As far as I know, it is possible to use Uint8Array as a texture in WebGL. But how is it possible to pass a large Float32Array or Float64Array to the shader efficiently? The float values are not in the (0.0, 1.0) range, and may be negative too.
I know, some devices does not support high precision float, and it is not a problem, if some precision is lost during the process.
Try to enable floating point textures with code like
var ext = gl.getExtension("OES_texture_float");
if (!ext) {
alert("sorry, no floating point textures");
return;
}
Now you can create textures with Float32Array using
gl.texImage2D(target, level, internalformat, width, height, 0,
format, gl.FLOAT, someFloat32Array);
Note that being able to filter a floating point texture is a separate extension so if you need to filter you'll have to check for that too
var ext = gl.getExtension("OES_texture_float_linear");
if (!ext) {
alert("sorry, can't filter floating point textures");
return;
}
Otherwise you need to set texture filtering to NEAREST
Note: in WebGL2 you can read from floating point textures by default. but you use an internal type of gl.RGBA32F. Filtering is still optional so you need to try to enable OES_texture_float_linear.
To render to them (attach them to a framebuffer). In WebGL1 you need to check for WEBGL_color_buffer_float. In WebGL2 you need to check for EXT_color_buffer_float.