c# - HLSL Computation - process pixels in order? -

imagine want to, say, compute first 1 million terms of fibonacci sequence using gpu. (i realize exceed precision limit of 32-bit data type - used example)

given gpu 40 shaders/stream processors, , cheating using reference book, can break million terms 40 blocks of 250,000 strips, , seed each shader 2 start values:

unit 0: 1,1 (which calculates 2,3,5,8,blah blah blah)

unit 1: 250,000th term

unit 2: 500,000th term

...

how, if possible, go ensuring pixels processed in order? if first few pixels in input texture have values (with rgba simplicity)

0,0,0,1 // initial condition 0,0,0,1 // initial condition 0,0,0,2 0,0,0,3 0,0,0,5 ... 

how can ensure don't try calculate 5th term before first 4 ready?

i realize done in multiple passes setting "ready" bit whenever value calculated, seems incredibly inefficient , sort of eliminates benefit of performing type of calculation on gpu.

opencl/cuda/etc provide nice ways this, i'm trying (for own edification) work xna/hlsl.

links or examples appreciated.

update/simplification

is possible write shader uses values 1 pixel influence values neighboring pixel?

you cannot determine order pixels processed. if could, break massive pixel throughput of shader pipelines. can calculating fibonacci sequence using non-recursive formula.

in question, trying serialize shader units run 1 after another. can use cpu right away , faster.

by way, multiple passes aren't slow might think, won't in case. cannot calculate next value without knowing previous ones, killing parallelization.