// read the PDF about this if you precompute it! it's got helpful info!

// can be optimized into lut (compute can gen it)
float GTTonemap(float x) {
	float m = 0.22; // linear section start
	float a = 1.0;  // contrast
	float c = 1.33; // black brightness
	float P = 1.0;  // maximum brightness
	float l = 0.4;  // linear section length
	float l0 = ((P-m)*l) / a; // 0.312
	float S0 = m + l0; // 0.532
	float S1 = m + a * l0; // 0.532
	float C2 = (a*P) / (P - S1); // 2.13675213675
	float L = m + a * (x - m);
	float T = m * pow(x/m, c);
	float S = P - (P - S1) * exp(-C2*(x - S0)/P);
	float w0 = 1 - smoothstep(0.0, m, x);
	float w2 = (x < m+l)?0:1;
	float w1 = 1 - w0 - w2;
	return float(T * w0 + L * w1 + S * w2);
}

// this costs about 0.2-0.3ms more than aces, as-is
vec3 GTTonemap(vec3 x) {
	return vec3(
		GTTonemap(x.r),
		GTTonemap(x.g),
		GTTonemap(x.b)
	);
}