openclonk/src/landscape/C4LandscapeRender.cpp

1102 lines
33 KiB
C++

#include "C4Include.h"
#include "C4LandscapeRender.h"
#include "C4Landscape.h"
#include "C4Texture.h"
#include "C4GroupSet.h"
#include "C4Components.h"
#include "C4DrawGL.h"
#include "StdColors.h"
#ifdef USE_GL
// Automatically reload shaders when changed at runtime?
#define AUTO_RELOAD_SHADERS
#ifdef _DEBUG
// Generate seperator textures into 3D texture so we can make sure that
// we are addressing textures using the right coordinates
#define DEBUG_SEPERATOR_TEXTURES
// Replace all textures by solid colors
//#define DEBUG_SOLID_COLOR_TEXTURES
#endif
// How much to look into each direction for bias
const int C4LR_BiasDistanceX = 8;
const int C4LR_BiasDistanceY = 8;
// Workarounds to try if shader fails to compile
const char *C4LR_ShaderWorkarounds[] = {
"#define NO_TEXTURE_LOD_IN_FRAGMENT\n",
"#define BROKEN_ARRAYS_WORKAROUND\n",
"#define SCALER_IN_GPU\n",
};
const int C4LR_ShaderWorkaroundCount = sizeof(C4LR_ShaderWorkarounds) / sizeof(*C4LR_ShaderWorkarounds);
// Name used for the seperator texture
const char *const SEPERATOR_TEXTURE = "--SEP--";
// Map of uniforms to names in shader
static const char *GetUniformName(int iUniform)
{
switch(iUniform)
{
case C4LRU_LandscapeTex: return "landscapeTex";
case C4LRU_ScalerTex: return "scalerTex";
case C4LRU_MaterialTex: return "materialTex";
case C4LRU_Resolution: return "resolution";
case C4LRU_MatMap: return "matMap";
case C4LRU_MatMapTex: return "matMapTex";
case C4LRU_MaterialDepth:return "materialDepth";
}
assert(false);
return "mysterious";
}
C4LandscapeRenderGL::C4LandscapeRenderGL()
: iLandscapeShaderTime(0),
hVert(0), hFrag(0), hProg(0)
{
ZeroMem(Surfaces, sizeof(Surfaces));
ZeroMem(hMaterialTexture, sizeof(hMaterialTexture));
ZeroMem(hUniforms, sizeof(hUniforms));
}
C4LandscapeRenderGL::~C4LandscapeRenderGL()
{
Clear();
}
bool C4LandscapeRenderGL::Init(int32_t iWidth, int32_t iHeight, C4TextureMap *pTexs, C4GroupSet *pGraphics)
{
Clear();
// Safe info
this->iWidth = iWidth;
this->iHeight = iHeight;
this->pTexs = pTexs;
// Allocate landscape textures
if (!InitLandscapeTexture())
{
LogFatal("[!] Could not initialize landscape texture!");
return false;
}
// Build texture, er, texture
if (!InitMaterialTexture(pTexs))
{
LogFatal("[!] Could not initialize landscape textures for rendering!");
return false;
}
// Load sclaer
if (!LoadScaler(pGraphics))
{
LogFatal("[!] Could not load scaler!");
return false;
}
// Load shader
if (!LoadShaders(pGraphics))
{
LogFatal("[!] Could not initialize landscape shader!");
return false;
}
return true;
}
void C4LandscapeRenderGL::Clear()
{
ClearShaders();
// free textures
int i;
for (i = 0; i < C4LR_SurfaceCount; i++)
{
delete Surfaces[i];
Surfaces[i] = NULL;
}
for (i = 0; i < C4LR_MipMapCount; i++)
{
glDeleteObjectARB(hMaterialTexture[i]);
hMaterialTexture[i] = 0;
}
LandscapeShader.Clear();
LandscapeShaderPath.Clear();
iLandscapeShaderTime = 0;
}
bool C4LandscapeRenderGL::InitLandscapeTexture()
{
// Round up to nearest power of two
int iSfcWdt = 1, iSfcHgt = 1;
while(iSfcWdt < iWidth) iSfcWdt *= 2;
while(iSfcHgt < iHeight) iSfcHgt *= 2;
// One bit more of information to safe more space
if(iSfcWdt * 3 / 4 >= iWidth) iSfcWdt = iSfcWdt * 3 / 4;
if(iSfcHgt * 3 / 4 >= iHeight) iSfcHgt = iSfcHgt * 3 / 4;
// Create our surfaces
for(int i = 0; i < C4LR_SurfaceCount; i++)
{
Surfaces[i] = new C4Surface();
if(!Surfaces[i]->Create(iSfcWdt, iSfcHgt))
return false;
}
return true;
}
bool C4LandscapeRenderGL::InitMaterialTexture(C4TextureMap *pTexs)
{
// Populate our map with all needed textures
MaterialTextureMap.push_back(StdCopyStrBuf(""));
AddTexturesFromMap(pTexs);
// Determine depth to use
iMaterialTextureDepth = 1;
while(iMaterialTextureDepth < int32_t(MaterialTextureMap.size()))
iMaterialTextureDepth <<= 1;
// Find first (actual) texture
int iRefTexIx = 0; C4Texture *pRefTex; C4Surface *pRefSfc = NULL;
for(; pRefTex = pTexs->GetTexture(pTexs->GetTexture(iRefTexIx)); iRefTexIx++)
if(pRefSfc = pRefTex->Surface32)
break;
if(!pRefSfc)
return false;
// Get size for our textures. We might be limited by hardware
int iTexWdt = pRefSfc->Wdt, iTexHgt = pRefSfc->Hgt;
GLint iMaxTexSize;
glGetIntegerv(GL_MAX_3D_TEXTURE_SIZE, &iMaxTexSize);
if (iTexWdt > iMaxTexSize || iTexHgt > iMaxTexSize)
{
iTexWdt = Min(iTexWdt, iMaxTexSize);
iTexHgt = Min(iTexHgt, iMaxTexSize);
LogF(" gl: Material textures too large, GPU only supports %dx%d! Cropping might occur!", iMaxTexSize, iMaxTexSize);
}
if(iMaterialTextureDepth >= iMaxTexSize)
{
LogF(" gl: Too many material textures! GPU only supports 3D texture depth of %d!", iMaxTexSize);
return false;
}
// Compose together data of all textures
const int iBytesPP = pRefSfc->byBytesPP;
const int iTexSize = iTexWdt * iTexHgt * iBytesPP;
const int iSize = iTexSize * iMaterialTextureDepth;
BYTE *pData = new BYTE [iSize];
for(int i = 0; i < iMaterialTextureDepth; i++)
{
BYTE *p = pData + i * iTexSize;
// Get texture at position
const char *szTexture;
if(i < int32_t(MaterialTextureMap.size()))
szTexture = MaterialTextureMap[i].getData();
else
szTexture = "";
// Try to find the texture
C4Texture *pTex; C4Surface *pSurface;
if((pTex = pTexs->GetTexture(szTexture)) && (pSurface = pTex->Surface32))
{
#ifdef DEBUG_SOLID_COLOR_TEXTURES
// Just write a solid color that depends on the texture index
DWORD *texdata = reinterpret_cast<DWORD *>(p);
for (int y = 0; y < iTexHgt; ++y)
for (int x = 0; x < iTexWdt; ++x)
*texdata++ = RGBA((iTex & 48), (iTex & 3) * 16, (i & 12) * 4, 255);
continue;
#else
if(pSurface->iTexX != 1 || pSurface->iTexY != 1)
Log(" gl: Halp! Material texture is fragmented!");
else
{
// Size recheck
if(pSurface->Wdt != iTexWdt || pSurface->Hgt != iTexHgt)
LogF(" gl: texture %s size mismatch (%dx%d vs %dx%d)!", szTexture, pSurface->Wdt, pSurface->Hgt, iTexWdt, iTexHgt);
// Copy bytes
DWORD *texdata = reinterpret_cast<DWORD *>(p);
pSurface->Lock();
for (int y = 0; y < iTexHgt; ++y)
for (int x = 0; x < iTexWdt; ++x)
*texdata++ = pSurface->GetPixDw(x % pSurface->Wdt, y % pSurface->Hgt, false);
pSurface->Unlock();
continue;
}
#endif
}
// Seperator texture?
if(SEqual(szTexture, SEPERATOR_TEXTURE))
{
// Make some ugly stripes
DWORD *texdata = reinterpret_cast<DWORD *>(p);
for (int y = 0; y < iTexHgt; ++y)
for (int x = 0; x < iTexWdt; ++x)
*texdata++ = ((x + y) % 32 < 16 ? RGBA(255, 0, 0, 255) : RGBA(0, 255, 255, 255));
continue;
}
// If we didn't "continue" yet, we haven't written the texture yet. Make it transparent.
memset(p, 0, iTexSize);
}
// Clear error error(s?)
while(glGetError()) {}
int iMMLevels = 3;
while(iTexWdt <= (1 >> iMMLevels) || iTexHgt <= (1 >> iMMLevels))
iMMLevels--;
// Alloc 3D textures
glEnable(GL_TEXTURE_3D);
glGenTextures(iMMLevels, hMaterialTexture);
// Generate textures (mipmaps too!)
int iFullWdt = iTexWdt, iFullHgt = iTexHgt;
int iSizeSum = 0;
BYTE *pLastData = new BYTE [iSize / 4];
for(int iMMLevel = 0; iMMLevel < C4LR_MipMapCount; iMMLevel++)
{
// Scale the texture down for mip-mapping
if(iMMLevel) {
BYTE *pOut = pData;
BYTE *pIn[4] = {
pLastData, pLastData + iBytesPP,
pLastData + iBytesPP * iTexWdt, pLastData + iBytesPP * iTexWdt + iBytesPP
};
for (int i = 0; i < iMaterialTextureDepth; ++i)
for (int y = 0; y < iTexHgt / 2; ++y)
{
for (int x = 0; x < iTexWdt / 2; ++x)
{
for (int j = 0; j < iBytesPP; j++)
{
unsigned int s = 0;
s += *pIn[0]++; s += 3 * *pIn[1]++; s += 3 * *pIn[2]++; s += *pIn[3]++;
*pOut++ = BYTE(s / 8);
}
pIn[0] += iBytesPP; pIn[1] += iBytesPP; pIn[2] += iBytesPP; pIn[3] += iBytesPP;
}
pIn[0] += iBytesPP * iTexWdt; pIn[1] += iBytesPP * iTexWdt;
pIn[2] += iBytesPP * iTexWdt; pIn[3] += iBytesPP * iTexWdt;
}
iTexWdt /= 2; iTexHgt /= 2;
}
// Select texture
glBindTexture(GL_TEXTURE_3D, hMaterialTexture[iMMLevel]);
glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
// We fully expect to tile these
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_REPEAT);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
// Make it happen!
glTexImage3D(GL_TEXTURE_3D, 0, 4, iTexWdt, iTexHgt, iMaterialTextureDepth, 0, GL_BGRA,
iBytesPP == 2 ? GL_UNSIGNED_SHORT_4_4_4_4_REV : GL_UNSIGNED_INT_8_8_8_8_REV,
pData);
// Exchange buffers
BYTE *tmp = pLastData;
pLastData = pData;
pData = tmp;
// Statistics
iSizeSum += iTexWdt * iTexHgt * iMaterialTextureDepth * iBytesPP;
}
// Dispose of data
delete [] pData;
delete [] pLastData;
glDisable(GL_TEXTURE_3D);
// Check whether we were successful
if(int err = glGetError())
{
LogF(" gl: Could not load textures (error %d)", err);
return false;
}
// Announce the good news
LogF(" gl: Texturing uses %d slots at %dx%d, %d levels (%d MB total)",
static_cast<int>(MaterialTextureMap.size()),
iFullWdt, iFullHgt,
C4LR_MipMapCount,
iSizeSum / 1000000);
return true;
}
C4Rect C4LandscapeRenderGL::GetAffectedRect(C4Rect Rect)
{
Rect.Enlarge(C4LR_BiasDistanceX, C4LR_BiasDistanceY);
return Rect;
}
void C4LandscapeRenderGL::Update(C4Rect To, C4Landscape *pSource)
{
// clip to landscape size
To.Intersect(C4Rect(0,0,iWidth,iHeight));
// everything clipped?
if (To.Wdt<=0 || To.Hgt<=0) return;
// Lock surfaces
// We clear the affected region here because ClearBoxDw allocates the
// main memory buffer for the box, so that only that box needs to be
// sent to the gpu, and not the whole texture, or every pixel
// separately. It's an important optimization.
for (int i = 0; i < C4LR_SurfaceCount; i++) {
if (!Surfaces[i]->Lock()) return;
Surfaces[i]->ClearBoxDw(To.x, To.y, To.Wdt, To.Hgt);
}
// Initialize up & down arrays
int x, y;
int *pUp = new int [To.Wdt * 2];
int *pDown = pUp + To.Wdt;
for(x = 0; x < To.Wdt; x++) {
int iSum = 0;
for(y = 1; y < Min(C4LR_BiasDistanceY, To.y+1); y++)
iSum += pSource->_GetPlacement(To.x+x, To.y-y);
pUp[x] = iSum;
iSum = 0;
for(y = 1; y < Min(C4LR_BiasDistanceY, iHeight - To.y); y++)
iSum += pSource->_GetPlacement(To.x+x, To.y+y);
pDown[x] = iSum;
}
// Get tex refs (shortcut, we will use them quite heavily)
C4TexRef *TexRefs[C4LR_SurfaceCount];
x = y = 0;
for(int i = 0; i < C4LR_SurfaceCount; i++)
Surfaces[i]->GetTexAt(&TexRefs[i], x, y);
// Go through it from top to bottom
for(y = 0; y < To.Hgt; y++) {
// Initialize left & right
int iLeft = 0;
int iRight = 0;
for(x = 1; x < Min(C4LR_BiasDistanceX, To.x+1); x++)
iLeft += pSource->_GetPlacement(To.x-x,To.y+y);
for(x = 1; x < Min(C4LR_BiasDistanceX, iWidth - To.x); x++)
iRight += pSource->_GetPlacement(To.x+x,To.y+y);
for(x = 0; x < To.Wdt; x++) {
// Biases
int iPix = pSource->_GetPix(To.x+x, To.y+y);
int iPlac = pSource->_GetPlacement(To.x+x, To.y+y);
int iMat = pSource->_GetMat(To.x+x, To.y+y);
int iHBias = Max(0, iPlac * (C4LR_BiasDistanceY-1) - iRight) -
Max(0, iPlac * (C4LR_BiasDistanceY-1) - iLeft);
int iVBias = Max(0, iPlac * (C4LR_BiasDistanceY-1) - pDown[x]) -
Max(0, iPlac * (C4LR_BiasDistanceY-1) - pUp[x]);
// Maximum placement differences that make a difference in the result,
// after which we are at the limits of what can be packed into a byte
const int iMaxPlacDiff = 40;
int iHBiasScaled = BoundBy(iHBias * 127 / iMaxPlacDiff / C4LR_BiasDistanceX + 128, 0, 255);
int iVBiasScaled = BoundBy(iVBias * 127 / iMaxPlacDiff / C4LR_BiasDistanceY + 128, 0, 255);
// Visit our neighbours
int iNeighbours[8] = { 0, 0, 0, 0, 0, 0, 0, 0 };
if(To.y+y > 0) {
if(To.x+x > 0)
iNeighbours[0] = pSource->_GetPix(To.x+x-1, To.y+y-1);
iNeighbours[1] = pSource->_GetPix(To.x+x, To.y+y-1);
if(To.x+x < iWidth-1)
iNeighbours[2] = pSource->_GetPix(To.x+x+1, To.y+y-1);
}
if(To.x+x > 0)
iNeighbours[3] = pSource->_GetPix(To.x+x-1, To.y+y);
if(To.x+x < iWidth-1)
iNeighbours[4] = pSource->_GetPix(To.x+x+1, To.y+y);
if(To.y+y+1 < iHeight-1) {
if(To.x+x > 0)
iNeighbours[5] = pSource->_GetPix(To.x+x-1, To.y+y+1);
iNeighbours[6] = pSource->_GetPix(To.x+x, To.y+y+1);
if(To.x+x < iWidth-1)
iNeighbours[7] = pSource->_GetPix(To.x+x+1, To.y+y+1);
}
// Look for highest-placement material in our surroundings
int iMaxPix = iPix, iMaxPlace = iPlac, i;
for(i = 0; i < 8; i++) {
int iTempPlace = MatPlacement(PixCol2Mat(iNeighbours[i]));
if(iTempPlace > iMaxPlace || (iTempPlace == iMaxPlace && iNeighbours[i] > iMaxPix) ) {
iMaxPix = iNeighbours[i]; iMaxPlace = iTempPlace;
}
}
// Scaler calculation depends on whether this is the highest-placement material around
int iScaler = 0;
if(iMaxPix == iPix) {
// If yes, we consider all other materials as "other"
for(i = 0; i < 8; i++)
if(iNeighbours[i] == iPix)
iScaler += (1<<i);
} else {
// Otherwise, we *only* consider the highest-placement material as "other"
for(i = 0; i < 8; i++)
if(iNeighbours[i] != iMaxPix)
iScaler += (1<<i);
}
// Collect data to save per pixel
unsigned char data[C4LR_SurfaceCount * 4];
memset(data, 0, sizeof(data));
data[C4LR_Material] = iPix;
data[C4LR_BiasX] = iHBiasScaled;
data[C4LR_BiasY] = iVBiasScaled;
data[C4LR_Scaler] = iScaler;
for(int i = 0; i < C4LR_SurfaceCount; i++)
TexRefs[i]->SetPix4(To.x+x, To.y+y,
RGBA(data[i*4+0], data[i*4+1], data[i*4+2], data[i*4+3]));
// Update left & right
if(To.x+x + 1 < iWidth)
iRight -= pSource->_GetPlacement(To.x+x + 1, To.y+y);
if(To.x+x + C4LR_BiasDistanceX < iWidth)
iRight += pSource->_GetPlacement(To.x+x + C4LR_BiasDistanceX, To.y+y);
iLeft += pSource->_GetPlacement(To.x+x, To.y+y);
if(To.x+x - C4LR_BiasDistanceX - 1 >= 0)
iLeft -= pSource->_GetPlacement(To.x+x - C4LR_BiasDistanceX - 1, To.y+y);
// Update up & down arrays
if(To.y+y + 1 < iHeight)
pDown[x] -= pSource->_GetPlacement(To.x+x, To.y+y + 1);
if(To.y+y + C4LR_BiasDistanceY < iHeight)
pDown[x] += pSource->_GetPlacement(To.x+x, To.y+y + C4LR_BiasDistanceY);
pUp[x] += pSource->_GetPlacement(To.x+x, To.y+y);
if(To.y+y - C4LR_BiasDistanceY + 1 >= 0) {
pUp[x] -= pSource->_GetPlacement(To.x+x, To.y+y - C4LR_BiasDistanceY + 1);
}
}
}
// done
delete[] pUp;
for (int i = 0; i < C4LR_SurfaceCount; i++)
Surfaces[i]->Unlock();
}
void C4LandscapeRenderGL::DumpInfoLog(const char *szWhat, GLhandleARB hShader, int32_t iWorkaround)
{
// Get length of info line
int iLength = 0;
glGetObjectParameterivARB(hShader, GL_OBJECT_INFO_LOG_LENGTH_ARB, &iLength);
if(iLength <= 1) return;
// Allocate buffer, get data
char *pBuf = new char [iLength + 1];
int iActualLength = 0;
glGetInfoLogARB(hShader, iLength, &iActualLength, pBuf);
if(iActualLength > iLength || iActualLength <= 0) return;
// Terminate, log
pBuf[iActualLength] = '\0';
LogSilentF(" gl: Compiling %s %d:", szWhat, iWorkaround);
LogSilent(pBuf);
delete[] pBuf;
}
int C4LandscapeRenderGL::GetObjectStatus(GLhandleARB hObj, GLenum type)
{
int iStatus = 0;
glGetObjectParameterivARB(hObj, type, &iStatus);
return iStatus;
}
GLhandleARB C4LandscapeRenderGL::CreateShader(GLenum iShaderType, const char *szWhat, const char *szCode, int32_t iWorkaround)
{
// Create shader
GLhandleARB hShader = glCreateShaderObjectARB(iShaderType);
// Find #version
StdStrBuf Version("");
const char *szCodeRest = szCode;
if (const char *szVersion = SSearch(szCode, "#version"))
{
while (*szVersion && *szVersion != '\n')
szVersion++;
if (*szVersion == '\n')
szVersion++;
Version.Copy(szCode, szVersion - szCode);
szCodeRest = szVersion;
}
// Build code
const char *szCodes[C4LR_ShaderWorkaroundCount + 2];
szCodes[0] = Version.getData();
for(int i = 0; i < C4LR_ShaderWorkaroundCount; i++)
if(iWorkaround & (1 << i))
szCodes[i+1] = C4LR_ShaderWorkarounds[i];
else
szCodes[i+1] = "";
szCodes[C4LR_ShaderWorkaroundCount+1] = szCodeRest;
// Compile
glShaderSourceARB(hShader, C4LR_ShaderWorkaroundCount + 2, szCodes, 0);
glCompileShaderARB(hShader);
// Dump any information to log
DumpInfoLog(szWhat, hShader, iWorkaround);
// Success?
if(GetObjectStatus(hShader, GL_OBJECT_COMPILE_STATUS_ARB) == 1)
return hShader;
// Did not work :/
glDeleteObjectARB(hShader);
return 0;
}
bool C4LandscapeRenderGL::InitShaders()
{
// Already initialized or no shader load?
if(hProg || LandscapeShader.getLength() <= 0)
return false;
// No support?
if(!GLEW_ARB_fragment_program)
{
Log(" gl: no shader support!");
return false;
}
// Try all workarounds until one works
int iWorkaround;
for(iWorkaround = 0; iWorkaround < (1 << C4LR_ShaderWorkaroundCount); iWorkaround++)
{
// Create trivial fragment shader
const char *szVert = "#version 110\nvoid main() { gl_TexCoord[0] = gl_MultiTexCoord0; gl_Position = ftransform(); } ";
hVert = CreateShader(GL_VERTEX_SHADER_ARB, "vertex shader", szVert, iWorkaround);
hFrag = CreateShader(GL_FRAGMENT_SHADER_ARB, "fragment shader", LandscapeShader.getData(), iWorkaround);
if(!hFrag || !hVert)
continue;
// Link program
hProg = glCreateProgramObjectARB();
glAttachObjectARB(hProg, hVert);
glAttachObjectARB(hProg, hFrag);
glLinkProgramARB(hProg);
// Link successful?
DumpInfoLog("shader program", hProg, iWorkaround);
if(GetObjectStatus(hProg, GL_OBJECT_LINK_STATUS_ARB) == 1)
break;
// Clear up
glDetachObjectARB(hProg, hVert);
glDetachObjectARB(hProg, hFrag);
glDeleteObjectARB(hVert);
glDeleteObjectARB(hFrag);
glDeleteObjectARB(hProg);
hProg = hVert = hFrag = 0;
}
// Did not get it to work?
if(!hProg)
{
Log(" gl: Failed to link shader!");
return false;
}
LogF(" gl: Shader %d linked successfully", iWorkaround);
// Get uniform locations. Note this is expected to fail for a few of them
// because the respective uniforms got optimized out!
for (int i = 0; i < C4LRU_Count; i++)
hUniforms[i] = glGetUniformLocationARB(hProg, GetUniformName(i));
// Success?
if(int err = glGetError())
{
LogF(" gl: Error code %d while linking shader!", err);
return false;
}
return true;
}
void C4LandscapeRenderGL::ClearShaders()
{
if (!hProg) return;
// Need to be detached, then deleted
glDetachObjectARB(hProg, hFrag);
glDetachObjectARB(hProg, hVert);
glDeleteObjectARB(hFrag);
glDeleteObjectARB(hVert);
glDeleteObjectARB(hProg);
hFrag = hVert = hProg = 0;
ZeroMem(hUniforms, sizeof(hUniforms));
}
bool C4LandscapeRenderGL::LoadShaders(C4GroupSet *pGroups)
{
// First, clear out all existing shaders
ClearShaders();
// Search for our shaders
C4Group *pGroup = pGroups->FindEntry(C4CFN_LandscapeShader);
if(!pGroup) return false;
// Load it, save the path for later reloading
if(!pGroup->LoadEntryString(C4CFN_LandscapeShader, &LandscapeShader))
return false;
// If it physically exists, save back file name
if(FileExists(pGroup->GetFullName().getData()))
{
LandscapeShaderPath.Format("%s" DirSep C4CFN_LandscapeShader, pGroup->GetFullName().getData());
iLandscapeShaderTime = FileTime(LandscapeShaderPath.getData());
}
// Initialize
return InitShaders();
}
bool C4LandscapeRenderGL::LoadScaler(C4GroupSet *pGroups)
{
// Search for scaler
C4Group *pGroup = pGroups->FindEntry(C4CFN_LandscapeScaler);
if(!pGroup) return false;
// Load scaler from group
if(!fctScaler.Load(*pGroup, C4CFN_LandscapeScaler))
return false;
// Check size
const int iOrigWdt = 8 * 3, iOrigHgt = 4 * 8 * 3;
const int iFactor = fctScaler.Wdt / iOrigWdt;
if(fctScaler.Wdt != iFactor * iOrigWdt || fctScaler.Hgt != iFactor * iOrigHgt)
{
LogF(" gl: Unexpected scaler size - should be multiple of %dx%d!", iOrigWdt, iOrigHgt);
return false;
}
// Walk through all lookups we have in the texture and decide where
// to look for the "other" pixel. This might not be unique later on,
// so it is a good idea to have a proper priority order here.
fctScaler.Surface->Lock();
int i;
for (i = 0; i < 8 * 4 * 8; i++) {
// Decode from ID what pixels are expected to be set in this case
enum Px { NW, N, NE, W, E, SW, S, SE, X };
int p_x[9] = { -1, 0, 1, -1, 1, -1, 0, 1, 0 };
int p_y[9] = { -1, -1, -1, 0, 0, 1, 1, 1, 0 };
bool pxAt[X];
for(int j = 0; j < X; j++)
pxAt[j] = !!(i & (1 << j));
// Oc = octant borders. Set up arrays to get righthand border
// of an octant, in a way that we can easily rotate further.
enum Oc { NWW, NEE, SWW, SEE, NNW, NNE, SSW, SSE };
int p2a[8] = { 5, 6, 7, 4, 0, 3, 2, 1 };
Oc a2o[8] = { SEE, SSE, SSW, SWW, NWW, NNW, NNE, NEE };
// Decide in which octant we want to interpolate towards
// which pixel. Pick the nearest unset pixel using a special
// priority order.
Px opx[8] = { X,X,X,X,X,X,X,X };
#define INTERPOLATE(x,da) do { \
int y = a2o[(8+p2a[x]+(da)) % 8];\
if (!pxAt[x] && opx[y] == X) opx[y] = x; \
} while(false)
for(int j = 0; j < 4; j++) {
// vertical
INTERPOLATE(N, j); INTERPOLATE(N, -j-1);
INTERPOLATE(S, j); INTERPOLATE(S, -j-1);
// horizontal
INTERPOLATE(W, j); INTERPOLATE(W, -j-1);
INTERPOLATE(E, j); INTERPOLATE(E, -j-1);
// diagonals
INTERPOLATE(NW, j); INTERPOLATE(NW, -j-1);
INTERPOLATE(SW, j); INTERPOLATE(SW, -j-1);
INTERPOLATE(NE, j); INTERPOLATE(NE, -j-1);
INTERPOLATE(SE, j); INTERPOLATE(SE, -j-1);
}
// Decide in which octants we will not interpolate normals.
// It doesn't make sense when there's another material in that
// general direction, as then the bias of that will factor into
// the interpolation, giving bright borders on dark shading,
// and vice-versa.
bool noNormals[8];
noNormals[NNW] = noNormals[NWW] = !pxAt[W] || !pxAt[NW] || !pxAt[N];
noNormals[NNE] = noNormals[NEE] = !pxAt[E] || !pxAt[NE] || !pxAt[N];
noNormals[SSW] = noNormals[SWW] = !pxAt[W] || !pxAt[SW] || !pxAt[S];
noNormals[SSE] = noNormals[SEE] = !pxAt[E] || !pxAt[SE] || !pxAt[S];
// Set blue and green components to relative coordinates of
// "other" pixel, and alpha to mix param for normals
const int x0 = (i % 8) * 3 * iFactor;
const int y0 = (i / 8) * 3 * iFactor;
const int iPxs = 3 * iFactor;
int y, x;
for(y = 0; y < iPxs; y++)
{
for(x = 0; x < iPxs; x++)
{
// Find out in which octagon we are
int oct = 0;
if(2 * x >= iPxs) oct+=1;
if(2 * y >= iPxs) oct+=2;
if((x >= y) != (x >= iPxs - y)) oct+=4;
// Get pixel, do processing
DWORD pix = fctScaler.Surface->GetPixDw(x0+x, y0+y, false);
BYTE val = GetGreenValue(pix);
if(val >= 250) val = 255;
BYTE bx = 64 * (p_x[opx[oct]] + 1);
BYTE by = 64 * (p_y[opx[oct]] + 1);
BYTE bn = (noNormals[oct] ? 255 : 1);
fctScaler.Surface->SetPixDw(x0+x, y0+y, RGBA(val, bx, by, bn));
}
}
}
return fctScaler.Surface->Unlock();
}
void C4LandscapeRenderGL::RefreshShaders()
{
// File changed?
if(!LandscapeShaderPath.isNull() &&
FileTime(LandscapeShaderPath.getData()) != iLandscapeShaderTime)
{
ClearShaders();
// Load new shader
char szParentPath[_MAX_PATH+1]; C4Group Group;
GetParentPath(LandscapeShaderPath.getData(),szParentPath);
if(!Group.Open(szParentPath) ||
!Group.LoadEntryString(GetFilename(LandscapeShaderPath.getData()),&LandscapeShader) ||
!Group.Close())
return;
// Reinitialize
InitShaders();
iLandscapeShaderTime = FileTime(LandscapeShaderPath.getData());
}
}
int32_t C4LandscapeRenderGL::LookupTextureTransition(const char *szFrom, const char *szTo)
{
// Is this actually a transition? Otherwise we're looking for a single texture
bool fTransit = !SEqual(szFrom, szTo);
// Look for a position in the map where the textures appear in sequence
uint32_t i;
for(i = 1; i < MaterialTextureMap.size(); i++)
{
if(SEqual(szFrom, MaterialTextureMap[i].getData()))
{
// Single texture: We're done
if(!fTransit) return i;
// Check next texture as well
if(i + 1 >= MaterialTextureMap.size())
return -1;
if(SEqual(szTo, MaterialTextureMap[i+1].getData()))
return i;
}
}
return -1;
}
void C4LandscapeRenderGL::AddTextureTransition(const char *szFrom, const char *szTo)
{
// Empty?
if (!szFrom || !szTo) return;
// First try the lookup (both directions)
if (LookupTextureTransition(szFrom, szTo) >= 0) return;
if (LookupTextureTransition(szTo, szFrom) >= 0) return;
// Single texture? Add it as single
if (SEqual(szTo, szFrom))
MaterialTextureMap.push_back(StdCopyStrBuf(szFrom));
// Have one of the textures at the end of the list?
else if(SEqual(MaterialTextureMap.back().getData(), szFrom))
MaterialTextureMap.push_back(StdCopyStrBuf(szTo));
else if(SEqual(MaterialTextureMap.back().getData(), szTo))
MaterialTextureMap.push_back(StdCopyStrBuf(szFrom));
else
{
// Otherwise add both
MaterialTextureMap.push_back(StdCopyStrBuf(szFrom));
MaterialTextureMap.push_back(StdCopyStrBuf(szTo));
}
}
void C4LandscapeRenderGL::AddTextureAnim(const char *szTextureAnim)
{
if(!szTextureAnim) return;
#ifdef DEBUG_SEPERATOR_TEXTURES
// Save back count of textures at start
uint32_t iStartTexCount = MaterialTextureMap.size();
#endif
// Add all individual transitions
const char *pFrom = szTextureAnim;
for(;;)
{
// Get next phase
const char *pTo = strchr(pFrom, '-');
if(!pTo) pTo = szTextureAnim; else pTo++;
// Add transition
StdStrBuf From, To;
From.CopyUntil(pFrom, '-');
To.CopyUntil(pTo, '-');
AddTextureTransition(From.getData(), To.getData());
// Advance
if(pTo == szTextureAnim) break;
pFrom = pTo;
}
#ifdef DEBUG_SEPERATOR_TEXTURES
// Add a seperator texture, if we added any new ones
if(MaterialTextureMap.size() > iStartTexCount)
MaterialTextureMap.push_back(StdCopyStrBuf(SEPERATOR_TEXTURE));
#endif
}
void C4LandscapeRenderGL::AddTexturesFromMap(C4TextureMap *pMap)
{
// Go through used texture (animations) and add all phases to our map
// Note: We can be smarter here, for example add longer animations
// first in order to make better reuse of 3D texture slots.
// We could even make a full-blown optimization problem out of it.
// Future work...
const C4TexMapEntry *pEntry;
for(int32_t i = 0; pEntry = pMap->GetEntry(i); i++)
// ToDo: Properly handle jumping back
AddTextureAnim(pEntry->GetTextureName());
}
void C4LandscapeRenderGL::BuildMatMap(GLfloat *pFMap, GLubyte *pIMap)
{
// TODO: Still merely an inefficient placeholder for things to come...
// Build material-texture map (depth parameter where to find appropriate texture)
for(int pix = 0; pix < 256; pix++)
{
// Look up indexed entry
const C4TexMapEntry *pEntry = pTexs->GetEntry(PixCol2Tex(BYTE(pix)));
if(!pEntry->GetTextureName())
{
// Undefined textures transparent
if(pFMap) pFMap[pix] = 0.5 / iMaterialTextureDepth;
if(pIMap) pIMap[pix] = 0;
continue;
}
// Got animation?
int iPhases = 1; const char *p = pEntry->GetTextureName();
while(p = strchr(p, '-')) { p++; iPhases++; }
// Hard-coded hack. Fix me!
const int iPhaseLength = 300;
float phase = (iPhases == 1 ? 0 : float(GetTime() % (iPhases * iPhaseLength)) / iPhaseLength);
// Find our transition
const char *pFrom = pEntry->GetTextureName();
float gTexCoo = 0;
for(int iP = 0;; iP++)
{
// Get next phase
const char *pTo = strchr(pFrom, '-');
if(!pTo) pTo = pEntry->GetTextureName(); else pTo++;
// Add transition
if(iP == int(phase))
{
StdStrBuf From, To;
From.CopyUntil(pFrom, '-');
To.CopyUntil(pTo, '-');
// Find transition
int iTrans;
if ((iTrans = LookupTextureTransition(From.getData(), To.getData())) >= 0)
gTexCoo = float(iTrans) + fmod(phase, 1.0f);
else if ((iTrans = LookupTextureTransition(To.getData(), From.getData())) >= 0)
gTexCoo = float(iTrans) + 1.0 - fmod(phase, 1.0f);
break;
}
// Advance
pFrom = pTo;
}
// Assign texture
if(pFMap) pFMap[pix] = (gTexCoo + 0.5) / iMaterialTextureDepth;
if(pIMap) pIMap[pix] = int((gTexCoo * 256.0 / iMaterialTextureDepth) + 0.5);
}
}
void C4LandscapeRenderGL::Draw(const C4TargetFacet &cgo)
{
// Must have GL and be initialized
if(!pGL && !hProg) return;
// prepare rendering to surface
C4Surface *sfcTarget = cgo.Surface;
if (!pGL->PrepareRendering(sfcTarget)) return;
#ifdef AUTO_RELOAD_SHADERS
RefreshShaders();
#endif // AUTO_RELOAD_SHADERS
// Clear error(s?)
while(glGetError()) {}
// Activate shader
glUseProgramObjectARB(hProg);
// Bind data
if (hUniforms[C4LRU_Resolution] != GLhandleARB(-1))
glUniform2fARB(hUniforms[C4LRU_Resolution], Surfaces[0]->Wdt, Surfaces[0]->Hgt);
if (hUniforms[C4LRU_MatMap] != GLhandleARB(-1))
{
GLfloat MatMap[256];
BuildMatMap(MatMap, NULL);
glUniform1fvARB(hUniforms[C4LRU_MatMap], 256, MatMap);
}
if (hUniforms[C4LRU_MaterialDepth] != GLhandleARB(-1))
glUniform1iARB(hUniforms[C4LRU_MaterialDepth], iMaterialTextureDepth);
// Setup facilities for texture unit allocation (gimme local functions...)
int iUnit = 0; int iUnitMap[32]; ZeroMem(iUnitMap, sizeof(iUnitMap));
#define ALLOC_UNIT(hUniform, iType) do { \
if(hUniform != GLhandleARB(-1)) glUniform1iARB(hUniform, iUnit); \
glActiveTexture(GL_TEXTURE0 + iUnit); \
iUnitMap[iUnit] = iType; \
glEnable(iType); \
iUnit++; \
assert(iUnit < 32); \
} while(false)
// Start binding textures
if(hUniforms[C4LRU_ScalerTex] != GLhandleARB(-1))
{
ALLOC_UNIT(hUniforms[C4LRU_ScalerTex], GL_TEXTURE_2D);
glBindTexture(GL_TEXTURE_2D, fctScaler.Surface->ppTex[0]->texName);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
}
if(hUniforms[C4LRU_MaterialTex] != GLhandleARB(-1))
{
ALLOC_UNIT(hUniforms[C4LRU_MaterialTex], GL_TEXTURE_3D);
// Decide which mip-map level to use
double z = 2.0; int iMM = 0;
while(pGL->Zoom < z && iMM + 1 <C4LR_MipMapCount)
{ z /= 2; iMM++; }
glBindTexture(GL_TEXTURE_3D, hMaterialTexture[iMM]);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
}
if(hUniforms[C4LRU_LandscapeTex] != GLhandleARB(-1))
{
GLint iLandscapeUnits[C4LR_SurfaceCount];
for(int i = 0; i < C4LR_SurfaceCount; i++)
{
iLandscapeUnits[i] = iUnit;
ALLOC_UNIT(GLhandleARB(-1), GL_TEXTURE_2D);
glBindTexture(GL_TEXTURE_2D, Surfaces[i]->ppTex[0]->texName);
if (pGL->Zoom != 1.0)
{
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
}
else
{
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
}
}
glUniform1ivARB(hUniforms[C4LRU_LandscapeTex], C4LR_SurfaceCount, iLandscapeUnits);
}
if(hUniforms[C4LRU_MatMapTex] != GLhandleARB(-1))
{
ALLOC_UNIT(hUniforms[C4LRU_MatMapTex], GL_TEXTURE_1D);
GLubyte MatMap[256];
BuildMatMap(NULL, MatMap);
glTexImage1D(GL_TEXTURE_1D, 0, 1, 256, 0, GL_RED, GL_UNSIGNED_BYTE, MatMap);
glTexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
}
// set up blit data as rect
FLOAT_RECT fTexBlt, tTexBlt;
float fx = float(cgo.TargetX), fy = float(cgo.TargetY);
fTexBlt.left = fx;
fTexBlt.top = fy;
fTexBlt.right = fx + float(cgo.Wdt);
fTexBlt.bottom= fy + float(cgo.Hgt);
// apply Zoom
float tx = float(cgo.X), ty = float(cgo.Y);
pGL->ApplyZoom(tx, ty);
tTexBlt.left = tx;
tTexBlt.top = ty;
tTexBlt.right = tx + float(cgo.Wdt) * pGL->Zoom;
tTexBlt.bottom= ty + float(cgo.Hgt) * pGL->Zoom;
// blit positions
C4BltVertex Vtx[4];
Vtx[0].ftx = tTexBlt.left; Vtx[0].fty = tTexBlt.top;
Vtx[1].ftx = tTexBlt.right; Vtx[1].fty = tTexBlt.top;
Vtx[2].ftx = tTexBlt.right; Vtx[2].fty = tTexBlt.bottom;
Vtx[3].ftx = tTexBlt.left; Vtx[3].fty = tTexBlt.bottom;
Vtx[0].tx = fTexBlt.left; Vtx[0].ty = fTexBlt.top;
Vtx[1].tx = fTexBlt.right; Vtx[1].ty = fTexBlt.top;
Vtx[2].tx = fTexBlt.right; Vtx[2].ty = fTexBlt.bottom;
Vtx[3].tx = fTexBlt.left; Vtx[3].ty = fTexBlt.bottom;
for (int i=0; i<4; ++i)
{
Vtx[i].tx /= float(Surfaces[0]->Wdt);
Vtx[i].ty /= float(Surfaces[0]->Hgt);
Vtx[i].ftz = 0;
Vtx[i].color[0] = 255;
Vtx[i].color[1] = 255;
Vtx[i].color[2] = 255;
Vtx[i].color[3] = 255;
//DwTo4UB(RGBA(255, 255, 255, 255), Vtx[i].color);
}
// color modulation?
//DWORD dwModClr = BlitModulated ? BlitModulateClr : 0xffffffff;
//for (int i=0; i<4; ++i)
// DwTo4UB(dwModClr | dwModMask, Vtx[i].color);
// Blend it
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
// Blit
glInterleavedArrays(GL_T2F_C4UB_V3F, sizeof(C4BltVertex), Vtx);
glDrawArrays(GL_QUADS, 0, 4);
// Remove shader
glUseProgramObjectARB(0);
// Unbind textures
while(iUnit > 0)
{
iUnit--;
glActiveTexture(GL_TEXTURE0 + iUnit);
glDisable(iUnitMap[iUnit]);
}
// Got an error?
if(int err = glGetError())
{
LogF("GL error: %d", err /*, gluErrorString(err)*/);
}
}
#endif // USE_GL