X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=libpcsxcore%2Fgte.c;h=991a4452c5b20c74c434d35d2887b339bbf44a91;hb=HEAD;hp=26d3d3470cf9fc3bc32cfb484f54ddb4f8d79883;hpb=b3254af90fafc88bd1b40566de65a0471771c8d5;p=pcsx_rearmed.git diff --git a/libpcsxcore/gte.c b/libpcsxcore/gte.c index 26d3d347..991a4452 100644 --- a/libpcsxcore/gte.c +++ b/libpcsxcore/gte.c @@ -149,7 +149,13 @@ #define gteBFC (((s32 *)regs->CP2C.r)[23]) #define gteOFX (((s32 *)regs->CP2C.r)[24]) #define gteOFY (((s32 *)regs->CP2C.r)[25]) -#define gteH (regs->CP2C.p[26].sw.l) +// senquack - gteH register is u16, not s16, and used in GTE that way. +// HOWEVER when read back by CPU using CFC2, it will be incorrectly +// sign-extended by bug in original hardware, according to Nocash docs +// GTE section 'Screen Offset and Distance'. The emulator does this +// sign extension when it is loaded to GTE by CTC2. +//#define gteH (regs->CP2C.p[26].sw.l) +#define gteH (regs->CP2C.p[26].w.l) #define gteDQA (regs->CP2C.p[27].sw.l) #define gteDQB (((s32 *)regs->CP2C.r)[28]) #define gteZSF3 (regs->CP2C.p[29].sw.l) @@ -170,7 +176,7 @@ #ifndef FLAGLESS -static inline s32 BOUNDS_(psxCP2Regs *regs, s64 n_value, s64 n_max, int n_maxflag, s64 n_min, int n_minflag) { +static inline s64 BOUNDS_(psxCP2Regs *regs, s64 n_value, s64 n_max, int n_maxflag, s64 n_min, int n_minflag) { if (n_value > n_max) { gteFLAG |= n_maxflag; } else if (n_value < n_min) { @@ -243,12 +249,59 @@ static inline u32 limE_(psxCP2Regs *regs, u32 result) { #define limG2(a) LIM((a), 0x3ff, -0x400, (1 << 31) | (1 << 13)) #define limH(a) LIM((a), 0x1000, 0x0000, (1 << 12)) +#ifndef __arm__ +#define A1U A1 +#define A2U A2 +#define A3U A3 +#else +/* these are unlikely to be hit and usually waste cycles, don't want them on ARM */ +#define A1U(x) (x) +#define A2U(x) (x) +#define A3U(x) (x) +#endif + + +//senquack - n param should be unsigned (will be 'gteH' reg which is u16) +#ifdef GTE_USE_NATIVE_DIVIDE +INLINE u32 DIVIDE(u16 n, u16 d) { + if (n < d * 2) { + return ((u32)n << 16) / d; + } + return 0xffffffff; +} +#else #include "gte_divider.h" +#endif // GTE_USE_NATIVE_DIVIDE #ifndef FLAGLESS -static inline u32 MFC2(int reg) { - psxCP2Regs *regs = (psxCP2Regs *)&psxRegs.CP2D; +const unsigned char gte_cycletab[64] = { + /* 1 2 3 4 5 6 7 8 9 a b c d e f */ + 0, 15, 0, 0, 0, 0, 8, 0, 0, 0, 0, 0, 6, 0, 0, 0, + 8, 8, 8, 19, 13, 0, 44, 0, 0, 0, 0, 17, 11, 0, 14, 0, + 30, 0, 0, 0, 0, 0, 0, 0, 5, 8, 17, 0, 0, 5, 6, 0, + 23, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5, 5, 39, +}; + +// warning: called by the dynarec +int gteCheckStallRaw(u32 op_cycles, psxRegisters *regs) { + u32 left = regs->gteBusyCycle - regs->cycle; + int stall = 0; + + if (left <= 44) { + //printf("c %2u stall %2u %u\n", op_cycles, left, regs->cycle); + regs->cycle = regs->gteBusyCycle; + stall = left; + } + regs->gteBusyCycle = regs->cycle + op_cycles; + return stall; +} + +void gteCheckStall(u32 op) { + gteCheckStallRaw(gte_cycletab[op], &psxRegs); +} + +u32 MFC2(struct psxCP2Regs *regs, int reg) { switch (reg) { case 1: case 3: @@ -257,7 +310,7 @@ static inline u32 MFC2(int reg) { case 9: case 10: case 11: - psxRegs.CP2D.r[reg] = (s32)psxRegs.CP2D.p[reg].sw.l; + regs->CP2D.r[reg] = (s32)regs->CP2D.p[reg].sw.l; break; case 7: @@ -265,25 +318,24 @@ static inline u32 MFC2(int reg) { case 17: case 18: case 19: - psxRegs.CP2D.r[reg] = (u32)psxRegs.CP2D.p[reg].w.l; + regs->CP2D.r[reg] = (u32)regs->CP2D.p[reg].w.l; break; case 15: - psxRegs.CP2D.r[reg] = gteSXY2; + regs->CP2D.r[reg] = gteSXY2; break; case 28: case 29: - psxRegs.CP2D.r[reg] = LIM(gteIR1 >> 7, 0x1f, 0, 0) | + regs->CP2D.r[reg] = LIM(gteIR1 >> 7, 0x1f, 0, 0) | (LIM(gteIR2 >> 7, 0x1f, 0, 0) << 5) | (LIM(gteIR3 >> 7, 0x1f, 0, 0) << 10); break; } - return psxRegs.CP2D.r[reg]; + return regs->CP2D.r[reg]; } -static inline void MTC2(u32 value, int reg) { - psxCP2Regs *regs = (psxCP2Regs *)&psxRegs.CP2D; +void MTC2(struct psxCP2Regs *regs, u32 value, int reg) { switch (reg) { case 15: gteSXY0 = gteSXY1; @@ -294,9 +346,10 @@ static inline void MTC2(u32 value, int reg) { case 28: gteIRGB = value; - gteIR1 = (value & 0x1f) << 7; - gteIR2 = (value & 0x3e0) << 2; - gteIR3 = (value & 0x7c00) >> 3; + // not gteIR1 etc. just to be consistent with dynarec + regs->CP2D.n.ir1 = (value & 0x1f) << 7; + regs->CP2D.n.ir2 = (value & 0x3e0) << 2; + regs->CP2D.n.ir3 = (value & 0x7c00) >> 3; break; case 30: @@ -324,11 +377,11 @@ static inline void MTC2(u32 value, int reg) { return; default: - psxRegs.CP2D.r[reg] = value; + regs->CP2D.r[reg] = value; } } -static inline void CTC2(u32 value, int reg) { +void CTC2(struct psxCP2Regs *regs, u32 value, int reg) { switch (reg) { case 4: case 12: @@ -346,35 +399,7 @@ static inline void CTC2(u32 value, int reg) { break; } - psxRegs.CP2C.r[reg] = value; -} - -void gteMFC2() { - if (!_Rt_) return; - psxRegs.GPR.r[_Rt_] = MFC2(_Rd_); -} - -void gteCFC2() { - if (!_Rt_) return; - psxRegs.GPR.r[_Rt_] = psxRegs.CP2C.r[_Rd_]; -} - -void gteMTC2() { - MTC2(psxRegs.GPR.r[_Rt_], _Rd_); -} - -void gteCTC2() { - CTC2(psxRegs.GPR.r[_Rt_], _Rd_); -} - -#define _oB_ (psxRegs.GPR.r[_Rs_] + _Imm_) - -void gteLWC2() { - MTC2(psxMemRead32(_oB_), _Rt_); -} - -void gteSWC2() { - psxMemWrite32(_oB_, MFC2(_Rt_)); + regs->CP2C.r[reg] = value; } #endif // FLAGLESS @@ -393,6 +418,7 @@ static u32 DIVIDE_(s16 n, u16 d) { void gteRTPS(psxCP2Regs *regs) { int quotient; + s64 tmp; #ifdef GTE_LOG GTE_LOG("GTE RTPS\n"); @@ -415,14 +441,16 @@ void gteRTPS(psxCP2Regs *regs) { gteSX2 = limG1(F((s64)gteOFX + ((s64)gteIR1 * quotient)) >> 16); gteSY2 = limG2(F((s64)gteOFY + ((s64)gteIR2 * quotient)) >> 16); - gteMAC0 = F((s64)gteDQB + ((s64)gteDQA * quotient)); - gteIR0 = limH(gteMAC0 >> 12); + tmp = (s64)gteDQB + ((s64)gteDQA * quotient); + gteMAC0 = F(tmp); + gteIR0 = limH(tmp >> 12); } void gteRTPT(psxCP2Regs *regs) { int quotient; int v; s32 vx, vy, vz; + s64 tmp; #ifdef GTE_LOG GTE_LOG("GTE RTPT\n"); @@ -445,8 +473,10 @@ void gteRTPT(psxCP2Regs *regs) { fSX(v) = limG1(F((s64)gteOFX + ((s64)gteIR1 * quotient)) >> 16); fSY(v) = limG2(F((s64)gteOFY + ((s64)gteIR2 * quotient)) >> 16); } - gteMAC0 = F((s64)gteDQB + ((s64)gteDQA * quotient)); - gteIR0 = limH(gteMAC0 >> 12); + + tmp = (s64)gteDQB + ((s64)gteDQA * quotient); + gteMAC0 = F(tmp); + gteIR0 = limH(tmp >> 12); } void gteMVMVA(psxCP2Regs *regs) { @@ -613,9 +643,9 @@ void gteNCDS(psxCP2Regs *regs) { gteIR1 = limB1(gteMAC1, 1); gteIR2 = limB2(gteMAC2, 1); gteIR3 = limB3(gteMAC3, 1); - gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(gteRFC - ((gteR * gteIR1) >> 8), 0))) >> 12; - gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(gteGFC - ((gteG * gteIR2) >> 8), 0))) >> 12; - gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(gteBFC - ((gteB * gteIR3) >> 8), 0))) >> 12; + gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(A1U((s64)gteRFC - ((gteR * gteIR1) >> 8)), 0))) >> 12; + gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(A2U((s64)gteGFC - ((gteG * gteIR2) >> 8)), 0))) >> 12; + gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(A3U((s64)gteBFC - ((gteB * gteIR3) >> 8)), 0))) >> 12; gteIR1 = limB1(gteMAC1, 1); gteIR2 = limB2(gteMAC2, 1); gteIR3 = limB3(gteMAC3, 1); @@ -653,9 +683,9 @@ void gteNCDT(psxCP2Regs *regs) { gteIR1 = limB1(gteMAC1, 1); gteIR2 = limB2(gteMAC2, 1); gteIR3 = limB3(gteMAC3, 1); - gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(gteRFC - ((gteR * gteIR1) >> 8), 0))) >> 12; - gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(gteGFC - ((gteG * gteIR2) >> 8), 0))) >> 12; - gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(gteBFC - ((gteB * gteIR3) >> 8), 0))) >> 12; + gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(A1U((s64)gteRFC - ((gteR * gteIR1) >> 8)), 0))) >> 12; + gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(A2U((s64)gteGFC - ((gteG * gteIR2) >> 8)), 0))) >> 12; + gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(A3U((s64)gteBFC - ((gteB * gteIR3) >> 8)), 0))) >> 12; gteRGB0 = gteRGB1; gteRGB1 = gteRGB2; @@ -698,9 +728,9 @@ void gteDCPL(psxCP2Regs *regs) { #endif gteFLAG = 0; - gteMAC1 = RIR1 + ((gteIR0 * limB1(gteRFC - RIR1, 0)) >> 12); - gteMAC2 = GIR2 + ((gteIR0 * limB1(gteGFC - GIR2, 0)) >> 12); - gteMAC3 = BIR3 + ((gteIR0 * limB1(gteBFC - BIR3, 0)) >> 12); + gteMAC1 = RIR1 + ((gteIR0 * limB1(A1U((s64)gteRFC - RIR1), 0)) >> 12); + gteMAC2 = GIR2 + ((gteIR0 * limB1(A2U((s64)gteGFC - GIR2), 0)) >> 12); + gteMAC3 = BIR3 + ((gteIR0 * limB1(A3U((s64)gteBFC - BIR3), 0)) >> 12); gteIR1 = limB1(gteMAC1, lm); gteIR2 = limB2(gteMAC2, lm); @@ -768,9 +798,9 @@ void gteDPCS(psxCP2Regs *regs) { #endif gteFLAG = 0; - gteMAC1 = ((gteR << 16) + (gteIR0 * limB1(A1((s64)gteRFC - (gteR << 4)) << (12 - shift), 0))) >> 12; - gteMAC2 = ((gteG << 16) + (gteIR0 * limB2(A2((s64)gteGFC - (gteG << 4)) << (12 - shift), 0))) >> 12; - gteMAC3 = ((gteB << 16) + (gteIR0 * limB3(A3((s64)gteBFC - (gteB << 4)) << (12 - shift), 0))) >> 12; + gteMAC1 = ((gteR << 16) + (gteIR0 * limB1(A1U(((s64)gteRFC - (gteR << 4)) << (12 - shift)), 0))) >> 12; + gteMAC2 = ((gteG << 16) + (gteIR0 * limB2(A2U(((s64)gteGFC - (gteG << 4)) << (12 - shift)), 0))) >> 12; + gteMAC3 = ((gteB << 16) + (gteIR0 * limB3(A3U(((s64)gteBFC - (gteB << 4)) << (12 - shift)), 0))) >> 12; gteIR1 = limB1(gteMAC1, 0); gteIR2 = limB2(gteMAC2, 0); @@ -792,9 +822,9 @@ void gteDPCT(psxCP2Regs *regs) { gteFLAG = 0; for (v = 0; v < 3; v++) { - gteMAC1 = ((gteR0 << 16) + (gteIR0 * limB1(gteRFC - (gteR0 << 4), 0))) >> 12; - gteMAC2 = ((gteG0 << 16) + (gteIR0 * limB1(gteGFC - (gteG0 << 4), 0))) >> 12; - gteMAC3 = ((gteB0 << 16) + (gteIR0 * limB1(gteBFC - (gteB0 << 4), 0))) >> 12; + gteMAC1 = ((gteR0 << 16) + (gteIR0 * limB1(A1U((s64)gteRFC - (gteR0 << 4)), 0))) >> 12; + gteMAC2 = ((gteG0 << 16) + (gteIR0 * limB1(A2U((s64)gteGFC - (gteG0 << 4)), 0))) >> 12; + gteMAC3 = ((gteB0 << 16) + (gteIR0 * limB1(A3U((s64)gteBFC - (gteB0 << 4)), 0))) >> 12; gteRGB0 = gteRGB1; gteRGB1 = gteRGB2; @@ -905,9 +935,9 @@ void gteINTPL(psxCP2Regs *regs) { #endif gteFLAG = 0; - gteMAC1 = ((gteIR1 << 12) + (gteIR0 * limB1(gteRFC - gteIR1, 0))) >> shift; - gteMAC2 = ((gteIR2 << 12) + (gteIR0 * limB2(gteGFC - gteIR2, 0))) >> shift; - gteMAC3 = ((gteIR3 << 12) + (gteIR0 * limB3(gteBFC - gteIR3, 0))) >> shift; + gteMAC1 = ((gteIR1 << 12) + (gteIR0 * limB1(A1U((s64)gteRFC - gteIR1), 0))) >> shift; + gteMAC2 = ((gteIR2 << 12) + (gteIR0 * limB2(A2U((s64)gteGFC - gteIR2), 0))) >> shift; + gteMAC3 = ((gteIR3 << 12) + (gteIR0 * limB3(A3U((s64)gteBFC - gteIR3), 0))) >> shift; gteIR1 = limB1(gteMAC1, lm); gteIR2 = limB2(gteMAC2, lm); gteIR3 = limB3(gteMAC3, lm); @@ -931,9 +961,9 @@ void gteCDP(psxCP2Regs *regs) { gteIR1 = limB1(gteMAC1, 1); gteIR2 = limB2(gteMAC2, 1); gteIR3 = limB3(gteMAC3, 1); - gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(gteRFC - ((gteR * gteIR1) >> 8), 0))) >> 12; - gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(gteGFC - ((gteG * gteIR2) >> 8), 0))) >> 12; - gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(gteBFC - ((gteB * gteIR3) >> 8), 0))) >> 12; + gteMAC1 = (((gteR << 4) * gteIR1) + (gteIR0 * limB1(A1U((s64)gteRFC - ((gteR * gteIR1) >> 8)), 0))) >> 12; + gteMAC2 = (((gteG << 4) * gteIR2) + (gteIR0 * limB2(A2U((s64)gteGFC - ((gteG * gteIR2) >> 8)), 0))) >> 12; + gteMAC3 = (((gteB << 4) * gteIR3) + (gteIR0 * limB3(A3U((s64)gteBFC - ((gteB * gteIR3) >> 8)), 0))) >> 12; gteIR1 = limB1(gteMAC1, 1); gteIR2 = limB2(gteMAC2, 1); gteIR3 = limB3(gteMAC3, 1); @@ -945,3 +975,131 @@ void gteCDP(psxCP2Regs *regs) { gteG2 = limC2(gteMAC2 >> 4); gteB2 = limC3(gteMAC3 >> 4); } + +/* decomposed/parametrized versions for the recompiler */ + +#ifndef FLAGLESS + +void gteSQR_part_noshift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = gteIR1 * gteIR1; + gteMAC2 = gteIR2 * gteIR2; + gteMAC3 = gteIR3 * gteIR3; +} + +void gteSQR_part_shift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = (gteIR1 * gteIR1) >> 12; + gteMAC2 = (gteIR2 * gteIR2) >> 12; + gteMAC3 = (gteIR3 * gteIR3) >> 12; +} + +void gteOP_part_noshift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = (gteR22 * gteIR3) - (gteR33 * gteIR2); + gteMAC2 = (gteR33 * gteIR1) - (gteR11 * gteIR3); + gteMAC3 = (gteR11 * gteIR2) - (gteR22 * gteIR1); +} + +void gteOP_part_shift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = ((gteR22 * gteIR3) - (gteR33 * gteIR2)) >> 12; + gteMAC2 = ((gteR33 * gteIR1) - (gteR11 * gteIR3)) >> 12; + gteMAC3 = ((gteR11 * gteIR2) - (gteR22 * gteIR1)) >> 12; +} + +void gteDCPL_part(psxCP2Regs *regs) { + s32 RIR1 = ((s32)gteR * gteIR1) >> 8; + s32 GIR2 = ((s32)gteG * gteIR2) >> 8; + s32 BIR3 = ((s32)gteB * gteIR3) >> 8; + + gteFLAG = 0; + + gteMAC1 = RIR1 + ((gteIR0 * limB1(A1U((s64)gteRFC - RIR1), 0)) >> 12); + gteMAC2 = GIR2 + ((gteIR0 * limB1(A2U((s64)gteGFC - GIR2), 0)) >> 12); + gteMAC3 = BIR3 + ((gteIR0 * limB1(A3U((s64)gteBFC - BIR3), 0)) >> 12); +} + +void gteGPF_part_noshift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = gteIR0 * gteIR1; + gteMAC2 = gteIR0 * gteIR2; + gteMAC3 = gteIR0 * gteIR3; +} + +void gteGPF_part_shift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = (gteIR0 * gteIR1) >> 12; + gteMAC2 = (gteIR0 * gteIR2) >> 12; + gteMAC3 = (gteIR0 * gteIR3) >> 12; +} + +#endif // !FLAGLESS + +void gteGPL_part_noshift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = A1((s64)gteMAC1 + (gteIR0 * gteIR1)); + gteMAC2 = A2((s64)gteMAC2 + (gteIR0 * gteIR2)); + gteMAC3 = A3((s64)gteMAC3 + (gteIR0 * gteIR3)); +} + +void gteGPL_part_shift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = A1((s64)gteMAC1 + ((gteIR0 * gteIR1) >> 12)); + gteMAC2 = A2((s64)gteMAC2 + ((gteIR0 * gteIR2) >> 12)); + gteMAC3 = A3((s64)gteMAC3 + ((gteIR0 * gteIR3) >> 12)); +} + +void gteDPCS_part_noshift(psxCP2Regs *regs) { + int shift = 0; + + gteFLAG = 0; + + gteMAC1 = ((gteR << 16) + (gteIR0 * limB1(A1U((s64)gteRFC - (gteR << 4)) << (12 - shift), 0))) >> 12; + gteMAC2 = ((gteG << 16) + (gteIR0 * limB2(A2U((s64)gteGFC - (gteG << 4)) << (12 - shift), 0))) >> 12; + gteMAC3 = ((gteB << 16) + (gteIR0 * limB3(A3U((s64)gteBFC - (gteB << 4)) << (12 - shift), 0))) >> 12; +} + +void gteDPCS_part_shift(psxCP2Regs *regs) { + int shift = 12; + + gteFLAG = 0; + + gteMAC1 = ((gteR << 16) + (gteIR0 * limB1(A1U((s64)gteRFC - (gteR << 4)) << (12 - shift), 0))) >> 12; + gteMAC2 = ((gteG << 16) + (gteIR0 * limB2(A2U((s64)gteGFC - (gteG << 4)) << (12 - shift), 0))) >> 12; + gteMAC3 = ((gteB << 16) + (gteIR0 * limB3(A3U((s64)gteBFC - (gteB << 4)) << (12 - shift), 0))) >> 12; +} + +void gteINTPL_part_noshift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = ((gteIR1 << 12) + (gteIR0 * limB1(A1U((s64)gteRFC - gteIR1), 0))); + gteMAC2 = ((gteIR2 << 12) + (gteIR0 * limB2(A2U((s64)gteGFC - gteIR2), 0))); + gteMAC3 = ((gteIR3 << 12) + (gteIR0 * limB3(A3U((s64)gteBFC - gteIR3), 0))); +} + +void gteINTPL_part_shift(psxCP2Regs *regs) { + gteFLAG = 0; + + gteMAC1 = ((gteIR1 << 12) + (gteIR0 * limB1(A1U((s64)gteRFC - gteIR1), 0))) >> 12; + gteMAC2 = ((gteIR2 << 12) + (gteIR0 * limB2(A2U((s64)gteGFC - gteIR2), 0))) >> 12; + gteMAC3 = ((gteIR3 << 12) + (gteIR0 * limB3(A3U((s64)gteBFC - gteIR3), 0))) >> 12; +} + +void gteMACtoRGB(psxCP2Regs *regs) { + gteRGB0 = gteRGB1; + gteRGB1 = gteRGB2; + gteCODE2 = gteCODE; + gteR2 = limC1(gteMAC1 >> 4); + gteG2 = limC2(gteMAC2 >> 4); + gteB2 = limC3(gteMAC3 >> 4); +} +