Created
April 15, 2012 04:28
-
-
Save rygorous/2390021 to your computer and use it in GitHub Desktop.
ye olde texgen code
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#include "types.h" | |
#include "opsys.h" | |
#include "texture.h" | |
#include "opdata.h" | |
#include "rtlib.h" | |
#include "math3d_2.h" | |
// for text stuff | |
#define WIN32_LEAN_AND_MEAN | |
#include <windows.h> | |
// ---- tool functions | |
static sF32 getGrayscale(const sU16* ptr) | |
{ | |
return (ptr[0] * 0.11f + ptr[1] * 0.59f + ptr[2] * 0.30f) / 32767.0f; | |
} | |
static sF32 calcHeights(const sU16* hmap, sInt x, sInt y, sInt w, sInt h, sInt xo, sInt yo, sF32& hx, sF32& hy) | |
{ | |
const sF32 hgt = getGrayscale(hmap + (y * w + x) * 4); | |
hx = getGrayscale(hmap + (y * w + ((x + xo) & (w - 1))) * 4) - hgt; | |
hy = getGrayscale(hmap + (((y + yo) & (h - 1)) * w + x) * 4) - hgt; | |
return hgt; | |
} | |
static void makeNormalMap(sU16* dst, const sU16* src, sInt w, sInt h, sF32 hScale) | |
{ | |
sInt x, y; | |
sF32 nx, ny, nz; | |
for (y = 0; y < h; y++) | |
{ | |
for (x = 0; x < w; x++) | |
{ | |
dst[3] = calcHeights(src, x, y, w, h, 1, 1, nx, ny) * 32767.0f; | |
nx = -nx * hScale; | |
ny = -ny * hScale; | |
nz = 16383.0f / sqrt(nx * nx + ny * ny + 1.0f); | |
dst[0] = nz + 16384; | |
dst[1] = ny * nz + 16384; | |
dst[2] = nx * nz + 16384; | |
dst += 4; | |
} | |
} | |
} | |
static void sampleBilinear(sU16* outPtr, const sU16* srcPtr, sS32 xp, sS32 yp, sInt w, sInt h) | |
{ | |
sU32 rxp = (xp >> 16) & (w - 1), rxpn=(rxp + 1) & (w - 1); | |
sU32 ryp = (yp >> 16) & (h - 1); | |
const sU16 *pt1 = srcPtr + ((ryp * w) + rxp) * 4; | |
const sU16 *pt2 = srcPtr + ((ryp * w) + rxpn) * 4; | |
ryp = (ryp + 1) & (h - 1); | |
const sU16 *pt3 = srcPtr + ((ryp * w) + rxp) * 4; | |
const sU16 *pt4 = srcPtr + ((ryp * w) + rxpn) * 4; | |
for (sInt k = 0; k < 4; k++) | |
{ | |
sInt xw = (xp >> 8) & 0xff; | |
sU32 t1 = ((sU32) pt1[k] << 8) + xw * (pt2[k] - pt1[k]); | |
sU32 t2 = ((sU32) pt3[k] << 8) + xw * (pt4[k] - pt3[k]); | |
sU32 t3 = (t1 << 8) + ((yp & 0xff00) >> 8) * ((sS32) (t2 - t1)); | |
outPtr[k] = t3 >> 16; | |
} | |
} | |
static const sU8* getColor(sU8* dest,const sU8* src,sInt code) | |
{ | |
switch(code & 3) | |
{ | |
case 0: // other | |
dest[0] = *src++; | |
dest[1] = *src++; | |
dest[2] = *src++; | |
break; | |
case 1: // gray | |
dest[0] = dest[1] = dest[2] = *src++; | |
break; | |
case 2: // black | |
dest[0] = dest[1] = dest[2] = 0; | |
break; | |
case 3: // white | |
dest[0] = dest[1] = dest[2] = 255; | |
break; | |
} | |
return src; | |
} | |
// ---- random dots | |
class frOpTGRandomDots: public frTextureOperator | |
{ | |
sU16 dotCount; | |
sU32 randomSeed; | |
sU16 color[9]; | |
void doProcess() | |
{ | |
frBitmap* data = getData(); | |
sU16* outData = data->data; | |
for (sInt i = 0; i < data->size; i++) | |
{ | |
outData[0] = color[0]; | |
outData[1] = color[1]; | |
outData[2] = color[2]; | |
outData[3] = 0; | |
outData += 4; | |
} | |
sU32 seed = randomSeed; | |
for (sInt i = 0; i < dotCount; i++) | |
{ | |
const sU16* pColor = color + (i & 1) * 3 + 3; | |
sInt x = (seed >> 16) & (data->w - 1); seed = seed * 375375621 + 1; | |
sInt y = (seed >> 16) & (data->h - 1); seed = seed * 375375621 + 1; | |
sU16* out = data->data + (y * data->w + x) * 4; | |
out[0] = pColor[0]; | |
out[1] = pColor[1]; | |
out[2] = pColor[2]; | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
dotCount = ((sU16 *) strm)[0]; | |
randomSeed = (((sU16 *) strm)[1] + 303) * 0x31337303; | |
randomSeed ^= randomSeed >> 5; | |
strm += 4; | |
for (sInt i = 0; i < 9; i++) | |
color[i] = *strm++ << 7; | |
return strm; | |
} | |
}; | |
frOperator* create_TGRandomDots() { return new frOpTGRandomDots; } | |
// ---- trivial ops | |
class frOpTOTrivial: public frTextureOperator | |
{ | |
sU8 color[6]; | |
void doProcess() | |
{ | |
sU16* dst = data->data; | |
switch(def->id) | |
{ | |
case 0x01: // rectangle | |
{ | |
for (sInt y = 0; y < data->h; y++) | |
{ | |
for (sInt x = 0; x < data->w; x++) | |
{ | |
const sInt tx = (x << 8) / data->w; | |
const sInt ty = (y << 8) / data->h; | |
const sU8* clr = color; | |
if (tx >= param[0] && tx <= param[1] && ty >= param[2] && ty <= param[3]) | |
clr += 3; | |
*dst++ = clr[0] << 7; | |
*dst++ = clr[1] << 7; | |
*dst++ = clr[2] << 7; | |
*dst++ = 0; | |
} | |
} | |
} | |
break; | |
case 0x02: // solid | |
{ | |
for (sInt i = 0; i < data->size; i++) | |
{ | |
*dst++ = param[0] << 7; | |
*dst++ = param[1] << 7; | |
*dst++ = param[2] << 7; | |
*dst++ = 0; | |
} | |
} | |
break; | |
case 0x07: // make normal map | |
makeNormalMap(dst, getInputData(0)->data, data->w, data->h, *reinterpret_cast<const sU16*>(param) / 1024.0f); | |
break; | |
case 0x10: // displace | |
{ | |
const sU16* source = getInputData(0)->data; | |
const sU16* height = getInputData(1)->data; | |
sS32 xa = reinterpret_cast<const sS16*>(param)[1] * 16.0f / 32767.0f * 65536.0f * 64.0f; | |
sS32 ya = reinterpret_cast<const sS16*>(param)[2] * 16.0f / 32767.0f * 65536.0f * 64.0f; | |
for (sInt y = 0; y < data->h; y++) | |
{ | |
for (sInt x = 0; x < data->w; x++) | |
{ | |
sF32 hx, hy; | |
calcHeights(height, x, y, data->w, data->h, param[0], param[1], hx, hy); | |
sampleBilinear(dst, source, (x << 16) + hx * xa, (y << 16) + hy * ya, data->w, data->h); | |
dst += 4; | |
} | |
} | |
} | |
break; | |
case 0x11: // set alpha | |
{ | |
const sU16* src1 = getInputData(0)->data; | |
const sU16* src2 = getInputData(1)->data; | |
for (sInt i = 0; i < data->size; i++) | |
{ | |
sU32 alpha = (src2[0] + src2[1] + src2[1] + src2[2]) >> 2; | |
if (param[0] & 1) | |
alpha ^= 32767; | |
dst[0] = src1[0]; | |
dst[1] = src1[1]; | |
dst[2] = src1[2]; | |
dst[3] = alpha; | |
src1 += 4; | |
src2 += 4; | |
dst += 4; | |
} | |
} | |
break; | |
case 0x12: // blend | |
{ | |
const sU16* src1 = getInputData(0)->data; | |
const sU16* src2 = getInputData(1)->data; | |
const sU16* src3 = getInputData(2)->data; | |
for (sInt i = 0; i < data->size; i++) | |
{ | |
sU32 alpha = (src3[0] + src3[1] + src3[1] + src3[2]) >> 2; | |
if (param[0] & 1) | |
alpha ^= 32767; | |
for (sInt c = 0; c < 4; c++) | |
dst[c] = src1[c] + ((((sS32) src2[c] - src1[c]) * alpha) >> 15); | |
src1 += 4; | |
src2 += 4; | |
src3 += 4; | |
dst += 4; | |
} | |
} | |
break; | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
if(def->id == 0x01) // rectangle | |
{ | |
param = strm; | |
strm += 4; | |
sU8 type = *strm++; | |
strm = getColor(color + 0, strm, type >> 0); | |
strm = getColor(color + 3, strm, type >> 2); | |
return strm; | |
} | |
else | |
return frOperator::readData(strm); | |
} | |
}; | |
frOperator* create_TOTrivial() { return new frOpTOTrivial; } | |
// ---- text | |
class frOpTGText: public frTextureOperator | |
{ | |
const sChar* strs[2]; | |
sInt xp, yp, height, style; | |
sU32 color[2]; | |
void doProcess() | |
{ | |
sU16* dst = data->data; | |
HDC hDC, hTempDC; | |
HFONT hFont, hOldFont; | |
HBITMAP hBitmap, hOldBitmap; | |
HBRUSH hBrush; | |
sU8* buf; | |
// initialize & allocate resources | |
hTempDC = GetDC(0); | |
hDC = CreateCompatibleDC(hTempDC); | |
static BITMAPINFOHEADER bih={ sizeof(BITMAPINFOHEADER), 0, 0, 1, 32, 0, 0, 0, 0, 0, 0 }; | |
bih.biWidth = data->w; | |
bih.biHeight = -data->h; | |
sInt realHeight = (height * data->h) >> 8; | |
hBitmap = CreateDIBSection(hTempDC, (BITMAPINFO *) &bih, DIB_RGB_COLORS, (void **) &buf, 0, 0); | |
hFont = CreateFont(realHeight, 0, 0, 0, (style & 1) ? FW_BOLD : FW_NORMAL, style & 2, 0, 0, DEFAULT_CHARSET, | |
OUT_DEFAULT_PRECIS, CLIP_DEFAULT_PRECIS, ANTIALIASED_QUALITY, DEFAULT_PITCH, strs[1]); | |
hBrush = CreateSolidBrush(color[1]); | |
hOldBitmap=(HBITMAP) SelectObject(hDC, hBitmap); | |
hOldFont=(HFONT) SelectObject(hDC, hFont); | |
// fill bitmap with background color | |
RECT rc; | |
rc.left = rc.top = 0; | |
rc.right = data->w; | |
rc.bottom = data->h; | |
FillRect(hDC, &rc, hBrush); | |
// write the text | |
SetTextColor(hDC, color[0]); | |
SetBkMode(hDC, TRANSPARENT); | |
sInt sp = 0, sl, sa; | |
const sChar* text = strs[0]; | |
while (text[sp]) | |
{ | |
sa = sl = 0; | |
while (1) | |
{ | |
if (text[sp + sl] == '\\') | |
{ | |
sa = 1; | |
break; | |
} | |
else | |
if (!text[sp + sl]) | |
break; | |
sl++; | |
} | |
TextOut(hDC, xp, yp, text + sp, sl); | |
sp += sl + sa; | |
yp += realHeight; | |
} | |
SelectObject(hDC, hOldBitmap); | |
SelectObject(hDC, hOldFont); | |
for (sInt i = 0; i < data->size; i++) | |
{ | |
dst[0] = buf[0] << 7; | |
dst[1] = buf[1] << 7; | |
dst[2] = buf[2] << 7; | |
dst[3] = 0; | |
dst += 4; | |
buf += 4; | |
} | |
DeleteObject(hBitmap); | |
DeleteObject(hFont); | |
DeleteObject(hBrush); | |
DeleteDC(hDC); | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
xp = reinterpret_cast<const sU16*>(strm)[0]; | |
yp = reinterpret_cast<const sU16*>(strm)[1]; | |
height = strm[4]; | |
style = strm[5]; | |
strm += 6; | |
for (sInt i = 0; i < 2; i++) | |
{ | |
strs[i] = reinterpret_cast<const sChar*>(strm); | |
strm += strlen(strs[i]) + 1; | |
} | |
for (sInt i = 0; i < 2; i++) | |
{ | |
color[i] = *reinterpret_cast<const sU32*>(strm) & 0xffffff; | |
strm += 3; | |
} | |
return strm; | |
} | |
}; | |
frOperator* create_TGText() { return new frOpTGText; } | |
// ---- rotozoom | |
class frOpTFRotoZoom: public frTextureOperator | |
{ | |
sF32 angle, zoom1, zoom2; | |
sS32 stx, sty; | |
void doProcess() | |
{ | |
const sU16* src = getInputData(0)->data; | |
sU16* dst = data->data; | |
sS32 xp, yp, yxp, yyp, xsx, ysx, xsy, ysy; | |
xsx = zoom1 * cos(angle); ysx = zoom2 * sin(angle); | |
xsy = -zoom1 * sin(angle); ysy = zoom2 * cos(angle); | |
yxp = stx; | |
yyp = sty; | |
for (sInt y = 0; y < data->h; y++) | |
{ | |
xp = yxp; | |
yp = yyp; | |
for (sInt x = 0; x < data->w; x++) | |
{ | |
sampleBilinear(dst, src, xp, yp, data->w, data->h); | |
xp += xsx; | |
yp += ysx; | |
dst += 4; | |
} | |
yxp += xsy; | |
yyp += ysy; | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
angle = *reinterpret_cast<const sU16*>(strm) * 3.1415926535f / 32768.0f; | |
strm += 2; | |
zoom1 = getPackedFloat(strm); | |
zoom2 = getPackedFloat(strm); | |
stx = reinterpret_cast<const sS32*>(strm)[0]; | |
sty = reinterpret_cast<const sS32*>(strm)[1]; | |
strm += 8; | |
return strm; | |
} | |
}; | |
frOperator* create_TFRotoZoom() { return new frOpTFRotoZoom; } | |
// ---- adjust | |
class frOpTFAdjust: public frTextureOperator | |
{ | |
static sInt __fastcall conBright(sInt v, sS32 cont, sS32 bright) | |
{ | |
v = ((v * cont) >> 11) + bright; | |
v = (v < 0) ? 0 : (v > 32767) ? 32767 : v; | |
return v; | |
} | |
void doProcess() | |
{ | |
sU16* pts = getInputData(0)->data; | |
sU16* ptd = data->data; | |
sInt cont, brgt, hue, sat; | |
cont = reinterpret_cast<const sS16*>(param)[0]; | |
brgt = reinterpret_cast<const sS16*>(param)[1] + 16384 - 8 * cont; | |
hue = reinterpret_cast<const sU16*>(param)[2] * 6; | |
sat = reinterpret_cast<const sS16*>(param)[3]; | |
sU32 cnt = data->size; | |
do | |
{ | |
// rgb->hsb | |
sU32 max = pts[0], min = pts[0]; | |
min = (pts[1] < min) ? pts[1] : min; | |
max = (pts[1] > max) ? pts[1] : max; | |
min = (pts[2] < min) ? pts[2] : min; | |
max = (pts[2] > max) ? pts[2] : max; | |
// calculate and adjust saturation | |
sS32 s = max - min; | |
if (sat < 0) | |
s = (s * (sat + 32768)) >> 15; | |
else | |
s += ((max - s) * sat) >> 15; | |
// calculate and shift hue, then convert back to rgb | |
if (s) | |
{ | |
sS32 h, d; | |
if (pts[2] == max) | |
h = 3, d = pts[1] - pts[0]; | |
else if (pts[1] == max) | |
h = 1, d = pts[0] - pts[2]; | |
else if (pts[0] == max) | |
h = 2, d = pts[2] - pts[1]; | |
sS32 hv = hue; | |
__asm | |
{ | |
mov eax, [d]; | |
mov edx, eax; | |
shl eax, 16; | |
sar edx, 16; | |
mov ebx, [max]; | |
sub ebx, [min]; | |
inc ebx; | |
idiv ebx; | |
mov ebx, [h]; | |
shl ebx, 17; | |
add eax, [hv]; | |
add eax, ebx; | |
mov [h], eax; | |
} | |
const sInt ih = h >> 16; | |
const sS32 fh = h & 0xffff; | |
sS32 vals[4]; | |
vals[0] = 0; | |
vals[1] = s; | |
vals[2] = (s * fh) >> 16; | |
vals[3] = (s * (65536 - fh)) >> 16; | |
static const sU8 st[] = { 1, 1, 3, 0, 0, 2, 1, 1, 3, 0, 0, 2, 1, 1, 3, 0, 0, 2 }; | |
ptd[0] = conBright(max - vals[st[ih]], cont, brgt); | |
ptd[1] = conBright(max - vals[st[ih + 2]], cont, brgt); | |
ptd[2] = conBright(max - vals[st[ih + 4]], cont, brgt); | |
} | |
else // plain grayscale | |
{ | |
sInt x = conBright(max, cont, brgt); | |
ptd[0] = x; | |
ptd[1] = x; | |
ptd[2] = x; | |
} | |
ptd[3] = conBright(pts[3], cont, brgt); // contrast+brightness alpha (but no color shift) | |
pts += 4; | |
ptd += 4; | |
} | |
while(--cnt); | |
} | |
}; | |
frOperator* create_TFAdjust() { return new frOpTFAdjust; } | |
// ---- Add | |
class frOpTCAdd: public frTextureOperator | |
{ | |
const sS8* gains; | |
void doProcess() | |
{ | |
sU16* dst = data->data; | |
const sInt cnt = data->size; | |
memset(dst, 0, data->size * 4 * sizeof(sU16)); | |
for (sInt i = 0; i < nInputs; i++) | |
{ | |
const sU16* src = getInputData(i)->data; | |
sInt gain = gains[i] * 32767 / 127; | |
if (gain > 0) | |
{ | |
__asm | |
{ | |
mov esi, [src]; | |
mov edi, [dst]; | |
mov ecx, [cnt]; | |
lea esi, [esi + ecx * 8]; | |
lea edi, [edi + ecx * 8]; | |
neg ecx; | |
movd mm2, [gain]; | |
punpcklwd mm2, mm2; | |
punpcklwd mm2, mm2; | |
lp1: | |
movq mm0, [edi + ecx * 8]; | |
movq mm1, [esi + ecx * 8]; | |
pmulhw mm1, mm2; | |
psllw mm1, 1; | |
paddsw mm0, mm1; | |
movq [edi + ecx * 8], mm0; | |
inc ecx; | |
jnz lp1; | |
} | |
} | |
else if (gain < 0) | |
{ | |
gain = -gain; | |
__asm | |
{ | |
mov esi, [src]; | |
mov edi, [dst]; | |
mov ecx, [cnt]; | |
lea esi, [esi + ecx * 8]; | |
lea edi, [edi + ecx * 8]; | |
neg ecx; | |
movd mm2, [gain]; | |
punpcklwd mm2, mm2; | |
punpcklwd mm2, mm2; | |
lp2: | |
movq mm0, [edi + ecx * 8]; | |
movq mm1, [esi + ecx * 8]; | |
pmulhw mm1, mm2; | |
psllw mm1, 1; | |
psubusw mm0, mm1; | |
movq [edi + ecx * 8], mm0; | |
dec ecx; | |
jns lp2; | |
} | |
} | |
} | |
__asm emms; | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
gains = reinterpret_cast<const sS8*>(strm); | |
return strm + nInputs; | |
} | |
}; | |
frOperator* create_TCAdd() { return new frOpTCAdd; } | |
// ---- bump mapping | |
class frOpTCBump: public frTextureOperator | |
{ | |
sF32 bump; | |
sU8 color[9]; | |
sInt lType; | |
fr::vector olr; | |
sF32 escale, falloff; | |
sF32 oca, ica; | |
sF32 lat, lon; | |
void doProcess() | |
{ | |
fr::vector c, d, lr, l[2]; | |
sF32 ocs; | |
if (lType != 1) | |
{ | |
l[0].x = sin(lon); | |
l[0].y = sin(lat) * cos(lon); | |
l[0].z = cos(lat) * cos(lon); | |
l[1].x = 0.0f; | |
l[1].y = 0.0f; | |
l[1].z = 0.0f; | |
} | |
lr = olr; | |
if (lType == 2) | |
{ | |
d = l[0]; | |
lr.x += lr.z * d.x; | |
lr.y += lr.z * d.y; | |
ocs = ica - oca; | |
if (ocs) | |
ocs = 1.0f / ocs; | |
} | |
sU16* dst = data->data; | |
makeNormalMap(dst, getInputData(1)->data, data->w, data->h, bump); | |
const sU16* src = getInputData(0)->data; | |
for (sInt y = 0; y < data->h; y++) | |
{ | |
for (sInt x = 0; x < data->w; x++) | |
{ | |
if (lType != 0) // point/spot light | |
{ | |
c.x = 1.0f - 2.0f * x / data->w; | |
c.y = 1.0f - 2.0f * y / data->h; | |
c.z = 1.0f; | |
l[0].x = lr.x + c.x; | |
l[0].y = lr.y + c.y; | |
l[0].z = lr.z; | |
l[0].norm(); | |
c.norm(); | |
l[1].add(l[0], c); | |
l[1].norm(); | |
} | |
sF32 dots[2]; | |
fr::vector normal(dst[2] - 16384, dst[1] - 16384, dst[0] - 16384); | |
for (sInt i = 0; i < 2; i++) | |
{ | |
sF32 dot = l[i].dot(normal) / 16384.0f; | |
if (dot < 0.0f) | |
dot = 0.0f; | |
dots[i]=dot; | |
} | |
dots[1] = pow(dots[1], escale); | |
sF32 spotTerm = 1.0f; | |
if (lType == 2) // spot light | |
{ | |
sF32 ca = l[0].dot(d); // cos alpha | |
if (ca <= oca) | |
spotTerm = 0.0f; | |
else if (ca <= ica) | |
spotTerm = pow((ca - oca) * ocs, falloff); | |
} | |
for (sInt i = 0; i < 3; i++) | |
{ | |
const sF32 scale = (color[i] + color[i + 3] * dots[0]) / 255.0f; | |
sInt v = (src[i] * scale + color[i + 6] * dots[1] * 256.0f) * spotTerm; | |
if (v > 32767) | |
v = 32767; | |
dst[i] = v; | |
} | |
dst[3] = src[3]; | |
dst += 4; | |
src += 4; | |
} | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
bump = *reinterpret_cast<const sU16*>(strm) / 1024.0f; | |
lType = strm[2]; | |
strm += 3; | |
for (sInt i = 0; i < ((lType != 0) ? 9 : 6); i++) | |
color[i] = *strm++; | |
if (lType != 0) // point&spot | |
{ | |
for (sInt i = 0; i < 3; i++) | |
{ | |
(&olr.x)[i] = *reinterpret_cast<const sS16*>(strm) / 512.0f; | |
strm += 2; | |
} | |
escale = pow(2.0f, *strm++ * 13.0f / 255.0f); | |
} | |
if (lType != 1) // dir&spot | |
{ | |
lat = reinterpret_cast<const sS8*>(strm)[0] * 3.1415926535f / 254.0f; | |
lon = reinterpret_cast<const sS8*>(strm)[1] * 3.1415926535f / 254.0f; | |
strm += 2; | |
} | |
if (lType == 2) // spot | |
{ | |
oca = cos(strm[0] * 3.1415926535f / 510.0f); | |
ica = cos(strm[1] * 3.1415926535f / 510.0f); | |
falloff = strm[2] * 10.0f / 255.0f; | |
strm += 3; | |
} | |
return strm; | |
} | |
}; | |
frOperator* create_TCBump() { return new frOpTCBump; } | |
// ---- Perlin noise | |
class frOpTGPerlin: public frTextureOperator | |
{ | |
sInt nOctaves; | |
sInt persistence; | |
sInt freq; | |
sU32 oSeed; | |
sInt startOctave; | |
sU16 scale; | |
sInt color[6]; | |
static sU32 seed; | |
static sU16 wtab[512]; | |
static sBool wTabThere; | |
static sS32 __forceinline inoise(sU32 x, sU32 y, sInt msk) | |
{ | |
sInt ix = x >> 10, iy = y >> 10; | |
sInt inx = (ix + 1) & msk, iny = (iy + 1) & msk; | |
ix += seed; | |
inx += seed; | |
iy *= 31337; | |
iny *= 31337; | |
sInt fx = wtab[(x >> 1) & 511]; | |
sS32 t1 = srandom2(ix + iy); | |
t1 = (t1 << 10) + (srandom2(inx + iy) - t1) * fx; | |
sS32 t2 = srandom2(ix + iny); | |
t2 = (t2 << 10) + (srandom2(inx + iny) - t2) * fx; | |
return ((t1 << 10) + (t2 - t1) * wtab[(y >> 1) & 511]) >> 20; | |
} | |
static sS32 imul16(sS32 a, sS32 b) | |
{ | |
__asm | |
{ | |
mov eax, [a]; | |
imul [b]; | |
shrd eax, edx, 16; | |
mov [a], eax; | |
} | |
return a; | |
} | |
static void calcWTab() | |
{ | |
for (sInt i = 0; i < 512; i++) | |
{ | |
const sF32 val = i / 512.0f; | |
wtab[i] = 1024.0f * val * val * val * (10.0f + val * (val * 6.0f - 15.0f)); | |
} | |
} | |
void doProcess() | |
{ | |
sU16* out = data->data; | |
sInt x, y, o, f; | |
sS32 amp, sc, sum, xv, yv, fx, fy; | |
if (!wTabThere) | |
calcWTab(); | |
sU32 oseed = oSeed + startOctave * 0xdeadbeef; | |
const sInt fbase = (1024 << startOctave) * freq; | |
fx = fbase / data->w; | |
fy = fbase / data->h; | |
sS32 samp; | |
sum = 0; | |
amp = 32768; | |
for (o = 0; o < nOctaves; o++) | |
{ | |
if (o >= startOctave) | |
sum+=amp; | |
if (o == startOctave) | |
samp = amp; | |
amp = (amp * persistence) >> 8; | |
} | |
sc = scale * 8 * 64 * 16 / sum; | |
sInt fb = freq << startOctave; | |
for (y = 0; y < data->h; y++) | |
{ | |
for (x = 0; x < data->w; x++) | |
{ | |
amp = samp; | |
sum = 0; | |
xv = x * fx; | |
yv = y * fy; | |
f = fb; | |
seed = oseed; | |
for (o = startOctave; o < nOctaves; o++) | |
{ | |
sum += amp * inoise(xv, yv, f - 1); | |
xv *= 2; | |
yv *= 2; | |
f *= 2; | |
seed += 0xdeadbeef; | |
amp = (amp * persistence) >> 8; | |
} | |
sS32 sm = imul16(sum, sc) + 32768; | |
sm = (sm < 0) ? 0 : (sm > 65536) ? 65536 : sm; | |
*out++ = color[0] + ((sm * ((sInt) color[3] - (sInt) color[0])) >> 16); | |
*out++ = color[1] + ((sm * ((sInt) color[4] - (sInt) color[1])) >> 16); | |
*out++ = color[2] + ((sm * ((sInt) color[5] - (sInt) color[2])) >> 16); | |
*out++ = 0; | |
} | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
sInt temp = *strm++; | |
nOctaves = temp & 15; | |
persistence = *strm++; | |
freq = 1 << (temp >> 4); | |
oSeed = (*reinterpret_cast<const sU16*>(strm) + 1) * 0xb37fa184 + 303 * 31337; | |
strm += 2; | |
temp = *strm++; | |
startOctave = temp & 15; | |
scale = getSmallInt(strm); | |
sU8 kolor[6]; | |
strm = getColor(kolor + 0,strm,temp >> 4); | |
strm = getColor(kolor + 3,strm,temp >> 6); | |
for(sInt i=0;i<6;i++) | |
color[i] = kolor[i]<<7; | |
return strm; | |
} | |
static sU32 srandom2(sU32 n) | |
{ | |
__asm | |
{ | |
mov ecx, [n]; | |
mov ebx, ecx; | |
shl ebx, 13; | |
xor ecx, ebx; | |
mov ebx, ecx; | |
mov eax, 15731; | |
imul ebx, ebx; | |
imul ebx, eax; | |
add ebx, 789221; | |
imul ecx, ebx; | |
add ecx, 1376312589; | |
mov eax, 2048; | |
and ecx, 07fffffffh; | |
shr ecx, 19; | |
sub eax, ecx; | |
mov [n], eax; | |
} | |
return n; | |
} | |
}; | |
sU32 frOpTGPerlin::seed; | |
sU16 frOpTGPerlin::wtab[512]; | |
sBool frOpTGPerlin::wTabThere=sFALSE; | |
frOperator* create_TGPerlin() { return new frOpTGPerlin; } | |
// ---- Gradient | |
class frOpTGGradient: public frTextureOperator | |
{ | |
sInt type; | |
sU8 color[6]; | |
sF32 iw, ih; | |
sF32 xc, yc, x2, y2, expon, isc; | |
sInt scv; | |
void doProcess() | |
{ | |
sF32 sc; | |
sF32 xd = x2 - xc, yd = y2 - yc; | |
if (type == 0) | |
sc = 1.0f / (xd * xd + yd * yd); | |
else | |
{ | |
sc = 100.0f / scv; | |
sc *= sc; | |
} | |
sU16* dst = data->data; | |
for (sInt y = 0; y < data->h; y++) | |
{ | |
const sF32 p = y * ih - yc, pp = p * p; | |
for (sInt x = 0; x < data->w; x++) | |
{ | |
sF32 f; | |
sF32 n = x * iw - xc; | |
if (type == 0) // linear | |
f = (n * xd + p * yd) * sc; | |
else if (type == 1) // glow | |
{ | |
f = 1.0f - sqrt((n * n + pp) * sc); | |
f = pow(f < 0.0f ? 0.0f : f > 1.0f ? 1.0f : f, expon) * isc; | |
} | |
f=f < 0.0f ? 0.0f : f > 1.0f ? 1.0f : f; | |
sS32 fd = f * 65536; | |
*dst++ = ((color[0] << 16) + (color[3] - color[0]) * fd) >> 9; | |
*dst++ = ((color[1] << 16) + (color[4] - color[1]) * fd) >> 9; | |
*dst++ = ((color[2] << 16) + (color[5] - color[2]) * fd) >> 9; | |
*dst++ = 0; | |
} | |
} | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
sU8 typeByte = *strm++; | |
type = typeByte & 1; | |
strm = getColor(color + 0, strm, typeByte >> 1); | |
strm = getColor(color + 3, strm, typeByte >> 3); | |
iw = 1.0f / data->w; | |
ih = 1.0f / data->h; | |
if (type == 0) | |
{ | |
xc = reinterpret_cast<const sS16*>(strm)[0] / 256.0f; | |
yc = reinterpret_cast<const sS16*>(strm)[1] / 256.0f; | |
x2 = reinterpret_cast<const sS16*>(strm)[2] / 256.0f; | |
y2 = reinterpret_cast<const sS16*>(strm)[3] / 256.0f; | |
strm += 8; | |
} | |
else | |
{ | |
scv = *strm++; | |
expon = getSmallInt(strm) / 8.0f; | |
isc = reinterpret_cast<const sU16*>(strm)[0] / 256.0f; | |
xc = reinterpret_cast<const sS16*>(strm)[1] / 256.0f; | |
yc = reinterpret_cast<const sS16*>(strm)[2] / 256.0f; | |
strm += 6; | |
} | |
return strm; | |
} | |
void setAnim(sInt parm, const sF32* val) | |
{ | |
switch (parm) | |
{ | |
case 1: // point 1/center | |
case 3: | |
xc = val[0]; | |
yc = val[1]; | |
break; | |
case 2: // point 2 | |
xc = val[0]; | |
yc = val[1]; | |
break; | |
} | |
} | |
}; | |
frOperator* create_TGGradient() { return new frOpTGGradient; } | |
// ---- Multiply | |
class frOpTCMultiply: public frTextureOperator | |
{ | |
void doProcess() | |
{ | |
sU16* dst = data->data; | |
sInt cnt = data->w * data->h; | |
memcpy(dst, getInputData(0)->data, cnt * 4 * 2); | |
for (sInt i = 1; i < nInputs; i++) | |
{ | |
const sU16* src = getInputData(i)->data; | |
__asm | |
{ | |
mov esi, [src]; | |
mov edi, [dst]; | |
mov ecx, [cnt]; | |
lea esi, [esi + ecx * 8]; | |
lea edi, [edi + ecx * 8]; | |
neg ecx; | |
lp: | |
movq mm0, [esi + ecx * 8]; | |
pmulhw mm0, [edi + ecx * 8]; | |
psllw mm0, 1; | |
movq [edi + ecx * 8], mm0; | |
inc ecx; | |
jnz lp; | |
} | |
} | |
__asm emms; | |
} | |
}; | |
frOperator* create_TCMultiply() { return new frOpTCMultiply; } | |
// ---- Crystal | |
class frOpTGCrystal: public frTextureOperator | |
{ | |
sU32 seed, count; | |
sU32 col[2]; | |
void doProcess() | |
{ | |
sU8 randBuf[512]; | |
sU16* dptr = data->data; | |
sInt width = data->w; | |
sInt height = data->h; | |
static const sU32 _MMMax = 0x7fffffff; | |
const sU32 _MM256 = (height << 16) | width, temp = (_MM256 >> 1) & 0x7fff7fff; | |
const sU64 _MM128 = ((sU64)temp << 32) | temp; | |
const sU64 _MMMask = ((height - 1) << 16) | (width - 1); | |
const sU32 col1 = col[0]; | |
const sU32 col2 = col[1]; | |
const sU32 cnt = count; | |
const sU32 sed = seed; | |
__asm | |
{ | |
mov edx, [sed]; | |
mov esi, [dptr]; | |
mov ecx, 511; | |
randomloop: | |
imul edx, 015a4e35h; | |
inc edx; | |
mov eax, edx; | |
shr eax, 16; | |
mov randBuf[ecx], al; | |
dec ecx; | |
jns randomloop; | |
emms; | |
pxor mm7, mm7; | |
xor ecx, ecx; | |
xor edx, edx; | |
yloop: | |
and ecx, 0ffff0000h; | |
xloop: | |
mov edi, [cnt]; | |
dec edi | |
movd mm2, ecx; | |
movd mm0, [_MMMax]; | |
movd mm6, [_MM256]; | |
movq mm5, mm2; | |
xor ebx, ebx; | |
dec ebx; | |
push esi; | |
tryloop: | |
movd mm1, randBuf[edi*2]; | |
movq mm2, mm5; | |
pand mm1, [_MMMask]; | |
psubw mm2, mm1; | |
movq mm1, mm0; | |
movq mm3, mm2; | |
psraw mm3, 15; | |
pxor mm2, mm3; | |
psubsw mm2, mm3; | |
movq mm3, mm2; | |
pcmpgtw mm3, [_MM128]; | |
movq mm4, mm6; | |
psubw mm4, mm2; | |
pand mm4, mm3; | |
pandn mm3, mm2; | |
por mm3, mm4; | |
pmaddwd mm3, mm3; | |
pcmpgtd mm0, mm3; | |
movd eax, mm3; | |
pand mm3, mm0; | |
pandn mm0, mm1; | |
por mm0, mm3; | |
cmp eax, ebx; | |
cmovb ebx, eax; | |
cmovb esi, edi; | |
dec edi; | |
jns tryloop; | |
mov ebx, esi; | |
pop esi; | |
movd eax, mm0; | |
cmp eax, edx; | |
cmova edx, eax; | |
mov [esi], eax; | |
add esi, 8; | |
inc ecx; | |
movzx eax, cx; | |
cmp eax, [width]; | |
jne xloop; | |
add ecx, 010000h; | |
mov eax, ecx; | |
shr eax, 16; | |
cmp eax, [height]; | |
jne yloop; | |
mov ecx, [width]; | |
imul ecx, [height]; | |
mov [width], ecx; | |
mov ecx, 0ffffh; | |
mov edi, edx; | |
movd mm0, [col1]; | |
movd mm1, [col2]; | |
punpcklbw mm0, mm7; | |
punpcklbw mm1, mm7; | |
psllw mm0, 7; | |
psllw mm1, 7; | |
psubw mm1, mm0; | |
contrastloop: | |
sub esi, 8; | |
mov eax, [esi]; | |
mov edx, eax; | |
shl eax, 16; | |
sar edx, 16; | |
idiv edi; | |
xor eax, -1; | |
add eax, ecx; | |
imul eax; | |
shrd eax, edx, 16; | |
xor eax, -1; | |
add eax, ecx; | |
shr eax, 1; | |
movd mm2, eax; | |
punpcklwd mm2, mm2; | |
punpcklwd mm2, mm2; | |
pmulhw mm2, mm1; | |
psllw mm2, 1; | |
paddw mm2, mm0; | |
movq [esi], mm2; | |
mov [esi+6], ax; | |
dec [width]; | |
jnz contrastloop; | |
emms; | |
}; | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
seed = reinterpret_cast<const sU16*>(strm)[0]; | |
count = strm[2]; | |
strm += 3; | |
for (sInt i = 0; i < 2; i++) | |
{ | |
col[i] = *reinterpret_cast<const sU32*>(strm) & 0xffffff; | |
strm += 3; | |
} | |
return strm; | |
} | |
}; | |
frOperator* create_TGCrystal() { return new frOpTGCrystal; } | |
// ---- blur v2 | |
class frOpTFBlurV2: public frTextureOperator | |
{ | |
sInt rx, ry, parm, intens; | |
public: | |
static void inplaceBlur(sU16* dst, sU32 w, sU32 h, sUInt radius, sInt wrap) | |
{ | |
static sU16 pbuf[(1024+256)*4]; | |
if (2*radius>=w) | |
radius=(w-1)/2; | |
sU32 sc=65536/(2*radius+1); | |
if (wrap==0) | |
{ | |
memset(pbuf, 0, radius*4*sizeof(sU16)); | |
memset(pbuf+(w+radius)*4, 0, radius*4*sizeof(sU16)); | |
} | |
__asm emms; | |
while (h--) | |
{ | |
memcpy(pbuf+radius*4, dst, w*4*sizeof(sU16)); | |
if (wrap==1) // clamp | |
{ | |
for (sUInt i=0; i<radius*4; i++) | |
{ | |
const sUInt ia=i&3; | |
pbuf[i]=dst[ia]; | |
pbuf[i+(w+radius)*4]=dst[w*4-4+ia]; | |
} | |
} | |
else if (wrap==2) // wrap | |
{ | |
memcpy(pbuf, pbuf+w*4, radius*4*sizeof(sU16)); | |
memcpy(pbuf+(w+radius)*4, pbuf+radius*4, radius*4*sizeof(sU16)); | |
} | |
const sU16 *src=pbuf+radius*4; | |
static const sU64 msk1=0x0000ffff0000ffff; | |
__asm | |
{ | |
mov esi, offset pbuf; | |
pxor mm0, mm0; | |
pxor mm1, mm1; | |
pxor mm3, mm3; | |
mov ecx, [radius]; | |
shl ecx, 1; | |
sumlp: | |
movq mm4, [esi]; | |
add esi, 8; | |
movq mm5, mm4; | |
punpcklwd mm4, mm3; | |
dec ecx; | |
punpckhwd mm5, mm3; | |
paddd mm0, mm4; | |
paddd mm1, mm5; | |
jns sumlp; | |
mov esi, [src]; | |
mov edi, [dst]; | |
mov ecx, [w]; | |
mov ebx, [radius]; | |
shl ebx, 3; | |
lea esi, [esi+ecx*8]; | |
lea edi, [edi+ecx*8]; | |
lea edx, [esi+ebx+8]; | |
sub esi, ebx; | |
neg ecx; | |
movd mm2, [sc]; // load scale reg | |
punpcklwd mm2, mm2; | |
punpcklwd mm2, mm2; | |
pixlp: | |
// calc color & store | |
movq mm4, mm0; // copy values | |
movq mm5, mm1; | |
movq mm6, mm0; | |
movq mm7, mm1; | |
// perform scaling | |
psrld mm4, 16; | |
psrld mm5, 16; | |
psrlw mm6, 1; | |
psrlw mm7, 1; | |
pmulhw mm6, mm2; | |
pmulhw mm7, mm2; | |
pmullw mm4, mm2; | |
pmullw mm5, mm2; | |
pand mm6, [msk1]; | |
pand mm7, [msk1]; | |
psllw mm6, 1; | |
psllw mm7, 1; | |
paddd mm4, mm6; | |
paddd mm5, mm7; | |
// re-pack & store | |
packssdw mm4, mm5; | |
movq [edi+ecx*8], mm4; | |
// update state: | |
movq mm4, [edx+ecx*8]; // add val | |
movq mm6, [esi+ecx*8]; // sub val | |
movq mm5, mm4; | |
movq mm7, mm6; | |
punpcklwd mm4, mm3; // splice values | |
punpckhwd mm5, mm3; | |
punpcklwd mm6, mm3; | |
punpckhwd mm7, mm3; | |
paddd mm0, mm4; // add to state | |
paddd mm1, mm5; | |
psubd mm0, mm6; // sub from state | |
psubd mm1, mm7; | |
// loop | |
inc ecx; | |
jnz pixlp; | |
mov [dst], edi; | |
} | |
} | |
__asm emms; | |
} | |
static void flipAxes(const sU16 *src, sU16 *dst, sInt w, sInt h) | |
{ | |
sInt ws=2, wt=w; | |
while (wt>>=1) | |
ws++; | |
for (sInt x=0; x<w; x++) | |
{ | |
const sU16 *s0=src+(x<<2); | |
for (sInt y=0; y<h; y++) | |
{ | |
const sU16 *so=s0+(y<<ws); | |
((sU32 *) dst)[0]=((const sU32 *) so)[0]; | |
((sU32 *) dst)[1]=((const sU32 *) so)[1]; | |
dst+=4; | |
} | |
} | |
} | |
static void scaleIntensity(sU16 *data, sUInt count, sU16 sf) | |
{ | |
while (count--) | |
{ | |
sU32 v=(data[0]*sf)>>8; | |
if (v>32767) | |
v=32767; | |
*data++=v; | |
} | |
} | |
void doProcess() | |
{ | |
const sInt wrap=parm>>2, steps=parm & 3; | |
const sInt dw = data->w, dh = data->h; | |
sU16* dptr = data->data; | |
memcpy(dptr, getInputData(0)->data, dw * dh * 8); // first, copy over the data. | |
if (rx) | |
{ | |
for (sInt i=0; i<=steps; i++) | |
inplaceBlur(dptr, dw, dh, rx, wrap); | |
} | |
if (ry) | |
{ | |
sU16 *temp=new sU16[dw*dh*4]; | |
flipAxes(dptr, temp, dw, dh); | |
for (sInt i=0; i<=steps; i++) | |
inplaceBlur(temp, dh, dw, ry, wrap); | |
flipAxes(temp, dptr, dh, dw); | |
delete[] temp; | |
} | |
if (intens != 256) | |
scaleIntensity(dptr, dw*dh*4, intens); | |
} | |
public: | |
const sU8* readData(const sU8* strm) | |
{ | |
rx = *strm++; | |
ry = *strm++; | |
parm = *strm++; | |
intens = *reinterpret_cast<const sU16*>(strm); | |
strm += 2; | |
return strm; | |
} | |
void setAnim(sInt index, const sF32* val) | |
{ | |
switch (index) | |
{ | |
case 1: // radius | |
rx = val[0] - 0.5f; | |
ry = val[1] - 0.5f; | |
break; | |
case 2: // intensity | |
intens = val[0] * 256.0f; | |
break; | |
} | |
} | |
}; | |
frOperator* create_TFBlurV2() { return new frOpTFBlurV2; } | |
// ---- install font | |
class frOpTFInstallFont: public frTextureOperator | |
{ | |
sChar fontFile[256]; | |
void doProcess() | |
{ | |
memcpy(data->data, getInputData(0)->data, data->w*data->h*4*sizeof(sU16)); | |
} | |
public: | |
~frOpTFInstallFont() | |
{ | |
RemoveFontResource(fontFile); | |
DeleteFile(fontFile); | |
} | |
const sU8* readData(const sU8* strm) | |
{ | |
sInt len = getSmallInt(strm); | |
lstrcpy(fontFile, "c:\\temp.ttf"); | |
DWORD dwt; | |
HANDLE hfl = CreateFile(fontFile, GENERIC_WRITE, 0, 0, CREATE_ALWAYS, 0, 0); | |
WriteFile(hfl, strm, len, &dwt, 0); | |
CloseHandle(hfl); | |
AddFontResource(fontFile); | |
strm += len; | |
return strm; | |
} | |
}; | |
frOperator* create_TFInstallFont() { return new frOpTFInstallFont; } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment