kernel.glsl

#version 300 es
precision mediump float;
precision mediump int;
precision mediump sampler2D;
precision mediump sampler2DArray;

const int LOOP_MAX = 1000;

uniform ivec3 uOutputDim;
uniform ivec2 uTexSize;

in vec2 vTexCoord;

const int BIT_COUNT = 32;
int modi(int x, int y) {
  return x - y * (x / y);
}

int bitwiseOr(int a, int b) {
  int result = 0;
  int n = 1;
  
  for (int i = 0; i < BIT_COUNT; i++) {
    if ((modi(a, 2) == 1) || (modi(b, 2) == 1)) {
      result += n;
    }
    a = a / 2;
    b = b / 2;
    n = n * 2;
    if(!(a > 0 || b > 0)) {
      break;
    }
  }
  return result;
}
int bitwiseXOR(int a, int b) {
  int result = 0;
  int n = 1;
  
  for (int i = 0; i < BIT_COUNT; i++) {
    if ((modi(a, 2) == 1) != (modi(b, 2) == 1)) {
      result += n;
    }
    a = a / 2;
    b = b / 2;
    n = n * 2;
    if(!(a > 0 || b > 0)) {
      break;
    }
  }
  return result;
}
int bitwiseAnd(int a, int b) {
  int result = 0;
  int n = 1;
  for (int i = 0; i < BIT_COUNT; i++) {
    if ((modi(a, 2) == 1) && (modi(b, 2) == 1)) {
      result += n;
    }
    a = a / 2;
    b = b / 2;
    n = n * 2;
    if(!(a > 0 && b > 0)) {
      break;
    }
  }
  return result;
}
int bitwiseNot(int a) {
  int result = 0;
  int n = 1;
  
  for (int i = 0; i < BIT_COUNT; i++) {
    if (modi(a, 2) == 0) {
      result += n;    
    }
    a = a / 2;
    n = n * 2;
  }
  return result;
}
int bitwiseZeroFillLeftShift(int n, int shift) {
  int maxBytes = BIT_COUNT;
  for (int i = 0; i < BIT_COUNT; i++) {
    if (maxBytes >= n) {
      break;
    }
    maxBytes *= 2;
  }
  for (int i = 0; i < BIT_COUNT; i++) {
    if (i >= shift) {
      break;
    }
    n *= 2;
  }

  int result = 0;
  int byteVal = 1;
  for (int i = 0; i < BIT_COUNT; i++) {
    if (i >= maxBytes) break;
    if (modi(n, 2) > 0) { result += byteVal; }
    n = int(n / 2);
    byteVal *= 2;
  }
  return result;
}

int bitwiseSignedRightShift(int num, int shifts) {
  return int(floor(float(num) / pow(2.0, float(shifts))));
}

int bitwiseZeroFillRightShift(int n, int shift) {
  int maxBytes = BIT_COUNT;
  for (int i = 0; i < BIT_COUNT; i++) {
    if (maxBytes >= n) {
      break;
    }
    maxBytes *= 2;
  }
  for (int i = 0; i < BIT_COUNT; i++) {
    if (i >= shift) {
      break;
    }
    n /= 2;
  }
  int result = 0;
  int byteVal = 1;
  for (int i = 0; i < BIT_COUNT; i++) {
    if (i >= maxBytes) break;
    if (modi(n, 2) > 0) { result += byteVal; }
    n = int(n / 2);
    byteVal *= 2;
  }
  return result;
}

vec2 integerMod(vec2 x, float y) {
  vec2 res = floor(mod(x, y));
  return res * step(1.0 - floor(y), -res);
}

vec3 integerMod(vec3 x, float y) {
  vec3 res = floor(mod(x, y));
  return res * step(1.0 - floor(y), -res);
}

vec4 integerMod(vec4 x, vec4 y) {
  vec4 res = floor(mod(x, y));
  return res * step(1.0 - floor(y), -res);
}

float integerMod(float x, float y) {
  float res = floor(mod(x, y));
  return res * (res > floor(y) - 1.0 ? 0.0 : 1.0);
}

int integerMod(int x, int y) {
  return x - (y * int(x/y));
}

float div_with_int_check(float x, float y) {
  if (floor(x) == x && floor(y) == y && integerMod(x, y) == 0.0) {
    return float(int(x)/int(y));
  }
  return x / y;
}
// Here be dragons!
// DO NOT OPTIMIZE THIS CODE
// YOU WILL BREAK SOMETHING ON SOMEBODY'S MACHINE
// LEAVE IT AS IT IS, LEST YOU WASTE YOUR OWN TIME
const vec2 MAGIC_VEC = vec2(1.0, -256.0);
const vec4 SCALE_FACTOR = vec4(1.0, 256.0, 65536.0, 0.0);
const vec4 SCALE_FACTOR_INV = vec4(1.0, 0.00390625, 0.0000152587890625, 0.0); // 1, 1/256, 1/65536
float decode32(vec4 texel) {
  texel *= 255.0;
  vec2 gte128;
  gte128.x = texel.b >= 128.0 ? 1.0 : 0.0;
  gte128.y = texel.a >= 128.0 ? 1.0 : 0.0;
  float exponent = 2.0 * texel.a - 127.0 + dot(gte128, MAGIC_VEC);
  float res = exp2(round(exponent));
  texel.b = texel.b - 128.0 * gte128.x;
  res = dot(texel, SCALE_FACTOR) * exp2(round(exponent-23.0)) + res;
  res *= gte128.y * -2.0 + 1.0;
  return res;
}

float decode16(vec4 texel, int index) {
  int channel = integerMod(index, 2);
  return texel[channel*2] * 255.0 + texel[channel*2 + 1] * 65280.0;
}

float decode8(vec4 texel, int index) {
  int channel = integerMod(index, 4);
  return texel[channel] * 255.0;
}

vec4 legacyEncode32(float f) {
  float F = abs(f);
  float sign = f < 0.0 ? 1.0 : 0.0;
  float exponent = floor(log2(F));
  float mantissa = (exp2(-exponent) * F);
  // exponent += floor(log2(mantissa));
  vec4 texel = vec4(F * exp2(23.0-exponent)) * SCALE_FACTOR_INV;
  texel.rg = integerMod(texel.rg, 256.0);
  texel.b = integerMod(texel.b, 128.0);
  texel.a = exponent*0.5 + 63.5;
  texel.ba += vec2(integerMod(exponent+127.0, 2.0), sign) * 128.0;
  texel = floor(texel);
  texel *= 0.003921569; // 1/255
  return texel;
}

// https://github.com/gpujs/gpu.js/wiki/Encoder-details
vec4 encode32(float value) {
  if (value == 0.0) return vec4(0, 0, 0, 0);

  float exponent;
  float mantissa;
  vec4  result;
  float sgn;

  sgn = step(0.0, -value);
  value = abs(value);

  exponent = floor(log2(value));

  mantissa = value*pow(2.0, -exponent)-1.0;
  exponent = exponent+127.0;
  result   = vec4(0,0,0,0);

  result.a = floor(exponent/2.0);
  exponent = exponent - result.a*2.0;
  result.a = result.a + 128.0*sgn;

  result.b = floor(mantissa * 128.0);
  mantissa = mantissa - result.b / 128.0;
  result.b = result.b + exponent*128.0;

  result.g = floor(mantissa*32768.0);
  mantissa = mantissa - result.g/32768.0;

  result.r = floor(mantissa*8388608.0);
  return result/255.0;
}
// Dragons end here

int index;
ivec3 threadId;

ivec3 indexTo3D(int idx, ivec3 texDim) {
  int z = int(idx / (texDim.x * texDim.y));
  idx -= z * int(texDim.x * texDim.y);
  int y = int(idx / texDim.x);
  int x = int(integerMod(idx, texDim.x));
  return ivec3(x, y, z);
}

float get32(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + texDim.x * (y + texDim.y * z);
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  vec4 texel = texture(tex, st / vec2(texSize));
  return decode32(texel);
}

float get16(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + (texDim.x * (y + (texDim.y * z)));
  int w = texSize.x * 2;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  vec4 texel = texture(tex, st / vec2(texSize.x * 2, texSize.y));
  return decode16(texel, index);
}

float get8(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + (texDim.x * (y + (texDim.y * z)));
  int w = texSize.x * 4;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  vec4 texel = texture(tex, st / vec2(texSize.x * 4, texSize.y));
  return decode8(texel, index);
}

float getMemoryOptimized32(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + (texDim.x * (y + (texDim.y * z)));
  int channel = integerMod(index, 4);
  index = index / 4;
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  index = index / 4;
  vec4 texel = texture(tex, st / vec2(texSize));
  return texel[channel];
}

vec4 getImage2D(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + texDim.x * (y + texDim.y * z);
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  return texture(tex, st / vec2(texSize));
}

vec4 getImage3D(sampler2DArray tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + texDim.x * (y + texDim.y * z);
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  return texture(tex, vec3(st / vec2(texSize), z));
}

float getFloatFromSampler2D(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  vec4 result = getImage2D(tex, texSize, texDim, z, y, x);
  return result[0];
}

vec2 getVec2FromSampler2D(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  vec4 result = getImage2D(tex, texSize, texDim, z, y, x);
  return vec2(result[0], result[1]);
}

vec2 getMemoryOptimizedVec2(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + texDim.x * (y + texDim.y * z);
  int channel = integerMod(index, 2);
  index = index / 2;
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  vec4 texel = texture(tex, st / vec2(texSize));
  if (channel == 0) return vec2(texel.r, texel.g);
  if (channel == 1) return vec2(texel.b, texel.a);
  return vec2(0.0, 0.0);
}

vec3 getVec3FromSampler2D(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  vec4 result = getImage2D(tex, texSize, texDim, z, y, x);
  return vec3(result[0], result[1], result[2]);
}

vec3 getMemoryOptimizedVec3(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int fieldIndex = 3 * (x + texDim.x * (y + texDim.y * z));
  int vectorIndex = fieldIndex / 4;
  int vectorOffset = fieldIndex - vectorIndex * 4;
  int readY = vectorIndex / texSize.x;
  int readX = vectorIndex - readY * texSize.x;
  vec4 tex1 = texture(tex, (vec2(readX, readY) + 0.5) / vec2(texSize));

  if (vectorOffset == 0) {
    return tex1.xyz;
  } else if (vectorOffset == 1) {
    return tex1.yzw;
  } else {
    readX++;
    if (readX >= texSize.x) {
      readX = 0;
      readY++;
    }
    vec4 tex2 = texture(tex, vec2(readX, readY) / vec2(texSize));
    if (vectorOffset == 2) {
      return vec3(tex1.z, tex1.w, tex2.x);
    } else {
      return vec3(tex1.w, tex2.x, tex2.y);
    }
  }
}

vec4 getVec4FromSampler2D(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  return getImage2D(tex, texSize, texDim, z, y, x);
}

vec4 getMemoryOptimizedVec4(sampler2D tex, ivec2 texSize, ivec3 texDim, int z, int y, int x) {
  int index = x + texDim.x * (y + texDim.y * z);
  int channel = integerMod(index, 2);
  int w = texSize.x;
  vec2 st = vec2(float(integerMod(index, w)), float(index / w)) + 0.5;
  vec4 texel = texture(tex, st / vec2(texSize));
  return vec4(texel.r, texel.g, texel.b, texel.a);
}

vec4 actualColor;
void color(float r, float g, float b, float a) {
  actualColor = vec4(r,g,b,a);
}

void color(float r, float g, float b) {
  color(r,g,b,1.0);
}

uniform mediump sampler2D user_input;
mediump ivec2 user_inputSize = ivec2(35, 36);
mediump ivec3 user_inputDim = ivec3(5000, 1, 1);
out vec4 data0;
float kernelResult;
void kernel() {
float user_accumulator=0.0;
for (int user_i=1;(user_i<=5000);user_i++){
user_accumulator+=div_with_int_check((getMemoryOptimized32(user_input, user_inputSize, user_inputDim, 0, 0, threadId.x)+1.0), float(user_i));}

kernelResult = user_accumulator;return;
}
void main(void) {
  index = int(vTexCoord.s * float(uTexSize.x)) + int(vTexCoord.t * float(uTexSize.y)) * uTexSize.x;
  threadId = indexTo3D(index, uOutputDim);
  kernel();
  data0[0] = kernelResult;

}