High Throughput Compression of Double-Precision Floating-Point Data

gFPC: A Self-Tuning Compression Algorithm

“…gFPC outperforms FPC size , especially for larger predictor sizes, because FPC size uses the same configuration for all inputs whereas gFPC individually tunes itself for each in- 1 The result is not guaranteed to be optimal but is assumed to be at least close. put.…”

“…gFPC is based on FPC [1,2] and compresses linear sequences of IEEE 754 doubleprecision floating-point values by sequentially predicting each value, xoring the true value with the predicted value, and leading-zero compressing the result. As illustrated in Figure 1, it uses variants of an fcm [3] and a dfcm [4] value predictor to predict the doubles.…”

“…The optimal setting of these parameters is often input dependent and may even change during compression. This paper presents an on-line, genetic-algorithm-based, self-tuning approach to automatically optimize such parameters and demonstrates how self tuning works on the example of the FPC compression algorithm [1].…”

Lossless Compression of Hexahedral Meshes

“…Each row holds the vertices of an element that span the three spatial dimensions. Though the example in Table 1 is overly simplistic, the runs of constant strides v A more general stride-based approach is the differential finite context method (DFCM) [13], which has been used successfully for trace file and floating-point compression [7,12]. The basic DFCM predictor is a hash table that maps a set of recent strides to the current, predicted stride.…”

“…For example, the methods proposed in [4,5] require memory on the order of several hundreds to over a thousand bytes per hexahedron. Second, for checkpointing purposes and accurate analysis, the compression scheme must be lossless [6,7]. Not only does this requirement rule out quantization, but it also implies that the geometry and connectivity arrays may not be reordered so that the simulation state can be perfectly recovered, e.g.…”