CODEWITHSUNDEEP
androidopengl-esopengl-es-2.0glsl
Timothy
Timothy

Reputation: 469

Is it possible to use all the 16 bits of a float to sample texture in GLSL

I am now trying to sample a texture using floating point vectors but I found that the floating point precision is only 2 ^ -10 bits and the resolution is very limited. So I am wondering whether it is possible for me to use all the 16 bits of a mediump floating variable to sample textures? Thank you.

Upvotes: 0

Views: 589

Answers (1)

Nicol Bolas
Nicol Bolas

Reputation: 473407

You are using all 16 bits of a 16-bit floating point number. That's what a 16-bit floating-point number is. They have ~11 bits of mantissa and 5 bits of exponent. You cannot make a half-precision floating-point value have more than ~11 bits of mantissa.

You could try multiplying the number by 4096 in the vertex shader, but this would only help if the texture coordinate was originally a 32-bit float (before converting it to half-precision).

In short: texture coordinates generally need more precision than this. So you're going to have to use a higher-precision floating-point value.

Upvotes: 2

Related Questions