Using graphics boards to compute holograms

Holographic Optical Tweezers

“…With eight "grayscale" phase levels and all other parameters set as before, we obtained e = 0.84 and u = 1.00 after 7N steps (that is, about three times longer than GS). When speed is an important issue, the time needed to compute an SR hologram, or equivalently to perform a single step in GS-based algorithms, can be efficiently reduced by relying on the Graphic Processing Unit (GPU) of a graphic board for both computational and rendering tasks [98]. For a better exploration of the gain function landscape, we could define more complex acceptance rules, allowing moves that temporarily decrease the gain (as in a Metropolis algorithm).…”

Combining rigorous diffraction calculation and GPU accelerated nonsequential raytracing for high precision simulation of a linear grating spectrometer

“…Therefore, these GPU cores certainly don't compare easily to CPU cores. Nevertheless, they have a strong potential for speeding up scientific computing in general [15]- [17] and ray tracing applications in particular [18]. Even though in principle all graphic cards can be used to do general purpose computations, it was nVidia who started to open this application of GPUs to a broad community of software developers with the release of CUDA in 2007 [14].…”

Applications of LCoS-based adaptive optical elements in microscopy

“…Today, the computational power of the graphics processing units is by far superior to that of the CPUs of standard PCs. Therefore the desired computational speed for the hologram update can now be obtained by implementation of tailored algorithms for HOT generation on the GPU [4], but also the implementation of the IFTA with sufficient speed has become possible. We have achieved hologram update rates of >20 Hz using current NVidia hardware and a tailored implementation of the IFTA.…”