This page contains samples of recent work by Omar and Sadiq, ahead of pre-prints.

For context, I suggest starting with:

• Benchmarking on real datasets (speedups, lossless & lossy compressibility)
  • Broad technical article benchmarking the compressibility of the datasets of interest, with a variety of compression schemes. Also does a bit of speedup benchmarking.
  • Focus on compression ratios and loss. Narrow perf stuff is in later articles.

Then, you have enough context for the low-level fusion and performance side:

• Fusion speedups - sum, single-band, all codecs
  • Fusion of fundamental computation (sum) into various SOTA decompressors (Frame of Reference, PFOR, bitpacking).
  • Includes single-threaded benchmark results (lossless & lossy).
• Fusion speedups - NDVI, multi-band
  • Fusion of a canonical multi-band computation (NDVI) into bit unpacking.
  • Includes multi-threaded benchmarks (only lossless).

If you’re interested in the surrounding literature and how our method fits:

• k2-acc/tree/raster lineage - comparison with our work