ShuriZma
/
project64
mirror of https://github.com/project64/project64.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

RSP: Clean up code for vector multiple ops

This commit is contained in:
zilmar 2023-09-07 11:54:36 +09:30
parent 4e9a692449
commit 002f2e17c3
4 changed files with 72 additions and 446 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

494
Source/Project64-rsp-core/cpu/RSPInterpreterOps.cpp
View File

@ -643,50 +643,26 @@ void RSP_COP2_VECTOR(void)
void RSP_Vector_VMULF(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
if (RSP_Vect[RSPOpC.vs].u16(el) != 0x8000 || RSP_Vect[RSPOpC.vt].u16(del) != 0x8000)
{
temp.W = ((int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)RSP_Vect[RSPOpC.vt].s16(del)) << 1;
temp.UW += 0x8000;
RSP_ACCUM[el].HW[2] = temp.HW[1];
RSP_ACCUM[el].HW[1] = temp.HW[0];
RSP_ACCUM[el].HW[3] = (RSP_ACCUM[el].HW[2] < 0) ? -1 : 0;
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
else
{
temp.W = 0x80000000;
RSP_ACCUM[el].UHW[3] = 0;
RSP_ACCUM[el].UHW[2] = 0x8000;
RSP_ACCUM[el].UHW[1] = 0x8000;
Result.s16(el) = 0x7FFF;
}
AccumulatorSet(el, ((int64_t)RSP_Vect[RSPOpC.vs].s16(el) * (int64_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e) * 2) + 0x8000);
Result.s16(el) = AccumulatorSaturate(el, true);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMULU(void)
{
uint8_t el, del;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
RSP_ACCUM[el].DW = (int64_t)(RSP_Vect[RSPOpC.vs].s16(el) * RSP_Vect[RSPOpC.vt].s16(del)) << 17;
RSP_ACCUM[el].DW += 0x80000000;
if (RSP_ACCUM[el].DW < 0)
AccumulatorSet(el, ((int64_t)RSP_Vect[RSPOpC.vs].s16(el) * (int64_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e) * 2) + 0x8000);
if (RSP_ACCUM[el].HW[3] < 0)
{
Result.s16(el) = 0;
}
else if ((int16_t)(RSP_ACCUM[el].UHW[3] ^ RSP_ACCUM[el].UHW[2]) < 0)
else if ((RSP_ACCUM[el].HW[3] ^ RSP_ACCUM[el].HW[2]) < 0)
{
Result.s16(el) = -1;
}
@ -721,17 +697,10 @@ void RSP_Vector_VRNDP(void)
void RSP_Vector_VMUDL(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)RSP_Vect[RSPOpC.vt].u16(del);
RSP_ACCUM[el].W[1] = 0;
RSP_ACCUM[el].HW[1] = temp.HW[1];
AccumulatorSet(el, (uint16_t)((uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)RSP_Vect[RSPOpC.vt].ue(el, RSPOpC.e) >> 16));
Result.s16(el) = RSP_ACCUM[el].HW[1];
}
RSP_Vect[RSPOpC.vd] = Result;
@ -739,25 +708,10 @@ void RSP_Vector_VMUDL(void)
void RSP_Vector_VMUDM(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)((int32_t)RSP_Vect[RSPOpC.vs].s16(el)) * (uint32_t)RSP_Vect[RSPOpC.vt].u16(del);
if (temp.W < 0)
{
RSP_ACCUM[el].HW[3] = -1;
}
else
{
RSP_ACCUM[el].HW[3] = 0;
}
RSP_ACCUM[el].HW[2] = temp.HW[1];
RSP_ACCUM[el].HW[1] = temp.HW[0];
AccumulatorSet(el, (int32_t)((int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (uint32_t)RSP_Vect[RSPOpC.vt].ue(el, RSPOpC.e)));
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
RSP_Vect[RSPOpC.vd] = Result;
@ -784,25 +738,10 @@ void RSP_Vector_VMULQ(void)
void RSP_Vector_VMUDN(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)((int32_t)RSP_Vect[RSPOpC.vt].s16(del));
if (temp.W < 0)
{
RSP_ACCUM[el].HW[3] = -1;
}
else
{
RSP_ACCUM[el].HW[3] = 0;
}
RSP_ACCUM[el].HW[2] = temp.HW[1];
RSP_ACCUM[el].HW[1] = temp.HW[0];
AccumulatorSet(el, (int32_t)((uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)((int32_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e))));
Result.s16(el) = RSP_ACCUM[el].HW[1];
}
RSP_Vect[RSPOpC.vd] = Result;
@ -810,154 +749,44 @@ void RSP_Vector_VMUDN(void)
void RSP_Vector_VMUDH(void)
{
uint8_t el, del;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
RSP_ACCUM[el].W[1] = (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)RSP_Vect[RSPOpC.vt].s16(del);
RSP_ACCUM[el].W[1] = (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e);
RSP_ACCUM[el].HW[1] = 0;
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0x8000;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0x8000;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0x7FFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.s16(el) = 0x7FFF;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
}
Result.u16(el) = AccumulatorSaturate(el, true);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMACF(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
/*temp.W = (long)RSP_Vect[RSPOpC.vs].s16(el) * (long)(DWORD)RSP_Vect[RSPOpC.vt].s16(del);
RSP_ACCUM[el].UHW[3] += (WORD)(temp.W >> 31);
temp.UW = temp.UW << 1;
temp2.UW = temp.UHW[0] + RSP_ACCUM[el].UHW[1];
RSP_ACCUM[el].HW[1] = temp2.HW[0];
temp2.UW = temp.UHW[1] + RSP_ACCUM[el].UHW[2] + temp2.UHW[1];
RSP_ACCUM[el].HW[2] = temp2.HW[0];
RSP_ACCUM[el].HW[3] += temp2.HW[1];*/
temp.W = (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)(uint32_t)RSP_Vect[RSPOpC.vt].s16(del);
RSP_ACCUM[el].DW += ((int64_t)temp.W) << 17;
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0x8000;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0x8000;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.s16(el) = 0x7FFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.s16(el) = 0x7FFF;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
}
AccumulatorSet(el, AccumulatorGet(el) + (((int64_t)RSP_Vect[RSPOpC.vs].s16(el) * (int64_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e)) << 1));
Result.u16(el) = AccumulatorSaturate(el, true);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMACU(void)
{
uint8_t el, del;
UWORD32 temp, temp2;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.W = (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)(uint32_t)RSP_Vect[RSPOpC.vt].s16(del);
RSP_ACCUM[el].UHW[3] = (RSP_ACCUM[el].UHW[3] + (uint16_t)(temp.W >> 31)) & 0xFFFF;
temp.UW = temp.UW << 1;
temp2.UW = temp.UHW[0] + RSP_ACCUM[el].UHW[1];
RSP_ACCUM[el].HW[1] = temp2.HW[0];
temp2.UW = temp.UHW[1] + RSP_ACCUM[el].UHW[2] + temp2.UHW[1];
RSP_ACCUM[el].HW[2] = temp2.HW[0];
RSP_ACCUM[el].HW[3] += temp2.HW[1];
AccumulatorSet(el, AccumulatorGet(el) + (((int64_t)RSP_Vect[RSPOpC.vs].s16(el) * (int64_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e)) << 1));
if (RSP_ACCUM[el].HW[3] < 0)
{
Result.s16(el) = 0;
}
else if (RSP_ACCUM[el].UHW[3] != 0 || RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
RSP_Vect[RSPOpC.vd] = Result;
@ -965,65 +794,21 @@ void RSP_Vector_VMACU(void)
void RSP_Vector_VMACQ(void)
{
uint8_t el, del;
UWORD32 temp;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
if (RSP_ACCUM[el].W[1] > 0x20)
int32_t Accum = (RSP_ACCUM[el].UHW[3] << 16) | RSP_ACCUM[el].UHW[2];
if (Accum < -0x20 && ((Accum & 0x20) == 0))
{
if ((RSP_ACCUM[el].W[1] & 0x20) == 0)
{
RSP_ACCUM[el].W[1] -= 0x20;
}
Accum += 0x20;
}
else if (RSP_ACCUM[el].W[1] < -0x20)
else if (Accum > 0x20 && (Accum & 0x20) == 0)
{
if ((RSP_ACCUM[el].W[1] & 0x20) == 0)
{
RSP_ACCUM[el].W[1] += 0x20;
}
}
temp.W = RSP_ACCUM[el].W[1] >> 1;
if (temp.HW[1] < 0)
{
if (temp.UHW[1] != 0xFFFF)
{
Result.u16(el) = 0x8000;
}
else
{
if (temp.HW[0] >= 0)
{
Result.u16(el) = 0x8000;
}
else
{
Result.u16(el) = (uint16_t)(temp.UW & 0xFFF0);
}
}
}
else
{
if (temp.UHW[1] != 0)
{
Result.u16(el) = 0x7FF0;
}
else
{
if (temp.HW[0] < 0)
{
Result.u16(el) = 0x7FF0;
}
else
{
Result.u16(el) = (uint16_t)(temp.UW & 0xFFF0);
}
}
Accum -= 0x20;
}
Result.u16(el) = clamp16(Accum >> 1) & 0xFFF0;
RSP_ACCUM[el].UHW[3] = (uint16_t)(Accum >> 16);
RSP_ACCUM[el].UHW[2] = (uint16_t)Accum;
}
RSP_Vect[RSPOpC.vd] = Result;
}
@ -1051,227 +836,46 @@ void RSP_Vector_VRNDN(void)
void RSP_Vector_VMADL(void)
{
uint8_t el, del;
UWORD32 temp, temp2;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)RSP_Vect[RSPOpC.vt].u16(del);
temp2.UW = temp.UHW[1] + RSP_ACCUM[el].UHW[1];
RSP_ACCUM[el].HW[1] = temp2.HW[0];
temp2.UW = RSP_ACCUM[el].UHW[2] + temp2.UHW[1];
RSP_ACCUM[el].HW[2] = temp2.HW[0];
RSP_ACCUM[el].HW[3] += temp2.HW[1];
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[1];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[1];
}
}
}
AccumulatorSet(el, AccumulatorGet(el) + (((uint32_t)(RSP_Vect[RSPOpC.vs].u16(el)) * (uint32_t)RSP_Vect[RSPOpC.vt].ue(el, RSPOpC.e)) >> 16));
Result.u16(el) = AccumulatorSaturate(el, false);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMADM(void)
{
uint8_t el, del;
UWORD32 temp, temp2;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)((int32_t)RSP_Vect[RSPOpC.vs].s16(el)) * (uint32_t)RSP_Vect[RSPOpC.vt].u16(del);
temp2.UW = temp.UHW[0] + RSP_ACCUM[el].UHW[1];
RSP_ACCUM[el].HW[1] = temp2.HW[0];
temp2.UW = temp.UHW[1] + RSP_ACCUM[el].UHW[2] + temp2.UHW[1];
RSP_ACCUM[el].HW[2] = temp2.HW[0];
RSP_ACCUM[el].HW[3] += temp2.HW[1];
if (temp.W < 0)
{
RSP_ACCUM[el].HW[3] -= 1;
}
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0x8000;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0x8000;
}
else
{
Result.s16(el) = RSP_ACCUM[el].HW[2];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0x7FFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0x7FFF;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[2];
}
}
}
//Result.s16(el) = RSP_ACCUM[el].HW[2];
AccumulatorSet(el, AccumulatorGet(el) + (RSP_Vect[RSPOpC.vs].s16(el) * RSP_Vect[RSPOpC.vt].ue(el, RSPOpC.e)));
Result.u16(el) = AccumulatorSaturate(el, true);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMADN(void)
{
uint8_t el, del;
UWORD32 temp, temp2;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
temp.UW = (uint32_t)RSP_Vect[RSPOpC.vs].u16(el) * (uint32_t)((int32_t)RSP_Vect[RSPOpC.vt].s16(del));
temp2.UW = temp.UHW[0] + RSP_ACCUM[el].UHW[1];
RSP_ACCUM[el].HW[1] = temp2.HW[0];
temp2.UW = temp.UHW[1] + RSP_ACCUM[el].UHW[2] + temp2.UHW[1];
RSP_ACCUM[el].HW[2] = temp2.HW[0];
RSP_ACCUM[el].HW[3] += temp2.HW[1];
if (temp.W < 0)
{
RSP_ACCUM[el].HW[3] -= 1;
}
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[1];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0xFFFF;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[1];
}
}
}
AccumulatorSet(el, AccumulatorGet(el) + (int64_t)(RSP_Vect[RSPOpC.vs].u16(el) * RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e)));
Result.u16(el) = AccumulatorSaturate(el, false);
}
RSP_Vect[RSPOpC.vd] = Result;
}
void RSP_Vector_VMADH(void)
{
uint8_t el, del;
RSPVector Result;
for (el = 0; el < 8; el++)
for (uint8_t el = 0; el < 8; el++)
{
del = EleSpec[RSPOpC.e].B[el];
RSP_ACCUM[el].W[1] += (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)RSP_Vect[RSPOpC.vt].s16(del);
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF)
{
Result.u16(el) = 0x8000;
}
else
{
if (RSP_ACCUM[el].HW[2] >= 0)
{
Result.u16(el) = 0x8000;
}
else
{
Result.u16(el) = RSP_ACCUM[el].HW[2];
}
}
}
else
{
if (RSP_ACCUM[el].UHW[3] != 0)
{
Result.u16(el) = 0x7FFF;
}
else
{
if (RSP_ACCUM[el].HW[2] < 0)
{
Result.u16(el) = 0x7FFF;
}
else
{
Result.u16(el) = RSP_ACCUM[el].UHW[2];
}
}
}
int32_t Value = (int32_t)((AccumulatorGet(el) >> 16) + (int32_t)RSP_Vect[RSPOpC.vs].s16(el) * (int32_t)RSP_Vect[RSPOpC.vt].se(el, RSPOpC.e));
RSP_ACCUM[el].HW[3] = (int16_t)(Value >> 16);
RSP_ACCUM[el].HW[2] = (int16_t)(Value >> 0);
Result.u16(el) = AccumulatorSaturate(el, true);
}
RSP_Vect[RSPOpC.vd] = Result;
}

20
Source/Project64-rsp-core/cpu/RSPRegister.cpp
View File

@ -55,3 +55,23 @@ void AccumulatorSet(uint8_t el, int64_t Accumulator)
RSP_ACCUM[el].HW[2] = (int16_t)(Accumulator >> 16);
RSP_ACCUM[el].HW[1] = (int16_t)(Accumulator);
}
uint16_t AccumulatorSaturate(uint8_t el, bool High)
{
if (RSP_ACCUM[el].HW[3] < 0)
{
if (RSP_ACCUM[el].UHW[3] != 0xFFFF || RSP_ACCUM[el].HW[2] >= 0)
{
return High ? 0x8000 : 0x0000;
}
else
{
return RSP_ACCUM[el].UHW[High ? 2 : 1];
}
}
if (RSP_ACCUM[el].UHW[3] != 0 || RSP_ACCUM[el].HW[2] < 0)
{
return High ? 0x7fff : 0xffff;
}
return RSP_ACCUM[el].UHW[High ? 2 : 1];
}

2
Source/Project64-rsp-core/cpu/RSPRegisters.h
View File

@ -101,6 +101,8 @@ void UpdateRSPRegistersScreen(void);
int64_t AccumulatorGet(uint8_t el);
void AccumulatorSet(uint8_t el, int64_t Accumulator);
uint16_t AccumulatorSaturate(uint8_t el, bool High);
// RSP registers
extern UWORD32 RSP_GPR[32], RSP_Flags[4];
extern UDWORD RSP_ACCUM[8];

2
Source/Project64-rsp-core/cpu/RspClamp.cpp
View File

@ -18,7 +18,7 @@ int64_t clip48(uint64_t Value)
enum : uint64_t
{
b = 1ull << (48 - 1),
m = b * 2 - 1
m = b * 2 - 1
};
return ((Value & m) ^ b) - b;
}