Skip to content

Commit

Permalink
Implement vertex caching
Browse files Browse the repository at this point in the history 
- Try using Blade_Arma's vertex cache to find untracked vertices
- Fix GTE_LOG
- Add more logging spew
- Update debug mode to track cached vertices
  • Loading branch information
@iCatButler
iCatButler committed May 15, 2016
1 parent fed5120 commit 2a24b74
Show file tree
Hide file tree
Showing 7 changed files with 230 additions and 35 deletions.
50 changes: 28 additions & 22 deletions libpcsxcore/gte.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -258,6 +258,9 @@ void gteMFC2() {
void gteCFC2() {
if (!_Rt_) return;
psxRegs.GPR.r[_Rt_] = psxRegs.CP2C.p[_Rd_].d;
#ifdef GTE_LOG
GTE_LOG("GTE_CFC2 [%x]<-[%x] %x (%u %u)|", _Rt_, _Rd_, psxRegs.CP2C.p[_Rd_].d);
#endif
}

void gteMTC2() {
Expand All @@ -266,6 +269,9 @@ void gteMTC2() {
}

void gteCTC2() {
#ifdef GTE_LOG
GTE_LOG("GTE_CFC2 [%x]->[%x] %x (%u %u)|", _Rt_, _Rd_, psxRegs.GPR.r[_Rt_]);
#endif
CTC2(psxRegs.GPR.r[_Rt_], _Rd_);
}

Expand Down Expand Up @@ -478,7 +484,7 @@ int docop2(int op) {
case 0x00:
case 0x01:
#ifdef GTE_LOG
GTELOG("%08x RTPS", op);
GTE_LOG("%08x RTPS", op);
#endif

MAC1 = A1(/*int44*/(s64)((s64) TRX << 12) + (R11 * VX0) + (R12 * VY0) + (R13 * VZ0));
Expand Down Expand Up @@ -509,7 +515,7 @@ int docop2(int op) {

case 0x06:
#ifdef GTE_LOG
GTELOG("%08x NCLIP", op);
GTE_LOG("%08x NCLIP", op);
#endif
if (PGXP_NLCIP_valid())
MAC0 = F(PGXP_NCLIP());
Expand All @@ -519,7 +525,7 @@ int docop2(int op) {

case 0x0c:
#ifdef GTE_LOG
GTELOG("%08x OP", op);
GTE_LOG("%08x OP", op);
#endif

MAC1 = A1((s64) (R22 * IR3) - (R33 * IR2));
Expand All @@ -532,7 +538,7 @@ int docop2(int op) {

case 0x10:
#ifdef GTE_LOG
GTELOG("%08x DPCS", op);
GTE_LOG("%08x DPCS", op);
#endif

MAC1 = A1((R << 16) + (IR0 * Lm_B1(A1(((s64) RFC << 12) - (R << 16)), 0)));
Expand All @@ -551,7 +557,7 @@ int docop2(int op) {

case 0x11:
#ifdef GTE_LOG
GTELOG("%08x INTPL", op);
GTE_LOG("%08x INTPL", op);
#endif

MAC1 = A1((IR1 << 12) + (IR0 * Lm_B1(A1(((s64) RFC << 12) - (IR1 << 12)), 0)));
Expand All @@ -570,7 +576,7 @@ int docop2(int op) {

case 0x12:
#ifdef GTE_LOG
GTELOG("%08x MVMVA", op);
GTE_LOG("%08x MVMVA", op);
#endif

mx = GTE_MX(gteop);
Expand Down Expand Up @@ -601,7 +607,7 @@ int docop2(int op) {

case 0x13:
#ifdef GTE_LOG
GTELOG("%08x NCDS", op);
GTE_LOG("%08x NCDS", op);
#endif

MAC1 = A1((s64) (L11 * VX0) + (L12 * VY0) + (L13 * VZ0));
Expand Down Expand Up @@ -632,7 +638,7 @@ int docop2(int op) {

case 0x14:
#ifdef GTE_LOG
GTELOG("%08x CDP", op);
GTE_LOG("%08x CDP", op);
#endif

MAC1 = A1(/*int44*/(s64)((s64) RBK << 12) + (LR1 * IR1) + (LR2 * IR2) + (LR3 * IR3));
Expand All @@ -657,7 +663,7 @@ int docop2(int op) {

case 0x16:
#ifdef GTE_LOG
GTELOG("%08x NCDT", op);
GTE_LOG("%08x NCDT", op);
#endif

for(v = 0; v < 3; v++) {
Expand Down Expand Up @@ -690,7 +696,7 @@ int docop2(int op) {

case 0x1b:
#ifdef GTE_LOG
GTELOG("%08x NCCS", op);
GTE_LOG("%08x NCCS", op);
#endif

MAC1 = A1((s64) (L11 * VX0) + (L12 * VY0) + (L13 * VZ0));
Expand Down Expand Up @@ -721,7 +727,7 @@ int docop2(int op) {

case 0x1c:
#ifdef GTE_LOG
GTELOG("%08x CC", op);
GTE_LOG("%08x CC", op);
#endif

MAC1 = A1(/*int44*/(s64)(((s64) RBK) << 12) + (LR1 * IR1) + (LR2 * IR2) + (LR3 * IR3));
Expand All @@ -746,7 +752,7 @@ int docop2(int op) {

case 0x1e:
#ifdef GTE_LOG
GTELOG("%08x NCS", op);
GTE_LOG("%08x NCS", op);
#endif

MAC1 = A1((s64) (L11 * VX0) + (L12 * VY0) + (L13 * VZ0));
Expand All @@ -771,7 +777,7 @@ int docop2(int op) {

case 0x20:
#ifdef GTE_LOG
GTELOG("%08x NCT", op);
GTE_LOG("%08x NCT", op);
#endif

for(v = 0; v < 3; v++) {
Expand All @@ -798,7 +804,7 @@ int docop2(int op) {

case 0x28:
#ifdef GTE_LOG
GTELOG("%08x SQR", op);
GTE_LOG("%08x SQR", op);
#endif

MAC1 = A1(IR1 * IR1);
Expand All @@ -811,7 +817,7 @@ int docop2(int op) {

case 0x29:
#ifdef GTE_LOG
GTELOG("%08x DPCL", op);
GTE_LOG("%08x DPCL", op);
#endif

MAC1 = A1(((R << 4) * IR1) + (IR0 * Lm_B1(A1(((s64) RFC << 12) - ((R << 4) * IR1)), 0)));
Expand All @@ -830,7 +836,7 @@ int docop2(int op) {

case 0x2a:
#ifdef GTE_LOG
GTELOG("%08x DPCT", op);
GTE_LOG("%08x DPCT", op);
#endif

for(v = 0; v < 3; v++) {
Expand All @@ -851,7 +857,7 @@ int docop2(int op) {

case 0x2d:
#ifdef GTE_LOG
GTELOG("%08x AVSZ3", op);
GTE_LOG("%08x AVSZ3", op);
#endif

MAC0 = F((s64) (ZSF3 * SZ1) + (ZSF3 * SZ2) + (ZSF3 * SZ3));
Expand All @@ -860,7 +866,7 @@ int docop2(int op) {

case 0x2e:
#ifdef GTE_LOG
GTELOG("%08x AVSZ4", op);
GTE_LOG("%08x AVSZ4", op);
#endif

MAC0 = F((s64) (ZSF4 * SZ0) + (ZSF4 * SZ1) + (ZSF4 * SZ2) + (ZSF4 * SZ3));
Expand All @@ -869,7 +875,7 @@ int docop2(int op) {

case 0x30:
#ifdef GTE_LOG
GTELOG("%08x RTPT", op);
GTE_LOG("%08x RTPT", op);
#endif

for(v = 0; v < 3; v++) {
Expand Down Expand Up @@ -909,7 +915,7 @@ int docop2(int op) {

case 0x3d:
#ifdef GTE_LOG
GTELOG("%08x GPF", op);
GTE_LOG("%08x GPF", op);
#endif

MAC1 = A1(IR0 * IR1);
Expand All @@ -928,7 +934,7 @@ int docop2(int op) {

case 0x3e:
#ifdef GTE_LOG
GTELOG("%08x GPL", op);
GTE_LOG("%08x GPL", op);
#endif

MAC1 = A1(gte_shift(MAC1, -m_sf) + (IR0 * IR1));
Expand All @@ -947,7 +953,7 @@ int docop2(int op) {

case 0x3f:
#ifdef GTE_LOG
GTELOG("%08x NCCT", op);
GTE_LOG("%08x NCCT", op);
#endif

for(v = 0; v < 3; v++) {
Expand Down
41 changes: 40 additions & 1 deletion libpcsxcore/pgxp_gte.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -60,6 +60,8 @@ const u32 ScratchOffset = 2048 * 1024 / 4;
const u32 RegisterOffset = 2 * 2048 * 1024 / 4;
const u32 InvalidAddress = 3 * 2048 * 1024 / 4;

unsigned int pgxp_debug = 0;

//precise_value Scratch[2048 * 1024 / 4]; // mirror 2MB in 32-bit words
//precise_value Registers[2048 * 1024 / 4]; // mirror 2MB in 32-bit words

Expand Down Expand Up @@ -220,6 +222,7 @@ precise_value PGXP_validateXY(precise_value *high, u32 low)
// high->x = temp.x;
// high->y = temp.y;
// high->valid = 1;
// high->value = low;
//}

return *high;
Expand All @@ -243,17 +246,20 @@ precise_value PGXP_copyXY(u32 low)
ret.valid = 0;
temp.word = low;

ret.z= 1;
ret.x = temp.x;
ret.y = temp.y;
ret.count = 0;
ret.valid = 1;
ret.value = low;

return ret;
}

void PGXP_pushSXYZ2f(float _x, float _y, float _z, unsigned int _v)
{
static unsigned int uCount = 0;
low_value temp;
// push values down FIFO
SXY0 = SXY1;
SXY1 = SXY2;
Expand All @@ -265,6 +271,10 @@ void PGXP_pushSXYZ2f(float _x, float _y, float _z, unsigned int _v)
SXY2.valid = 1;
SXY2.count = uCount++;

// cache value in GPU plugin
temp.word = _v;
GPU_pgxpCacheVertex(temp.x, temp.y, &SXY2);

#ifdef GTE_LOG
GTE_LOG("PGPR_PUSH (%f, %f) %u %u|", SXY2.x, SXY2.y, SXY2.valid, SXY2.count);
#endif
Expand Down Expand Up @@ -483,7 +493,7 @@ void PGPR_L32(u32 addr, u32 code, u32 value)
case 35: //LW
CPU_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
break;
case 37: //LWR
case 38: //LWR
CPU_reg[reg] = PGXP_validateXY(ReadMem(addr), value);
break;
case 50: //LWC2 (GTE vertex reads)
Expand Down Expand Up @@ -603,25 +613,54 @@ void PGXP_psxMemWrite32Trace(u32 mem, u32 value, u32 code)
u16 PGXP_psxMemRead16Trace(u32 mem, u32 code)
{
u16 value = psxMemRead16(mem);
#ifdef GTE_LOG
u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register
GTE_LOG("PGPR_L16 %x %x[%x %x]|", mem, value, code, reg);
#endif
PGPR_InvalidLoad(mem, code, 116);
return value;
}

void PGXP_psxMemWrite16Trace(u32 mem, u16 value, u32 code)
{
PGPR_InvalidStore(mem, code, 216);
#ifdef GTE_LOG
u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register
GTE_LOG("PGPR_S16 %x %x[%x %x]|", mem, value, code, reg);
#endif
psxMemWrite16(mem, value);
}

u8 PGXP_psxMemRead8Trace(u32 mem, u32 code)
{
u8 value = psxMemRead8(mem);
#ifdef GTE_LOG
u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register
GTE_LOG("PGPR_L8 %x %x[%x %x]|", mem, value, code, reg);
#endif
PGPR_InvalidLoad(mem, code, 18);
return value;
}

void PGXP_psxMemWrite8Trace(u32 mem, u8 value, u32 code)
{
PGPR_InvalidStore(mem, code, 28);
#ifdef GTE_LOG
u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register
GTE_LOG("PGPR_S8 %x %x[%x %x]|", mem, value, code, reg);
#endif
psxMemWrite8(mem, value);
}

void PGXP_psxTrace(u32 code, u32 rtv)
{
#ifdef GTE_LOG
//u32 reg = ((code >> 16) & 0x1F); // The rt part of the instruction register
u32 op = ((code >> 26));
u32 func = ((code ) & 0x3F); // The funct part of the instruction register
u32 rd = ((code >> 11) & 0x1F); // The rd part of the instruction register
u32 rt = ((code >> 16) & 0x1F); // The rt part of the instruction register
u32 rs = ((code >> 21) & 0x1F); // The rs part of the instruction register
GTE_LOG("PGPR_Trace op:%u func:%u [rt:%x (%x) rs:%x rd:%x] %x|", op, func, rt, rtv, rs, rd, code);
#endif
}
1 change: 1 addition & 0 deletions libpcsxcore/pgxp_gte.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -65,4 +65,5 @@ void PGXP_psxMemWrite16Trace(u32 mem, u16 value, u32 code);
u8 PGXP_psxMemRead8Trace(u32 mem, u32 code);
void PGXP_psxMemWrite8Trace(u32 mem, u8 value, u32 code);

void PGXP_psxTrace(u32 code, u32 rtv);
#endif /* _PGXP_GTE_H_ */
3 changes: 3 additions & 0 deletions libpcsxcore/plugins.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ GPUcursor GPU_cursor;
GPUaddVertex GPU_addVertex;
GPUsetSpeed GPU_setSpeed;
GPUpgxpMemory GPU_pgxpMemory;
GPUpgxpCacheVertex GPU_pgxpCacheVertex;

CDRinit CDR_init;
CDRshutdown CDR_shutdown;
Expand Down Expand Up @@ -220,6 +221,7 @@ void CALLBACK GPU__cursor(int player, int x, int y) {}
void CALLBACK GPU__addVertex(short sx,short sy,s64 fx,s64 fy,s64 fz) {}
void CALLBACK GPU__setSpeed(float newSpeed) {}
void CALLBACK GPU__pgxpMemory(unsigned int addr, unsigned char* pVRAM) {}
void CALLBACK GPU__pgxpCacheVertex(short sx, short sy, const unsigned char* _pVertex) {}

#define LoadGpuSym1(dest, name) \
LoadSym(GPU_##dest, GPU##dest, name, TRUE);
Expand Down Expand Up @@ -268,6 +270,7 @@ static int LoadGPUplugin(const char *GPUdll) {
LoadGpuSym0(addVertex, "GPUaddVertex");
LoadGpuSym0(setSpeed, "GPUsetSpeed");
LoadGpuSym0(pgxpMemory, "GPUpgxpMemory");
LoadGpuSym0(pgxpCacheVertex, "GPUpgxpCacheVertex");
LoadGpuSym0(configure, "GPUconfigure");
LoadGpuSym0(test, "GPUtest");
LoadGpuSym0(about, "GPUabout");
Expand Down
2 changes: 2 additions & 0 deletions libpcsxcore/plugins.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -99,6 +99,7 @@ typedef void (CALLBACK* GPUcursor)(int, int, int);
typedef void (CALLBACK* GPUaddVertex)(short,short,s64,s64,s64);
typedef void (CALLBACK* GPUsetSpeed)(float); // 1.0 = natural speed
typedef void (CALLBACK* GPUpgxpMemory)(unsigned int, unsigned char*);
typedef void (CALLBACK* GPUpgxpCacheVertex)(short sx, short sy, const unsigned char* _pVertex);

// GPU function pointers
extern GPUupdateLace GPU_updateLace;
Expand Down Expand Up @@ -131,6 +132,7 @@ extern GPUcursor GPU_cursor;
extern GPUaddVertex GPU_addVertex;
extern GPUsetSpeed GPU_setSpeed;
extern GPUpgxpMemory GPU_pgxpMemory;
extern GPUpgxpCacheVertex GPU_pgxpCacheVertex;

// CD-ROM Functions
typedef long (CALLBACK* CDRinit)(void);
Expand Down
Loading

0 comments on commit 2a24b74

Please sign in to comment.