Reducing memory usage by the lifting-based discrete wavelet transform with a unified buffer on a GPU

Abstract

h i g h l i g h t s• We propose a memory saving method for the lifting-based discrete wavelet transform on a GPU. • Our method reduces the memory usage by unifying the input buffer and output buffer. • Our compact data representation interprets a sequence of memory addresses as a circular permutation.• Experimental results on four GPU architectures are presented. • Our unified method is capable of transforming a twice large problem size with a maximum speedup of 3.9. a b s t r a c tIn this study, to improve the speed of the lifting-based discrete wavelet transform (DWT) for large-scale data, we propose a parallel method that achieves low memory usage and highly efficient memory access on a graphics processing unit (GPU). The proposed method reduces the memory usage by unifying the input buffer and output buffer but at the cost of a working memory region that is smaller than the data size n. The method partitions the input data into small chunks, which are then rearranged into groups so different groups of chunks can be processed in parallel. This data rearrangement scheme classifies chunks in terms of data dependency but it also facilitates transformation via simultaneous access to contiguous memory regions, which can be handled efficiently by the GPU. In addition, this data rearrangement is interpreted as a product of circular permutations such that a sequence of seeds, which is an order of magnitude shorter than input data, allows the GPU threads to compute the complicated memory indexes needed for parallel rearrangement. Because the DWT is usually part of a processing pipeline in an application, we believe that the proposed method is useful for retaining the amount of memory for use by other pipeline stages.

Multi‐band joint local sparse tracking via wavelet transforms

“…In the next step, we will try using graphics processing unit (GPU) to achieve the wavelet transform and solution of sparse representation with low cost and fast process in the proposed algorithm. Similar work has been studied by some scholars [38][39][40][41], but these methods are not quite suitable for our situation. Modern GPUs mostly use the pipeline structure with altitudinal parallelism, which can implement parallel processing on large amount of calculation.…”

Lifting-Based Fractional Wavelet Filter: Energy-Efficient DWT Architecture for Low-Cost Wearable Sensors

“…However, bit truncation degrades the quality of the reconstructed image when the inverse DWT is applied. ere have been some DWT implementations on graphics processing units (GPUs) [72][73][74][75][76][77][78]; however, GPUs are relatively expensive for low-cost sensing platforms.…”

A generic parallel computational framework of lifting wavelet transform for online engineering surface filtration

“…The GPCF-LWT integrates the devised LBB, improved lifting scheme, the three implementation optimizations, and the improved data -oriented DLB. This analysis evaluates the SURFSTAND that is a surface characterization system employing modern CPU core [10], and two other multi-GPU implementations of RC&LB and the method proposed by Ikuzawa [46] [45]. It should be noted that the pure dataset division method was applied for the two multi-GPU implementations.…”

“…It is worth mentioning that CUDA supports the single float (32-bit, abbreviated as FP32) and the half float (16-bit, FP16) precisions [19,45] [19,46]. This experiment tested both of these two precisions.…”