Reproduce “Performance evaluation of accelerated complex multiple-precision LU decomposition” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tomonori Kouya
This research evaluates strategies to accelerate complex LU decomposition using multi-component multiple-precision arithmetic, revealing that an AVX2 SIMDized "Normal LU" decomposition offers the fastest computation times and strong parallel scalability while maintaining high numerical accuracy. It achieves up to 726 times faster performance than MPLAPACK for double-double precision.
Key insights
- The AVX2 SIMDized "Normal LU" decomposition, which relies on element-by-element elimination, demonstrated superior parallel scalability and overall performance for complex LU decomposition compared to methods based on fast matrix multiplication algorithms like Strassen or Ozaki.
- The 3M method consistently provides a speed improvement of approximately 1.3 times for serial complex matrix multiplication over the traditional 4M method.
- Effective utilization of modern hardware features, specifically AVX2 SIMD instructions and OpenMP parallelization, is critical for achieving substantial performance gains in multiple-precision complex linear algebra operations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.