X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=Pico%2Fcarthw%2Fsvp%2Fcompiler.c;h=2db22b1ff0d668e27a0f21145af2f1e7d906d40c;hb=43e6eaad0b73b37907be3342e0fd4cf65919e9f6;hp=0cdd6ad44739d72fdc641b4d9aa83880f3c2dca4;hpb=e122fae6f47c4ca4fce5a178d6d2055b3cef6423;p=picodrive.git diff --git a/Pico/carthw/svp/compiler.c b/Pico/carthw/svp/compiler.c index 0cdd6ad..2db22b1 100644 --- a/Pico/carthw/svp/compiler.c +++ b/Pico/carthw/svp/compiler.c @@ -1,497 +1,52 @@ -// 187 blocks, 12072 bytes -// 14 IRAM blocks +// SSP1601 to ARM recompiler + +// (c) Copyright 2008, Grazvydas "notaz" Ignotas +// Free for non-commercial use. #include "../../PicoInt.h" #include "compiler.h" -static unsigned int *block_table[0x5090/2]; -static unsigned int *block_table_iram[15][0x800/2]; -static unsigned int *tcache_ptr = NULL; +#define u32 unsigned int + +static u32 *tcache_ptr = NULL; static int nblocks = 0; -static int iram_context = 0; +static int n_in_ops = 0; -#ifndef ARM -#define DUMP_BLOCK 0x2018 -unsigned int tcache[512*1024]; -void regfile_load(void){} -void regfile_store(void){} -#endif +extern ssp1601_t *ssp; -#define EMBED_INTERPRETER -#define ssp1601_reset ssp1601_reset_local -#define ssp1601_run ssp1601_run_local +#define rPC ssp->gr[SSP_PC].h +#define rPMC ssp->gr[SSP_PMC] -#define GET_PC() rPC -#define GET_PPC_OFFS() (GET_PC()*2 - 2) -#define SET_PC(d) { rPC = d; } /* must return to dispatcher after this */ -//#define GET_PC() (PC - (unsigned short *)svp->iram_rom) -//#define GET_PPC_OFFS() ((unsigned int)PC - (unsigned int)svp->iram_rom - 2) -//#define SET_PC(d) PC = (unsigned short *)svp->iram_rom + d +#define SSP_FLAG_Z (1<<0xd) +#define SSP_FLAG_N (1<<0xf) + +#ifndef ARM +#define DUMP_BLOCK 0x0c9a +u32 *ssp_block_table[0x5090/2]; +u32 *ssp_block_table_iram[15][0x800/2]; +u32 tcache[SSP_TCACHE_SIZE/4]; +void ssp_drc_next(void){} +void ssp_drc_next_patch(void){} +void ssp_drc_end(void){} +#endif -#include "ssp16.c" #include "gen_arm.c" // ----------------------------------------------------- -// ld d, s -static void op00(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - PC = ((unsigned short *)(void *)&op) + 1; /* FIXME: needed for interpreter */ - if (op == 0) return; // nop - if (op == ((SSP_A<<4)|SSP_P)) { // A <- P - // not sure. MAME claims that only hi word is transfered. - read_P(); // update P - rA32 = rP.v; - } - else - { - tmpv = REG_READ(op & 0x0f); - REG_WRITE((op & 0xf0) >> 4, tmpv); - } -} - -// ld d, (ri) -static void op01(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); REG_WRITE((op & 0xf0) >> 4, tmpv); -} - -// ld (ri), s -static void op02(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = REG_READ((op & 0xf0) >> 4); ptr1_write(op, tmpv); -} - -// ldi d, imm -static void op04(unsigned int op, unsigned int imm) -{ - REG_WRITE((op & 0xf0) >> 4, imm); -} - -// ld d, ((ri)) -static void op05(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); REG_WRITE((op & 0xf0) >> 4, tmpv); -} - -// ldi (ri), imm -static void op06(unsigned int op, unsigned int imm) -{ - ptr1_write(op, imm); -} - -// ld adr, a -static void op07(unsigned int op, unsigned int imm) -{ - ssp->RAM[op & 0x1ff] = rA; -} - -// ld d, ri -static void op09(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[(op&3)|((op>>6)&4)]; REG_WRITE((op & 0xf0) >> 4, tmpv); -} - -// ld ri, s -static void op0a(unsigned int op, unsigned int imm) -{ - rIJ[(op&3)|((op>>6)&4)] = REG_READ((op & 0xf0) >> 4); -} - -// ldi ri, simm (also op0d op0e op0f) -static void op0c(unsigned int op, unsigned int imm) -{ - rIJ[(op>>8)&7] = op; -} - -// call cond, addr -static void op24(unsigned int op, unsigned int imm) -{ - int cond = 0; - do { - COND_CHECK - if (cond) { int new_PC = imm; write_STACK(GET_PC()); SET_PC(new_PC); } - } - while (0); -} - -// ld d, (a) -static void op25(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ((unsigned short *)svp->iram_rom)[rA]; REG_WRITE((op & 0xf0) >> 4, tmpv); -} - -// bra cond, addr -static void op26(unsigned int op, unsigned int imm) -{ - do - { - int cond = 0; - COND_CHECK - if (cond) SET_PC(imm); - } - while (0); -} - -// mod cond, op -static void op48(unsigned int op, unsigned int imm) -{ - do - { - int cond = 0; - COND_CHECK - if (cond) { - switch (op & 7) { - case 2: rA32 = (signed int)rA32 >> 1; break; // shr (arithmetic) - case 3: rA32 <<= 1; break; // shl - case 6: rA32 = -(signed int)rA32; break; // neg - case 7: if ((int)rA32 < 0) rA32 = -(signed int)rA32; break; // abs - default: elprintf(EL_SVP|EL_ANOMALY, "ssp FIXME: unhandled mod %i @ %04x", - op&7, GET_PPC_OFFS()); - } - UPD_ACC_ZN // ? - } - } - while(0); -} - -// mpys? -static void op1b(unsigned int op, unsigned int imm) -{ - read_P(); // update P - rA32 -= rP.v; // maybe only upper word? - UPD_ACC_ZN // there checking flags after this - rX = ptr1_read_(op&3, 0, (op<<1)&0x18); // ri (maybe rj?) - rY = ptr1_read_((op>>4)&3, 4, (op>>3)&0x18); // rj -} - -// mpya (rj), (ri), b -static void op4b(unsigned int op, unsigned int imm) -{ - read_P(); // update P - rA32 += rP.v; // confirmed to be 32bit - UPD_ACC_ZN // ? - rX = ptr1_read_(op&3, 0, (op<<1)&0x18); // ri (maybe rj?) - rY = ptr1_read_((op>>4)&3, 4, (op>>3)&0x18); // rj -} - -// mld (rj), (ri), b -static void op5b(unsigned int op, unsigned int imm) -{ - rA32 = 0; - rST &= 0x0fff; // ? - rX = ptr1_read_(op&3, 0, (op<<1)&0x18); // ri (maybe rj?) - rY = ptr1_read_((op>>4)&3, 4, (op>>3)&0x18); // rj -} - -// OP a, s -static void op10(unsigned int op, unsigned int imm) -{ - do - { - unsigned int tmpv; - OP_CHECK32(OP_SUBA32); tmpv = REG_READ(op & 0x0f); OP_SUBA(tmpv); - } - while(0); -} - -static void op30(unsigned int op, unsigned int imm) -{ - do - { - unsigned int tmpv; - OP_CHECK32(OP_CMPA32); tmpv = REG_READ(op & 0x0f); OP_CMPA(tmpv); - } - while(0); -} - -static void op40(unsigned int op, unsigned int imm) -{ - do - { - unsigned int tmpv; - OP_CHECK32(OP_ADDA32); tmpv = REG_READ(op & 0x0f); OP_ADDA(tmpv); - } - while(0); -} - -static void op50(unsigned int op, unsigned int imm) -{ - do - { - unsigned int tmpv; - OP_CHECK32(OP_ANDA32); tmpv = REG_READ(op & 0x0f); OP_ANDA(tmpv); - } - while(0); -} - -static void op60(unsigned int op, unsigned int imm) -{ - do - { - unsigned int tmpv; - OP_CHECK32(OP_ORA32 ); tmpv = REG_READ(op & 0x0f); OP_ORA (tmpv); - } - while(0); -} - -static void op70(unsigned int op, unsigned int imm) +static int get_inc(int mode) { - do - { - unsigned int tmpv; - OP_CHECK32(OP_EORA32); tmpv = REG_READ(op & 0x0f); OP_EORA(tmpv); + int inc = (mode >> 11) & 7; + if (inc != 0) { + if (inc != 7) inc--; + inc = 1 << inc; // 0 1 2 4 8 16 32 128 + if (mode & 0x8000) inc = -inc; // decrement mode } - while(0); -} - -// OP a, (ri) -static void op11(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_SUBA(tmpv); -} - -static void op31(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_CMPA(tmpv); -} - -static void op41(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_ADDA(tmpv); -} - -static void op51(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_ANDA(tmpv); -} - -static void op61(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_ORA (tmpv); -} - -static void op71(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr1_read(op); OP_EORA(tmpv); -} - -// OP a, adr -static void op03(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_LDA (tmpv); -} - -static void op13(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_SUBA(tmpv); -} - -static void op33(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_CMPA(tmpv); -} - -static void op43(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_ADDA(tmpv); -} - -static void op53(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_ANDA(tmpv); -} - -static void op63(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_ORA (tmpv); -} - -static void op73(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ssp->RAM[op & 0x1ff]; OP_EORA(tmpv); -} - -// OP a, imm -static void op14(unsigned int op, unsigned int imm) -{ - OP_SUBA(imm); -} - -static void op34(unsigned int op, unsigned int imm) -{ - OP_CMPA(imm); -} - -static void op44(unsigned int op, unsigned int imm) -{ - OP_ADDA(imm); -} - -static void op54(unsigned int op, unsigned int imm) -{ - OP_ANDA(imm); -} - -static void op64(unsigned int op, unsigned int imm) -{ - OP_ORA (imm); -} - -static void op74(unsigned int op, unsigned int imm) -{ - OP_EORA(imm); -} - -// OP a, ((ri)) -static void op15(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_SUBA(tmpv); -} - -static void op35(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_CMPA(tmpv); -} - -static void op45(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_ADDA(tmpv); -} - -static void op55(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_ANDA(tmpv); -} - -static void op65(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_ORA (tmpv); + return inc; } -static void op75(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = ptr2_read(op); OP_EORA(tmpv); -} - -// OP a, ri -static void op19(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_SUBA(tmpv); -} - -static void op39(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_CMPA(tmpv); -} - -static void op49(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_ADDA(tmpv); -} - -static void op59(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_ANDA(tmpv); -} - -static void op69(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_ORA (tmpv); -} - -static void op79(unsigned int op, unsigned int imm) -{ - unsigned int tmpv; - tmpv = rIJ[IJind]; OP_EORA(tmpv); -} - -// OP simm -static void op1c(unsigned int op, unsigned int imm) -{ - OP_SUBA(op & 0xff); -} - -static void op3c(unsigned int op, unsigned int imm) -{ - OP_CMPA(op & 0xff); -} - -static void op4c(unsigned int op, unsigned int imm) -{ - OP_ADDA(op & 0xff); -} - -static void op5c(unsigned int op, unsigned int imm) -{ - OP_ANDA(op & 0xff); -} - -static void op6c(unsigned int op, unsigned int imm) -{ - OP_ORA (op & 0xff); -} - -static void op7c(unsigned int op, unsigned int imm) -{ - OP_EORA(op & 0xff); -} - -typedef void (in_func)(unsigned int op, unsigned int imm); - -static in_func *in_funcs[0x80] = -{ - op00, op01, op02, op03, op04, op05, op06, op07, - NULL, op09, op0a, NULL, op0c, op0c, op0c, op0c, - op10, op11, NULL, op13, op14, op15, NULL, NULL, - NULL, op19, NULL, op1b, op1c, NULL, NULL, NULL, - NULL, NULL, NULL, NULL, op24, op25, op26, NULL, - NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, - op30, op31, NULL, op33, op34, op35, NULL, NULL, - NULL, op39, NULL, NULL, op3c, NULL, NULL, NULL, - op40, op41, NULL, op43, op44, op45, NULL, NULL, - op48, op49, NULL, op4b, op4c, NULL, NULL, NULL, - op50, op51, NULL, op53, op54, op55, NULL, NULL, - NULL, op59, NULL, op5b, op5c, NULL, NULL, NULL, - op60, op61, NULL, op63, op64, op65, NULL, NULL, - NULL, op69, NULL, NULL, op6c, NULL, NULL, NULL, - op70, op71, NULL, op73, op74, op75, NULL, NULL, - NULL, op79, NULL, NULL, op7c, NULL, NULL, NULL, -}; - - -static u32 ssp_pm_read(int reg) +u32 ssp_pm_read(int reg) { u32 d = 0, mode; @@ -499,11 +54,9 @@ static u32 ssp_pm_read(int reg) { ssp->pmac_read[reg] = rPMC.v; ssp->emu_status &= ~SSP_PMC_SET; - //elprintf("set PM%i %08x", ssp->pmac_read[reg]); return 0; } - //elprintf("rd PM%i %08x", ssp->pmac_read[reg]); // just in case ssp->emu_status &= ~SSP_PMC_HAVE_ADDR; @@ -527,7 +80,15 @@ static u32 ssp_pm_read(int reg) return d; } -static void ssp_pm_write(u32 d, int reg) +#define overwrite_write(dst, d) \ +{ \ + if (d & 0xf000) { dst &= ~0xf000; dst |= d & 0xf000; } \ + if (d & 0x0f00) { dst &= ~0x0f00; dst |= d & 0x0f00; } \ + if (d & 0x00f0) { dst &= ~0x00f0; dst |= d & 0x00f0; } \ + if (d & 0x000f) { dst &= ~0x000f; dst |= d & 0x000f; } \ +} + +void ssp_pm_write(u32 d, int reg) { unsigned short *dram; int mode, addr; @@ -558,7 +119,7 @@ static void ssp_pm_write(u32 d, int reg) if (mode & 0x0400) { overwrite_write(dram[addr], d); } else dram[addr] = d; - ssp->pmac_write[reg] += (addr&1) ? 31 : 1; + ssp->pmac_write[reg] += (addr&1) ? 0x1f : 1; } else if ((mode & 0x47ff) == 0x001c) // IRAM { @@ -574,6 +135,7 @@ static void ssp_pm_write(u32 d, int reg) // ----------------------------------------------------- +// 14 IRAM blocks static unsigned char iram_context_map[] = { 0, 0, 0, 0, 1, 0, 0, 0, // 04 @@ -586,30 +148,22 @@ static unsigned char iram_context_map[] = 13,14, 0, 0, 0, 0, 0, 0 // 38 39 }; -static int get_iram_context(void) +int ssp_get_iram_context(void) { unsigned char *ir = (unsigned char *)svp->iram_rom; int val1, val = ir[0x083^1] + ir[0x4FA^1] + ir[0x5F7^1] + ir[0x47B^1]; val1 = iram_context_map[(val>>1)&0x3f]; if (val1 == 0) { - printf("val: %02x PC=%04x\n", (val>>1)&0x3f, rPC); + elprintf(EL_ANOMALY, "svp: iram ctx val: %02x PC=%04x\n", (val>>1)&0x3f, rPC); //debug_dump2file(name, svp->iram_rom, 0x800); - exit(1); + //exit(1); } -// elprintf(EL_ANOMALY, "iram_context: %02i", val1); return val1; } // ----------------------------------------------------- -/* -enum { - SSP_GR0, SSP_X, SSP_Y, SSP_A, - SSP_ST, SSP_STACK, SSP_PC, SSP_P, - SSP_PM0, SSP_PM1, SSP_PM2, SSP_XST, - SSP_PM4, SSP_gr13, SSP_PMC, SSP_AL -}; -*/ + /* regs with known values */ static struct { @@ -680,13 +234,13 @@ static void hostreg_sspreg_changed(int sspreg) #define PROGRAM(x) ((unsigned short *)svp->iram_rom)[x] #define PROGRAM_P(x) ((unsigned short *)svp->iram_rom + (x)) -static void tr_unhandled(void) +void tr_unhandled(void) { - FILE *f = fopen("tcache.bin", "wb"); - fwrite(tcache, 1, (tcache_ptr - tcache)*4, f); - fclose(f); - printf("unhandled @ %04x\n", known_regs.gr[SSP_PC].h<<1); - exit(1); + //FILE *f = fopen("tcache.bin", "wb"); + //fwrite(tcache, 1, (tcache_ptr - tcache)*4, f); + //fclose(f); + elprintf(EL_ANOMALY, "unhandled @ %04x\n", known_regs.gr[SSP_PC].h<<1); + //exit(1); } /* update P, if needed. Trashes r0 */ @@ -726,6 +280,14 @@ static void tr_flush_dirty_prs(void) { int i, ror = 0, reg; int dirty = dirty_regb >> 8; + if ((dirty&7) == 7) { + emit_mov_const(A_COND_AL, 8, known_regs.r[0]|(known_regs.r[1]<<8)|(known_regs.r[2]<<16)); + dirty &= ~7; + } + if ((dirty&0x70) == 0x70) { + emit_mov_const(A_COND_AL, 9, known_regs.r[4]|(known_regs.r[5]<<8)|(known_regs.r[6]<<16)); + dirty &= ~0x70; + } /* r0-r7 */ for (i = 0; dirty && i < 8; i++, dirty >>= 1) { @@ -793,7 +355,7 @@ static void tr_mov16_cond(int cond, int r, int val) hostreg_r[r] = -1; } -/* trashes r0 */ +/* trashes r1 */ static void tr_flush_dirty_pmcrs(void) { u32 i, val = (u32)-1; @@ -805,7 +367,7 @@ static void tr_flush_dirty_pmcrs(void) EOP_STR_IMM(1,7,0x400+SSP_PMC*4); if (known_regs.emu_status & (SSP_PMC_SET|SSP_PMC_HAVE_ADDR)) { - printf("!! SSP_PMC_SET|SSP_PMC_HAVE_ADDR set on flush\n"); + elprintf(EL_ANOMALY, "!! SSP_PMC_SET|SSP_PMC_HAVE_ADDR set on flush\n"); tr_unhandled(); } } @@ -1007,6 +569,69 @@ static void tr_rX_read2(int op) hostreg_r[0] = hostreg_r[2] = -1; } +// check if AL is going to be used later in block +static int tr_predict_al_need(void) +{ + int tmpv, tmpv2, op, pc = known_regs.gr[SSP_PC].h; + + while (1) + { + op = PROGRAM(pc); + switch (op >> 9) + { + // ld d, s + case 0x00: + tmpv2 = (op >> 4) & 0xf; // dst + tmpv = op & 0xf; // src + if ((tmpv2 == SSP_A && tmpv == SSP_P) || tmpv2 == SSP_AL) // ld A, P; ld AL, * + return 0; + break; + + // ld (ri), s + case 0x02: + // ld ri, s + case 0x0a: + // OP a, s + case 0x10: case 0x30: case 0x40: case 0x60: case 0x70: + tmpv = op & 0xf; // src + if (tmpv == SSP_AL) // OP *, AL + return 1; + break; + + case 0x04: + case 0x06: + case 0x14: + case 0x34: + case 0x44: + case 0x64: + case 0x74: pc++; break; + + // call cond, addr + case 0x24: + // bra cond, addr + case 0x26: + // mod cond, op + case 0x48: + // mpys? + case 0x1b: + // mpya (rj), (ri), b + case 0x4b: return 1; + + // mld (rj), (ri), b + case 0x5b: return 0; // cleared anyway + + // and A, * + case 0x50: + tmpv = op & 0xf; // src + if (tmpv == SSP_AL) return 1; + case 0x51: case 0x53: case 0x54: case 0x55: case 0x59: case 0x5c: + return 0; + } + pc++; + } +} + + /* get ARM cond which would mean that SSP cond is satisfied. No trash. */ static int tr_cond_check(int op) { @@ -1022,7 +647,7 @@ static int tr_cond_check(int op) EOP_TST_IMM(6, 0, 8); return f ? A_COND_NE : A_COND_EQ; default: - printf("unimplemented cond?\n"); + elprintf(EL_ANOMALY, "unimplemented cond?\n"); tr_unhandled(); return 0; } @@ -1031,12 +656,12 @@ static int tr_cond_check(int op) static int tr_neg_cond(int cond) { switch (cond) { - case A_COND_AL: printf("neg for AL?\n"); exit(1); + case A_COND_AL: elprintf(EL_ANOMALY, "neg for AL?\n"); exit(1); case A_COND_EQ: return A_COND_NE; case A_COND_NE: return A_COND_EQ; case A_COND_MI: return A_COND_PL; case A_COND_PL: return A_COND_MI; - default: printf("bad cond for neg\n"); exit(1); + default: elprintf(EL_ANOMALY, "bad cond for neg\n"); exit(1); } return 0; } @@ -1058,8 +683,6 @@ static int tr_aop_ssp2arm(int op) // ----------------------------------------------------- -// SSP_GR0, SSP_X, SSP_Y, SSP_A, -// SSP_ST, SSP_STACK, SSP_PC, SSP_P, //@ r4: XXYY //@ r5: A //@ r6: STACK and emu flags @@ -1082,7 +705,6 @@ static void tr_X_to_r0(int op) static void tr_Y_to_r0(int op) { - // TODO.. if (hostreg_r[0] != (SSP_Y<<16)) { EOP_MOV_REG_SIMPLE(0, 4); // mov r0, r4 hostreg_r[0] = SSP_Y<<16; @@ -1182,7 +804,7 @@ static void tr_PMX_to_r0(int reg) tr_flush_dirty_ST(); EOP_LDR_IMM(1,7,0x484); // ldr r1, [r7, #0x484] // emu_status EOP_TST_REG_SIMPLE(0,0); - EOP_C_DOP_IMM(A_COND_EQ,A_OP_ADD,0,11,11,22/2,1); // addeq r11, r11, #1024 + EOP_C_DOP_IMM(A_COND_EQ,A_OP_SUB,0,11,11,22/2,1); // subeq r11, r11, #1024 EOP_C_DOP_IMM(A_COND_EQ,A_OP_ORR,0, 1, 1,24/2,flag>>8); // orreq r1, r1, #SSP_WAIT_30FE08 EOP_STR_IMM(1,7,0x484); // str r1, [r7, #0x484] // emu_status } @@ -1209,7 +831,7 @@ static void tr_PMX_to_r0(int reg) tr_flush_dirty_ST(); //tr_flush_dirty_pmcrs(); tr_mov16(0, reg); - emit_call(ssp_pm_read); + emit_call(A_COND_AL, ssp_pm_read); hostreg_clear(); } @@ -1329,9 +951,13 @@ static void tr_r0_to_Y(int const_val) static void tr_r0_to_A(int const_val) { - EOP_MOV_REG_LSL(5, 5, 16); // mov r5, r5, lsl #16 - EOP_MOV_REG_LSR(5, 5, 16); // mov r5, r5, lsr #16 @ AL - EOP_ORR_REG_LSL(5, 5, 0, 16); // orr r5, r5, r0, lsl #16 + if (tr_predict_al_need()) { + EOP_MOV_REG_LSL(5, 5, 16); // mov r5, r5, lsl #16 + EOP_MOV_REG_LSR(5, 5, 16); // mov r5, r5, lsr #16 @ AL + EOP_ORR_REG_LSL(5, 5, 0, 16); // orr r5, r5, r0, lsl #16 + } + else + EOP_MOV_REG_LSL(5, 0, 16); TR_WRITE_R0_TO_REG(SSP_A); } @@ -1359,9 +985,12 @@ static void tr_r0_to_STACK(int const_val) static void tr_r0_to_PC(int const_val) { +/* + * do nothing - dispatcher will take care of this EOP_MOV_REG_LSL(1, 0, 16); // mov r1, r0, lsl #16 EOP_STR_IMM(1,7,0x400+6*4); // str r1, [r7, #(0x400+6*8)] hostreg_r[1] = -1; +*/ } static void tr_r0_to_AL(int const_val) @@ -1379,7 +1008,6 @@ static void tr_r0_to_AL(int const_val) static void tr_r0_to_PMX(int reg) { -#if 1 if ((known_regb & KRREG_PMC) && (known_regs.emu_status & SSP_PMC_SET)) { known_regs.pmac_write[reg] = known_regs.pmc.v; @@ -1388,8 +1016,7 @@ static void tr_r0_to_PMX(int reg) dirty_regb |= 1 << (25+reg); return; } -#endif -#if 1 + if ((known_regb & KRREG_PMC) && (known_regb & (1 << (25+reg)))) { int mode, addr; @@ -1440,21 +1067,12 @@ static void tr_r0_to_PMX(int reg) dirty_regb &= ~KRREG_PMC; known_regb &= ~(1 << (25+reg)); dirty_regb &= ~(1 << (25+reg)); -#else -tr_flush_dirty_pmcrs(); -hostreg_clear(); -known_regb &= ~KRREG_PMC; -dirty_regb &= ~KRREG_PMC; -known_regb &= ~(1 << (25+reg)); -dirty_regb &= ~(1 << (25+reg)); -#endif - // call the C code to handle this tr_flush_dirty_ST(); //tr_flush_dirty_pmcrs(); tr_mov16(1, reg); - emit_call(ssp_pm_write); + emit_call(A_COND_AL, ssp_pm_write); hostreg_clear(); } @@ -1564,6 +1182,7 @@ static int tr_detect_set_pm(unsigned int op, int *pc, int imm) known_regb |= KRREG_PMC; dirty_regb |= KRREG_PMC; known_regs.emu_status |= SSP_PMC_SET; + n_in_ops++; // check for possible reg programming tmpv = PROGRAM(*pc); @@ -1578,6 +1197,7 @@ static int tr_detect_set_pm(unsigned int op, int *pc, int imm) dirty_regb |= is_write ? (1 << (reg+25)) : (1 << (reg+20)); known_regs.emu_status &= ~SSP_PMC_SET; (*pc)++; + n_in_ops++; return 5; } @@ -1605,6 +1225,7 @@ static int tr_detect_pm0_block(unsigned int op, int *pc, int imm) known_regb |= 1 << SSP_ST; dirty_regb &= ~KRREG_ST; (*pc) += 3*2; + n_in_ops += 3; return 4*2; } @@ -1621,12 +1242,13 @@ static int tr_detect_rotate(unsigned int op, int *pc, int imm) EOP_ORR_REG_LSR(0, 0, 0, 16); tr_bank_write(0); (*pc) += 2; + n_in_ops += 2; return 3; } // ----------------------------------------------------- -static int translate_op(unsigned int op, int *pc, int imm) +static int translate_op(unsigned int op, int *pc, int imm, int *end_cond, int *jump_pc) { u32 tmpv, tmpv2, tmpv3; int ret = 0; @@ -1639,7 +1261,6 @@ static int translate_op(unsigned int op, int *pc, int imm) if (op == 0) { ret++; break; } // nop tmpv = op & 0xf; // src tmpv2 = (op >> 4) & 0xf; // dst - //if (tmpv2 >= 8) return -1; // TODO if (tmpv2 == SSP_A && tmpv == SSP_P) { // ld A, P tr_flush_dirty_P(); EOP_MOV_REG_SIMPLE(5, 10); @@ -1649,23 +1270,34 @@ static int translate_op(unsigned int op, int *pc, int imm) } tr_read_funcs[tmpv](op); tr_write_funcs[tmpv2]((known_regb & (1 << tmpv)) ? known_regs.gr[tmpv].h : -1); - if (tmpv2 == SSP_PC) ret |= 0x10000; + if (tmpv2 == SSP_PC) { + ret |= 0x10000; + *end_cond = -A_COND_AL; + } ret++; break; // ld d, (ri) case 0x01: { - // tmpv = ptr1_read(op); REG_WRITE((op & 0xf0) >> 4, tmpv); break; int r = (op&3) | ((op>>6)&4); int mod = (op>>2)&3; tmpv = (op >> 4) & 0xf; // dst ret = tr_detect_rotate(op, pc, imm); if (ret > 0) break; - if (tmpv >= 8) return -1; // TODO if (tmpv != 0) - tr_rX_read(r, mod); - else tr_ptrr_mod(r, mod, 1, 1); + tr_rX_read(r, mod); + else { + int cnt = 1; + while (PROGRAM(*pc) == op) { + (*pc)++; cnt++; ret++; + n_in_ops++; + } + tr_ptrr_mod(r, mod, 1, cnt); // skip + } tr_write_funcs[tmpv](-1); - if (tmpv == SSP_PC) ret |= 0x10000; + if (tmpv == SSP_PC) { + ret |= 0x10000; + *end_cond = -A_COND_AL; + } ret++; break; } @@ -1687,24 +1319,25 @@ static int translate_op(unsigned int op, int *pc, int imm) tmpv = (op & 0xf0) >> 4; // dst ret = tr_detect_pm0_block(op, pc, imm); if (ret > 0) break; - if (tmpv < 8) - { - tr_mov16(0, imm); - tr_write_funcs[tmpv](imm); - ret += 2; break; - } ret = tr_detect_set_pm(op, pc, imm); if (ret > 0) break; - if (tmpv == SSP_PC) ret |= 0x10000; - return -1; /* TODO.. */ + tr_mov16(0, imm); + tr_write_funcs[tmpv](imm); + if (tmpv == SSP_PC) { + ret |= 0x10000; + *jump_pc = imm; + } + ret += 2; break; // ld d, ((ri)) case 0x05: tmpv2 = (op >> 4) & 0xf; // dst - if (tmpv2 >= 8) return -1; // TODO tr_rX_read2(op); tr_write_funcs[tmpv2](-1); - if (tmpv2 == SSP_PC) ret |= 0x10000; + if (tmpv2 == SSP_PC) { + ret |= 0x10000; + *end_cond = -A_COND_AL; + } ret += 3; break; // ldi (ri), imm @@ -1724,7 +1357,6 @@ static int translate_op(unsigned int op, int *pc, int imm) int r; r = (op&3) | ((op>>6)&4); // src tmpv2 = (op >> 4) & 0xf; // dst - if (tmpv2 >= 8) tr_unhandled(); if ((r&3) == 3) tr_unhandled(); if (known_regb & (1 << (r+8))) { @@ -1766,7 +1398,7 @@ static int translate_op(unsigned int op, int *pc, int imm) } // ldi ri, simm - case 0x0c ... 0x0f: + case 0x0c: case 0x0d: case 0x0e: case 0x0f: tmpv = (op>>8)&7; known_regs.r[tmpv] = op; known_regb |= 1 << (tmpv + 8); @@ -1790,37 +1422,40 @@ static int translate_op(unsigned int op, int *pc, int imm) tcache_ptr = real_ptr; } tr_mov16_cond(tmpv, 0, imm); - if (tmpv != A_COND_AL) { + if (tmpv != A_COND_AL) tr_mov16_cond(tr_neg_cond(tmpv), 0, *pc); - } tr_r0_to_PC(tmpv == A_COND_AL ? imm : -1); ret |= 0x10000; + *end_cond = tmpv; + *jump_pc = imm; ret += 2; break; } // ld d, (a) case 0x25: tmpv2 = (op >> 4) & 0xf; // dst - if (tmpv2 >= 8) return -1; // TODO - tr_A_to_r0(op); EOP_LDR_IMM(1,7,0x48c); // ptr_iram_rom EOP_ADD_REG_LSL(0,1,0,1); // add r0, r1, r0, lsl #1 EOP_LDRH_SIMPLE(0,0); // ldrh r0, [r0] hostreg_r[0] = hostreg_r[1] = -1; tr_write_funcs[tmpv2](-1); - if (tmpv2 == SSP_PC) ret |= 0x10000; + if (tmpv2 == SSP_PC) { + ret |= 0x10000; + *end_cond = -A_COND_AL; + } ret += 3; break; // bra cond, addr case 0x26: tmpv = tr_cond_check(op); tr_mov16_cond(tmpv, 0, imm); - if (tmpv != A_COND_AL) { + if (tmpv != A_COND_AL) tr_mov16_cond(tr_neg_cond(tmpv), 0, *pc); - } tr_r0_to_PC(tmpv == A_COND_AL ? imm : -1); ret |= 0x10000; + *end_cond = tmpv; + *jump_pc = imm; ret += 2; break; // mod cond, op @@ -1829,6 +1464,7 @@ static int translate_op(unsigned int op, int *pc, int imm) tmpv = 1; // count while (PROGRAM(*pc) == op && (op & 7) != 6) { (*pc)++; tmpv++; + n_in_ops++; } if ((op&0xf0) != 0) // !always tr_make_dirty_ST(); @@ -2016,19 +1652,61 @@ static int translate_op(unsigned int op, int *pc, int imm) ret++; break; } + n_in_ops++; + return ret; } -static void *translate_block(int pc) +static void emit_block_prologue(void) +{ + // check if there are enough cycles.. + // note: r0 must contain PC of current block + EOP_CMP_IMM(11,0,0); // cmp r11, #0 + emit_call(A_COND_LE, ssp_drc_end); +} + +/* cond: + * >0: direct (un)conditional jump + * <0: indirect jump + */ +static void emit_block_epilogue(int cycles, int cond, int pc, int end_pc) +{ + if (cycles > 0xff) { elprintf(EL_ANOMALY, "large cycle count: %i\n", cycles); cycles = 0xff; } + EOP_SUB_IMM(11,11,0,cycles); // sub r11, r11, #cycles + + if (cond < 0 || (end_pc >= 0x400 && pc < 0x400)) { + // indirect jump, or rom -> iram jump, must use dispatcher + emit_jump(A_COND_AL, ssp_drc_next); + } + else if (cond == A_COND_AL) { + u32 *target = (pc < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][pc] : ssp_block_table[pc]; + if (target != NULL) + emit_jump(A_COND_AL, target); + else { + emit_jump(A_COND_AL, ssp_drc_next); + // cause the next block to be emitted over jump instrction + tcache_ptr--; + } + } + else { + u32 *target1 = (pc < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][pc] : ssp_block_table[pc]; + u32 *target2 = (end_pc < 0x400) ? ssp_block_table_iram[ssp->drc.iram_context][end_pc] : ssp_block_table[end_pc]; + if (target1 != NULL) + emit_jump(cond, target1); + else emit_call(cond, ssp_drc_next_patch); + if (target2 != NULL) + emit_jump(tr_neg_cond(cond), target2); // neg_cond, to be able to swap jumps if needed + else emit_call(tr_neg_cond(cond), ssp_drc_next_patch); + } +} + +void *ssp_translate_block(int pc) { unsigned int op, op1, imm, ccount = 0; unsigned int *block_start; - int ret, ret_prev = -1, tpc; - - // create .pool - //*tcache_ptr++ = (u32) in_funcs; // -1 func pool + int ret, end_cond = A_COND_AL, jump_pc = -1; - printf("translate %04x -> %04x\n", pc<<1, (tcache_ptr-tcache)<<2); + //printf("translate %04x -> %04x\n", pc<<1, (tcache_ptr-tcache)<<2); block_start = tcache_ptr; known_regb = 0; dirty_regb = KRREG_P; @@ -2039,72 +1717,45 @@ static void *translate_block(int pc) for (; ccount < 100;) { - //printf(" insn #%i\n", icount); op = PROGRAM(pc++); op1 = op >> 9; imm = (u32)-1; if ((op1 & 0xf) == 4 || (op1 & 0xf) == 6) imm = PROGRAM(pc++); // immediate - tpc = pc; - ret = translate_op(op, &pc, imm); + ret = translate_op(op, &pc, imm, &end_cond, &jump_pc); if (ret <= 0) { - tr_flush_dirty_prs(); - tr_flush_dirty_ST(); - tr_flush_dirty_pmcrs(); - known_regs.emu_status = 0; - - emit_mov_const(A_COND_AL, 0, op); - - // need immediate? - if (imm != (u32)-1) - emit_mov_const(A_COND_AL, 1, imm); - - // dump PC - emit_pc_dump(pc); - - if (ret_prev > 0) emit_call(regfile_store); - emit_call(in_funcs[op1]); - emit_call(regfile_load); - - if (in_funcs[op1] == NULL) { - printf("NULL func! op=%08x (%02x)\n", op, op1); - exit(1); - } - ccount++; - hostreg_clear(); - dirty_regb |= KRREG_P; - known_regb = 0; - } - else - { - ccount += ret & 0xffff; - if (ret & 0x10000) break; + elprintf(EL_ANOMALY, "NULL func! op=%08x (%02x)\n", op, op1); + //exit(1); } - ret_prev = ret; + ccount += ret & 0xffff; + if (ret & 0x10000) break; + } + + if (ccount >= 100) { + end_cond = A_COND_AL; + jump_pc = pc; + emit_mov_const(A_COND_AL, 0, pc); } tr_flush_dirty_prs(); tr_flush_dirty_ST(); tr_flush_dirty_pmcrs(); - emit_block_epilogue(ccount + 1); - *tcache_ptr++ = 0xffffffff; // end of block - //printf(" %i inst\n", icount); + emit_block_epilogue(ccount, end_cond, jump_pc, pc); - if (tcache_ptr - tcache > TCACHE_SIZE/4) { - printf("tcache overflow!\n"); + if (tcache_ptr - tcache > SSP_TCACHE_SIZE/4) { + elprintf(EL_ANOMALY, "tcache overflow!\n"); fflush(stdout); exit(1); } // stats nblocks++; - //if (pc >= 0x400) - printf("%i blocks, %i bytes\n", nblocks, (tcache_ptr - tcache)*4); - //printf("%p %p\n", tcache_ptr, emit_block_epilogue); + //printf("%i blocks, %i bytes, k=%.3f\n", nblocks, (tcache_ptr - tcache)*4, + // (double)(tcache_ptr - tcache) / (double)n_in_ops); #ifdef DUMP_BLOCK { @@ -2112,6 +1763,7 @@ static void *translate_block(int pc) fwrite(tcache, 1, (tcache_ptr - tcache)*4, f); fclose(f); } + printf("dumped tcache.bin\n"); exit(0); #endif @@ -2124,71 +1776,63 @@ static void *translate_block(int pc) // ----------------------------------------------------- +static void ssp1601_state_load(void) +{ + ssp->drc.iram_dirty = 1; + ssp->drc.iram_context = 0; +} + int ssp1601_dyn_startup(void) { - memset(tcache, 0, TCACHE_SIZE); - memset(block_table, 0, sizeof(block_table)); - memset(block_table_iram, 0, sizeof(block_table_iram)); + memset(tcache, 0, SSP_TCACHE_SIZE); + memset(ssp_block_table, 0, sizeof(ssp_block_table)); + memset(ssp_block_table_iram, 0, sizeof(ssp_block_table_iram)); tcache_ptr = tcache; - *tcache_ptr++ = 0xffffffff; + PicoLoadStateHook = ssp1601_state_load; + + n_in_ops = 0; #ifdef ARM // hle'd blocks - block_table[0x400] = (void *) ssp_hle_800; + ssp_block_table[0x800/2] = (void *) ssp_hle_800; + ssp_block_table[0x902/2] = (void *) ssp_hle_902; + ssp_block_table_iram[ 7][0x030/2] = (void *) ssp_hle_07_030; + ssp_block_table_iram[ 7][0x036/2] = (void *) ssp_hle_07_036; + ssp_block_table_iram[ 7][0x6d6/2] = (void *) ssp_hle_07_6d6; + ssp_block_table_iram[11][0x12c/2] = (void *) ssp_hle_11_12c; + ssp_block_table_iram[11][0x384/2] = (void *) ssp_hle_11_384; + ssp_block_table_iram[11][0x38a/2] = (void *) ssp_hle_11_38a; #endif -// TODO: rm -{ -static unsigned short dummy = 0; -PC = &dummy; -} return 0; } void ssp1601_dyn_reset(ssp1601_t *ssp) { - ssp1601_reset_local(ssp); - ssp->drc.ptr_rom = (unsigned int) Pico.rom; - ssp->drc.ptr_iram_rom = (unsigned int) svp->iram_rom; - ssp->drc.ptr_dram = (unsigned int) svp->dram; - ssp->drc.iram_dirty = 0; + ssp1601_reset(ssp); + ssp->drc.iram_dirty = 1; + ssp->drc.iram_context = 0; + // must do this here because ssp is not available @ startup() + ssp->drc.ptr_rom = (u32) Pico.rom; + ssp->drc.ptr_iram_rom = (u32) svp->iram_rom; + ssp->drc.ptr_dram = (u32) svp->dram; + ssp->drc.ptr_btable = (u32) ssp_block_table; + ssp->drc.ptr_btable_iram = (u32) ssp_block_table_iram; + + // prevent new versions of IRAM from appearing + memset(svp->iram_rom, 0, 0x800); } void ssp1601_dyn_run(int cycles) { if (ssp->emu_status & SSP_WAIT_MASK) return; - //{ printf("%i wait\n", Pico.m.frame_count); return; } - //printf("%i %04x\n", Pico.m.frame_count, rPC<<1); #ifdef DUMP_BLOCK - rPC = DUMP_BLOCK >> 1; + ssp_translate_block(DUMP_BLOCK >> 1); +#endif +#ifdef ARM + ssp_drc_entry(cycles); #endif - while (cycles > 0) - { - int (*trans_entry)(void); - if (rPC < 0x800/2) - { - if (ssp->drc.iram_dirty) { - iram_context = get_iram_context(); - ssp->drc.iram_dirty--; - } - if (block_table_iram[iram_context][rPC] == NULL) - block_table_iram[iram_context][rPC] = translate_block(rPC); - trans_entry = (void *) block_table_iram[iram_context][rPC]; - } - else - { - if (block_table[rPC] == NULL) - block_table[rPC] = translate_block(rPC); - trans_entry = (void *) block_table[rPC]; - } - - //printf("enter %04x\n", rPC<<1); - cycles -= trans_entry(); - //printf("leave %04x\n", rPC<<1); - } -// debug_dump2file("tcache.bin", tcache, (tcache_ptr - tcache) << 1); -// exit(1); }