ShuriZma
/
dolphin
mirror of https://github.com/dolphin-emu/dolphin.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
dolphin/Source/Core/VideoCommon/Src/TextureCacheBase.cpp

798 lines
27 KiB
C++
Raw Blame History

// Copyright (C) 2003 Dolphin Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#include "MemoryUtil.h"
#include "VideoConfig.h"
#include "Statistics.h"
#include "HiresTextures.h"
#include "RenderBase.h"
#include "FileUtil.h"
#include "TextureCacheBase.h"
#include "Debugger.h"
#include "ConfigManager.h"
#include "HW/Memmap.h"
// ugly
extern int frameCount;
enum
{
TEXTURE_KILL_THRESHOLD = 200,
};
TextureCache *g_texture_cache;
GC_ALIGNED16(u8 *TextureCache::temp) = NULL;
unsigned int TextureCache::temp_size;
TextureCache::TexCache TextureCache::textures;
TextureCache::BackupConfig TextureCache::backup_config;
TextureCache::TCacheEntryBase::~TCacheEntryBase()
{
}
TextureCache::TextureCache()
{
temp_size = 2048 * 2048 * 4;
if (!temp)
temp = (u8*)AllocateAlignedMemory(temp_size, 16);
TexDecoder_SetTexFmtOverlayOptions(g_ActiveConfig.bTexFmtOverlayEnable, g_ActiveConfig.bTexFmtOverlayCenter);
if(g_ActiveConfig.bHiresTextures && !g_ActiveConfig.bDumpTextures)
HiresTextures::Init(SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
SetHash64Function(g_ActiveConfig.bHiresTextures || g_ActiveConfig.bDumpTextures);
}
void TextureCache::Invalidate()
{
TexCache::iterator
iter = textures.begin(),
tcend = textures.end();
for (; iter != tcend; ++iter)
delete iter->second;
textures.clear();
}
TextureCache::~TextureCache()
{
Invalidate();
if (temp)
{
FreeAlignedMemory(temp);
temp = NULL;
}
}
void TextureCache::OnConfigChanged(VideoConfig& config)
{
if (!g_texture_cache)
goto skip_checks;
// TODO: Invalidating texcache is really stupid in some of these cases
if (config.iSafeTextureCache_ColorSamples != backup_config.s_colorsamples ||
config.bTexFmtOverlayEnable != backup_config.s_texfmt_overlay ||
config.bTexFmtOverlayCenter != backup_config.s_texfmt_overlay_center ||
config.bHiresTextures != backup_config.s_hires_textures)
{
g_texture_cache->Invalidate();
if(g_ActiveConfig.bHiresTextures)
HiresTextures::Init(SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str());
SetHash64Function(g_ActiveConfig.bHiresTextures || g_ActiveConfig.bDumpTextures);
TexDecoder_SetTexFmtOverlayOptions(g_ActiveConfig.bTexFmtOverlayEnable, g_ActiveConfig.bTexFmtOverlayCenter);
}
// TODO: Probably shouldn't clear all render targets here, just mark them dirty or something.
if (config.bEFBCopyCacheEnable != backup_config.s_copy_cache_enable || // TODO: not sure if this is needed?
config.bCopyEFBToTexture != backup_config.s_copy_efb_to_texture ||
config.bCopyEFBScaled != backup_config.s_copy_efb_scaled ||
config.bEFBCopyEnable != backup_config.s_copy_efb ||
config.iEFBScale != backup_config.s_efb_scale)
{
g_texture_cache->ClearRenderTargets();
}
skip_checks:
backup_config.s_colorsamples = config.iSafeTextureCache_ColorSamples;
backup_config.s_copy_efb_to_texture = config.bCopyEFBToTexture;
backup_config.s_copy_efb_scaled = config.bCopyEFBScaled;
backup_config.s_copy_efb = config.bEFBCopyEnable;
backup_config.s_efb_scale = config.iEFBScale;
backup_config.s_texfmt_overlay = config.bTexFmtOverlayEnable;
backup_config.s_texfmt_overlay_center = config.bTexFmtOverlayCenter;
backup_config.s_hires_textures = config.bHiresTextures;
backup_config.s_copy_cache_enable = config.bEFBCopyCacheEnable;
}
void TextureCache::Cleanup()
{
TexCache::iterator iter = textures.begin();
TexCache::iterator tcend = textures.end();
while (iter != tcend)
{
if (frameCount > TEXTURE_KILL_THRESHOLD + iter->second->frameCount) // TODO: Deleting EFB copies might not be a good idea here...
{
delete iter->second;
textures.erase(iter++);
}
else
++iter;
}
}
void TextureCache::InvalidateRange(u32 start_address, u32 size)
{
TexCache::iterator
iter = textures.begin(),
tcend = textures.end();
while (iter != tcend)
{
const int rangePosition = iter->second->IntersectsMemoryRange(start_address, size);
if (0 == rangePosition)
{
delete iter->second;
textures.erase(iter++);
}
else
++iter;
}
}
void TextureCache::MakeRangeDynamic(u32 start_address, u32 size)
{
TexCache::iterator
iter = textures.lower_bound(start_address),
tcend = textures.upper_bound(start_address + size);
if (iter != textures.begin())
iter--;
for (; iter != tcend; ++iter)
{
const int rangePosition = iter->second->IntersectsMemoryRange(start_address, size);
if (0 == rangePosition)
{
iter->second->SetHashes(TEXHASH_INVALID);
}
}
}
bool TextureCache::Find(u32 start_address, u64 hash)
{
TexCache::iterator iter = textures.lower_bound(start_address);
if (iter->second->hash == hash)
return true;
return false;
}
int TextureCache::TCacheEntryBase::IntersectsMemoryRange(u32 range_address, u32 range_size) const
{
if (addr + size_in_bytes < range_address)
return -1;
if (addr >= range_address + range_size)
return 1;
return 0;
}
void TextureCache::ClearRenderTargets()
{
TexCache::iterator
iter = textures.begin(),
tcend = textures.end();
for (; iter!=tcend; ++iter)
if (iter->second->type != TCET_EC_DYNAMIC)
iter->second->type = TCET_NORMAL;
}
bool TextureCache::CheckForCustomTextureLODs(u64 tex_hash, int texformat, unsigned int levels)
{
if (levels == 1)
return false;
// Just checking if the necessary files exist, if they can't be loaded or have incorrect dimensions LODs will be black
char texBasePathTemp[MAX_PATH];
char texPathTemp[MAX_PATH];
sprintf(texBasePathTemp, "%s_%08x_%i", SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), (u32) (tex_hash & 0x00000000FFFFFFFFLL), texformat);
for (unsigned int level = 1; level < levels; ++level)
{
sprintf(texPathTemp, "%s_mip%i", texBasePathTemp, level);
if (!HiresTextures::HiresTexExists(texPathTemp))
{
if (level > 1)
WARN_LOG(VIDEO, "Couldn't find custom texture LOD with index %i (filename: %s), disabling custom LODs for this texture", level, texPathTemp);
return false;
}
}
return true;
}
PC_TexFormat TextureCache::LoadCustomTexture(u64 tex_hash, int texformat, unsigned int level, unsigned int& width, unsigned int& height)
{
char texPathTemp[MAX_PATH];
unsigned int newWidth = 0;
unsigned int newHeight = 0;
if (level == 0)
sprintf(texPathTemp, "%s_%08x_%i", SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), (u32) (tex_hash & 0x00000000FFFFFFFFLL), texformat);
else
sprintf(texPathTemp, "%s_%08x_%i_mip%i", SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), (u32) (tex_hash & 0x00000000FFFFFFFFLL), texformat, level);
unsigned int required_size = 0;
PC_TexFormat ret = HiresTextures::GetHiresTex(texPathTemp, &newWidth, &newHeight, &required_size, texformat, temp_size, temp);
if (ret == PC_TEX_FMT_NONE && temp_size < required_size)
{
// Allocate more memory and try again
// TODO: Should probably check if newWidth and newHeight are texture dimensions which are actually supported by the current video backend
temp_size = required_size;
FreeAlignedMemory(temp);
temp = (u8*)AllocateAlignedMemory(temp_size, 16);
ret = HiresTextures::GetHiresTex(texPathTemp, &newWidth, &newHeight, &required_size, texformat, temp_size, temp);
}
if (ret != PC_TEX_FMT_NONE)
{
width = newWidth;
height = newHeight;
}
return ret;
}
void TextureCache::DumpTexture(TCacheEntryBase* entry, unsigned int level)
{
char szTemp[MAX_PATH];
std::string szDir = File::GetUserPath(D_DUMPTEXTURES_IDX) +
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID;
// make sure that the directory exists
if (false == File::Exists(szDir) || false == File::IsDirectory(szDir))
File::CreateDir(szDir.c_str());
// For compatibility with old texture packs, don't print the LOD index for level 0.
// TODO: TLUT format should actually be stored in filename? :/
if (level == 0)
{
sprintf(szTemp, "%s/%s_%08x_%i.png", szDir.c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(),
(u32) (entry->hash & 0x00000000FFFFFFFFLL), entry->format & 0xFFFF);
}
else
{
sprintf(szTemp, "%s/%s_%08x_%i_mip%i.png", szDir.c_str(),
SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(),
(u32) (entry->hash & 0x00000000FFFFFFFFLL), entry->format & 0xFFFF, level);
}
if (false == File::Exists(szTemp))
entry->Save(szTemp, level);
}
TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage,
u32 address, unsigned int width, unsigned int height, int texformat,
unsigned int tlutaddr, int tlutfmt, bool use_mipmaps, unsigned int maxlevel, bool from_tmem)
{
if (0 == address)
return NULL;
// TexelSizeInNibbles(format)*width*height/16;
const unsigned int bsw = TexDecoder_GetBlockWidthInTexels(texformat) - 1;
const unsigned int bsh = TexDecoder_GetBlockHeightInTexels(texformat) - 1;
unsigned int expandedWidth = (width + bsw) & (~bsw);
unsigned int expandedHeight = (height + bsh) & (~bsh);
const unsigned int nativeW = width;
const unsigned int nativeH = height;
bool using_custom_texture = false;
bool using_custom_lods = false;
u32 texID = address;
u64 tex_hash = TEXHASH_INVALID; // Hash assigned to texcache entry (also used to generate filenames used for texture dumping and custom texture lookup)
u64 tlut_hash = TEXHASH_INVALID;
u32 full_format = texformat;
PC_TexFormat pcfmt = PC_TEX_FMT_NONE;
const bool isPaletteTexture = (texformat == GX_TF_C4 || texformat == GX_TF_C8 || texformat == GX_TF_C14X2);
if (isPaletteTexture)
full_format = texformat | (tlutfmt << 16);
const u32 texture_size = TexDecoder_GetTextureSizeInBytes(expandedWidth, expandedHeight, texformat);
u8* src_data;
if (from_tmem) src_data = &texMem[bpmem.tex[stage/4].texImage1[stage%4].tmem_even * TMEM_LINE_SIZE];
else src_data = Memory::GetPointer(address);
tex_hash = GetHash64(src_data, texture_size, g_ActiveConfig.iSafeTextureCache_ColorSamples);
if (isPaletteTexture)
{
const u32 palette_size = TexDecoder_GetPaletteSize(texformat);
tlut_hash = GetHash64(&texMem[tlutaddr], palette_size, g_ActiveConfig.iSafeTextureCache_ColorSamples);
// NOTE: For non-paletted textures, texID is equal to the texture address.
// A paletted texture, however, may have multiple texIDs assigned though depending on the currently used tlut.
// This (changing texID depending on the tlut_hash) is a trick to get around
// an issue with Metroid Prime's fonts (it has multiple sets of fonts on each other
// stored in a single texture and uses the palette to make different characters
// visible or invisible. Thus, unless we want to recreate the textures for every drawn character,
// we must make sure that a paletted texture gets assigned multiple IDs for each tlut used.
//
// TODO: Because texID isn't always the same as the address now, CopyRenderTargetToTexture might be broken now
texID ^= ((u32)tlut_hash) ^(u32)(tlut_hash >> 32);
tex_hash ^= tlut_hash;
}
TCacheEntryBase *entry = textures[texID];
if (entry)
{
// 1. Calculate reference hash:
// calculated from RAM texture data for normal textures. Hashes for paletted textures are modified by tlut_hash. 0 for virtual EFB copies.
if (g_ActiveConfig.bCopyEFBToTexture && entry->IsEfbCopy())
tex_hash = TEXHASH_INVALID;
// 2. a) For EFB copies, only the hash and the texture address need to match
if (entry->IsEfbCopy() && tex_hash == entry->hash && address == entry->addr)
{
entry->type = TCET_EC_VRAM;
// TODO: Print a warning if the format changes! In this case, we could reinterpret the internal texture object data to the new pixel format (similiar to what is already being done in Renderer::ReinterpretPixelFormat())
goto return_entry;
}
// 2. b) For normal textures, all texture parameters need to match
if (address == entry->addr && tex_hash == entry->hash && full_format == entry->format &&
entry->num_mipmaps > maxlevel && entry->native_width == nativeW && entry->native_height == nativeH)
{
goto return_entry;
}
// 3. If we reach this line, we'll have to upload the new texture data to VRAM.
// If we're lucky, the texture parameters didn't change and we can reuse the internal texture object instead of destroying and recreating it.
//
// TODO: Don't we need to force texture decoding to RGBA8 for dynamic EFB copies?
// TODO: Actually, it should be enough if the internal texture format matches...
if ((entry->type == TCET_NORMAL && width == entry->native_width && height == entry->native_height && full_format == entry->format && entry->num_mipmaps > maxlevel)
|| (entry->type == TCET_EC_DYNAMIC && entry->native_width == width && entry->native_height == height))
{
// reuse the texture
}
else
{
// delete the texture and make a new one
delete entry;
entry = NULL;
}
}
if (g_ActiveConfig.bHiresTextures)
{
pcfmt = LoadCustomTexture(tex_hash, texformat, 0, width, height);
if (pcfmt != PC_TEX_FMT_NONE)
{
expandedWidth = width;
expandedHeight = height;
using_custom_texture = true;
}
}
// TODO: RGBA8 textures are stored non-continuously in tmem, that might cause problems when preloading is enabled
if (pcfmt == PC_TEX_FMT_NONE)
pcfmt = TexDecoder_Decode(temp, src_data, expandedWidth,
expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
unsigned int texLevels;
bool use_native_mips;
texLevels = use_mipmaps ? (maxlevel + 1) : 1;
using_custom_lods = using_custom_texture && CheckForCustomTextureLODs(tex_hash, texformat, texLevels);
use_native_mips = use_mipmaps && !using_custom_lods && (width == nativeW && height == nativeH); // Only load native mips if their dimensions fit to our virtual texture dimensions
texLevels = (use_native_mips || using_custom_lods) ? texLevels : 1;
// create the entry/texture
if (NULL == entry) {
textures[texID] = entry = g_texture_cache->CreateTexture(width, height, expandedWidth, texLevels, pcfmt);
// Sometimes, we can get around recreating a texture if only the number of mip levels changes
// e.g. if our texture cache entry got too many mipmap levels we can limit the number of used levels by setting the appropriate render states
// Thus, we don't update this member for every Load, but just whenever the texture gets recreated
// TODO: D3D9 doesn't support min_lod. We should add a workaround for that here!
entry->num_mipmaps = maxlevel + 1;
entry->type = TCET_NORMAL;
GFX_DEBUGGER_PAUSE_AT(NEXT_NEW_TEXTURE, true);
}
entry->SetGeneralParameters(address, texture_size, full_format, entry->num_mipmaps);
entry->SetDimensions(nativeW, nativeH, width, height);
entry->hash = tex_hash;
if (entry->IsEfbCopy() && !g_ActiveConfig.bCopyEFBToTexture) entry->type = TCET_EC_DYNAMIC;
else entry->type = TCET_NORMAL;
// load texture
entry->Load(width, height, expandedWidth, 0);
if (g_ActiveConfig.bDumpTextures && !using_custom_texture)
DumpTexture(entry, 0);
// load mips - TODO: Loading mipmaps from tmem is untested!
if (texLevels > 1 && pcfmt != PC_TEX_FMT_NONE && use_native_mips)
{
const unsigned int bsdepth = TexDecoder_GetTexelSizeInNibbles(texformat);
unsigned int level = 1;
unsigned int mipWidth = (width + 1) >> 1;
unsigned int mipHeight = (height + 1) >> 1;
u8* ptr_even = NULL, *ptr_odd = NULL;
if (from_tmem)
{
ptr_even = &texMem[bpmem.tex[stage/4].texImage1[stage%4].tmem_even * TMEM_LINE_SIZE + texture_size];
ptr_odd = &texMem[bpmem.tex[stage/4].texImage2[stage%4].tmem_odd * TMEM_LINE_SIZE];
}
src_data += texture_size;
while ((mipHeight || mipWidth) && (level < texLevels))
{
u8** ptr;
if (from_tmem) ptr = (level % 2) ? &ptr_odd : &ptr_even;
else ptr = &src_data;
const unsigned int currentWidth = (mipWidth > 0) ? mipWidth : 1;
const unsigned int currentHeight = (mipHeight > 0) ? mipHeight : 1;
expandedWidth = (currentWidth + bsw) & (~bsw);
expandedHeight = (currentHeight + bsh) & (~bsh);
TexDecoder_Decode(temp, *ptr, expandedWidth, expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures);
entry->Load(currentWidth, currentHeight, expandedWidth, level);
if (g_ActiveConfig.bDumpTextures)
DumpTexture(entry, level);
*ptr += ((std::max(mipWidth, bsw) * std::max(mipHeight, bsh) * bsdepth) >> 1);
mipWidth >>= 1;
mipHeight >>= 1;
++level;
}
}
else if (texLevels > 1 && pcfmt != PC_TEX_FMT_NONE && using_custom_lods)
{
unsigned int level = 1;
unsigned int mipWidth = (width + 1) >> 1;
unsigned int mipHeight = (height + 1) >> 1;
while ((mipHeight || mipWidth) && (level < texLevels))
{
unsigned int currentWidth = (mipWidth > 0) ? mipWidth : 1;
unsigned int currentHeight = (mipHeight > 0) ? mipHeight : 1;
LoadCustomTexture(tex_hash, texformat, level, currentWidth, currentHeight);
entry->Load(currentWidth, currentHeight, currentWidth, level);
mipWidth >>= 1;
mipHeight >>= 1;
++level;
}
}
INCSTAT(stats.numTexturesCreated);
SETSTAT(stats.numTexturesAlive, textures.size());
return_entry:
entry->frameCount = frameCount;
entry->Bind(stage);
GFX_DEBUGGER_PAUSE_AT(NEXT_TEXTURE_CHANGE, true);
return entry;
}
void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, unsigned int 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;
if (srcFormat == PIXELFMT_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 = 24;
break;
case 11: // Z16 (reverse order)
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 2;
break;
case 6: // Z24X8
colmat[0] = colmat[5] = colmat[10] = 1.0f;
cbufid = 3;
break;
case 9: // Z8M
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 4;
break;
case 10: // Z8L
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 5;
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 = 6;
break;
default:
ERROR_LOG(VIDEO, "Unknown copy zbuf format: 0x%x", dstFormat);
colmat[2] = colmat[5] = colmat[8] = 1.0f;
cbufid = 7;
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 = 8;
}
else
{
cbufid = 9;
}
}
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 = 10;
}
else
{
cbufid = 11;
}
}
break;
default:
ERROR_LOG(VIDEO, "Unknown copy intensity format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 23;
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 = 12;
break;
case 1: // R8
case 8: // R8
colmat[0] = colmat[4] = colmat[8] = colmat[12] = 1;
cbufid = 13;
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 = 14;
break;
case 3: // RA8
colmat[0] = colmat[4] = colmat[8] = colmat[15] = 1.0f;
cbufid = 15;
break;
case 7: // A8
colmat[3] = colmat[7] = colmat[11] = colmat[15] = 1.0f;
cbufid = 16;
break;
case 9: // G8
colmat[1] = colmat[5] = colmat[9] = colmat[13] = 1.0f;
cbufid = 17;
break;
case 10: // B8
colmat[2] = colmat[6] = colmat[10] = colmat[14] = 1.0f;
cbufid = 18;
break;
case 11: // RG8
colmat[0] = colmat[4] = colmat[8] = colmat[13] = 1.0f;
cbufid = 19;
break;
case 12: // GB8
colmat[1] = colmat[5] = colmat[9] = colmat[14] = 1.0f;
cbufid = 20;
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 = 21;
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 = 22;
break;
case 6: // RGBA8
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 23;
break;
default:
ERROR_LOG(VIDEO, "Unknown copy color format: 0x%x", dstFormat);
colmat[0] = colmat[5] = colmat[10] = colmat[15] = 1.0f;
cbufid = 23;
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 = NULL;
}
}
if (NULL == 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, 0);
entry->SetDimensions(tex_w, tex_h, scaled_tex_w, scaled_tex_h);
entry->SetHashes(TEXHASH_INVALID);
entry->type = TCET_EC_VRAM;
}
entry->frameCount = frameCount;
g_renderer->ResetAPIState(); // reset any game specific settings
entry->FromRenderTarget(dstAddr, dstFormat, srcFormat, srcRect, isIntensity, scaleByHalf, cbufid, colmat);
g_renderer->RestoreAPIState();
}
Reference in New Issue View Git Blame Copy Permalink