void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, u32 dstStride, PEControl::PixelFormat srcFormat,
const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf)
{
// Emulation methods:
//
// - EFB to RAM:
// Encodes the requested EFB data at its native resolution to the emulated RAM using shaders.
// Load() decodes the data from there again (using TextureDecoder) if the EFB copy is being used as a texture again.
// Advantage: CPU can read data from the EFB copy and we don't lose any important updates to the texture
// Disadvantage: Encoding+decoding steps often are redundant because only some games read or modify EFB copies before using them as textures.
//
// - EFB to texture:
// Copies the requested EFB data to a texture object in VRAM, performing any color conversion using shaders.
// Advantage: Works for many games, since in most cases EFB copies aren't read or modified at all before being used as a texture again.
// Since we don't do any further encoding or decoding here, this method is much faster.
// It also allows enhancing the visual quality by doing scaled EFB copies.
//
// - Hybrid EFB copies:
// 1a) Whenever this function gets called, encode the requested EFB data to RAM (like EFB to RAM)
// 1b) Set type to TCET_EC_DYNAMIC for all texture cache entries in the destination address range.
// If EFB copy caching is enabled, further checks will (try to) prevent redundant EFB copies.
// 2) Check if a texture cache entry for the specified dstAddr already exists (i.e. if an EFB copy was triggered to that address before):
// 2a) Entry doesn't exist:
// - Also copy the requested EFB data to a texture object in VRAM (like EFB to texture)
// - Create a texture cache entry for the target (type = TCET_EC_VRAM)
// - Store a hash of the encoded RAM data in the texcache entry.
// 2b) Entry exists AND type is TCET_EC_VRAM:
// - Like case 2a, but reuse the old texcache entry instead of creating a new one.
// 2c) Entry exists AND type is TCET_EC_DYNAMIC:
// - Only encode the texture to RAM (like EFB to RAM) and store a hash of the encoded data in the existing texcache entry.
// - Do NOT copy the requested EFB data to a VRAM object. Reason: the texture is dynamic, i.e. the CPU is modifying it. Storing a VRAM copy is useless, because we'd always end up deleting it and reloading the data from RAM anyway.
// 3) If the EFB copy gets used as a texture, compare the source RAM hash with the hash you stored when encoding the EFB data to RAM.
// 3a) If the two hashes match AND type is TCET_EC_VRAM, reuse the VRAM copy you created
// 3b) If the two hashes differ AND type is TCET_EC_VRAM, screw your existing VRAM copy. Set type to TCET_EC_DYNAMIC.
// Redecode the source RAM data to a VRAM object. The entry basically behaves like a normal texture now.
// 3c) If type is TCET_EC_DYNAMIC, treat the EFB copy like a normal texture.
// Advantage: Non-dynamic EFB copies can be visually enhanced like with EFB to texture.
// Compatibility is as good as EFB to RAM.
// Disadvantage: Slower than EFB to texture and often even slower than EFB to RAM.
// EFB copy cache depends on accurate texture hashing being enabled. However, with accurate hashing you end up being as slow as without a copy cache anyway.
//
// Disadvantage of all methods: Calling this function requires the GPU to perform a pipeline flush which stalls any further CPU processing.
//
// For historical reasons, Dolphin doesn't actually implement "pure" EFB to RAM emulation, but only EFB to texture and hybrid EFB copies.
float colmat[28] = { 0 };
float *const fConstAdd = colmat + 16;
float *const ColorMask = colmat + 20;
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 255.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 255.0f;
unsigned int cbufid = -1;
bool efbHasAlpha = bpmem.zcontrol.pixel_format == PEControl::RGBA6_Z24;
if (srcFormat == PEControl::Z24)
{
switch (dstFormat)
{
case 0: // Z4
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 0;
dstFormat |= _GX_TF_CTF;
break;
case 8: // Z8H
dstFormat |= _GX_TF_CTF;
case 1: // Z8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1.0f;
cbufid = 1;
break;
case 3: // Z16
colmat[1] = colmat[5] = colmat[9] = colmat[12] = 1.0f;
cbufid = 2;
break;
case 11: // Z16 (reverse order)
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 3;
dstFormat |= _GX_TF_CTF;
break;
case 6: // Z24X8
colmat[0] = colmat[5] = colmat[10] = 1.0f;
cbufid = 4;
break;
case 9: // Z8M
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 5;
dstFormat |= _GX_TF_CTF;
break;
case 10: // Z8L
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 6;
dstFormat |= _GX_TF_CTF;
break;
case 12: // Z16L - copy lower 16 depth bits
// expected to be used as an IA8 texture (upper 8 bits stored as intensity, lower 8 bits stored as alpha)
// Used e.g. in Zelda: Skyward Sword
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 7;
dstFormat |= _GX_TF_CTF;
break;
default:
ERROR_LOG(VIDEO, "Unknown copy zbuf format: 0x%x", dstFormat);
colmat[2] = colmat[5] = colmat[8] = 1.0f;
cbufid = 8;
break;
}
dstFormat |= _GX_TF_ZTF;
}
else if (isIntensity)
{
fConstAdd[0] = fConstAdd[1] = fConstAdd[2] = 16.0f / 255.0f;
switch (dstFormat)
{
case 0: // I4
case 1: // I8
case 2: // IA4
case 3: // IA8
case 8: // I8
// TODO - verify these coefficients
colmat[0] = 0.257f; colmat[1] = 0.504f; colmat[2] = 0.098f;
colmat[4] = 0.257f; colmat[5] = 0.504f; colmat[6] = 0.098f;
colmat[8] = 0.257f; colmat[9] = 0.504f; colmat[10] = 0.098f;
if (dstFormat < 2 || dstFormat == 8)
{
colmat[12] = 0.257f; colmat[13] = 0.504f; colmat[14] = 0.098f;
fConstAdd[3] = 16.0f / 255.0f;
if (dstFormat == 0)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 15.0f;
cbufid = 9;
}
else
{
cbufid = 10;
}
}
else// alpha
{
colmat[15] = 1;
if (dstFormat == 2)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 11;
}
else
{
cbufid = 12;
}
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy intensity format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 13;
break;
}
}
else
{
switch (dstFormat)
{
case 0: // R4
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
ColorMask[0] = 15.0f;
ColorMask[4] = 1.0f / 15.0f;
cbufid = 14;
dstFormat |= _GX_TF_CTF;
break;
case 1: // R8
case 8: // R8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
cbufid = 15;
dstFormat |= _GX_TF_CTF;
break;
case 2: // RA4
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 16;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 17;
}
dstFormat |= _GX_TF_CTF;
break;
case 3: // RA8
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
cbufid = 18;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 19;
}
dstFormat |= _GX_TF_CTF;
break;
case 7: // A8
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 20;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[0] = 1.0f;
fConstAdd[1] = 1.0f;
fConstAdd[2] = 1.0f;
fConstAdd[3] = 1.0f;
cbufid = 21;
}
dstFormat |= _GX_TF_CTF;
break;
case 9: // G8
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 22;
dstFormat |= _GX_TF_CTF;
break;
case 10: // B8
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 23;
dstFormat |= _GX_TF_CTF;
break;
case 11: // RG8
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 24;
dstFormat |= _GX_TF_CTF;
break;
case 12: // GB8
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 25;
dstFormat |= _GX_TF_CTF;
break;
case 4: // RGB565
colmat[0] = colmat[5] = colmat[10] = 1.0f;
ColorMask[0] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[1] = 63.0f;
ColorMask[5] = 1.0f / 63.0f;
fConstAdd[3] = 1.0f; // set alpha to 1
cbufid = 26;
break;
case 5: // RGB5A3
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[1] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[3] = 7.0f;
ColorMask[7] = 1.0f / 7.0f;
cbufid = 27;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 28;
}
break;
case 6: // RGBA8
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 29;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 30;
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy color format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 31;
break;
}
}
u8* dst = Memory::GetPointer(dstAddr);
if (dst == nullptr)
{
ERROR_LOG(VIDEO, "Trying to copy from EFB to invalid address 0x%8x", dstAddr);
return;
}
const unsigned int tex_w = scaleByHalf ? srcRect.GetWidth() / 2 : srcRect.GetWidth();
const unsigned int tex_h = scaleByHalf ? srcRect.GetHeight() / 2 : srcRect.GetHeight();
unsigned int scaled_tex_w = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledX(tex_w) : tex_w;
unsigned int scaled_tex_h = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledY(tex_h) : tex_h;
// remove all texture cache entries at dstAddr
{
std::pair<TexCache::iterator, TexCache::iterator> iter_range = textures_by_address.equal_range((u64)dstAddr);
TexCache::iterator iter = iter_range.first;
while (iter != iter_range.second)
{
iter = FreeTexture(iter);
}
}
// create the texture
TCacheEntryConfig config;
config.rendertarget = true;
config.width = scaled_tex_w;
config.height = scaled_tex_h;
config.layers = FramebufferManagerBase::GetEFBLayers();
TCacheEntryBase* entry = AllocateTexture(config);
entry->SetGeneralParameters(dstAddr, 0, dstFormat);
entry->SetDimensions(tex_w, tex_h, 1);
entry->frameCount = FRAMECOUNT_INVALID;
entry->SetEfbCopy(dstStride);
entry->is_custom_tex = false;
entry->FromRenderTarget(dst, dstFormat, dstStride, srcFormat, srcRect, isIntensity, scaleByHalf, cbufid, colmat);
u64 hash = entry->CalculateHash();
entry->SetHashes(hash, hash);
// Invalidate all textures that overlap the range of our efb copy.
// Unless our efb copy has a weird stride, then we want avoid invalidating textures which
// we might be able to do a partial texture update on.
if (entry->memory_stride == entry->CacheLinesPerRow() * 32)
{
TexCache::iterator iter = textures_by_address.begin();
while (iter != textures_by_address.end())
{
if (iter->second->OverlapsMemoryRange(dstAddr, entry->size_in_bytes))
iter = FreeTexture(iter);
else
++iter;
}
}
if (g_ActiveConfig.bDumpEFBTarget)
{
static int count = 0;
entry->Save(StringFromFormat("%sefb_frame_%i.png", File::GetUserPath(D_DUMPTEXTURES_IDX).c_str(),
count++), 0);
}
if (g_bRecordFifoData)
{
// Mark the memory behind this efb copy as dynamicly generated for the Fifo log
u32 address = dstAddr;
for (u32 i = 0; i < entry->NumBlocksY(); i++)
{
FifoRecorder::GetInstance().UseMemory(address, entry->CacheLinesPerRow() * 32, MemoryUpdate::TEXTURE_MAP, true);
address += entry->memory_stride;
}
}
textures_by_address.emplace((u64)dstAddr, entry);
}
void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, PEControl::PixelFormat srcFormat,
const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf)
{
// Emulation methods:
//
// - EFB to RAM:
// Encodes the requested EFB data at its native resolution to the emulated RAM using shaders.
// Load() decodes the data from there again (using TextureDecoder) if the EFB copy is being used as a texture again.
// Advantage: CPU can read data from the EFB copy and we don't lose any important updates to the texture
// Disadvantage: Encoding+decoding steps often are redundant because only some games read or modify EFB copies before using them as textures.
//
// - EFB to texture:
// Copies the requested EFB data to a texture object in VRAM, performing any color conversion using shaders.
// Advantage: Works for many games, since in most cases EFB copies aren't read or modified at all before being used as a texture again.
// Since we don't do any further encoding or decoding here, this method is much faster.
// It also allows enhancing the visual quality by doing scaled EFB copies.
//
// - Hybrid EFB copies:
// 1a) Whenever this function gets called, encode the requested EFB data to RAM (like EFB to RAM)
// 1b) Set type to TCET_EC_DYNAMIC for all texture cache entries in the destination address range.
// If EFB copy caching is enabled, further checks will (try to) prevent redundant EFB copies.
// 2) Check if a texture cache entry for the specified dstAddr already exists (i.e. if an EFB copy was triggered to that address before):
// 2a) Entry doesn't exist:
// - Also copy the requested EFB data to a texture object in VRAM (like EFB to texture)
// - Create a texture cache entry for the target (type = TCET_EC_VRAM)
// - Store a hash of the encoded RAM data in the texcache entry.
// 2b) Entry exists AND type is TCET_EC_VRAM:
// - Like case 2a, but reuse the old texcache entry instead of creating a new one.
// 2c) Entry exists AND type is TCET_EC_DYNAMIC:
// - Only encode the texture to RAM (like EFB to RAM) and store a hash of the encoded data in the existing texcache entry.
// - Do NOT copy the requested EFB data to a VRAM object. Reason: the texture is dynamic, i.e. the CPU is modifying it. Storing a VRAM copy is useless, because we'd always end up deleting it and reloading the data from RAM anyway.
// 3) If the EFB copy gets used as a texture, compare the source RAM hash with the hash you stored when encoding the EFB data to RAM.
// 3a) If the two hashes match AND type is TCET_EC_VRAM, reuse the VRAM copy you created
// 3b) If the two hashes differ AND type is TCET_EC_VRAM, screw your existing VRAM copy. Set type to TCET_EC_DYNAMIC.
// Redecode the source RAM data to a VRAM object. The entry basically behaves like a normal texture now.
// 3c) If type is TCET_EC_DYNAMIC, treat the EFB copy like a normal texture.
// Advantage: Non-dynamic EFB copies can be visually enhanced like with EFB to texture.
// Compatibility is as good as EFB to RAM.
// Disadvantage: Slower than EFB to texture and often even slower than EFB to RAM.
// EFB copy cache depends on accurate texture hashing being enabled. However, with accurate hashing you end up being as slow as without a copy cache anyway.
//
// Disadvantage of all methods: Calling this function requires the GPU to perform a pipeline flush which stalls any further CPU processing.
//
// For historical reasons, Dolphin doesn't actually implement "pure" EFB to RAM emulation, but only EFB to texture and hybrid EFB copies.
float colmat[28] = {0};
float *const fConstAdd = colmat + 16;
float *const ColorMask = colmat + 20;
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 255.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 255.0f;
unsigned int cbufid = -1;
bool efbHasAlpha = bpmem.zcontrol.pixel_format == PEControl::RGBA6_Z24;
if (srcFormat == PEControl::Z24)
{
switch (dstFormat)
{
case 0: // Z4
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 0;
break;
case 1: // Z8
case 8: // Z8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1.0f;
cbufid = 1;
break;
case 3: // Z16
colmat[1] = colmat[5] = colmat[9] = colmat[12] = 1.0f;
cbufid = 2;
break;
case 11: // Z16 (reverse order)
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 3;
break;
case 6: // Z24X8
colmat[0] = colmat[5] = colmat[10] = 1.0f;
cbufid = 4;
break;
case 9: // Z8M
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 5;
break;
case 10: // Z8L
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 6;
break;
case 12: // Z16L - copy lower 16 depth bits
// expected to be used as an IA8 texture (upper 8 bits stored as intensity, lower 8 bits stored as alpha)
// Used e.g. in Zelda: Skyward Sword
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 7;
break;
default:
ERROR_LOG(VIDEO, "Unknown copy zbuf format: 0x%x", dstFormat);
colmat[2] = colmat[5] = colmat[8] = 1.0f;
cbufid = 8;
break;
}
}
else if (isIntensity)
{
fConstAdd[0] = fConstAdd[1] = fConstAdd[2] = 16.0f/255.0f;
switch (dstFormat)
{
case 0: // I4
case 1: // I8
case 2: // IA4
case 3: // IA8
case 8: // I8
// TODO - verify these coefficients
colmat[0] = 0.257f; colmat[1] = 0.504f; colmat[2] = 0.098f;
colmat[4] = 0.257f; colmat[5] = 0.504f; colmat[6] = 0.098f;
colmat[8] = 0.257f; colmat[9] = 0.504f; colmat[10] = 0.098f;
if (dstFormat < 2 || dstFormat == 8)
{
colmat[12] = 0.257f; colmat[13] = 0.504f; colmat[14] = 0.098f;
fConstAdd[3] = 16.0f/255.0f;
if (dstFormat == 0)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 15.0f;
cbufid = 9;
}
else
{
cbufid = 10;
}
}
else// alpha
{
colmat[15] = 1;
if (dstFormat == 2)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 11;
}
else
{
cbufid = 12;
}
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy intensity format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 13;
break;
}
}
else
{
switch (dstFormat)
{
case 0: // R4
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
ColorMask[0] = 15.0f;
ColorMask[4] = 1.0f / 15.0f;
cbufid = 14;
break;
case 1: // R8
case 8: // R8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
cbufid = 15;
break;
case 2: // RA4
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 16;
if (!efbHasAlpha) {
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 17;
}
break;
case 3: // RA8
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
cbufid = 18;
if (!efbHasAlpha) {
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 19;
}
break;
case 7: // A8
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 20;
if (!efbHasAlpha) {
ColorMask[3] = 0.0f;
fConstAdd[0] = 1.0f;
fConstAdd[1] = 1.0f;
fConstAdd[2] = 1.0f;
fConstAdd[3] = 1.0f;
cbufid = 21;
}
break;
case 9: // G8
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 22;
break;
case 10: // B8
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 23;
break;
case 11: // RG8
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 24;
break;
case 12: // GB8
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 25;
break;
case 4: // RGB565
colmat[0] = colmat[5] = colmat[10] = 1.0f;
ColorMask[0] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[1] = 63.0f;
ColorMask[5] = 1.0f / 63.0f;
fConstAdd[3] = 1.0f; // set alpha to 1
cbufid = 26;
break;
case 5: // RGB5A3
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[1] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[3] = 7.0f;
ColorMask[7] = 1.0f / 7.0f;
cbufid = 27;
if (!efbHasAlpha) {
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 28;
}
break;
case 6: // RGBA8
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 29;
if (!efbHasAlpha) {
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 30;
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy color format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 31;
break;
}
}
const unsigned int tex_w = scaleByHalf ? srcRect.GetWidth()/2 : srcRect.GetWidth();
const unsigned int tex_h = scaleByHalf ? srcRect.GetHeight()/2 : srcRect.GetHeight();
unsigned int scaled_tex_w = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledX(tex_w) : tex_w;
unsigned int scaled_tex_h = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledY(tex_h) : tex_h;
TCacheEntryBase *entry = textures[dstAddr];
if (entry)
{
if (entry->type == TCET_EC_DYNAMIC && entry->native_width == tex_w && entry->native_height == tex_h)
{
scaled_tex_w = tex_w;
scaled_tex_h = tex_h;
}
else if (!(entry->type == TCET_EC_VRAM && entry->virtual_width == scaled_tex_w && entry->virtual_height == scaled_tex_h))
{
// remove it and recreate it as a render target
delete entry;
entry = nullptr;
}
}
if (nullptr == entry)
{
// create the texture
textures[dstAddr] = entry = g_texture_cache->CreateRenderTargetTexture(scaled_tex_w, scaled_tex_h);
// TODO: Using the wrong dstFormat, dumb...
entry->SetGeneralParameters(dstAddr, 0, dstFormat, 1);
entry->SetDimensions(tex_w, tex_h, scaled_tex_w, scaled_tex_h);
entry->SetHashes(TEXHASH_INVALID);
entry->type = TCET_EC_VRAM;
}
entry->frameCount = frameCount;
entry->FromRenderTarget(dstAddr, dstFormat, srcFormat, srcRect, isIntensity, scaleByHalf, cbufid, colmat);
}
void TextureCacheBase::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, u32 dstStride, PEControl::PixelFormat srcFormat,
const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf)
{
// Emulation methods:
//
// - EFB to RAM:
// Encodes the requested EFB data at its native resolution to the emulated RAM using shaders.
// Load() decodes the data from there again (using TextureDecoder) if the EFB copy is being used as a texture again.
// Advantage: CPU can read data from the EFB copy and we don't lose any important updates to the texture
// Disadvantage: Encoding+decoding steps often are redundant because only some games read or modify EFB copies before using them as textures.
//
// - EFB to texture:
// Copies the requested EFB data to a texture object in VRAM, performing any color conversion using shaders.
// Advantage: Works for many games, since in most cases EFB copies aren't read or modified at all before being used as a texture again.
// Since we don't do any further encoding or decoding here, this method is much faster.
// It also allows enhancing the visual quality by doing scaled EFB copies.
//
// - Hybrid EFB copies:
// 1a) Whenever this function gets called, encode the requested EFB data to RAM (like EFB to RAM)
// 1b) Set type to TCET_EC_DYNAMIC for all texture cache entries in the destination address range.
// If EFB copy caching is enabled, further checks will (try to) prevent redundant EFB copies.
// 2) Check if a texture cache entry for the specified dstAddr already exists (i.e. if an EFB copy was triggered to that address before):
// 2a) Entry doesn't exist:
// - Also copy the requested EFB data to a texture object in VRAM (like EFB to texture)
// - Create a texture cache entry for the target (type = TCET_EC_VRAM)
// - Store a hash of the encoded RAM data in the texcache entry.
// 2b) Entry exists AND type is TCET_EC_VRAM:
// - Like case 2a, but reuse the old texcache entry instead of creating a new one.
// 2c) Entry exists AND type is TCET_EC_DYNAMIC:
// - Only encode the texture to RAM (like EFB to RAM) and store a hash of the encoded data in the existing texcache entry.
// - Do NOT copy the requested EFB data to a VRAM object. Reason: the texture is dynamic, i.e. the CPU is modifying it. Storing a VRAM copy is useless, because we'd always end up deleting it and reloading the data from RAM anyway.
// 3) If the EFB copy gets used as a texture, compare the source RAM hash with the hash you stored when encoding the EFB data to RAM.
// 3a) If the two hashes match AND type is TCET_EC_VRAM, reuse the VRAM copy you created
// 3b) If the two hashes differ AND type is TCET_EC_VRAM, screw your existing VRAM copy. Set type to TCET_EC_DYNAMIC.
// Redecode the source RAM data to a VRAM object. The entry basically behaves like a normal texture now.
// 3c) If type is TCET_EC_DYNAMIC, treat the EFB copy like a normal texture.
// Advantage: Non-dynamic EFB copies can be visually enhanced like with EFB to texture.
// Compatibility is as good as EFB to RAM.
// Disadvantage: Slower than EFB to texture and often even slower than EFB to RAM.
// EFB copy cache depends on accurate texture hashing being enabled. However, with accurate hashing you end up being as slow as without a copy cache anyway.
//
// Disadvantage of all methods: Calling this function requires the GPU to perform a pipeline flush which stalls any further CPU processing.
//
// For historical reasons, Dolphin doesn't actually implement "pure" EFB to RAM emulation, but only EFB to texture and hybrid EFB copies.
float colmat[28] = { 0 };
float *const fConstAdd = colmat + 16;
float *const ColorMask = colmat + 20;
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 255.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 255.0f;
unsigned int cbufid = -1;
bool efbHasAlpha = bpmem.zcontrol.pixel_format == PEControl::RGBA6_Z24;
if (srcFormat == PEControl::Z24)
{
switch (dstFormat)
{
case 0: // Z4
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 0;
dstFormat |= _GX_TF_CTF;
break;
case 8: // Z8H
dstFormat |= _GX_TF_CTF;
case 1: // Z8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1.0f;
cbufid = 1;
break;
case 3: // Z16
colmat[1] = colmat[5] = colmat[9] = colmat[12] = 1.0f;
cbufid = 2;
break;
case 11: // Z16 (reverse order)
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 3;
dstFormat |= _GX_TF_CTF;
break;
case 6: // Z24X8
colmat[0] = colmat[5] = colmat[10] = 1.0f;
cbufid = 4;
break;
case 9: // Z8M
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 5;
dstFormat |= _GX_TF_CTF;
break;
case 10: // Z8L
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 6;
dstFormat |= _GX_TF_CTF;
break;
case 12: // Z16L - copy lower 16 depth bits
// expected to be used as an IA8 texture (upper 8 bits stored as intensity, lower 8 bits stored as alpha)
// Used e.g. in Zelda: Skyward Sword
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 7;
dstFormat |= _GX_TF_CTF;
break;
default:
ERROR_LOG(VIDEO, "Unknown copy zbuf format: 0x%x", dstFormat);
colmat[2] = colmat[5] = colmat[8] = 1.0f;
cbufid = 8;
break;
}
dstFormat |= _GX_TF_ZTF;
}
else if (isIntensity)
{
fConstAdd[0] = fConstAdd[1] = fConstAdd[2] = 16.0f / 255.0f;
switch (dstFormat)
{
case 0: // I4
case 1: // I8
case 2: // IA4
case 3: // IA8
case 8: // I8
// TODO - verify these coefficients
colmat[0] = 0.257f; colmat[1] = 0.504f; colmat[2] = 0.098f;
colmat[4] = 0.257f; colmat[5] = 0.504f; colmat[6] = 0.098f;
colmat[8] = 0.257f; colmat[9] = 0.504f; colmat[10] = 0.098f;
if (dstFormat < 2 || dstFormat == 8)
{
colmat[12] = 0.257f; colmat[13] = 0.504f; colmat[14] = 0.098f;
fConstAdd[3] = 16.0f / 255.0f;
if (dstFormat == 0)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 15.0f;
cbufid = 9;
}
else
{
cbufid = 10;
}
}
else// alpha
{
colmat[15] = 1;
if (dstFormat == 2)
{
ColorMask[0] = ColorMask[1] = ColorMask[2] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 11;
}
else
{
cbufid = 12;
}
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy intensity format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 13;
break;
}
}
else
{
switch (dstFormat)
{
case 0: // R4
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
ColorMask[0] = 15.0f;
ColorMask[4] = 1.0f / 15.0f;
cbufid = 14;
dstFormat |= _GX_TF_CTF;
break;
case 1: // R8
case 8: // R8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
cbufid = 15;
dstFormat = GX_CTF_R8;
break;
case 2: // RA4
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[3] = 15.0f;
ColorMask[4] = ColorMask[7] = 1.0f / 15.0f;
cbufid = 16;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 17;
}
dstFormat |= _GX_TF_CTF;
break;
case 3: // RA8
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
cbufid = 18;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 19;
}
dstFormat |= _GX_TF_CTF;
break;
case 7: // A8
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 20;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[0] = 1.0f;
fConstAdd[1] = 1.0f;
fConstAdd[2] = 1.0f;
fConstAdd[3] = 1.0f;
cbufid = 21;
}
dstFormat |= _GX_TF_CTF;
break;
case 9: // G8
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 22;
dstFormat |= _GX_TF_CTF;
break;
case 10: // B8
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 23;
dstFormat |= _GX_TF_CTF;
break;
case 11: // RG8
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 24;
dstFormat |= _GX_TF_CTF;
break;
case 12: // GB8
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 25;
dstFormat |= _GX_TF_CTF;
break;
case 4: // RGB565
colmat[0] = colmat[5] = colmat[10] = 1.0f;
ColorMask[0] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[1] = 63.0f;
ColorMask[5] = 1.0f / 63.0f;
fConstAdd[3] = 1.0f; // set alpha to 1
cbufid = 26;
break;
case 5: // RGB5A3
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
ColorMask[0] = ColorMask[1] = ColorMask[2] = 31.0f;
ColorMask[4] = ColorMask[5] = ColorMask[6] = 1.0f / 31.0f;
ColorMask[3] = 7.0f;
ColorMask[7] = 1.0f / 7.0f;
cbufid = 27;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 28;
}
break;
case 6: // RGBA8
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 29;
if (!efbHasAlpha)
{
ColorMask[3] = 0.0f;
fConstAdd[3] = 1.0f;
cbufid = 30;
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy color format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 31;
break;
}
}
u8* dst = Memory::GetPointer(dstAddr);
if (dst == nullptr)
{
ERROR_LOG(VIDEO, "Trying to copy from EFB to invalid address 0x%8x", dstAddr);
return;
}
const unsigned int tex_w = scaleByHalf ? srcRect.GetWidth() / 2 : srcRect.GetWidth();
const unsigned int tex_h = scaleByHalf ? srcRect.GetHeight() / 2 : srcRect.GetHeight();
unsigned int scaled_tex_w = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledX(tex_w) : tex_w;
unsigned int scaled_tex_h = g_ActiveConfig.bCopyEFBScaled ? Renderer::EFBToScaledY(tex_h) : tex_h;
// Remove all texture cache entries at dstAddr
// It's not possible to have two EFB copies at the same address, this makes sure any old efb copies
// (or normal textures) are removed from texture cache. They are also un-linked from any partially
// updated textures, which forces that partially updated texture to be updated.
// TODO: This also wipes out non-efb copies, which is counterproductive.
{
std::pair<TexCache::iterator, TexCache::iterator> iter_range = textures_by_address.equal_range((u64)dstAddr);
TexCache::iterator iter = iter_range.first;
while (iter != iter_range.second)
{
iter = InvalidateTexture(iter);
}
}
// Get the base (in memory) format of this efb copy.
int baseFormat = TexDecoder_GetEfbCopyBaseFormat(dstFormat);
u32 blockH = TexDecoder_GetBlockHeightInTexels(baseFormat);
const u32 blockW = TexDecoder_GetBlockWidthInTexels(baseFormat);
// Round up source height to multiple of block size
u32 actualHeight = ROUND_UP(tex_h, blockH);
const u32 actualWidth = ROUND_UP(tex_w, blockW);
u32 num_blocks_y = actualHeight / blockH;
const u32 num_blocks_x = actualWidth / blockW;
// RGBA takes two cache lines per block; all others take one
const u32 bytes_per_block = baseFormat == GX_TF_RGBA8 ? 64 : 32;
u32 bytes_per_row = num_blocks_x * bytes_per_block;
bool copy_to_ram = !g_ActiveConfig.bSkipEFBCopyToRam;
bool copy_to_vram = true;
if (copy_to_ram)
{
g_texture_cache->CopyEFB(
dst,
dstFormat,
tex_w,
bytes_per_row,
num_blocks_y,
dstStride,
srcFormat,
srcRect,
isIntensity,
scaleByHalf);
}
else
{
// Hack: Most games don't actually need the correct texture data in RAM
// and we can just keep a copy in VRAM. We zero the memory so we
// can check it hasn't changed before using our copy in VRAM.
u8* ptr = dst;
for (u32 i = 0; i < num_blocks_y; i++)
{
memset(ptr, 0, bytes_per_row);
ptr += dstStride;
}
}
if (g_bRecordFifoData)
{
// Mark the memory behind this efb copy as dynamicly generated for the Fifo log
u32 address = dstAddr;
for (u32 i = 0; i < num_blocks_y; i++)
{
FifoRecorder::GetInstance().UseMemory(address, bytes_per_row, MemoryUpdate::TEXTURE_MAP, true);
address += dstStride;
}
}
if (dstStride < bytes_per_row)
{
// This kind of efb copy results in a scrambled image.
// I'm pretty sure no game actually wants to do this, it might be caused by a
// programming bug in the game, or a CPU/Bounding box emulation issue with dolphin.
// The copy_to_ram code path above handles this "correctly" and scrambles the image
// but the copy_to_vram code path just saves and uses unscrambled texture instead.
// To avoid a "incorrect" result, we simply skip doing the copy_to_vram code path
// so if the game does try to use the scrambled texture, dolphin will grab the scrambled
// texture (or black if copy_to_ram is also disabled) out of ram.
ERROR_LOG(VIDEO, "Memory stride too small (%i < %i)", dstStride, bytes_per_row);
copy_to_vram = false;
}
// Invalidate all textures that overlap the range of our efb copy.
// Unless our efb copy has a weird stride, then we want avoid invalidating textures which
// we might be able to do a partial texture update on.
// TODO: This also invalidates partial overlaps, which we currently don't have a better way
// of dealing with.
if (dstStride == bytes_per_row || !copy_to_vram)
{
TexCache::iterator iter = textures_by_address.begin();
while (iter != textures_by_address.end())
{
if (iter->second->addr + iter->second->size_in_bytes <= dstAddr || iter->second->addr >= dstAddr + num_blocks_y * dstStride)
++iter;
else
iter = InvalidateTexture(iter);
}
}
if (copy_to_vram)
{
// create the texture
TCacheEntryConfig config;
config.rendertarget = true;
config.width = scaled_tex_w;
config.height = scaled_tex_h;
config.layers = FramebufferManagerBase::GetEFBLayers();
TCacheEntryBase* entry = AllocateTexture(config);
if (entry)
{
entry->SetGeneralParameters(dstAddr, 0, baseFormat);
entry->SetDimensions(tex_w, tex_h, 1);
entry->frameCount = FRAMECOUNT_INVALID;
entry->SetEfbCopy(dstStride);
entry->is_custom_tex = false;
entry->FromRenderTarget(dst, srcFormat, srcRect, scaleByHalf, cbufid, colmat);
u64 hash = entry->CalculateHash();
entry->SetHashes(hash, hash);
if (g_ActiveConfig.bDumpEFBTarget)
{
static int count = 0;
entry->Save(StringFromFormat("%sefb_frame_%i.png", File::GetUserPath(D_DUMPTEXTURES_IDX).c_str(),
count++), 0);
}
textures_by_address.emplace((u64)dstAddr, entry);
}
}
}