The choice of formats for OpenGL ES is very limited, so you have to choose between GL_R16F and GL_R8. Both formats are available for desktop.

GL_R16F is probably enough for your purposes.. or is it? Values won't be normalized and it will require you to generate data on the CPU. The reason behind CPU-only restriction that is because GL_R16F is not a color-renderable format, so you can't attach it to the framebuffer. But at least it will let you to use filtering.

Personally I think that 8 bits wouldn't be a bad choice, because it is filterable and for simplex noise you don't really need too much precision when in the end results are being interpolated in the shader. And it's an unsigned normalized integer format, meaning you won't have to deal with conversions to utilize full precision. And you can attach it to the framebuffer to use shader for noise computation.

The choice of formats for OpenGL ES is very limited, so you have to choose between GL_R16F and GL_R8. Both formats are available for desktop.

GL_R16F is probably enough for your purposes.. or is it? Values won't be normalized and it will require you to generate data on the CPU. The reason behind CPU-only restriction is because GL_R16F is not a color-renderable format, so you can't attach it to the framebuffer. But at least it will let you to use filtering.

Personally I think that 8 bits wouldn't be a bad choice, because it is filterable and for simplex noise you don't really need too much precision when in the end results are being interpolated in the shader. And it's an unsigned normalized integer format, meaning you won't have to deal with conversions to utilize full precision. And you can attach it to the framebuffer to use shader for noise computation.

The choice of formats for OpenGL ES is very limited, so you have to choose between GL_R16F and GL_R8. Both formats are available for desktop.

GL_R16F is probably enough for your purposes.. or is it? Values won't be normalized and it will require you to generate data on the CPU. The reason behind framebuffer restriction that is because GL_R16F is not a color-renderable format, so you can't attach it to the framebuffer, but it will let you use filtering.

Personally I think that 8 bits wouldn't be a bad choice, because it is filterable and for simplex noise you don't really need too much precision when in the end results are being interpolated in the shader. And it's an unsigned normalized integer format, meaning you won't have to deal with conversions to utilize full precision. And you can attach it to the framebuffer to use shader for noise computation.

The choice of formats for OpenGL ES is very limited, so you have to choose between GL_R16F and GL_R8. Both formats are available for desktop.

GL_R16F is probably enough for your purposes.. or is it? Values won't be normalized and it will require you to generate data on the CPU. The reason behind CPU-only restriction that is because GL_R16F is not a color-renderable format, so you can't attach it to the framebuffer. But at least it will let you to use filtering.

Personally I think that 8 bits wouldn't be a bad choice, because it is filterable and for simplex noise you don't really need too much precision when in the end results are being interpolated in the shader. And it's an unsigned normalized integer format, meaning you won't have to deal with conversions to utilize full precision. And you can attach it to the framebuffer to use shader for noise computation.