diff options
| author | Shawn Hoffman <godisgovernment@gmail.com> | 2012-02-27 22:36:41 -0800 |
|---|---|---|
| committer | Shawn Hoffman <godisgovernment@gmail.com> | 2012-02-27 22:36:41 -0800 |
| commit | 6eb2cb266663d6a7a70f6e9e233cb1bf2967cd4d (patch) | |
| tree | fe60ac36b66199f775b9606aaea259bcc448c3d5 /Source/Core/VideoCommon | |
| parent | 6ccb30ac5fd7dfb51f41adf2d7fce159d4d0c174 (diff) | |
| parent | ee09def802f9883056a2dc43136e65922dcccee2 (diff) | |
Merge branch 'master' into wii-network
Diffstat (limited to 'Source/Core/VideoCommon')
| -rw-r--r-- | Source/Core/VideoCommon/Src/AVIDump.cpp | 2 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/BPMemory.cpp | 2 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/BPMemory.h | 41 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/BPStructs.cpp | 27 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/CommandProcessor.cpp | 13 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/CommandProcessor.h | 1 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/ImageWrite.h | 1 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/PixelShaderGen.cpp | 14 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/PixelShaderGen.h | 19 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/PixelShaderManager.cpp | 42 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/PixelShaderManager.h | 6 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/TextureCacheBase.cpp | 322 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/TextureCacheBase.h | 97 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/TextureDecoder.h | 4 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/VideoConfig.cpp | 12 | ||||
| -rw-r--r-- | Source/Core/VideoCommon/Src/VideoConfig.h | 1 |
16 files changed, 345 insertions, 259 deletions
diff --git a/Source/Core/VideoCommon/Src/AVIDump.cpp b/Source/Core/VideoCommon/Src/AVIDump.cpp index 469e00d9f5..670693690b 100644 --- a/Source/Core/VideoCommon/Src/AVIDump.cpp +++ b/Source/Core/VideoCommon/Src/AVIDump.cpp @@ -263,7 +263,7 @@ bool AVIDump::CreateFile() s_Stream->codec->bit_rate = 400000; s_Stream->codec->width = s_width; s_Stream->codec->height = s_height; - s_Stream->codec->time_base = (AVRational){1, VideoInterface::TargetRefreshRate}; + s_Stream->codec->time_base = (AVRational){1, static_cast<int>(VideoInterface::TargetRefreshRate)}; s_Stream->codec->gop_size = 12; s_Stream->codec->pix_fmt = g_Config.bUseFFV1 ? PIX_FMT_BGRA : PIX_FMT_YUV420P; diff --git a/Source/Core/VideoCommon/Src/BPMemory.cpp b/Source/Core/VideoCommon/Src/BPMemory.cpp index da09893f35..5c4ed457e3 100644 --- a/Source/Core/VideoCommon/Src/BPMemory.cpp +++ b/Source/Core/VideoCommon/Src/BPMemory.cpp @@ -61,7 +61,7 @@ void BPReload() // Cases in which we DON'T want to reload the BP continue; default: - BPCmd bp = {i, 0xFFFFFF, ((u32*)&bpmem)[i]}; + BPCmd bp = {i, 0xFFFFFF, static_cast<int>(((u32*)&bpmem)[i])}; BPWritten(bp); } } diff --git a/Source/Core/VideoCommon/Src/BPMemory.h b/Source/Core/VideoCommon/Src/BPMemory.h index 1f6f453a77..0bb2bf9013 100644 --- a/Source/Core/VideoCommon/Src/BPMemory.h +++ b/Source/Core/VideoCommon/Src/BPMemory.h @@ -65,10 +65,10 @@ #define BPMEM_UNKOWN_57 0x57 #define BPMEM_REVBITS 0x58 #define BPMEM_SCISSOROFFSET 0x59 -#define BPMEM_UNKNOWN_60 0x60 -#define BPMEM_UNKNOWN_61 0x61 -#define BPMEM_UNKNOWN_62 0x62 -#define BPMEM_TEXMODESYNC 0x63 +#define BPMEM_PRELOAD_ADDR 0x60 +#define BPMEM_PRELOAD_TMEMEVEN 0x61 +#define BPMEM_PRELOAD_TMEMODD 0x62 +#define BPMEM_PRELOAD_MODE 0x63 #define BPMEM_LOADTLUT0 0x64 #define BPMEM_LOADTLUT1 0x65 #define BPMEM_TEXINVALIDATE 0x66 @@ -487,10 +487,10 @@ union TexImage1 { struct { - u32 tmem_offset : 15; // we ignore texture caching for now, we do it ourselves - u32 cache_width : 3; + u32 tmem_even : 15; // tmem line index for even LODs + u32 cache_width : 3; u32 cache_height : 3; - u32 image_type : 1; + u32 image_type : 1; // 1 if this texture is managed manually (0 means we'll autofetch the texture data whenever it changes) }; u32 hex; }; @@ -499,7 +499,7 @@ union TexImage2 { struct { - u32 tmem_offset : 15; // we ignore texture caching for now, we do it ourselves + u32 tmem_odd : 15; // tmem line index for odd LODs u32 cache_width : 3; u32 cache_height : 3; }; @@ -893,6 +893,25 @@ union UPE_Copy } }; +union BPU_PreloadTileInfo +{ + u32 hex; + struct { + u32 count : 15; + u32 type : 2; + }; +}; + +struct BPS_TmemConfig +{ + u32 preload_addr; + u32 preload_tmem_even; + u32 preload_tmem_odd; + BPU_PreloadTileInfo preload_tile_info; + u32 tlut_src; + u32 tlut_dest; + u32 texinvalidate; +}; // All of BP memory @@ -951,10 +970,8 @@ struct BPMemory u32 boundbox1;//56 u32 unknown7[2];//57,58 X10Y10 scissorOffset; //59 - u32 unknown8[10]; //5a,5b,5c,5d, 5e,5f,60,61, 62, 63 (GXTexModeSync), 0x60-0x63 have to do with preloaded textures? - u32 tlutXferSrc; //64 - u32 tlutXferDest; //65 - u32 texinvalidate;//66 + u32 unknown8[6]; //5a,5b,5c,5d, 5e,5f + BPS_TmemConfig tmem_config; // 60-66 u32 metric; //67 FieldMode fieldmode;//68 u32 unknown10[7];//69-6F diff --git a/Source/Core/VideoCommon/Src/BPStructs.cpp b/Source/Core/VideoCommon/Src/BPStructs.cpp index d06f1de8b9..6cb102555b 100644 --- a/Source/Core/VideoCommon/Src/BPStructs.cpp +++ b/Source/Core/VideoCommon/Src/BPStructs.cpp @@ -30,6 +30,7 @@ #include "VertexLoader.h" #include "VertexShaderManager.h" #include "Thread.h" +#include "HW/Memmap.h" using namespace BPFunctions; @@ -301,14 +302,14 @@ void BPWritten(const BPCmd& bp) // TODO - figure out a cleaner way. if (GetConfig(CONFIG_ISWII)) - ptr = GetPointer(bpmem.tlutXferSrc << 5); + ptr = GetPointer(bpmem.tmem_config.tlut_src << 5); else - ptr = GetPointer((bpmem.tlutXferSrc & 0xFFFFF) << 5); + ptr = GetPointer((bpmem.tmem_config.tlut_src & 0xFFFFF) << 5); if (ptr) memcpy_gc(texMem + tlutTMemAddr, ptr, tlutXferCount); else - PanicAlert("Invalid palette pointer %08x %08x %08x", bpmem.tlutXferSrc, bpmem.tlutXferSrc << 5, (bpmem.tlutXferSrc & 0xFFFFF)<< 5); + PanicAlert("Invalid palette pointer %08x %08x %08x", bpmem.tmem_config.tlut_src, bpmem.tmem_config.tlut_src << 5, (bpmem.tmem_config.tlut_src & 0xFFFFF)<< 5); // TODO(ector) : kill all textures that use this palette // Not sure if it's a good idea, though. For now, we hash texture palettes @@ -466,14 +467,22 @@ void BPWritten(const BPCmd& bp) DEBUG_LOG(VIDEO, "Uknown BP Reg 0x57: %08x", bp.newvalue); break; - case BPMEM_UNKNOWN_60: - case BPMEM_UNKNOWN_61: - case BPMEM_UNKNOWN_62: - // Cases added due to: http://code.google.com/p/dolphin-emu/issues/detail?id=360#c90 - // Are these related to BBox? + case BPMEM_PRELOAD_ADDR: + case BPMEM_PRELOAD_TMEMEVEN: + case BPMEM_PRELOAD_TMEMODD: // Used when PRELOAD_MODE is set break; - case BPMEM_TEXMODESYNC: // Always set to 0 when GX_TexModeSync() is called. + case BPMEM_PRELOAD_MODE: // Set to 0 when GX_TexModeSync() is called. + // if this is different from 0, manual TMEM management is used. + if (bp.newvalue != 0) + { + // NOTE(neobrain): Apparently tmemodd doesn't affect hardware behavior at all (libogc uses it just as a buffer and switches its contents with tmemeven whenever this is called) + BPS_TmemConfig& tmem_cfg = bpmem.tmem_config; + u8* ram_ptr = Memory::GetPointer(tmem_cfg.preload_addr << 5); + u32 tmem_addr = tmem_cfg.preload_tmem_even * TMEM_LINE_SIZE; + u32 size = tmem_cfg.preload_tile_info.count * 32; + memcpy(texMem + tmem_addr, ram_ptr, size); + } break; // ------------------------------------------------ diff --git a/Source/Core/VideoCommon/Src/CommandProcessor.cpp b/Source/Core/VideoCommon/Src/CommandProcessor.cpp index 67c8f411fa..fc85b8e39d 100644 --- a/Source/Core/VideoCommon/Src/CommandProcessor.cpp +++ b/Source/Core/VideoCommon/Src/CommandProcessor.cpp @@ -311,6 +311,7 @@ void Write16(const u16 _Value, const u32 _Address) UCPClearReg tmpCtrl(_Value); m_CPClearReg.Hex = tmpCtrl.Hex; DEBUG_LOG(COMMANDPROCESSOR,"\t write to CLEAR_REGISTER : %04x", _Value); + SetCpClearRegister(); } break; @@ -677,13 +678,15 @@ void SetCpControlRegister() } +// NOTE: The implementation of this function should be correct, but we intentionally aren't using it at the moment. +// We don't emulate proper GP timing anyway at the moment, so this code would just slow down emulation. void SetCpClearRegister() { - if (IsOnThread()) - { - if (!m_CPClearReg.ClearFifoUnderflow && m_CPClearReg.ClearFifoOverflow) - bProcessFifoToLoWatermark = true; - } +// if (IsOnThread()) +// { +// if (!m_CPClearReg.ClearFifoUnderflow && m_CPClearReg.ClearFifoOverflow) +// bProcessFifoToLoWatermark = true; +// } } } // end of namespace CommandProcessor diff --git a/Source/Core/VideoCommon/Src/CommandProcessor.h b/Source/Core/VideoCommon/Src/CommandProcessor.h index 7fb4e5a45f..db6772d66d 100644 --- a/Source/Core/VideoCommon/Src/CommandProcessor.h +++ b/Source/Core/VideoCommon/Src/CommandProcessor.h @@ -158,6 +158,7 @@ void UpdateInterruptsFromVideoBackend(u64 userdata); bool AllowIdleSkipping(); +void SetCpClearRegister(); void SetCpControlRegister(); void SetCpStatusRegister(); void SetOverflowStatusFromGatherPipe(); diff --git a/Source/Core/VideoCommon/Src/ImageWrite.h b/Source/Core/VideoCommon/Src/ImageWrite.h index 962ee5b5e7..9ea6cab859 100644 --- a/Source/Core/VideoCommon/Src/ImageWrite.h +++ b/Source/Core/VideoCommon/Src/ImageWrite.h @@ -21,7 +21,6 @@ #include "Common.h" bool SaveTGA(const char* filename, int width, int height, void* pdata); -bool SaveTexture(const char* filename, u32 textarget, u32 tex, int width, int height); bool SaveData(const char* filename, const char* pdata); #endif // _IMAGEWRITE_H diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp index 1726855d54..fe5827de05 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.cpp @@ -560,6 +560,10 @@ const char *GeneratePixelShaderCode(DSTALPHA_MODE dstAlphaMode, API_TYPE ApiType WRITE(p, "uniform float4 "I_KCOLORS"[4] : register(c%d);\n", C_KCOLORS); WRITE(p, "uniform float4 "I_ALPHA"[1] : register(c%d);\n", C_ALPHA); WRITE(p, "uniform float4 "I_TEXDIMS"[8] : register(c%d);\n", C_TEXDIMS); + if (ApiType & API_D3D9) + { + WRITE(p, "uniform float4 "I_VTEXSCALE"[4] : register(c%d);\n", C_VTEXSCALE); + } WRITE(p, "uniform float4 "I_ZBIAS"[2] : register(c%d);\n", C_ZBIAS); WRITE(p, "uniform float4 "I_INDTEXSCALE"[2] : register(c%d);\n", C_INDTEXSCALE); WRITE(p, "uniform float4 "I_INDTEXMTX"[6] : register(c%d);\n", C_INDTEXMTX); @@ -1099,9 +1103,15 @@ static void WriteStage(char *&p, int n, API_TYPE ApiType) void SampleTexture(char *&p, const char *destination, const char *texcoords, const char *texswap, int texmap, API_TYPE ApiType) { if (ApiType == API_D3D11) - WRITE(p, "%s=Tex%d.Sample(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + WRITE(p, "%s=Tex%d.Sample(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap,texmap, texcoords, texmap, texswap); + else if (ApiType & API_D3D9) + { + // D3D9 uses different pixel to texel mapping, so we need to offset our sampling address by half a pixel (assuming native and virtual texture dimensions match each other, otherwise some math is involved). + // Read the MSDN article "Directly Mapping Texels to Pixels (Direct3D 9)" for further info. + WRITE(p, "%s=tex2D(samp%d, (%s.xy + 0.5f*"I_VTEXSCALE"[%d].%s) * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap/2, (texmap&1)?"zw":"xy", texmap, texswap); + } else - WRITE(p, "%s=tex2D(samp%d,%s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); + WRITE(p, "%s=tex2D(samp%d, %s.xy * "I_TEXDIMS"[%d].xy).%s;\n", destination, texmap, texcoords, texmap, texswap); } static const char *tevAlphaFuncsTable[] = diff --git a/Source/Core/VideoCommon/Src/PixelShaderGen.h b/Source/Core/VideoCommon/Src/PixelShaderGen.h index 31242a916e..7374251c96 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderGen.h +++ b/Source/Core/VideoCommon/Src/PixelShaderGen.h @@ -24,26 +24,27 @@ #define I_KCOLORS "k" #define I_ALPHA "alphaRef" #define I_TEXDIMS "texdim" +#define I_VTEXSCALE "vtexscale" #define I_ZBIAS "czbias" #define I_INDTEXSCALE "cindscale" #define I_INDTEXMTX "cindmtx" #define I_FOG "cfog" #define I_PLIGHTS "cLights" -#define I_PMATERIALS "cmtrl" +#define I_PMATERIALS "cmtrl" #define C_COLORMATRIX 0 // 0 #define C_COLORS 0 // 0 #define C_KCOLORS (C_COLORS + 4) // 4 #define C_ALPHA (C_KCOLORS + 4) // 8 #define C_TEXDIMS (C_ALPHA + 1) // 9 -#define C_ZBIAS (C_TEXDIMS + 8) //17 -#define C_INDTEXSCALE (C_ZBIAS + 2) //19 -#define C_INDTEXMTX (C_INDTEXSCALE + 2) //21 -#define C_FOG (C_INDTEXMTX + 6) //27 - -#define C_PLIGHTS (C_FOG + 3) -#define C_PMATERIALS (C_PLIGHTS + 40) -#define C_PENVCONST_END (C_PMATERIALS + 4) +#define C_VTEXSCALE (C_TEXDIMS + 8) //17 - virtual texture scaling factor (e.g. custom textures, scaled EFB copies) +#define C_ZBIAS (C_VTEXSCALE + 4) //21 +#define C_INDTEXSCALE (C_ZBIAS + 2) //23 +#define C_INDTEXMTX (C_INDTEXSCALE + 2) //25 +#define C_FOG (C_INDTEXMTX + 6) //31 +#define C_PLIGHTS (C_FOG + 3) //34 +#define C_PMATERIALS (C_PLIGHTS + 40) //74 +#define C_PENVCONST_END (C_PMATERIALS + 4) //78 #define PIXELSHADERUID_MAX_VALUES 70 #define PIXELSHADERUID_MAX_VALUES_SAFE 120 diff --git a/Source/Core/VideoCommon/Src/PixelShaderManager.cpp b/Source/Core/VideoCommon/Src/PixelShaderManager.cpp index 2521f80500..3e0f6d73c4 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderManager.cpp +++ b/Source/Core/VideoCommon/Src/PixelShaderManager.cpp @@ -36,9 +36,11 @@ static bool s_bFogRangeAdjustChanged; static int nLightsChanged[2]; // min,max static float lastRGBAfull[2][4][4]; static u8 s_nTexDimsChanged; +static u8 s_nVirtualTexScalesChanged; static u8 s_nIndTexScaleChanged; static u32 lastAlpha; static u32 lastTexDims[8]; // width | height << 16 | wrap_s << 28 | wrap_t << 30 +static float lastVirtualTexScales[16]; // even fields: width ratio; odd fields: height ratio static u32 lastZBias; static int nMaterialsChanged; @@ -61,6 +63,7 @@ void PixelShaderManager::Init() { lastAlpha = 0; memset(lastTexDims, 0, sizeof(lastTexDims)); + memset(lastVirtualTexScales, 0, sizeof(lastVirtualTexScales)); lastZBias = 0; memset(lastRGBAfull, 0, sizeof(lastRGBAfull)); Dirty(); @@ -70,6 +73,7 @@ void PixelShaderManager::Dirty() { s_nColorsChanged[0] = s_nColorsChanged[1] = 15; s_nTexDimsChanged = 0xFF; + s_nVirtualTexScalesChanged = 0xFF; s_nIndTexScaleChanged = 0xFF; s_nIndTexMtxChanged = 15; s_bAlphaChanged = s_bZBiasChanged = s_bZTextureTypeChanged = s_bDepthRangeChanged = true; @@ -83,7 +87,7 @@ void PixelShaderManager::Shutdown() } -void PixelShaderManager::SetConstants() +void PixelShaderManager::SetConstants(API_TYPE api_type) { for (int i = 0; i < 2; ++i) { @@ -109,6 +113,16 @@ void PixelShaderManager::SetConstants() s_nTexDimsChanged = 0; } + if ((api_type & API_D3D9) && s_nVirtualTexScalesChanged) + { + for (int i = 0; i < 8; i += 2) + { + if (s_nVirtualTexScalesChanged & (3<<i)) + SetPSVirtualTexScalePair(i/2); + } + s_nVirtualTexScalesChanged = 0; + } + if (s_bAlphaChanged) { SetPSConstant4f(C_ALPHA, (lastAlpha&0xff)/255.0f, ((lastAlpha>>8)&0xff)/255.0f, 0, ((lastAlpha>>16)&0xff)/255.0f); @@ -338,6 +352,13 @@ void PixelShaderManager::SetPSTextureDims(int texid) SetPSConstant4fv(C_TEXDIMS + texid, fdims); } +void PixelShaderManager::SetPSVirtualTexScalePair(int texpairid) +{ + PRIM_LOG("vtexscale%d: %f %f %f %f\n", texpairid, lastVirtualTexScales[texpairid*4], lastVirtualTexScales[texpairid*4+1], + lastVirtualTexScales[texpairid*4+2], lastVirtualTexScales[texpairid*4+3]); + SetPSConstant4fv(C_VTEXSCALE + texpairid, &lastVirtualTexScales[texpairid*4]); +} + // This one is high in profiles (0.5%). TODO: Move conversion out, only store the raw color value // and update it when the shader constant is set, only. void PixelShaderManager::SetColorChanged(int type, int num, bool high) @@ -376,14 +397,25 @@ void PixelShaderManager::SetDestAlpha(const ConstantAlpha& alpha) } } -void PixelShaderManager::SetTexDims(int texmapid, u32 width, u32 height, u32 wraps, u32 wrapt) +void PixelShaderManager::SetTexDims(int texmapid, u32 width, u32 height, u32 virtual_width, u32 virtual_height, u32 wraps, u32 wrapt, API_TYPE api_type) { u32 wh = width | (height << 16) | (wraps << 28) | (wrapt << 30); - if (lastTexDims[texmapid] != wh) + + bool refresh = lastTexDims[texmapid] != wh; + if (api_type & API_D3D9) { - lastTexDims[texmapid] = wh; + refresh |= (lastVirtualTexScales[texmapid*2] != (float)width / (float)virtual_width); + refresh |= (lastVirtualTexScales[texmapid*2+1] != (float)height / (float)virtual_height); + } + + if (refresh) + { + lastTexDims[texmapid] = wh; + lastVirtualTexScales[texmapid*2] = (float)width / (float)virtual_width; + lastVirtualTexScales[texmapid*2+1] = (float)height / (float)virtual_height; s_nTexDimsChanged |= 1 << texmapid; - } + s_nVirtualTexScalesChanged |= 1 << texmapid; + } } void PixelShaderManager::SetZTextureBias(u32 bias) diff --git a/Source/Core/VideoCommon/Src/PixelShaderManager.h b/Source/Core/VideoCommon/Src/PixelShaderManager.h index 2d1c01cad6..5cbe86cae4 100644 --- a/Source/Core/VideoCommon/Src/PixelShaderManager.h +++ b/Source/Core/VideoCommon/Src/PixelShaderManager.h @@ -26,18 +26,20 @@ class PixelShaderManager { static void SetPSTextureDims(int texid); + static void SetPSVirtualTexScalePair(int texpairid); + public: static void Init(); static void Dirty(); static void Shutdown(); - static void SetConstants(); // sets pixel shader constants + static void SetConstants(API_TYPE api_type); // sets pixel shader constants // constant management, should be called after memory is committed static void SetColorChanged(int type, int index, bool high); static void SetAlpha(const AlphaFunc& alpha); static void SetDestAlpha(const ConstantAlpha& alpha); - static void SetTexDims(int texmapid, u32 width, u32 height, u32 wraps, u32 wrapt); + static void SetTexDims(int texmapid, u32 width, u32 height, u32 virtual_width, u32 virtual_height, u32 wraps, u32 wrapt, API_TYPE api_type); static void SetZTextureBias(u32 bias); static void SetViewportChanged(); static void SetIndMatrixChanged(int matrixidx); diff --git a/Source/Core/VideoCommon/Src/TextureCacheBase.cpp b/Source/Core/VideoCommon/Src/TextureCacheBase.cpp index 3550227389..cc14499330 100644 --- a/Source/Core/VideoCommon/Src/TextureCacheBase.cpp +++ b/Source/Core/VideoCommon/Src/TextureCacheBase.cpp @@ -1,3 +1,19 @@ +// 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" @@ -23,28 +39,19 @@ enum TextureCache *g_texture_cache; - GC_ALIGNED16(u8 *TextureCache::temp) = NULL; +GC_ALIGNED16(u8 *TextureCache::temp) = NULL; TextureCache::TexCache TextureCache::textures; bool TextureCache::DeferredInvalidate; TextureCache::TCacheEntryBase::~TCacheEntryBase() { - if (0 == addr) - return; - - if (!isRenderTarget && !g_ActiveConfig.bSafeTextureCache) - { - u32 *const ptr = (u32*)Memory::GetPointer(addr); - if (ptr && *ptr == hash) - *ptr = oldpixel; - } } TextureCache::TextureCache() { if (!temp) - temp =(u8*) AllocateAlignedMemory(TEMP_SIZE,16); + 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()); @@ -88,12 +95,11 @@ TextureCache::~TextureCache() void TextureCache::Cleanup() { - TexCache::iterator - iter = textures.begin(), - tcend = textures.end(); + TexCache::iterator iter = textures.begin(); + TexCache::iterator tcend = textures.end(); while (iter != tcend) { - if (frameCount > TEXTURE_KILL_THRESHOLD + iter->second->frameCount) + 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++); @@ -135,7 +141,7 @@ void TextureCache::MakeRangeDynamic(u32 start_address, u32 size) const int rangePosition = iter->second->IntersectsMemoryRange(start_address, size); if (0 == rangePosition) { - iter->second->hash = 0; + iter->second->SetHashes(TEXHASH_INVALID); } } } @@ -167,19 +173,16 @@ void TextureCache::ClearRenderTargets() iter = textures.begin(), tcend = textures.end(); for (; iter!=tcend; ++iter) - iter->second->isRenderTarget = false; + iter->second->type = TCET_NORMAL; } TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, u32 address, unsigned int width, unsigned int height, int texformat, - unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel) + unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel, bool from_tmem) { - // necessary? if (0 == address) return NULL; - u8* ptr = Memory::GetPointer(address); - // TexelSizeInNibbles(format)*width*height/16; const unsigned int bsw = TexDecoder_GetBlockWidthInTexels(texformat) - 1; const unsigned int bsh = TexDecoder_GetBlockHeightInTexels(texformat) - 1; @@ -188,111 +191,82 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, unsigned int expandedHeight = (height + bsh) & (~bsh); const unsigned int nativeW = width; const unsigned int nativeH = height; - bool isPow2; - u64 hash_value = 0; - u64 texHash = 0; 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; - const u32 texture_size = TexDecoder_GetTextureSizeInBytes(expandedWidth, expandedHeight, texformat); - const u32 palette_size = TexDecoder_GetPaletteSize(texformat); - bool texture_is_dynamic = false; - unsigned int texLevels; 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); - // hires texture loading and texture dumping require accurate hashes - if (g_ActiveConfig.bSafeTextureCache || g_ActiveConfig.bHiresTextures || g_ActiveConfig.bDumpTextures) - { - texHash = GetHash64(ptr, texture_size, g_ActiveConfig.iSafeTextureCache_ColorSamples); - - if (isPaletteTexture) - { - // WARNING! texID != address now => may break CopyRenderTargetToTexture (cf. TODO up) - // tlut size can be up to 32768B (GX_TF_C14X2) but Safer == Slower. - // This trick (to change the texID depending on the TLUT addr) is a trick to get around - // an issue with metroid prime's fonts, where it has multiple sets of fonts on top of - // 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 texture with different tluts get different IDs. - - const u64 tlutHash = GetHash64(texMem + tlutaddr, palette_size, - g_ActiveConfig.iSafeTextureCache_ColorSamples); - - texHash ^= tlutHash; - - if (g_ActiveConfig.bSafeTextureCache) - texID ^= ((u32)tlutHash) ^ (u32)(tlutHash >> 32); - } + 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); - if (g_ActiveConfig.bSafeTextureCache) - hash_value = texHash; + 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) { - if (!g_ActiveConfig.bSafeTextureCache) - { - if (entry->isRenderTarget || entry->isDynamic) - { - if (!g_ActiveConfig.bCopyEFBToTexture) - { - hash_value = GetHash64(ptr, texture_size, g_ActiveConfig.iSafeTextureCache_ColorSamples); + // 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; - if (isPaletteTexture) - { - hash_value ^= GetHash64(&texMem[tlutaddr], palette_size, - g_ActiveConfig.iSafeTextureCache_ColorSamples); - } - } - else - { - hash_value = 0; - } - } - else - { - hash_value = *(u32*)ptr; - } - } - else if ((entry->isRenderTarget || entry->isDynamic) && g_ActiveConfig.bCopyEFBToTexture) + // 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) { - hash_value = 0; + // 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; } - if (((entry->isRenderTarget || entry->isDynamic) && hash_value == entry->hash && address == entry->addr) - || ((address == entry->addr) && (hash_value == entry->hash) && full_format == entry->format && entry->mipLevels == maxlevel)) + // 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) { - entry->isDynamic = false; 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 { - // Let's reload the new texture data into the same texture, - // instead of destroying it and having to create a new one. - // Might speed up movie playback very, very slightly. - texture_is_dynamic = (entry->isRenderTarget || entry->isDynamic) && !g_ActiveConfig.bCopyEFBToTexture; - - if (!entry->isRenderTarget && - ((!entry->isDynamic && width == entry->realW && height == entry->realH && full_format == entry->format && entry->mipLevels == maxlevel) - || (entry->isDynamic && entry->realW == width && entry->realH == height))) - { - // reuse the texture - } - else - { - // delete the texture and make a new one - delete entry; - entry = NULL; - } + // delete the texture and make a new one + delete entry; + entry = NULL; } } - + if (g_ActiveConfig.bHiresTextures) { // Load Custom textures @@ -301,23 +275,24 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, unsigned int newWidth = width; unsigned int newHeight = height; - sprintf(texPathTemp, "%s_%08x_%i", SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), (u32) (texHash & 0x00000000FFFFFFFFLL), texformat); + sprintf(texPathTemp, "%s_%08x_%i", SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), (u32) (tex_hash & 0x00000000FFFFFFFFLL), texformat); pcfmt = HiresTextures::GetHiresTex(texPathTemp, &newWidth, &newHeight, texformat, temp); if (pcfmt != PC_TEX_FMT_NONE) { expandedWidth = width = newWidth; expandedHeight = height = newHeight; - - // TODO: shouldn't generating mips be forced on now? - // native mips with a custom texture wouldn't make sense } } + // 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, ptr, expandedWidth, - expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures); + pcfmt = TexDecoder_Decode(temp, src_data, expandedWidth, + expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures); + bool isPow2; + unsigned int texLevels; + UseNativeMips = UseNativeMips && (width == nativeW && height == nativeH); // Only load native mips if their dimensions fit to our virtual texture dimensions isPow2 = !((width & (width - 1)) || (height & (height - 1))); texLevels = (isPow2 && UseNativeMips && maxlevel) ? GetPow2(std::max(width, height)) : !isPow2; @@ -329,40 +304,27 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, 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 gets changes + // 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 - entry->mipLevels = maxlevel; + // + // TODO: Won't we end up recreating textures all the time because maxlevel doesn't necessarily equal texLevels? + entry->num_mipmaps = maxlevel; // TODO: Does this actually work? We can't really adjust mipmap settings per-stage... + entry->type = TCET_NORMAL; GFX_DEBUGGER_PAUSE_AT(NEXT_NEW_TEXTURE, true); } - entry->addr = address; - entry->format = full_format; - entry->size_in_bytes = texture_size; - - entry->virtualW = width; - entry->virtualH = height; - - entry->realW = nativeW; - entry->realH = nativeH; - - entry->isRenderTarget = false; - entry->isNonPow2 = false; - entry->isDynamic = texture_is_dynamic; - - entry->oldpixel = *(u32*)ptr; - - if (g_ActiveConfig.bSafeTextureCache || entry->isDynamic) - entry->hash = hash_value; - else - // don't like rand() here - entry->hash = *(u32*)ptr = (u32)(((double)rand() / RAND_MAX) * 0xFFFFFFFF); + entry->SetGeneralParameters(address, texture_size, full_format, entry->num_mipmaps); + entry->SetDimensions(nativeW, nativeH, width, height); + entry->hash = tex_hash; + if (g_ActiveConfig.bCopyEFBToTexture) entry->type = TCET_NORMAL; + else if (entry->IsEfbCopy()) entry->type = TCET_EC_DYNAMIC; // load texture entry->Load(width, height, expandedWidth, 0, (texLevels == 0)); - // load mips + // load mips - TODO: Loading mipmaps from tmem is untested! if (texLevels > 1 && pcfmt != PC_TEX_FMT_NONE) { const unsigned int bsdepth = TexDecoder_GetTexelSizeInNibbles(texformat); @@ -370,20 +332,31 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, unsigned int level = 1; unsigned int mipWidth = (width + 1) >> 1; unsigned int mipHeight = (height + 1) >> 1; - ptr += texture_size; + + 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); + TexDecoder_Decode(temp, *ptr, expandedWidth, expandedHeight, texformat, tlutaddr, tlutfmt, g_ActiveConfig.backend_info.bUseRGBATextures); + entry->Load(currentWidth, currentHeight, expandedWidth, level, false); - ptr += ((std::max(mipWidth, bsw) * std::max(mipHeight, bsh) * bsdepth) >> 1); + *ptr += ((std::max(mipWidth, bsw) * std::max(mipHeight, bsh) * bsdepth) >> 1); mipWidth >>= 1; mipHeight >>= 1; ++level; @@ -404,7 +377,7 @@ TextureCache::TCacheEntryBase* TextureCache::Load(unsigned int stage, sprintf(szTemp, "%s/%s_%08x_%i.png", szDir.c_str(), SConfig::GetInstance().m_LocalCoreStartupParameter.m_strUniqueID.c_str(), - (u32) (texHash & 0x00000000FFFFFFFFLL), texformat); + (u32) (tex_hash & 0x00000000FFFFFFFFLL), texformat); if (false == File::Exists(szTemp)) entry->Save(szTemp); @@ -426,6 +399,45 @@ 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; @@ -472,9 +484,9 @@ void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat 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
+ 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; @@ -628,15 +640,15 @@ void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat 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; - bool texture_is_dynamic = false; TCacheEntryBase *entry = textures[dstAddr]; if (entry) { - if ((entry->isRenderTarget && entry->virtualW == scaled_tex_w && entry->virtualH == scaled_tex_h) - || (entry->isDynamic && entry->realW == tex_w && entry->realH == tex_h)) + if ((entry->type == TCET_EC_VRAM && entry->virtual_width == scaled_tex_w && entry->virtual_height == scaled_tex_h) + || (entry->type == TCET_EC_DYNAMIC && entry->native_width == tex_w && entry->native_height == tex_h)) { - texture_is_dynamic = entry->isDynamic; + scaled_tex_w = tex_w; + scaled_tex_h = tex_h; } else { @@ -646,32 +658,16 @@ void TextureCache::CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat } } - if (texture_is_dynamic) - { - scaled_tex_w = tex_w; - scaled_tex_h = tex_h; - } - if (NULL == entry) { // create the texture textures[dstAddr] = entry = g_texture_cache->CreateRenderTargetTexture(scaled_tex_w, scaled_tex_h); - entry->addr = dstAddr; - entry->hash = 0; - - entry->realW = tex_w; - entry->realH = tex_h; - - entry->virtualW = scaled_tex_w; - entry->virtualH = scaled_tex_h; - - entry->format = dstFormat; - entry->mipLevels = 0; - - entry->isRenderTarget = true; - entry->isNonPow2 = true; - entry->isDynamic = false; + // 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; diff --git a/Source/Core/VideoCommon/Src/TextureCacheBase.h b/Source/Core/VideoCommon/Src/TextureCacheBase.h index 1c86e64006..2924c5871f 100644 --- a/Source/Core/VideoCommon/Src/TextureCacheBase.h +++ b/Source/Core/VideoCommon/Src/TextureCacheBase.h @@ -1,3 +1,19 @@ +// 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/ #ifndef _TEXTURECACHEBASE_H #define _TEXTURECACHEBASE_H @@ -14,47 +30,56 @@ class TextureCache { public: + enum TexCacheEntryType + { + TCET_NORMAL, + TCET_EC_VRAM, // EFB copy which sits in VRAM and is ready to be used + TCET_EC_DYNAMIC, // EFB copy which sits in RAM and needs to be decoded before being used + }; + struct TCacheEntryBase { - // TODO: organize +#define TEXHASH_INVALID 0 + + // common members u32 addr; u32 size_in_bytes; u64 hash; - //u32 paletteHash; - u32 oldpixel; + //u32 pal_hash; u32 format; - + + enum TexCacheEntryType type; + + unsigned int num_mipmaps; + unsigned int native_width, native_height; // Texture dimensions from the GameCube's point of view + unsigned int virtual_width, virtual_height; // Texture dimensions from OUR point of view - for hires textures or scaled EFB copies + + // used to delete textures which haven't been used for TEXTURE_KILL_THRESHOLD frames int frameCount; - unsigned int realW, realH; // Texture dimensions from the GameCube's point of view - unsigned int virtualW, virtualH; // Texture dimensions from OUR point of view - // Real and virtual dimensions are usually the same, but may be - // different if e.g. we use high-res textures. Then, realW,realH will - // be the dimensions of the original GameCube texture and - // virtualW,virtualH will be the dimensions of the high-res texture. - - unsigned int mipLevels; - - bool isRenderTarget; // copied from EFB - bool isDynamic; // Used for hybrid EFB copies to enable checks for CPU modifications - bool isNonPow2; // doesn't seem to be used anywhere - - //TCacheEntryBase() - //{ - // // TODO: remove these - // isRenderTarget = 0; - // hash = 0; - // //paletteHash = 0; - // oldpixel = 0; - // addr = 0; - // size_in_bytes = 0; - // frameCount = 0; - // isNonPow2 = true; - // w = 0; - // h = 0; - // scaledW = 0; - // scaledH = 0; - //} + + void SetGeneralParameters(u32 addr, u32 size, u32 format, unsigned int num_mipmaps) + { + this->addr = addr; + this->size_in_bytes = size; + this->format = format; + this->num_mipmaps = num_mipmaps; + } + + void SetDimensions(unsigned int native_width, unsigned int native_height, unsigned int virtual_width, unsigned int virtual_height) + { + this->native_width = native_width; + this->native_height = native_height; + this->virtual_width = virtual_width; + this->virtual_height = virtual_height; + } + + void SetHashes(u64 hash/*, u32 pal_hash*/) + { + this->hash = hash; + //this->pal_hash = pal_hash; + } + virtual ~TCacheEntryBase(); @@ -62,13 +87,15 @@ public: virtual bool Save(const char filename[]) = 0; virtual void Load(unsigned int width, unsigned int height, - unsigned int expanded_width, unsigned int level, bool autogen_mips = false) = 0; + unsigned int expanded_width, unsigned int level, bool autogen_mips) = 0; virtual void FromRenderTarget(u32 dstAddr, unsigned int dstFormat, unsigned int srcFormat, const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf, unsigned int cbufid, const float *colmat) = 0; int IntersectsMemoryRange(u32 range_address, u32 range_size) const; + + bool IsEfbCopy() { return (type == TCET_EC_VRAM || type == TCET_EC_DYNAMIC); } }; virtual ~TextureCache(); // needs virtual for DX11 dtor @@ -87,7 +114,7 @@ public: virtual TCacheEntryBase* CreateRenderTargetTexture(unsigned int scaled_tex_w, unsigned int scaled_tex_h) = 0; static TCacheEntryBase* Load(unsigned int stage, u32 address, unsigned int width, unsigned int height, - int format, unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel); + int format, unsigned int tlutaddr, int tlutfmt, bool UseNativeMips, unsigned int maxlevel, bool from_tmem); static void CopyRenderTargetToTexture(u32 dstAddr, unsigned int dstFormat, unsigned int srcFormat, const EFBRectangle& srcRect, bool isIntensity, bool scaleByHalf); diff --git a/Source/Core/VideoCommon/Src/TextureDecoder.h b/Source/Core/VideoCommon/Src/TextureDecoder.h index d990d21545..1901c86d27 100644 --- a/Source/Core/VideoCommon/Src/TextureDecoder.h +++ b/Source/Core/VideoCommon/Src/TextureDecoder.h @@ -20,8 +20,8 @@ #include "Hash.h" enum { - TMEM_SIZE = 1024*1024, - HALFTMEM_SIZE = 512*1024 + TMEM_SIZE = 1024*1024, + TMEM_LINE_SIZE = 32, }; extern GC_ALIGNED16(u8 texMem[TMEM_SIZE]); diff --git a/Source/Core/VideoCommon/Src/VideoConfig.cpp b/Source/Core/VideoCommon/Src/VideoConfig.cpp index b6ba76f45e..5bb6e4dd99 100644 --- a/Source/Core/VideoCommon/Src/VideoConfig.cpp +++ b/Source/Core/VideoCommon/Src/VideoConfig.cpp @@ -57,12 +57,8 @@ void VideoConfig::Load(const char *ini_file) iniFile.Get("Settings", "Crop", &bCrop, false); iniFile.Get("Settings", "UseXFB", &bUseXFB, 0); iniFile.Get("Settings", "UseRealXFB", &bUseRealXFB, 0); - iniFile.Get("Settings", "UseNativeMips", &bUseNativeMips, false); - - iniFile.Get("Settings", "SafeTextureCache", &bSafeTextureCache, true); // Settings - //Safe texture cache params + iniFile.Get("Settings", "UseNativeMips", &bUseNativeMips, false); iniFile.Get("Settings", "SafeTextureCacheColorSamples", &iSafeTextureCache_ColorSamples,128); - iniFile.Get("Settings", "ShowFPS", &bShowFPS, false); // Settings iniFile.Get("Settings", "ShowInputDisplay", &bShowInputDisplay, false); iniFile.Get("Settings", "OverlayStats", &bOverlayStats, false); @@ -134,7 +130,6 @@ void VideoConfig::GameIniLoad(const char *ini_file) iniFile.GetIfExists("Video_Settings", "UseXFB", &bUseXFB); iniFile.GetIfExists("Video_Settings", "UseRealXFB", &bUseRealXFB); iniFile.GetIfExists("Video_Settings", "UseNativeMips", &bUseNativeMips); - iniFile.GetIfExists("Video_Settings", "SafeTextureCache", &bSafeTextureCache); iniFile.GetIfExists("Video_Settings", "SafeTextureCacheColorSamples", &iSafeTextureCache_ColorSamples); iniFile.GetIfExists("Video_Settings", "DLOptimize", &iCompileDLsLevel); iniFile.GetIfExists("Video_Settings", "HiresTextures", &bHiresTextures); @@ -196,11 +191,7 @@ void VideoConfig::Save(const char *ini_file) iniFile.Set("Settings", "UseXFB", bUseXFB); iniFile.Set("Settings", "UseRealXFB", bUseRealXFB); iniFile.Set("Settings", "UseNativeMips", bUseNativeMips); - - iniFile.Set("Settings", "SafeTextureCache", bSafeTextureCache); - //safe texture cache params iniFile.Set("Settings", "SafeTextureCacheColorSamples", iSafeTextureCache_ColorSamples); - iniFile.Set("Settings", "ShowFPS", bShowFPS); iniFile.Set("Settings", "ShowInputDisplay", bShowInputDisplay); iniFile.Set("Settings", "OverlayStats", bOverlayStats); @@ -279,7 +270,6 @@ void VideoConfig::GameIniSave(const char* default_ini, const char* game_ini) SET_IF_DIFFERS("Video_Settings", "UseXFB", bUseXFB); SET_IF_DIFFERS("Video_Settings", "UseRealXFB", bUseRealXFB); SET_IF_DIFFERS("Video_Settings", "UseNativeMips", bUseNativeMips); - SET_IF_DIFFERS("Video_Settings", "SafeTextureCache", bSafeTextureCache); SET_IF_DIFFERS("Video_Settings", "SafeTextureCacheColorSamples", iSafeTextureCache_ColorSamples); SET_IF_DIFFERS("Video_Settings", "DLOptimize", iCompileDLsLevel); SET_IF_DIFFERS("Video_Settings", "HiresTextures", bHiresTextures); diff --git a/Source/Core/VideoCommon/Src/VideoConfig.h b/Source/Core/VideoCommon/Src/VideoConfig.h index 3bd833cc73..7c846676bb 100644 --- a/Source/Core/VideoCommon/Src/VideoConfig.h +++ b/Source/Core/VideoCommon/Src/VideoConfig.h @@ -129,7 +129,6 @@ struct VideoConfig bool bOSDHotKey; bool bCopyEFBToTexture; bool bCopyEFBScaled; - bool bSafeTextureCache; int iSafeTextureCache_ColorSamples; int iPhackvalue[4]; std::string sPhackvalue[2]; |
