Pack 4 8-bit unsigned integer to a 32-bit float

Question

I'm developing a webgl-based renderer and as the title says I need to pack 4 8-bit unsigned integer to a 32-bit float, I wrote the following code :

//pack 4 8-bit integer to a float
function packIntsToFloat(i1, i2, i3, i4) {

    //ensure 32-bit allocation
    var ints = new Uint32Array(4);

    ints[0] = i1;
    ints[1] = i2;
    ints[2] = i3;
    ints[3] = i4;

    ints[0] <<= 24;
    ints[1] <<= 16;
    ints[2] <<= 8;

    ints[3] |= ints[0] | ints[1] | ints[2];      

    //convert to float
    var f = new Float32Array(1);
    f[0] = ints[3];

    return f[0];
}

//unpack 4 8-bit integer from a float
function unPackIntsFromFloat(f) {

    var i = new Uint32Array(4);
    i[3] = f;

    var mask_7_to_0 = 255;
    var mask_15_to_8 = mask_7_to_0 << 8;
    var mask_23_to_16 = mask_15_to_8 << 8;
    var mask_31_to_24 = mask_23_to_16 << 8;

    i[0] = (i[3] & mask_31_to_24 ) >>> 24;
    i[1] = (i[3] & mask_23_to_16 ) >>> 16;
    i[2] = (i[3] & mask_15_to_8 ) >>> 8;
    i[3] = (i[3] & mask_7_to_0);        

    return new Uint8Array(i);
}

but it won't work unless skipping what I need :

//convert to float
var f = new Float32Array(1);
f[0] = ints[3];

I'm aware of the IEEE standard but there shouldn't be any change of the bits, only their interpretation as a value. Thank you in advance.

score 3 · Accepted Answer · answered Oct 01 '14 at 23:15

3

All TypedArrays are just views on a type agnostic ArrayBuffer, so you can simply use this:

Packing: new Float32Array((new Uint8Array([i1,i2,i3,i4])).buffer)[0];

Unpacking: new Uint8Array((new Float32Array([f])).buffer);

See the documentation of ArrayBufferView for more information.

answered Oct 01 '14 at 23:15

LJᛃ

7,655
2
24
35

Thank you very much! Very very helpful, can you also explain the bevahior of the code I wrote ? – user2302585 Oct 02 '14 at 07:07

score 1 · Answer 2 · answered Oct 02 '14 at 09:12

What are you trying to do exactly?

For example are you trying to put both float positions and unsigned byte colors in the same buffer? In that case make 2 views to the same buffer. Example:

var numVerts = 10;
var bytesPerPosition = 3 * 4;  // x,y,z * 4 bytes per float
var bytesPerColor = 4;  // r,g,b,a 1 byte each
var bytesPerVertex = bytesPerPosition * bytesPerColor;
var sizeOfBuffer = numVertex * bytesPerVertex;
var offsetOfPositions = 0;
var offsetOfColor = bytesPerPosition;

// now make the buffer.
var asUint8 = new Uint8Array(sizeOfBuffer);
var asFloat = new FloatArray(asUint8.buffer);

You now have 2 views to the same buffer. So for example to set a position you'd effectively do

var strideInFloats = bytesPerVertex / 4;

function setPosition(index, x, y, z) {
  var offset = strideInFloats * index;
  asFloat[offset    ] = x;
  asFloat[offset + 1] = y;
  asFloat[offset + 2] = z;
}

To set a color would be effectively

function setColor(index, r, g, b, a) {
  var offset = strideInBytes * index + offsetToColor;
  asUint8[offset    ] = r;
  asUint8[offset + 1] = g;
  asUint8[offset + 2] = b;
  asUint8[offset + 3] = a;
}

When setting up attributes, you'd do something like

gl.vertexAttribPointer(positionLoc, 3, gl.FLOAT, false, bytesPerVertex, 
                       offsetOfPosition);
gl.vertexAttribPointer(colorLoc, 4, gl.UNSIGNED_BYTE, true, bytesPerVertex, 
                       offsetOfColor);

Pack 4 8-bit unsigned integer to a 32-bit float

2 Answers2