File melonds_PR2127.patch of Package melonds
diff --git a/src/ARMInterpreter_LoadStore.cpp b/src/ARMInterpreter_LoadStore.cpp
index 580c66fc..a0078041 100644
--- a/src/ARMInterpreter_LoadStore.cpp
+++ b/src/ARMInterpreter_LoadStore.cpp
@@ -454,11 +454,7 @@ void A_LDM(ARM* cpu)
if (!(cpu->CurInstr & (1<<23))) // decrement
{
// decrement is actually an increment starting from the end address
- for (int i = 0; i < 16; i++)
- {
- if (cpu->CurInstr & (1<<i))
- base -= 4;
- }
+ base -= 4 * __builtin_popcount(cpu->CurInstr & 0xFFFF);
if (cpu->CurInstr & (1<<21))
{
@@ -473,20 +469,17 @@ void A_LDM(ARM* cpu)
if ((cpu->CurInstr & (1<<22)) && !(cpu->CurInstr & (1<<15)))
cpu->UpdateMode(cpu->CPSR, (cpu->CPSR&~0x1F)|0x10, true);
- for (int i = 0; i < 15; i++)
+ u16 reglist = cpu->CurInstr & 0x7FFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (preinc) base += 4;
- if (!(first ? cpu->DataRead32 (base, &cpu->R[i])
- : cpu->DataRead32S(base, &cpu->R[i])))
- {
- goto dataabort;
- }
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
- first = false;
- if (!preinc) base += 4;
- }
+ if (preinc) base += 4;
+ if (first) cpu->DataRead32 (base, &cpu->R[i]);
+ else cpu->DataRead32S(base, &cpu->R[i]);
+ first = false;
+ if (!preinc) base += 4;
}
u32 pc;
@@ -559,11 +552,7 @@ void A_STM(ARM* cpu)
if (!(cpu->CurInstr & (1<<23)))
{
- for (u32 i = 0; i < 16; i++)
- {
- if (cpu->CurInstr & (1<<i))
- base -= 4;
- }
+ base -= 4 * __builtin_popcount(cpu->CurInstr & 0xFFFF);
if (cpu->CurInstr & (1<<21))
cpu->R[baseid] = base;
@@ -583,32 +572,27 @@ void A_STM(ARM* cpu)
cpu->UpdateMode(cpu->CPSR, (cpu->CPSR&~0x1F)|0x10, true);
}
- for (u32 i = 0; i < 16; i++)
+ u16 reglist = cpu->CurInstr & 0xFFFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (preinc) base += 4;
-
- u32 val;
- if (i == baseid && !isbanked)
- {
- if ((cpu->Num == 0) || (!(cpu->CurInstr & ((1<<i)-1))))
- val = oldbase;
- else val = base;
- }
- else val = cpu->R[i];
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
- if (i == 15) val+=4;
+ if (preinc) base += 4;
- if (!(first ? cpu->DataWrite32 (base, val)
- : cpu->DataWrite32S(base, val)))
- {
- goto dataabort;
- }
+ if (i == baseid && !isbanked)
+ {
+ if ((cpu->Num == 0) || (!(cpu->CurInstr & ((1<<i)-1))))
+ first ? cpu->DataWrite32(base, oldbase) : cpu->DataWrite32S(base, oldbase);
+ else
+ first ? cpu->DataWrite32(base, base) : cpu->DataWrite32S(base, base); // checkme
+ }
+ else
+ first ? cpu->DataWrite32(base, cpu->R[i]) : cpu->DataWrite32S(base, cpu->R[i]);
- first = false;
+ first = false;
- if (!preinc) base += 4;
+ if (!preinc) base += 4;
}
}
@@ -799,34 +783,23 @@ void T_LDR_SPREL(ARM* cpu)
void T_PUSH(ARM* cpu)
{
- int nregs = 0;
bool first = true;
- for (int i = 0; i < 8; i++)
- {
- if (cpu->CurInstr & (1<<i))
- nregs++;
- }
-
- if (cpu->CurInstr & (1<<8))
- nregs++;
-
u32 base = cpu->R[13];
- base -= (nregs<<2);
+ base -= 4 * __builtin_popcount(cpu->CurInstr & 0x1FF);
+
u32 wbbase = base;
- for (int i = 0; i < 8; i++)
+ u8 reglist = cpu->CurInstr & 0xFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (!(first ? cpu->DataWrite32 (base, cpu->R[i])
- : cpu->DataWrite32S(base, cpu->R[i])))
- {
- goto dataabort;
- }
- first = false;
- base += 4;
- }
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
+
+ if (first) cpu->DataWrite32 (base, cpu->R[i]);
+ else cpu->DataWrite32S(base, cpu->R[i]);
+ first = false;
+ base += 4;
}
if (cpu->CurInstr & (1<<8))
@@ -849,18 +822,16 @@ void T_POP(ARM* cpu)
u32 base = cpu->R[13];
bool first = true;
- for (int i = 0; i < 8; i++)
+ u8 reglist = cpu->CurInstr & 0xFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (!(first ? cpu->DataRead32 (base, &cpu->R[i])
- : cpu->DataRead32S(base, &cpu->R[i])))
- {
- goto dataabort;
- }
- first = false;
- base += 4;
- }
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
+
+ if (first) cpu->DataRead32 (base, &cpu->R[i]);
+ else cpu->DataRead32S(base, &cpu->R[i]);
+ first = false;
+ base += 4;
}
if (cpu->CurInstr & (1<<8))
@@ -887,18 +858,16 @@ void T_STMIA(ARM* cpu)
u32 base = cpu->R[(cpu->CurInstr >> 8) & 0x7];
bool first = true;
- for (int i = 0; i < 8; i++)
+ u8 reglist = cpu->CurInstr & 0xFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (!(first ? cpu->DataWrite32 (base, cpu->R[i])
- : cpu->DataWrite32S(base, cpu->R[i])))
- {
- goto dataabort;
- }
- first = false;
- base += 4;
- }
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
+
+ if (first) cpu->DataWrite32 (base, cpu->R[i]);
+ else cpu->DataWrite32S(base, cpu->R[i]);
+ first = false;
+ base += 4;
}
// TODO: check "Rb included in Rlist" case
@@ -912,18 +881,16 @@ void T_LDMIA(ARM* cpu)
u32 base = cpu->R[(cpu->CurInstr >> 8) & 0x7];
bool first = true;
- for (int i = 0; i < 8; i++)
+ u8 reglist = cpu->CurInstr & 0xFF;
+ while (reglist)
{
- if (cpu->CurInstr & (1<<i))
- {
- if (!(first ? cpu->DataRead32 (base, &cpu->R[i])
- : cpu->DataRead32S(base, &cpu->R[i])))
- {
- goto dataabort;
- }
- first = false;
- base += 4;
- }
+ int i = __builtin_ctz(reglist);
+ reglist ^= 1<<i;
+
+ if (first) cpu->DataRead32 (base, &cpu->R[i]);
+ else cpu->DataRead32S(base, &cpu->R[i]);
+ first = false;
+ base += 4;
}
if (!(cpu->CurInstr & (1<<((cpu->CurInstr >> 8) & 0x7))))
diff --git a/src/GPU3D_Soft.cpp b/src/GPU3D_Soft.cpp
index 1221ed59..03c48f99 100644
--- a/src/GPU3D_Soft.cpp
+++ b/src/GPU3D_Soft.cpp
@@ -143,6 +143,8 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
u32 vramaddr = (texparam & 0xFFFF) << 3;
s32 width = 8 << ((texparam >> 20) & 0x7);
+ // since width is always a multple of 8 (thus a multiple of 2) we can replace all multiplications by width with a bitshift
+ s32 widthshift = 3 + ((texparam >> 20) & 0x7);
s32 height = 8 << ((texparam >> 23) & 0x7);
s >>= 4;
@@ -192,7 +194,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
{
case 1: // A3I5
{
- vramaddr += ((t * width) + s);
+ vramaddr += ((t << widthshift) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu);
texpal <<= 4;
@@ -203,7 +205,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 2: // 4-color
{
- vramaddr += (((t * width) + s) >> 2);
+ vramaddr += (((t << widthshift) + s) >> 2);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu);
pixel >>= ((s & 0x3) << 1);
pixel &= 0x3;
@@ -216,7 +218,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 3: // 16-color
{
- vramaddr += (((t * width) + s) >> 1);
+ vramaddr += (((t << widthshift) + s) >> 1);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu);
if (s & 0x1) pixel >>= 4;
else pixel &= 0xF;
@@ -229,7 +231,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 4: // 256-color
{
- vramaddr += ((t * width) + s);
+ vramaddr += ((t << widthshift) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu);
texpal <<= 4;
@@ -240,7 +242,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 5: // compressed
{
- vramaddr += ((t & 0x3FC) * (width>>2)) + (s & 0x3FC);
+ vramaddr += ((t & 0x3FC) << (widthshift-2)) + (s & 0x3FC);
vramaddr += (t & 0x3);
vramaddr &= 0x7FFFF; // address used for all calcs wraps around after slot 3
@@ -352,7 +354,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 6: // A5I3
{
- vramaddr += ((t * width) + s);
+ vramaddr += ((t << widthshift) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu);
texpal <<= 4;
@@ -363,7 +365,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 7: // direct color
{
- vramaddr += (((t * width) + s) << 1);
+ vramaddr += (((t << widthshift) + s) << 1);
*color = ReadVRAM_Texture<u16>(vramaddr, gpu);
*alpha = (*color & 0x8000) ? 31 : 0;
}
diff --git a/src/GPU3D_Soft.h b/src/GPU3D_Soft.h
index 55a698b0..e2c2fa1b 100644
--- a/src/GPU3D_Soft.h
+++ b/src/GPU3D_Soft.h
@@ -101,15 +101,15 @@ private:
if ((w0 & 0x1) && !(w1 & 0x1))
{
- this->w0n = w0 - 1;
- this->w0d = w0 + 1;
- this->w1d = w1;
+ this->w0n = w0 - 1 >> 1;
+ this->w0d = w0 + 1 >> 1;
+ this->w1d = w1 >> 1;
}
else
{
- this->w0n = w0 & 0xFFFE;
- this->w0d = w0 & 0xFFFE;
- this->w1d = w1 & 0xFFFE;
+ this->w0n = w0 >> 1;
+ this->w0d = w0 >> 1;
+ this->w1d = w1 >> 1;
}
this->shift = 9;
@@ -138,7 +138,7 @@ private:
// this seems to be a proper division on hardware :/
// I haven't been able to find cases that produce imperfect output
if (den == 0) yfactor = 0;
- else yfactor = (s32)(num / den);
+ else yfactor = ((u32)num / den);
}
}
