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

PGXP: Treat add rd, rs, zero as moves 

Fixes exploding vertices in BIOS intro.

Needs further investigation as to why this fixes it - clearly the actual
oepration is incorrect.
This commit is contained in:
Connor McLaughlin 2020-12-04 01:07:34 +10:00
parent 59ac365b52
commit 52dbcbaca5
1 changed files with 44 additions and 34 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
src/core/pgxp.cpp
View File

@ -901,17 +901,20 @@ void CPU_ADDI(u32 instr, u32 rtVal, u32 rsVal)
tempImm.d = imm(instr);
tempImm.sd = (tempImm.sd << 16) >> 16; // sign extend
ret.x = (float)f16Unsign(ret.x);
ret.x += (float)tempImm.w.l;
if (tempImm.d != 0)
{
ret.x = (float)f16Unsign(ret.x);
ret.x += (float)tempImm.w.l;
// carry on over/underflow
float of = (ret.x > USHRT_MAX) ? 1.f : (ret.x < 0) ? -1.f : 0.f;
ret.x = (float)f16Sign(ret.x);
// ret.x -= of * (USHRT_MAX + 1);
ret.y += tempImm.sw.h + of;
// carry on over/underflow
float of = (ret.x > USHRT_MAX) ? 1.f : (ret.x < 0) ? -1.f : 0.f;
ret.x = (float)f16Sign(ret.x);
// ret.x -= of * (USHRT_MAX + 1);
ret.y += tempImm.sw.h + of;
// truncate on overflow/underflow
ret.y += (ret.y > SHRT_MAX) ? -(USHRT_MAX + 1) : (ret.y < SHRT_MIN) ? USHRT_MAX + 1 : 0.f;
// truncate on overflow/underflow
ret.y += (ret.y > SHRT_MAX) ? -(USHRT_MAX + 1) : (ret.y < SHRT_MIN) ? USHRT_MAX + 1 : 0.f;
}
CPU_reg[rt(instr)] = ret;
CPU_reg[rt(instr)].value = rtVal;
@ -1069,33 +1072,40 @@ void CPU_ADD(u32 instr, u32 rdVal, u32 rsVal, u32 rtVal)
Validate(&CPU_reg[rs(instr)], rsVal);
Validate(&CPU_reg[rt(instr)], rtVal);
// iCB: Only require one valid input
if (((CPU_reg[rt(instr)].flags & VALID_01) != VALID_01) != ((CPU_reg[rs(instr)].flags & VALID_01) != VALID_01))
if (rtVal != 0)
{
MakeValid(&CPU_reg[rs(instr)], rsVal);
MakeValid(&CPU_reg[rt(instr)], rtVal);
// iCB: Only require one valid input
if (((CPU_reg[rt(instr)].flags & VALID_01) != VALID_01) != ((CPU_reg[rs(instr)].flags & VALID_01) != VALID_01))
{
MakeValid(&CPU_reg[rs(instr)], rsVal);
MakeValid(&CPU_reg[rt(instr)], rtVal);
}
ret = CPU_reg[rs(instr)];
ret.x = (float)f16Unsign(ret.x);
ret.x += (float)f16Unsign(CPU_reg[rt(instr)].x);
// carry on over/underflow
float of = (ret.x > USHRT_MAX) ? 1.f : (ret.x < 0) ? -1.f : 0.f;
ret.x = (float)f16Sign(ret.x);
// ret.x -= of * (USHRT_MAX + 1);
ret.y += CPU_reg[rt(instr)].y + of;
// truncate on overflow/underflow
ret.y += (ret.y > SHRT_MAX) ? -(USHRT_MAX + 1) : (ret.y < SHRT_MIN) ? USHRT_MAX + 1 : 0.f;
// TODO: decide which "z/w" component to use
ret.halfFlags[0] &= CPU_reg[rt(instr)].halfFlags[0];
ret.gFlags |= CPU_reg[rt(instr)].gFlags;
ret.lFlags |= CPU_reg[rt(instr)].lFlags;
ret.hFlags |= CPU_reg[rt(instr)].hFlags;
}
else
{
ret = CPU_reg[rs(instr)];
}
ret = CPU_reg[rs(instr)];
ret.x = (float)f16Unsign(ret.x);
ret.x += (float)f16Unsign(CPU_reg[rt(instr)].x);
// carry on over/underflow
float of = (ret.x > USHRT_MAX) ? 1.f : (ret.x < 0) ? -1.f : 0.f;
ret.x = (float)f16Sign(ret.x);
// ret.x -= of * (USHRT_MAX + 1);
ret.y += CPU_reg[rt(instr)].y + of;
// truncate on overflow/underflow
ret.y += (ret.y > SHRT_MAX) ? -(USHRT_MAX + 1) : (ret.y < SHRT_MIN) ? USHRT_MAX + 1 : 0.f;
// TODO: decide which "z/w" component to use
ret.halfFlags[0] &= CPU_reg[rt(instr)].halfFlags[0];
ret.gFlags |= CPU_reg[rt(instr)].gFlags;
ret.lFlags |= CPU_reg[rt(instr)].lFlags;
ret.hFlags |= CPU_reg[rt(instr)].hFlags;
ret.value = rdVal;