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

handle address wrap around in texture cache 

fixes out of bounds access in Mario 64
also slightly optimise paletted texture conversion
This commit is contained in:
RSDuck 2024-10-27 23:32:05 +01:00
parent b60f42b281
commit 58ab33210a
5 changed files with 136 additions and 114 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
src/GPU.h
View File

@ -499,6 +499,17 @@ public:
OAMDirty |= 1 << (addr / 1024); OAMDirty |= 1 << (addr / 1024);
} }
template <typename T>
inline T ReadVRAMFlat_Texture(u32 addr) const
{
return *(T*)&VRAMFlat_Texture[addr & 0x7FFFF];
}
template <typename T>
inline T ReadVRAMFlat_TexPal(u32 addr) const
{
return *(T*)&VRAMFlat_TexPal[addr & 0x1FFFF];
}
void SetPowerCnt(u32 val) noexcept; void SetPowerCnt(u32 val) noexcept;
void StartFrame() noexcept; void StartFrame() noexcept;

50
src/GPU3D_Soft.cpp
View File

@ -193,10 +193,10 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 1: // A3I5 case 1: // A3I5
{ {
vramaddr += ((t * width) + s); vramaddr += ((t * width) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu); u8 pixel = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
texpal <<= 4; texpal <<= 4;
*color = ReadVRAM_TexPal<u16>(texpal + ((pixel&0x1F)<<1), gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + ((pixel&0x1F)<<1));
*alpha = ((pixel >> 3) & 0x1C) + (pixel >> 6); *alpha = ((pixel >> 3) & 0x1C) + (pixel >> 6);
} }
break; break;
@ -204,12 +204,12 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 2: // 4-color case 2: // 4-color
{ {
vramaddr += (((t * width) + s) >> 2); vramaddr += (((t * width) + s) >> 2);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu); u8 pixel = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
pixel >>= ((s & 0x3) << 1); pixel >>= ((s & 0x3) << 1);
pixel &= 0x3; pixel &= 0x3;
texpal <<= 3; texpal <<= 3;
*color = ReadVRAM_TexPal<u16>(texpal + (pixel<<1), gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + (pixel<<1));
*alpha = (pixel==0) ? alpha0 : 31; *alpha = (pixel==0) ? alpha0 : 31;
} }
break; break;
@ -217,12 +217,12 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 3: // 16-color case 3: // 16-color
{ {
vramaddr += (((t * width) + s) >> 1); vramaddr += (((t * width) + s) >> 1);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu); u8 pixel = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
if (s & 0x1) pixel >>= 4; if (s & 0x1) pixel >>= 4;
else pixel &= 0xF; else pixel &= 0xF;
texpal <<= 4; texpal <<= 4;
*color = ReadVRAM_TexPal<u16>(texpal + (pixel<<1), gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + (pixel<<1));
*alpha = (pixel==0) ? alpha0 : 31; *alpha = (pixel==0) ? alpha0 : 31;
} }
break; break;
@ -230,10 +230,10 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 4: // 256-color case 4: // 256-color
{ {
vramaddr += ((t * width) + s); vramaddr += ((t * width) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu); u8 pixel = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
texpal <<= 4; texpal <<= 4;
*color = ReadVRAM_TexPal<u16>(texpal + (pixel<<1), gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + (pixel<<1));
*alpha = (pixel==0) ? alpha0 : 31; *alpha = (pixel==0) ? alpha0 : 31;
} }
break; break;
@ -253,31 +253,31 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
val = 0; val = 0;
else else
{ {
val = ReadVRAM_Texture<u8>(vramaddr, gpu); val = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
val >>= (2 * (s & 0x3)); val >>= (2 * (s & 0x3));
} }
u16 palinfo = ReadVRAM_Texture<u16>(slot1addr, gpu); u16 palinfo = gpu.ReadVRAMFlat_Texture<u16>(slot1addr);
u32 paloffset = (palinfo & 0x3FFF) << 2; u32 paloffset = (palinfo & 0x3FFF) << 2;
texpal <<= 4; texpal <<= 4;
switch (val & 0x3) switch (val & 0x3)
{ {
case 0: case 0:
*color = ReadVRAM_TexPal<u16>(texpal + paloffset, gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset);
*alpha = 31; *alpha = 31;
break; break;
case 1: case 1:
*color = ReadVRAM_TexPal<u16>(texpal + paloffset + 2, gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 2);
*alpha = 31; *alpha = 31;
break; break;
case 2: case 2:
if ((palinfo >> 14) == 1) if ((palinfo >> 14) == 1)
{ {
u16 color0 = ReadVRAM_TexPal<u16>(texpal + paloffset, gpu); u16 color0 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset);
u16 color1 = ReadVRAM_TexPal<u16>(texpal + paloffset + 2, gpu); u16 color1 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 2);
u32 r0 = color0 & 0x001F; u32 r0 = color0 & 0x001F;
u32 g0 = color0 & 0x03E0; u32 g0 = color0 & 0x03E0;
@ -294,8 +294,8 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
} }
else if ((palinfo >> 14) == 3) else if ((palinfo >> 14) == 3)
{ {
u16 color0 = ReadVRAM_TexPal<u16>(texpal + paloffset, gpu); u16 color0 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset);
u16 color1 = ReadVRAM_TexPal<u16>(texpal + paloffset + 2, gpu); u16 color1 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 2);
u32 r0 = color0 & 0x001F; u32 r0 = color0 & 0x001F;
u32 g0 = color0 & 0x03E0; u32 g0 = color0 & 0x03E0;
@ -311,20 +311,20 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
*color = r | g | b; *color = r | g | b;
} }
else else
*color = ReadVRAM_TexPal<u16>(texpal + paloffset + 4, gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 4);
*alpha = 31; *alpha = 31;
break; break;
case 3: case 3:
if ((palinfo >> 14) == 2) if ((palinfo >> 14) == 2)
{ {
*color = ReadVRAM_TexPal<u16>(texpal + paloffset + 6, gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 6);
*alpha = 31; *alpha = 31;
} }
else if ((palinfo >> 14) == 3) else if ((palinfo >> 14) == 3)
{ {
u16 color0 = ReadVRAM_TexPal<u16>(texpal + paloffset, gpu); u16 color0 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset);
u16 color1 = ReadVRAM_TexPal<u16>(texpal + paloffset + 2, gpu); u16 color1 = gpu.ReadVRAMFlat_TexPal<u16>(texpal + paloffset + 2);
u32 r0 = color0 & 0x001F; u32 r0 = color0 & 0x001F;
u32 g0 = color0 & 0x03E0; u32 g0 = color0 & 0x03E0;
@ -353,10 +353,10 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 6: // A5I3 case 6: // A5I3
{ {
vramaddr += ((t * width) + s); vramaddr += ((t * width) + s);
u8 pixel = ReadVRAM_Texture<u8>(vramaddr, gpu); u8 pixel = gpu.ReadVRAMFlat_Texture<u8>(vramaddr);
texpal <<= 4; texpal <<= 4;
*color = ReadVRAM_TexPal<u16>(texpal + ((pixel&0x7)<<1), gpu); *color = gpu.ReadVRAMFlat_TexPal<u16>(texpal + ((pixel&0x7)<<1));
*alpha = (pixel >> 3); *alpha = (pixel >> 3);
} }
break; break;
@ -364,7 +364,7 @@ void SoftRenderer::TextureLookup(const GPU& gpu, u32 texparam, u32 texpal, s16 s
case 7: // direct color case 7: // direct color
{ {
vramaddr += (((t * width) + s) << 1); vramaddr += (((t * width) + s) << 1);
*color = ReadVRAM_Texture<u16>(vramaddr, gpu); *color = gpu.ReadVRAMFlat_Texture<u16>(vramaddr);
*alpha = (*color & 0x8000) ? 31 : 0; *alpha = (*color & 0x8000) ? 31 : 0;
} }
break; break;
@ -1659,8 +1659,8 @@ void SoftRenderer::ClearBuffers(const GPU& gpu)
{ {
for (int x = 0; x < 256; x++) for (int x = 0; x < 256; x++)
{ {
u16 val2 = ReadVRAM_Texture<u16>(0x40000 + (yoff << 9) + (xoff << 1), gpu); u16 val2 = gpu.ReadVRAMFlat_Texture<u16>(0x40000 + (yoff << 9) + (xoff << 1));
u16 val3 = ReadVRAM_Texture<u16>(0x60000 + (yoff << 9) + (xoff << 1), gpu); u16 val3 = gpu.ReadVRAMFlat_Texture<u16>(0x60000 + (yoff << 9) + (xoff << 1));
// TODO: confirm color conversion // TODO: confirm color conversion
u32 r = (val2 << 1) & 0x3E; if (r) r++; u32 r = (val2 << 1) & 0x3E; if (r) r++;

10
src/GPU3D_Soft.h
View File

@ -430,16 +430,6 @@ private:
s32 ycoverage, ycov_incr; s32 ycoverage, ycov_incr;
}; };
template <typename T>
inline T ReadVRAM_Texture(u32 addr, const GPU& gpu) const
{
return *(T*)&gpu.VRAMFlat_Texture[addr & 0x7FFFF];
}
template <typename T>
inline T ReadVRAM_TexPal(u32 addr, const GPU& gpu) const
{
return *(T*)&gpu.VRAMFlat_TexPal[addr & 0x1FFFF];
}
u32 AlphaBlend(const GPU3D& gpu3d, u32 srccolor, u32 dstcolor, u32 alpha) const noexcept; u32 AlphaBlend(const GPU3D& gpu3d, u32 srccolor, u32 dstcolor, u32 alpha) const noexcept;
struct RendererPolygon struct RendererPolygon

61
src/GPU3D_Texcache.cpp
View File

@ -75,11 +75,11 @@ inline u32 ConvertRGB5ToRGB6(u16 val)
} }
template <int outputFmt> template <int outputFmt>
void ConvertBitmapTexture(u32 width, u32 height, u32* output, u8* texData) void ConvertBitmapTexture(u32 width, u32 height, u32* output, u32 addr, GPU& gpu)
{ {
for (u32 i = 0; i < width*height; i++) for (u32 i = 0; i < width*height; i++)
{ {
u16 value = *(u16*)&texData[i * 2]; u16 value = gpu.ReadVRAMFlat_Texture<u16>(addr + i * 2);
switch (outputFmt) switch (outputFmt)
{ {
@ -96,28 +96,28 @@ void ConvertBitmapTexture(u32 width, u32 height, u32* output, u8* texData)
} }
} }
template void ConvertBitmapTexture<outputFmt_RGB6A5>(u32 width, u32 height, u32* output, u8* texData); template void ConvertBitmapTexture<outputFmt_RGB6A5>(u32 width, u32 height, u32* output, u32 addr, GPU& gpu);
template <int outputFmt> template <int outputFmt>
void ConvertCompressedTexture(u32 width, u32 height, u32* output, u8* texData, u8* texAuxData, u16* palData) void ConvertCompressedTexture(u32 width, u32 height, u32* output, u32 addr, u32 addrAux, u32 palAddr, GPU& gpu)
{ {
// we process a whole block at the time // we process a whole block at the time
for (int y = 0; y < height / 4; y++) for (int y = 0; y < height / 4; y++)
{ {
for (int x = 0; x < width / 4; x++) for (int x = 0; x < width / 4; x++)
{ {
u32 data = ((u32*)texData)[x + y * (width / 4)]; u32 data = gpu.ReadVRAMFlat_Texture<u32>(addr + (x + y * (width / 4))*4);
u16 auxData = ((u16*)texAuxData)[x + y * (width / 4)]; u16 auxData = gpu.ReadVRAMFlat_Texture<u16>(addrAux + (x + y * (width / 4))*2);
u32 paletteOffset = auxData & 0x3FFF; u32 paletteOffset = palAddr + (auxData & 0x3FFF) * 4;
u16 color0 = palData[paletteOffset*2] | 0x8000; u16 color0 = gpu.ReadVRAMFlat_TexPal<u16>(paletteOffset) | 0x8000;
u16 color1 = palData[paletteOffset*2+1] | 0x8000; u16 color1 = gpu.ReadVRAMFlat_TexPal<u16>(paletteOffset+2) | 0x8000;
u16 color2, color3; u16 color2 = gpu.ReadVRAMFlat_TexPal<u16>(paletteOffset+4) | 0x8000;
u16 color3 = gpu.ReadVRAMFlat_TexPal<u16>(paletteOffset+6) | 0x8000;
switch ((auxData >> 14) & 0x3) switch ((auxData >> 14) & 0x3)
{ {
case 0: case 0:
color2 = palData[paletteOffset*2+2] | 0x8000;
color3 = 0; color3 = 0;
break; break;
case 1: case 1:
@ -137,8 +137,6 @@ void ConvertCompressedTexture(u32 width, u32 height, u32* output, u8* texData, u
color3 = 0; color3 = 0;
break; break;
case 2: case 2:
color2 = palData[paletteOffset*2+2] | 0x8000;
color3 = palData[paletteOffset*2+3] | 0x8000;
break; break;
case 3: case 3:
{ {
@ -179,7 +177,8 @@ void ConvertCompressedTexture(u32 width, u32 height, u32* output, u8* texData, u
{ {
for (int i = 0; i < 4; i++) for (int i = 0; i < 4; i++)
{ {
u16 color = (packed >> 16 * (data >> 2 * (i + j * 4))) & 0xFFFF; u32 colorIdx = 16 * ((data >> 2 * (i + j * 4)) & 0x3);
u16 color = (packed >> colorIdx) & 0xFFFF;
u32 res; u32 res;
switch (outputFmt) switch (outputFmt)
{ {
@ -197,20 +196,20 @@ void ConvertCompressedTexture(u32 width, u32 height, u32* output, u8* texData, u
} }
} }
template void ConvertCompressedTexture<outputFmt_RGB6A5>(u32, u32, u32*, u8*, u8*, u16*); template void ConvertCompressedTexture<outputFmt_RGB6A5>(u32, u32, u32*, u32, u32, u32, GPU&);
template <int outputFmt, int X, int Y> template <int outputFmt, int X, int Y>
void ConvertAXIYTexture(u32 width, u32 height, u32* output, u8* texData, u16* palData) void ConvertAXIYTexture(u32 width, u32 height, u32* output, u32 addr, u32 palAddr, GPU& gpu)
{ {
for (int y = 0; y < height; y++) for (int y = 0; y < height; y++)
{ {
for (int x = 0; x < width; x++) for (int x = 0; x < width; x++)
{ {
u8 val = texData[x + y * width]; u8 val = gpu.ReadVRAMFlat_Texture<u8>(addr + x + y * width);
u32 idx = val & ((1 << Y) - 1); u32 idx = val & ((1 << Y) - 1);
u16 color = palData[idx]; u16 color = gpu.ReadVRAMFlat_TexPal<u16>(palAddr + idx * 2);
u32 alpha = (val >> Y) & ((1 << X) - 1); u32 alpha = (val >> Y) & ((1 << X) - 1);
if (X != 5) if (X != 5)
alpha = alpha * 4 + alpha / 2; alpha = alpha * 4 + alpha / 2;
@ -228,22 +227,24 @@ void ConvertAXIYTexture(u32 width, u32 height, u32* output, u8* texData, u16* pa
} }
} }
template void ConvertAXIYTexture<outputFmt_RGB6A5, 5, 3>(u32, u32, u32*, u8*, u16*); template void ConvertAXIYTexture<outputFmt_RGB6A5, 5, 3>(u32, u32, u32*, u32, u32, GPU&);
template void ConvertAXIYTexture<outputFmt_RGB6A5, 3, 5>(u32, u32, u32*, u8*, u16*); template void ConvertAXIYTexture<outputFmt_RGB6A5, 3, 5>(u32, u32, u32*, u32, u32, GPU&);
template <int outputFmt, int colorBits> template <int outputFmt, int colorBits>
void ConvertNColorsTexture(u32 width, u32 height, u32* output, u8* texData, u16* palData, bool color0Transparent) void ConvertNColorsTexture(u32 width, u32 height, u32* output, u32 addr, u32 palAddr, bool color0Transparent, GPU& gpu)
{ {
for (int y = 0; y < height; y++) for (int y = 0; y < height; y++)
{ {
for (int x = 0; x < width / (8 / colorBits); x++) for (int x = 0; x < width / (16 / colorBits); x++)
{ {
u8 val = texData[x + y * (width / (8 / colorBits))]; // smallest possible row is 8 pixels with 2bpp => fits in u16
u16 val = gpu.ReadVRAMFlat_Texture<u16>(addr + 2 * (x + y * (width / (16 / colorBits))));
for (int i = 0; i < 8 / colorBits; i++) for (int i = 0; i < 16 / colorBits; i++)
{ {
u32 index = (val >> (i * colorBits)) & ((1 << colorBits) - 1); u32 index = val & ((1 << colorBits) - 1);
u16 color = palData[index]; val >>= colorBits;
u16 color = gpu.ReadVRAMFlat_TexPal<u16>(palAddr + index * 2);
bool transparent = color0Transparent && index == 0; bool transparent = color0Transparent && index == 0;
u32 res; u32 res;
@ -256,14 +257,14 @@ void ConvertNColorsTexture(u32 width, u32 height, u32* output, u8* texData, u16*
case outputFmt_BGRA8: res = ConvertRGB5ToBGR8(color) case outputFmt_BGRA8: res = ConvertRGB5ToBGR8(color)
| (transparent ? 0 : 0xFF000000); break; | (transparent ? 0 : 0xFF000000); break;
} }
output[x * (8 / colorBits) + y * width + i] = res; output[x * (16 / colorBits) + y * width + i] = res;
} }
} }
} }
} }
template void ConvertNColorsTexture<outputFmt_RGB6A5, 2>(u32, u32, u32*, u8*, u16*, bool); template void ConvertNColorsTexture<outputFmt_RGB6A5, 2>(u32, u32, u32*, u32, u32, bool, GPU&);
template void ConvertNColorsTexture<outputFmt_RGB6A5, 4>(u32, u32, u32*, u8*, u16*, bool); template void ConvertNColorsTexture<outputFmt_RGB6A5, 4>(u32, u32, u32*, u32, u32, bool, GPU&);
template void ConvertNColorsTexture<outputFmt_RGB6A5, 8>(u32, u32, u32*, u8*, u16*, bool); template void ConvertNColorsTexture<outputFmt_RGB6A5, 8>(u32, u32, u32*, u32, u32, bool, GPU&);
} }

118
src/GPU3D_Texcache.h
View File

@ -32,13 +32,13 @@ enum
}; };
template <int outputFmt> template <int outputFmt>
void ConvertBitmapTexture(u32 width, u32 height, u32* output, u8* texData); void ConvertBitmapTexture(u32 width, u32 height, u32* output, u32 addr, GPU& gpu);
template <int outputFmt> template <int outputFmt>
void ConvertCompressedTexture(u32 width, u32 height, u32* output, u8* texData, u8* texAuxData, u16* palData); void ConvertCompressedTexture(u32 width, u32 height, u32* output, u32 addr, u32 addrAux, u32 palAddr, GPU& gpu);
template <int outputFmt, int X, int Y> template <int outputFmt, int X, int Y>
void ConvertAXIYTexture(u32 width, u32 height, u32* output, u8* texData, u16* palData); void ConvertAXIYTexture(u32 width, u32 height, u32* output, u32 addr, u32 palAddr, GPU& gpu);
template <int outputFmt, int colorBits> template <int outputFmt, int colorBits>
void ConvertNColorsTexture(u32 width, u32 height, u32* output, u8* texData, u16* palData, bool color0Transparent); void ConvertNColorsTexture(u32 width, u32 height, u32* output, u32 addr, u32 palAddr, bool color0Transparent, GPU& gpu);
template <typename TexLoaderT, typename TexHandleT> template <typename TexLoaderT, typename TexHandleT>
class Texcache class Texcache
@ -48,6 +48,50 @@ public:
: TexLoader(texloader) // probably better if this would be a move constructor??? : TexLoader(texloader) // probably better if this would be a move constructor???
{} {}
u64 MaskedHash(u8* vram, u32 vramSize, u32 addr, u32 size)
{
u64 hash = 0;
while (size > 0)
{
u32 pieceSize;
if (addr + size > vramSize)
// wraps around, only do the part inside
pieceSize = vramSize - addr;
else
// fits completely inside
pieceSize = size;
hash = XXH64(&vram[addr], pieceSize, hash);
addr += pieceSize;
addr &= (vramSize - 1);
assert(size >= pieceSize);
size -= pieceSize;
}
return hash;
}
bool CheckInvalid(u32 start, u32 size, u64 oldHash, u64* dirty, u8* vram, u32 vramSize)
{
u32 startBit = start / VRAMDirtyGranularity;
u32 bitsCount = ((start + size + VRAMDirtyGranularity - 1) / VRAMDirtyGranularity) - startBit;
u32 startEntry = startBit >> 6;
u64 entriesCount = ((startBit + bitsCount + 0x3F) >> 6) - startEntry;
for (u32 j = startEntry; j < startEntry + entriesCount; j++)
{
if (GetRangedBitMask(j, startBit, bitsCount) & dirty[j & ((vramSize / VRAMDirtyGranularity)-1)])
{
if (MaskedHash(vram, vramSize, start, size) != oldHash)
return true;
}
}
return false;
}
bool Update(GPU& gpu) bool Update(GPU& gpu)
{ {
auto textureDirty = gpu.VRAMDirty_Texture.DeriveState(gpu.VRAMMap_Texture, gpu); auto textureDirty = gpu.VRAMDirty_Texture.DeriveState(gpu.VRAMMap_Texture, gpu);
@ -66,40 +110,21 @@ public:
{ {
for (u32 i = 0; i < 2; i++) for (u32 i = 0; i < 2; i++)
{ {
u32 startBit = entry.TextureRAMStart[i] / VRAMDirtyGranularity; if (CheckInvalid(entry.TextureRAMStart[i], entry.TextureRAMSize[i],
u32 bitsCount = ((entry.TextureRAMStart[i] + entry.TextureRAMSize[i] + VRAMDirtyGranularity - 1) / VRAMDirtyGranularity) - startBit; entry.TextureHash[i],
textureDirty.Data,
u32 startEntry = startBit >> 6; gpu.VRAMFlat_Texture, sizeof(gpu.VRAMFlat_Texture)))
u64 entriesCount = ((startBit + bitsCount + 0x3F) >> 6) - startEntry; goto invalidate;
for (u32 j = startEntry; j < startEntry + entriesCount; j++)
{
if (GetRangedBitMask(j, startBit, bitsCount) & textureDirty.Data[j])
{
u64 newTexHash = XXH3_64bits(&gpu.VRAMFlat_Texture[entry.TextureRAMStart[i]], entry.TextureRAMSize[i]);
if (newTexHash != entry.TextureHash[i])
goto invalidate;
}
}
} }
} }
if (texPalChanged && entry.TexPalSize > 0) if (texPalChanged && entry.TexPalSize > 0)
{ {
u32 startBit = entry.TexPalStart / VRAMDirtyGranularity; if (CheckInvalid(entry.TexPalStart, entry.TexPalSize,
u32 bitsCount = ((entry.TexPalStart + entry.TexPalSize + VRAMDirtyGranularity - 1) / VRAMDirtyGranularity) - startBit; entry.TexPalHash,
texPalDirty.Data,
u32 startEntry = startBit >> 6; gpu.VRAMFlat_TexPal, sizeof(gpu.VRAMFlat_TexPal)))
u64 entriesCount = ((startBit + bitsCount + 0x3F) >> 6) - startEntry; goto invalidate;
for (u32 j = startEntry; j < startEntry + entriesCount; j++)
{
if (GetRangedBitMask(j, startBit, bitsCount) & texPalDirty.Data[j])
{
u64 newPalHash = XXH3_64bits(&gpu.VRAMFlat_TexPal[entry.TexPalStart], entry.TexPalSize);
if (newPalHash != entry.TexPalHash)
goto invalidate;
}
}
} }
it++; it++;
@ -163,17 +188,13 @@ public:
{ {
entry.TextureRAMSize[0] = width*height*2; entry.TextureRAMSize[0] = width*height*2;
ConvertBitmapTexture<outputFmt_RGB6A5>(width, height, DecodingBuffer, &gpu.VRAMFlat_Texture[addr]); ConvertBitmapTexture<outputFmt_RGB6A5>(width, height, DecodingBuffer, addr, gpu);
} }
else if (fmt == 5) else if (fmt == 5)
{ {
u8* texData = &gpu.VRAMFlat_Texture[addr];
u32 slot1addr = 0x20000 + ((addr & 0x1FFFC) >> 1); u32 slot1addr = 0x20000 + ((addr & 0x1FFFC) >> 1);
if (addr >= 0x40000) if (addr >= 0x40000)
slot1addr += 0x10000; slot1addr += 0x10000;
u8* texAuxData = &gpu.VRAMFlat_Texture[slot1addr];
u16* palData = (u16*)(gpu.VRAMFlat_TexPal + palBase*16);
entry.TextureRAMSize[0] = width*height/16*4; entry.TextureRAMSize[0] = width*height/16*4;
entry.TextureRAMStart[1] = slot1addr; entry.TextureRAMStart[1] = slot1addr;
@ -181,7 +202,7 @@ public:
entry.TexPalStart = palBase*16; entry.TexPalStart = palBase*16;
entry.TexPalSize = 0x10000; entry.TexPalSize = 0x10000;
ConvertCompressedTexture<outputFmt_RGB6A5>(width, height, DecodingBuffer, texData, texAuxData, palData); ConvertCompressedTexture<outputFmt_RGB6A5>(width, height, DecodingBuffer, addr, slot1addr, entry.TexPalStart, gpu);
} }
else else
{ {
@ -204,30 +225,29 @@ public:
entry.TexPalStart = palAddr; entry.TexPalStart = palAddr;
entry.TexPalSize = numPalEntries*2; entry.TexPalSize = numPalEntries*2;
u8* texData = &gpu.VRAMFlat_Texture[addr];
u16* palData = (u16*)(gpu.VRAMFlat_TexPal + palAddr);
//assert(entry.TexPalStart+entry.TexPalSize <= 128*1024*1024); //assert(entry.TexPalStart+entry.TexPalSize <= 128*1024*1024);
bool color0Transparent = texParam & (1 << 29); bool color0Transparent = texParam & (1 << 29);
switch (fmt) switch (fmt)
{ {
case 1: ConvertAXIYTexture<outputFmt_RGB6A5, 3, 5>(width, height, DecodingBuffer, texData, palData); break; case 1: ConvertAXIYTexture<outputFmt_RGB6A5, 3, 5>(width, height, DecodingBuffer, addr, palAddr, gpu); break;
case 6: ConvertAXIYTexture<outputFmt_RGB6A5, 5, 3>(width, height, DecodingBuffer, texData, palData); break; case 6: ConvertAXIYTexture<outputFmt_RGB6A5, 5, 3>(width, height, DecodingBuffer, addr, palAddr, gpu); break;
case 2: ConvertNColorsTexture<outputFmt_RGB6A5, 2>(width, height, DecodingBuffer, texData, palData, color0Transparent); break; case 2: ConvertNColorsTexture<outputFmt_RGB6A5, 2>(width, height, DecodingBuffer, addr, palAddr, color0Transparent, gpu); break;
case 3: ConvertNColorsTexture<outputFmt_RGB6A5, 4>(width, height, DecodingBuffer, texData, palData, color0Transparent); break; case 3: ConvertNColorsTexture<outputFmt_RGB6A5, 4>(width, height, DecodingBuffer, addr, palAddr, color0Transparent, gpu); break;
case 4: ConvertNColorsTexture<outputFmt_RGB6A5, 8>(width, height, DecodingBuffer, texData, palData, color0Transparent); break; case 4: ConvertNColorsTexture<outputFmt_RGB6A5, 8>(width, height, DecodingBuffer, addr, palAddr, color0Transparent, gpu); break;
} }
} }
for (int i = 0; i < 2; i++) for (int i = 0; i < 2; i++)
{ {
if (entry.TextureRAMSize[i]) if (entry.TextureRAMSize[i])
entry.TextureHash[i] = XXH3_64bits(&gpu.VRAMFlat_Texture[entry.TextureRAMStart[i]], entry.TextureRAMSize[i]); entry.TextureHash[i] = MaskedHash(gpu.VRAMFlat_Texture, sizeof(gpu.VRAMFlat_Texture),
entry.TextureRAMStart[i], entry.TextureRAMSize[i]);
} }
if (entry.TexPalSize) if (entry.TexPalSize)
entry.TexPalHash = XXH3_64bits(&gpu.VRAMFlat_TexPal[entry.TexPalStart], entry.TexPalSize); entry.TexPalHash = MaskedHash(gpu.VRAMFlat_TexPal, sizeof(gpu.VRAMFlat_TexPal),
entry.TexPalStart, entry.TexPalSize);
auto& texArrays = TexArrays[widthLog2][heightLog2]; auto& texArrays = TexArrays[widthLog2][heightLog2];
auto& freeTextures = FreeTextures[widthLog2][heightLog2]; auto& freeTextures = FreeTextures[widthLog2][heightLog2];