0xC0000000) return 0; - #endif // Precompute load address *value=constmap[i][hr]+imm[i+1]; //printf("c=%x imm=%x\n",(int)constmap[i][hr],imm[i+1]); @@ -4493,17 +3949,6 @@ 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); - emit_storereg(i_regmap[hr]|64,hr); - #else - emit_sarimm(hr,31,HOST_TEMPREG); - 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); @@ -4529,17 +3974,6 @@ 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); - emit_storereg(i_regmap[hr]|64,hr); - #else - emit_sarimm(hr,31,HOST_TEMPREG); - 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); @@ -4757,7 +4191,7 @@ int match_bt(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr) { return 0; } - else + else if((i_dirty>>hr)&1) { if(i_regmap[hr]0&&(itype[t-1]==RJUMP||itype[t-1]==UJUMP||itype[t-1]==CJUMP||itype[t-1]==SJUMP||itype[t-1]==FJUMP)) return 0; // Delay slots require additional processing, so do not match @@ -5003,7 +4434,7 @@ void do_ccstub(int n) if(rs1[i]) { if((branch_regs[i].dirty>>s1l)&(branch_regs[i].is32>>rs1[i])&1) emit_loadreg(rs1[i],s1l); - } + } else { if((branch_regs[i].dirty>>s1l)&(branch_regs[i].is32>>rs2[i])&1) emit_loadreg(rs2[i],s1l); @@ -5194,7 +4625,7 @@ void do_ccstub(int n) load_all_regs(branch_regs[i].regmap); } emit_jmp(stubs[n][2]); // return address - + /* This works but uses a lot of memory... emit_readword((int)&last_count,ECX); emit_add(HOST_CCREG,ECX,EAX); @@ -5224,11 +4655,11 @@ void do_ccstub(int n) emit_jmpreg(EAX);*/ } -add_to_linker(int addr,int target,int ext) +static void add_to_linker(int addr,int target,int ext) { link_addr[linkcount][0]=addr; link_addr[linkcount][1]=target; - link_addr[linkcount][2]=ext; + link_addr[linkcount][2]=ext; linkcount++; } @@ -5254,7 +4685,7 @@ static void ujump_assemble_write_ra(int i) #endif { #ifdef REG_PREFETCH - if(temp>=0) + if(temp>=0) { if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp); } @@ -5269,16 +4700,16 @@ static void ujump_assemble_write_ra(int i) void ujump_assemble(int i,struct regstat *i_regs) { - signed char *i_regmap=i_regs->regmap; int ra_done=0; if(i==(ba[i]-start)>>2) assem_debug("idle loop\n"); address_generation(i+1,i_regs,regs[i].regmap_entry); #ifdef REG_PREFETCH int temp=get_reg(branch_regs[i].regmap,PTEMP); - if(rt1[i]==31&&temp>=0) + if(rt1[i]==31&&temp>=0) { + signed char *i_regmap=i_regs->regmap; int return_address=start+i*4+8; - if(get_reg(branch_regs[i].regmap,31)>0) + if(get_reg(branch_regs[i].regmap,31)>0) if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp); } #endif @@ -5329,7 +4760,7 @@ static void rjump_assemble_write_ra(int i) assert(rt>=0); return_address=start+i*4+8; #ifdef REG_PREFETCH - if(temp>=0) + if(temp>=0) { if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp); } @@ -5342,9 +4773,8 @@ static void rjump_assemble_write_ra(int i) void rjump_assemble(int i,struct regstat *i_regs) { - signed char *i_regmap=i_regs->regmap; int temp; - int rs,cc,adj; + int rs,cc; int ra_done=0; rs=get_reg(branch_regs[i].regmap,rs1[i]); assert(rs>=0); @@ -5358,9 +4788,10 @@ void rjump_assemble(int i,struct regstat *i_regs) } address_generation(i+1,i_regs,regs[i].regmap_entry); #ifdef REG_PREFETCH - if(rt1[i]==31) + if(rt1[i]==31) { if((temp=get_reg(branch_regs[i].regmap,PTEMP))>=0) { + signed char *i_regmap=i_regs->regmap; int return_address=start+i*4+8; if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp); } @@ -5389,6 +4820,7 @@ void rjump_assemble(int i,struct regstat *i_regs) rjump_assemble_write_ra(i); cc=get_reg(branch_regs[i].regmap,CCREG); assert(cc==HOST_CCREG); + (void)cc; #ifdef USE_MINI_HT int rh=get_reg(branch_regs[i].regmap,RHASH); int ht=get_reg(branch_regs[i].regmap,RHTBL); @@ -5419,13 +4851,11 @@ void rjump_assemble(int i,struct regstat *i_regs) //assert(adj==0); emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG); add_stub(CC_STUB,(int)out,jump_vaddr_reg[rs],0,i,-1,TAKEN,0); -#ifdef PCSX if(itype[i+1]==COP0&&(source[i+1]&0x3f)==0x10) // special case for RFE emit_jmp(0); else -#endif - emit_jns(0); + emit_jns(0); //load_regs_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,-1); #ifdef USE_MINI_HT if(rs1[i]==31) { @@ -5492,7 +4922,7 @@ void cjump_assemble(int i,struct regstat *i_regs) #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK if(i>(ba[i]-start)>>2) invert=1; #endif - + if(ooo[i]) { s1l=get_reg(branch_regs[i].regmap,rs1[i]); s1h=get_reg(branch_regs[i].regmap,rs1[i]|64); @@ -5547,7 +4977,7 @@ void cjump_assemble(int i,struct regstat *i_regs) load_regs(regs[i].regmap,branch_regs[i].regmap,regs[i].was32,CCREG,CCREG); cc=get_reg(branch_regs[i].regmap,CCREG); assert(cc==HOST_CCREG); - if(unconditional) + if(unconditional) store_regs_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]); //do_cc(i,branch_regs[i].regmap,&adj,unconditional?ba[i]:-1,unconditional); //assem_debug("cycle count (adj)\n"); @@ -5619,7 +5049,7 @@ void cjump_assemble(int i,struct regstat *i_regs) emit_jne(0); } } // if(!only32) - + //printf("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(s1l>=0); if(opcode[i]==4) // BEQ @@ -5749,7 +5179,7 @@ void cjump_assemble(int i,struct regstat *i_regs) emit_jne(1); } } // if(!only32) - + //printf("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(s1l>=0); if((opcode[i]&0x2f)==4) // BEQ @@ -5932,7 +5362,7 @@ void sjump_assemble(int i,struct regstat *i_regs) } cc=get_reg(branch_regs[i].regmap,CCREG); assert(cc==HOST_CCREG); - if(unconditional) + if(unconditional) store_regs_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]); //do_cc(i,branch_regs[i].regmap,&adj,unconditional?ba[i]:-1,unconditional); assem_debug("cycle count (adj)\n"); @@ -6019,7 +5449,7 @@ void sjump_assemble(int i,struct regstat *i_regs) } } } // if(!only32) - + if(invert) { #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK if(match&&(!internal||!is_ds[(ba[i]-start)>>2])) { @@ -6267,7 +5697,7 @@ void fjump_assemble(int i,struct regstat *i_regs) { } } // if(!only32) - + if(invert) { if(adj) emit_addimm(cc,-CLOCK_ADJUST(adj),cc); #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK @@ -6396,7 +5826,6 @@ static void pagespan_assemble(int i,struct regstat *i_regs) int s1h=get_reg(i_regs->regmap,rs1[i]|64); int s2l=get_reg(i_regs->regmap,rs2[i]); int s2h=get_reg(i_regs->regmap,rs2[i]|64); - void *nt_branch=NULL; int taken=0; int nottaken=0; int unconditional=0; @@ -6413,7 +5842,7 @@ static void pagespan_assemble(int i,struct regstat *i_regs) s1h=s2h=-1; } int hr=0; - int addr,alt,ntaddr; + int addr=-1,alt=-1,ntaddr=-1; if(i_regs->regmap[HOST_BTREG]<0) {addr=HOST_BTREG;} else { while(hr=(start+slen*4)) { // Branch out of this block, flush all regs u=1; uu=1; gte_u=gte_u_unknown; - /* Hexagon hack + /* Hexagon hack if(itype[i]==UJUMP&&rt1[i]==31) { uu=u=0x300C00F; // Discard at, v0-v1, t6-t9 @@ -7003,389 +6432,35 @@ void unneeded_registers(int istart,int iend,int r) uu&=~(tdep<>r)&1) { - if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - printf(" UU:"); - for(r=1;r<=CCREG;r++) { - if(((unneeded_reg_upper[i]&~unneeded_reg[i])>>r)&1) { - if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",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 -// This is used to predict whether any branches will jump to a -// location with 64-bit values in registers. -static void provisional_32bit() -{ - int i,j; - uint64_t is32=1; - uint64_t lastbranch=1; - - for(i=0;i0) { - if(itype[i-1]==CJUMP||itype[i-1]==SJUMP||itype[i-1]==FJUMP) { - if(i>1) is32=lastbranch; - else is32=1; - } - } - if(i>1) - { - if(itype[i-2]==CJUMP||itype[i-2]==SJUMP||itype[i-2]==FJUMP) { - if(likely[i-2]) { - if(i>2) is32=lastbranch; - else is32=1; - } - } - if((opcode[i-2]&0x2f)==0x05) // BNE/BNEL - { - if(rs1[i-2]==0||rs2[i-2]==0) - { - if(rs1[i-2]) { - is32|=1LL<=0;j--) - { - if(ba[j]==start+i*4) - //temp_is32&=branch_regs[j].is32; - temp_is32&=p32[j]; - } - for(j=i;j>s1)&1LL)<>s1)&1LL); - is32&=~(1LL<=0x20&&op2<=0x23) { // ADD/ADDU/SUB/SUBU - is32|=1LL<=0x24&&op2<=0x27) { // AND/OR/XOR/NOR - uint64_t sr=((is32>>s1)&(is32>>s2)&1LL); - is32&=~(1LL<=0x2c&&op2<=0x2d) { // DADD/DADDU - if(s1==0&&s2==0) { - is32|=1LL<>s1)&1LL); - is32&=~(1LL<>s2)&1LL); - is32&=~(1LL<=0x2e&&op2<=0x2f) { // DSUB/DSUBU - if(s1==0&&s2==0) { - is32|=1LL<>s1)&1LL); - is32&=~(1LL<=0x1c&&op2<=0x1f) { // DMULT/DMULTU/DDIV/DDIVU - is32&=~((1LL<>s1)&1LL); - is32&=~(1LL<=0x14&&op2<=0x17) is32&=~(1LL<=0x38&&op2<0x3f) is32&=~(1LL<0) - { - if(itype[i-1]==UJUMP||itype[i-1]==RJUMP||(source[i-1]>>16)==0x1000) - { - if(rt1[i-1]==31) // JAL/JALR - { - // Subroutine call will return here, don't alloc any registers - is32=1; - } - else if(i+1=0;i--) - { - int hr; - if(itype[i]==RJUMP||itype[i]==UJUMP||itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP) - { - if(ba[i]=(start+slen*4)) - { - // Branch out of this block, don't need anything - r32=0; - } - else - { - // Internal branch - // Need whatever matches the target - // (and doesn't get overwritten by the delay slot instruction) - r32=0; - int t=(ba[i]-start)>>2; - if(ba[i]>start+i*4) { - // Forward branch - //if(!(requires_32bit[t]&~regs[i].was32)) - // r32|=requires_32bit[t]&(~(1LL<>16)!=0x1000) - { - if(i0) - { - if((regs[i].was32>>us1[i+1])&1) r32|=1LL<0) - { - if((regs[i].was32>>us2[i+1])&1) r32|=1LL<>dep1[i+1])&1)) - { - if((regs[i].was32>>dep1[i+1])&1) r32|=1LL<>dep2[i+1])&1)) - { - if((regs[i].was32>>dep2[i+1])&1) r32|=1LL<0) - { - if((regs[i].was32>>us1[i])&1) r32|=1LL<0) - { - if((regs[i].was32>>us2[i])&1) r32|=1LL<>dep1[i])&1)) - { - if((regs[i].was32>>dep1[i])&1) r32|=1LL<>dep2[i])&1)) - { - if((regs[i].was32>>dep2[i])&1) r32|=1LL<>r)&1) { + if(r==HIREG) printf(" HI"); + else if(r==LOREG) printf(" LO"); + else printf(" r%d",r); + } } - //requires_32bit[i]=r32; - pr32[i]=r32; - - // Dirty registers which are 32-bit, require 32-bit input - // as they will be written as 32-bit values - for(hr=0;hr0&®s[i].regmap_entry[hr]<64) { - if((regs[i].was32>>regs[i].regmap_entry[hr])&(regs[i].wasdirty>>hr)&1) { - if(!((unneeded_reg_upper[i]>>regs[i].regmap_entry[hr])&1)) - pr32[i]|=1LL<>r)&1) { + if(r==HIREG) printf(" HI"); + else if(r==LOREG) printf(" LO"); + else printf(" r%d",r); } } + printf("\n");*/ + } + for (i=iend;i>=istart;i--) + { + unneeded_reg_upper[i]=branch_unneeded_reg_upper[i]=-1LL; } } @@ -7726,7 +6801,7 @@ void clean_registers(int istart,int iend,int wr) if((regs[i].regmap[r]&63)==rt2[i]) wont_dirty_i|=1<istart) { - if(itype[i]!=RJUMP&&itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=FJUMP) + if(itype[i]!=RJUMP&&itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=FJUMP) { // Don't store a register immediately after writing it, // may prevent dual-issue. @@ -7760,7 +6835,7 @@ void clean_registers(int istart,int iend,int wr) if(r!=EXCLUDE_REG) { if(regs[i].regmap[r]==regmap_pre[i+2][r]) { regs[i+2].wasdirty&=wont_dirty_i|~(1<>r)&1));*/} + }else {/*printf("i: %x (%d) mismatch(+2): %d\n",start+i*4,i,r);assert(!((wont_dirty_i>>r)&1));*/} } } } @@ -7772,7 +6847,7 @@ void clean_registers(int istart,int iend,int wr) if(r!=EXCLUDE_REG) { if(regs[i].regmap[r]==regmap_pre[i+1][r]) { regs[i+1].wasdirty&=wont_dirty_i|~(1<>r)&1));*/} + }else {/*printf("i: %x (%d) mismatch(+1): %d\n",start+i*4,i,r);assert(!((wont_dirty_i>>r)&1));*/} } } } @@ -7815,7 +6890,7 @@ void clean_registers(int istart,int iend,int wr) wont_dirty_i|=((unneeded_reg[i]>>(regmap_pre[i][r]&63))&1)<>r)&1));*/ + /*printf("i: %x (%d) mismatch: %d\n",start+i*4,i,r);assert(!((will_dirty>>r)&1));*/ } } } @@ -7921,13 +6996,14 @@ static void disassemble_inst(int i) {} static int new_dynarec_test(void) { int (*testfunc)(void) = (void *)out; + void *beginning; int ret; + + beginning = start_block(); emit_movimm(DRC_TEST_VAL,0); // test emit_jmpreg(14); literal_pool(0); -#ifdef __arm__ - __clear_cache((void *)testfunc, out); -#endif + end_block(beginning); SysPrintf("testing if we can run recompiled code..\n"); ret = testfunc(); if (ret == DRC_TEST_VAL) @@ -7956,15 +7032,6 @@ void new_dynarec_clear_full() stop_after_jal=0; inv_code_start=inv_code_end=~0; // TLB -#ifndef DISABLE_TLB - using_tlb=0; - for(n=0;n<524288;n++) // 0 .. 0x7FFFFFFF - memory_map[n]=-1; - for(n=524288;n<526336;n++) // 0x80000000 .. 0x807FFFFF - memory_map[n]=((u_int)rdram-0x80000000)>>2; - for(n=526336;n<1048576;n++) // 0x80800000 .. 0xFFFFFFFF - memory_map[n]=-1; -#endif for(n=0;n<4096;n++) ll_clear(jump_in+n); for(n=0;n<4096;n++) ll_clear(jump_out+n); for(n=0;n<4096;n++) ll_clear(jump_dirty+n); @@ -7973,71 +7040,49 @@ void new_dynarec_clear_full() void new_dynarec_init() { SysPrintf("Init new dynarec\n"); - out=(u_char *)BASE_ADDR; -#if BASE_ADDR_FIXED - if (mmap (out, 1< %x\n", (int)addr, (int)out); //printf("NOTCOMPILED: addr = %x -> %x\n", (int)addr, (int)out); //printf("TRACE: count=%d next=%d (compile %x)\n",Count,next_interupt,addr); - //if(debug) + //if(debug) //printf("TRACE: count=%d next=%d (checksum %x)\n",Count,next_interupt,mchecksum()); //printf("fpu mapping=%x enabled=%x\n",(Status & 0x04000000)>>26,(Status & 0x20000000)>>29); /*if(Count>=312978186) { @@ -8210,16 +7263,15 @@ int new_recompile_block(int addr) if (Config.HLE && start == 0x80001000) // hlecall { // XXX: is this enough? Maybe check hleSoftCall? - u_int beginning=(u_int)out; + void *beginning=start_block(); u_int page=get_page(start); + invalid_code[start>>12]=0; emit_movimm(start,0); emit_writeword(0,(int)&pcaddr); emit_jmp((int)new_dyna_leave); literal_pool(0); -#ifdef __arm__ - __clear_cache((void *)beginning,out); -#endif + end_block(beginning); ll_add_flags(jump_in+page,start,state_rflags,(void *)beginning); return 0; } @@ -8246,7 +7298,7 @@ int new_recompile_block(int addr) unsigned int type,op,op2; //printf("addr = %x source = %x %x\n", addr,source,source[0]); - + /* Pass 1 disassembly */ for(i=0;!done;i++) { @@ -8294,7 +7346,7 @@ int new_recompile_block(int addr) case 0x33: strcpy(insn[i],"TLTU"); type=NI; break; case 0x34: strcpy(insn[i],"TEQ"); type=NI; break; case 0x36: strcpy(insn[i],"TNE"); type=NI; break; -#ifndef FORCE32 +#if 0 case 0x14: strcpy(insn[i],"DSLLV"); type=SHIFT; break; case 0x16: strcpy(insn[i],"DSRLV"); type=SHIFT; break; case 0x17: strcpy(insn[i],"DSRAV"); type=SHIFT; break; @@ -8362,11 +7414,8 @@ int new_recompile_block(int addr) case 0x02: strcpy(insn[i],"TLBWI"); type=COP0; break; case 0x06: strcpy(insn[i],"TLBWR"); type=COP0; break; case 0x08: strcpy(insn[i],"TLBP"); type=COP0; break; -#ifdef PCSX case 0x10: strcpy(insn[i],"RFE"); type=COP0; break; -#else - case 0x18: strcpy(insn[i],"ERET"); type=COP0; break; -#endif + //case 0x18: strcpy(insn[i],"ERET"); type=COP0; break; } } break; @@ -8485,7 +7534,7 @@ int new_recompile_block(int addr) break; } break; -#ifndef FORCE32 +#if 0 case 0x14: strcpy(insn[i],"BEQL"); type=CJUMP; break; case 0x15: strcpy(insn[i],"BNEL"); type=CJUMP; break; case 0x16: strcpy(insn[i],"BLEZL"); type=CJUMP; break; @@ -8502,14 +7551,14 @@ int new_recompile_block(int addr) case 0x24: strcpy(insn[i],"LBU"); type=LOAD; break; case 0x25: strcpy(insn[i],"LHU"); type=LOAD; break; case 0x26: strcpy(insn[i],"LWR"); type=LOADLR; break; -#ifndef FORCE32 +#if 0 case 0x27: strcpy(insn[i],"LWU"); type=LOAD; break; #endif case 0x28: strcpy(insn[i],"SB"); type=STORE; break; case 0x29: strcpy(insn[i],"SH"); type=STORE; break; case 0x2A: strcpy(insn[i],"SWL"); type=STORELR; break; case 0x2B: strcpy(insn[i],"SW"); type=STORE; break; -#ifndef FORCE32 +#if 0 case 0x2C: strcpy(insn[i],"SDL"); type=STORELR; break; case 0x2D: strcpy(insn[i],"SDR"); type=STORELR; break; #endif @@ -8517,19 +7566,18 @@ int new_recompile_block(int addr) case 0x2F: strcpy(insn[i],"CACHE"); type=NOP; break; case 0x30: strcpy(insn[i],"LL"); type=NI; break; case 0x31: strcpy(insn[i],"LWC1"); type=C1LS; break; -#ifndef FORCE32 +#if 0 case 0x34: strcpy(insn[i],"LLD"); type=NI; break; case 0x35: strcpy(insn[i],"LDC1"); type=C1LS; break; case 0x37: strcpy(insn[i],"LD"); type=LOAD; break; #endif case 0x38: strcpy(insn[i],"SC"); type=NI; break; case 0x39: strcpy(insn[i],"SWC1"); type=C1LS; break; -#ifndef FORCE32 +#if 0 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; #endif -#ifdef PCSX case 0x12: strcpy(insn[i],"COP2"); type=NI; op2=(source[i]>>21)&0x1f; //if (op2 & 0x10) { @@ -8553,7 +7601,6 @@ int new_recompile_block(int addr) 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; SysPrintf("NI %08x @%08x (%08x)\n", source[i], addr + i*4, addr); break; @@ -8824,7 +7871,6 @@ int new_recompile_block(int addr) else if(type==CJUMP||type==SJUMP||type==FJUMP) ba[i]=start+i*4+4+((signed int)((unsigned int)source[i]<<16)>>14); else ba[i]=-1; -#ifdef PCSX if(i>0&&(itype[i-1]==RJUMP||itype[i-1]==UJUMP||itype[i-1]==CJUMP||itype[i-1]==SJUMP||itype[i-1]==FJUMP)) { int do_in_intrp=0; // branch in delay slot? @@ -8858,7 +7904,6 @@ int new_recompile_block(int addr) i--; // don't compile the DS } } -#endif /* 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]==0) { // Continue past subroutine call (JAL) @@ -8906,7 +7951,7 @@ int new_recompile_block(int addr) /* Pass 2 - Register dependencies and branch targets */ unneeded_registers(0,slen-1,0); - + /* Pass 3 - Register allocation */ struct regstat current; // Current register allocations/status @@ -8924,9 +7969,6 @@ int new_recompile_block(int addr) int cc=0; int hr=-1; -#ifndef FORCE32 - provisional_32bit(); -#endif if((u_int)addr&1) { // First instruction is delay slot cc=-1; @@ -8936,7 +7978,7 @@ int new_recompile_block(int addr) unneeded_reg_upper[0]=1; current.regmap[HOST_BTREG]=BTREG; } - + for(i=0;i=0;j--) - { - if(ba[j]==start+i*4) - temp_is32&=branch_regs[j].is32; - } - for(j=i;j0&&r<64) - { - if((current.dirty>>hr)&((current.is32&~temp_is32)>>r)&1) { - temp_is32|=1LL<=0;j--) - { - if(ba[j]==start+i*4+4) - temp_is32&=branch_regs[j].is32; - } - for(j=i;j0) - { - if((current.dirty>>hr)&((current.is32&~temp_is32)>>(r&63))&1) { - if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP&&itype[i]!=FJUMP) - { - if(rs1[i]!=(r&63)&&rs2[i]!=(r&63)) - { - //printf("dump %d/r%d\n",hr,r); - current.regmap[hr]=-1; - if(get_reg(current.regmap,r|64)>=0) - current.regmap[get_reg(current.regmap,r|64)]=-1; - } - } - } - } - } - } - } - else if(i>16)!=0x1000&&(itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP)) - { - uint64_t temp_is32=current.is32; - for(j=i-1;j>=0;j--) - { - if(ba[j]==start+i*4+8) - temp_is32&=branch_regs[j].is32; - } - for(j=i;j0) - { - if((current.dirty>>hr)&((current.is32&~temp_is32)>>(r&63))&1) { - if(rs1[i]!=(r&63)&&rs2[i]!=(r&63)&&rs1[i+1]!=(r&63)&&rs2[i+1]!=(r&63)) - { - //printf("dump %d/r%d\n",hr,r); - current.regmap[hr]=-1; - if(get_reg(current.regmap,r|64)>=0) - current.regmap[get_reg(current.regmap,r|64)]=-1; - } - } - } - } - } - } - #endif if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP&&itype[i]!=FJUMP) { if(i+1=0) { if(r!=regmap_pre[i][hr]) { @@ -9568,7 +8497,7 @@ int new_recompile_block(int addr) } } else { // Branches expect CCREG to be allocated at the target - if(regmap_pre[i][hr]==CCREG) + if(regmap_pre[i][hr]==CCREG) regs[i].regmap_entry[hr]=CCREG; else regs[i].regmap_entry[hr]=-1; @@ -9878,7 +8807,7 @@ int new_recompile_block(int addr) { cc=0; } -#if defined(PCSX) && !defined(DRC_DBG) +#if !defined(DRC_DBG) else if(itype[i]==C2OP&>e_cycletab[source[i]&0x3f]>2) { // GTE runs in parallel until accessed, divide by 2 for a rough guess @@ -9919,11 +8848,11 @@ int new_recompile_block(int addr) if(current.regmap[HOST_BTREG]==BTREG) current.regmap[HOST_BTREG]=-1; regs[i].waswritten=current.waswritten; } - + /* Pass 4 - Cull unused host registers */ - + uint64_t nr=0; - + for (i=slen-1;i>=0;i--) { int hr; @@ -10065,7 +8994,7 @@ int new_recompile_block(int addr) } // Save it needed_reg[i]=nr; - + // Deallocate unneeded registers for(hr=0;hr=start && ba[i]<(start+i*4)) + if(ba[i]>=start && ba[i]<(start+i*4)) if(itype[i+1]==NOP||itype[i+1]==MOV||itype[i+1]==ALU ||itype[i+1]==SHIFTIMM||itype[i+1]==IMM16||itype[i+1]==LOAD ||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS @@ -10255,10 +9173,10 @@ int new_recompile_block(int addr) } } if(ooo[i]) { - if(count_free_regs(regs[i].regmap)<=minimum_free_regs[i+1]) + if(count_free_regs(regs[i].regmap)<=minimum_free_regs[i+1]) f_regmap[hr]=branch_regs[i].regmap[hr]; }else{ - if(count_free_regs(branch_regs[i].regmap)<=minimum_free_regs[i+1]) + if(count_free_regs(branch_regs[i].regmap)<=minimum_free_regs[i+1]) f_regmap[hr]=branch_regs[i].regmap[hr]; } // Avoid dirty->clean transition @@ -10428,10 +9346,10 @@ int new_recompile_block(int addr) if(itype[j]==CJUMP||itype[j]==SJUMP||itype[j]==FJUMP) { if(ooo[j]) { - if(count_free_regs(regs[j].regmap)<=minimum_free_regs[j+1]) + if(count_free_regs(regs[j].regmap)<=minimum_free_regs[j+1]) break; }else{ - if(count_free_regs(branch_regs[j].regmap)<=minimum_free_regs[j+1]) + if(count_free_regs(branch_regs[j].regmap)<=minimum_free_regs[j+1]) break; } if(get_reg(branch_regs[j].regmap,f_regmap[hr])>=0) { @@ -10504,7 +9422,7 @@ int new_recompile_block(int addr) regs[k].isconst&=~(1<i&&f_regmap[HOST_CCREG]==CCREG) @@ -10546,18 +9464,18 @@ int new_recompile_block(int addr) } } } - + // Cache memory offset or tlb map pointer if a register is available #ifndef HOST_IMM_ADDR32 #ifndef RAM_OFFSET - if(using_tlb) + if(0) #endif { int earliest_available[HOST_REGS]; int loop_start[HOST_REGS]; int score[HOST_REGS]; int end[HOST_REGS]; - int reg=using_tlb?MMREG:ROREG; + int reg=ROREG; // Init for(hr=0;hr=0) { @@ -10804,46 +9722,6 @@ int new_recompile_block(int addr) } } } - // Preload map address - #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||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]); - if(sr>=0&&((regs[i+1].wasconst>>sr)&1)) { - int nr; - if(regs[i].regmap[hr]<0&®s[i+1].regmap_entry[hr]<0) - { - regs[i].regmap[hr]=MGEN1+((i+1)&1); - regmap_pre[i+1][hr]=MGEN1+((i+1)&1); - regs[i+1].regmap_entry[hr]=MGEN1+((i+1)&1); - regs[i].isconst&=~(1<=0) - { - // move it to another register - regs[i+1].regmap[hr]=-1; - regmap_pre[i+2][hr]=-1; - regs[i+1].regmap[nr]=TLREG; - regmap_pre[i+2][nr]=TLREG; - regs[i].regmap[nr]=MGEN1+((i+1)&1); - regmap_pre[i+1][nr]=MGEN1+((i+1)&1); - regs[i+1].regmap_entry[nr]=MGEN1+((i+1)&1); - regs[i].isconst&=~(1<=0;i--) - { - int hr; - if(itype[i]==RJUMP||itype[i]==UJUMP||itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP) - { - if(ba[i]=(start+slen*4)) - { - // Branch out of this block, don't need anything - r32=0; - } - else - { - // Internal branch - // Need whatever matches the target - // (and doesn't get overwritten by the delay slot instruction) - r32=0; - int t=(ba[i]-start)>>2; - if(ba[i]>start+i*4) { - // Forward branch - if(!(requires_32bit[t]&~regs[i].was32)) - r32|=requires_32bit[t]&(~(1LL<>16)!=0x1000) - { - if(i0) - { - if((regs[i].was32>>us1[i+1])&1) r32|=1LL<0) - { - if((regs[i].was32>>us2[i+1])&1) r32|=1LL<>dep1[i+1])&1)) - { - if((regs[i].was32>>dep1[i+1])&1) r32|=1LL<>dep2[i+1])&1)) - { - if((regs[i].was32>>dep2[i+1])&1) r32|=1LL<0) - { - if((regs[i].was32>>us1[i])&1) r32|=1LL<0) - { - if((regs[i].was32>>us2[i])&1) r32|=1LL<>dep1[i])&1)) - { - if((regs[i].was32>>dep1[i])&1) r32|=1LL<>dep2[i])&1)) - { - if((regs[i].was32>>dep2[i])&1) r32|=1LL<0&®s[i].regmap_entry[hr]<64) { - if((regs[i].was32>>regs[i].regmap_entry[hr])&(regs[i].wasdirty>>hr)&1) { - if(!((unneeded_reg_upper[i]>>regs[i].regmap_entry[hr])&1)) - requires_32bit[i]|=1LL<=0;i--) { if(itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP) @@ -11057,7 +9820,6 @@ int new_recompile_block(int addr) } } } -#endif if(itype[slen-1]==SPAN) { bt[slen-1]=1; // Mark as a branch target so instruction can restart after exception @@ -11076,26 +9838,6 @@ int new_recompile_block(int addr) else printf(" r%d",r); } } -#ifndef FORCE32 - printf(" UU:"); - for(r=1;r<=CCREG;r++) { - if(((unneeded_reg_upper[i]&~unneeded_reg[i])>>r)&1) { - if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - printf(" 32:"); - for(r=0;r<=CCREG;r++) { - //if(((is32[i]>>r)&(~unneeded_reg[i]>>r))&1) { - if((regs[i].was32>>r)&1) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - 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]); @@ -11111,29 +9853,7 @@ int new_recompile_block(int addr) if((needed_reg[i]>>5)&1) printf("ebp "); if((needed_reg[i]>>6)&1) printf("esi "); if((needed_reg[i]>>7)&1) printf("edi "); - printf("r:"); - for(r=0;r<=CCREG;r++) { - //if(((requires_32bit[i]>>r)&(~unneeded_reg[i]>>r))&1) { - if((requires_32bit[i]>>r)&1) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } printf("\n"); - /*printf("pr:"); - for(r=0;r<=CCREG;r++) { - //if(((requires_32bit[i]>>r)&(~unneeded_reg[i]>>r))&1) { - if((pr32[i]>>r)&1) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - if(pr32[i]!=requires_32bit[i]) printf(" OOPS"); - printf("\n");*/ #if defined(__i386__) || defined(__x86_64__) printf("entry: eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",regs[i].regmap_entry[0],regs[i].regmap_entry[1],regs[i].regmap_entry[2],regs[i].regmap_entry[3],regs[i].regmap_entry[5],regs[i].regmap_entry[6],regs[i].regmap_entry[7]); printf("dirty: "); @@ -11217,29 +9937,6 @@ 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) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - printf("\n"); -#endif - /*printf(" p32:"); - for(r=0;r<=CCREG;r++) { - if((p32[i]>>r)&1) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - if(p32[i]!=regs[i].is32) printf(" NO MATCH\n"); - else printf("\n");*/ if(itype[i]==RJUMP||itype[i]==UJUMP||itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP) { #if defined(__i386__) || defined(__x86_64__) printf("branch(%d): eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d dirty: ",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]); @@ -11266,18 +9963,6 @@ 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) { - if(r==CCREG) printf(" CC"); - else if(r==HIREG) printf(" HI"); - else if(r==LOREG) printf(" LO"); - else printf(" r%d",r); - } - } - printf("\n"); -#endif } } #endif // DISASM @@ -11288,14 +9973,13 @@ int new_recompile_block(int addr) cop1_usable=0; uint64_t is32_pre=0; u_int dirty_pre=0; - u_int beginning=(u_int)out; + void *beginning=start_block(); if((u_int)addr&1) { ds=1; pagespan_ds(); } u_int instr_addr0_override=0; -#ifdef PCSX if (start == 0x80030000) { // nasty hack for fastbios thing // override block entry to this code @@ -11309,7 +9993,6 @@ int new_recompile_block(int addr) emit_cmp(0,1); emit_jne((int)new_dyna_leave); } -#endif for(i=0;i>16)!=0x1000)) { - wb_sx(regmap_pre[i],regs[i].regmap_entry,regs[i].wasdirty,is32_pre,regs[i].was32, - unneeded_reg[i],unneeded_reg_upper[i]); wb_valid(regmap_pre[i],regs[i].regmap_entry,dirty_pre,regs[i].wasdirty,is32_pre, unneeded_reg[i],unneeded_reg_upper[i]); } @@ -11565,7 +10246,7 @@ int new_recompile_block(int addr) // replace it with the new address. // Don't add new entries. We'll insert the // ones that actually get used in check_addr(). - int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; + u_int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF]; if(ht_bin[0]==vaddr) { ht_bin[1]=entry_point; } @@ -11582,45 +10263,32 @@ int new_recompile_block(int addr) // Align code if(((u_int)out)&7) emit_addnop(13); #endif - assert((u_int)out-beginning(u_int)BASE_ADDR+(1<>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); - j=(((u_int)i<<12)+(memory_map[i]<<2)-(u_int)rdram+(u_int)0x80000000)>>12; - invalid_code[j]=0; - memory_map[j]|=0x40000000; - //printf("write protect physical page: %x (virtual %x)\n",j<<12,start); - } -#endif } inv_code_start=inv_code_end=~0; -#ifdef PCSX + // for PCSX we need to mark all mirrors too if(get_page(start)<(RAM_SIZE>>12)) for(i=start>>12;i<=(start+slen*4)>>12;i++) invalid_code[((u_int)0x00000000>>12)|(i&0x1ff)]= invalid_code[((u_int)0x80000000>>12)|(i&0x1ff)]= invalid_code[((u_int)0xa0000000>>12)|(i&0x1ff)]=0; -#endif - + /* Pass 10 - Free memory by expiring oldest blocks */ - + int end=((((int)out-(int)BASE_ADDR)>>(TARGET_SIZE_2-16))+16384)&65535; while(expirep!=end) { @@ -11644,7 +10312,7 @@ int new_recompile_block(int addr) case 2: // Clear hash table for(i=0;i<32;i++) { - int *ht_bin=hash_table[((expirep&2047)<<5)+i]; + u_int *ht_bin=hash_table[((expirep&2047)<<5)+i]; if((ht_bin[3]>>shift)==(base>>shift) || ((ht_bin[3]-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(base>>shift)) { inv_debug("EXP: Remove hash %x -> %x\n",ht_bin[2],ht_bin[3]); @@ -11662,7 +10330,7 @@ int new_recompile_block(int addr) case 3: // Clear jump_out #ifdef __arm__ - if((expirep&2047)==0) + if((expirep&2047)==0) do_clear_cache(); #endif ll_remove_matching_addrs(jump_out+(expirep&2047),base,shift); diff --git a/libpcsxcore/new_dynarec/new_dynarec_config.h b/libpcsxcore/new_dynarec/new_dynarec_config.h new file mode 100644 index 00000000..fbd08ac2 --- /dev/null +++ b/libpcsxcore/new_dynarec/new_dynarec_config.h @@ -0,0 +1,12 @@ + + +#define CORTEX_A8_BRANCH_PREDICTION_HACK 1 +#define USE_MINI_HT 1 +//#define REG_PREFETCH 1 + +#if defined(__MACH__) || defined(VITA) +#define NO_WRITE_EXEC 1 +#endif +#ifdef VITA +#define BASE_ADDR_DYNAMIC 1 +#endif