Tailored Software Optimization Services

Are you looking to optimize performance, computing cost or power consumption of your algorithm?

Talk to an Expert
What we offer

Huxelerate, thanks to its proprietary acceleration technology, provides premium software optimization services, in a fraction of the time compared to standard consulting options.

  • 1. ADVANCED PERFORMANCE ESTIMATION REPORTS
  • 2. SOFTWARE OTPIMIZATION SERVICES ON FPGA, CPU & GPU
Up to 30x Performance Improvements

Achieve outstanding performance

Have a look at some previous works

Sequence Aligner

Comparison between Huxelerate Optimized version of Seqan 2.4 on FPGA and non optimized version on CPU. Baseline runs performed using localAlignement function parallelized with OpenMP on multiple threads.

Experimental Settings

  • Pairwise local alignment of multiple pairs of sequences (query/target length: 8000)
  • Target Architectures:
    • Intel Xeon E5-2666 v3 (36 threads)
    • Intel Xeon Platinum (72 threads)
    • AMD EPYC 7002 (96 threads)
    • Xilinx VU9P+ Intel Xeon E5-2686 v4 (8 threads)
  • Reference hourly cost are expressed in $, and are taken from AWS pricing.

Signal Aligner

Comparison between Huxelerate Optimized version of lbimproved on FPGA and non-optimized version on CPU. Baseline run performed with lbimproved using the fastdtw function parallelized with OpenMP.

Experimental Settings

  • Dynamic Time Warping alignment of multiple pairs of sequences (query/target length: 2048)
  • Target Architectures:
    • Intel Xeon E5-2666 v3 (36 threads)
    • Intel Xeon Platinum (72 threads)
    • AMD EPYC 7002 (96 threads)
    • Xilinx VU9P+ Intel Xeon E5-2686 v4 (8 threads)
  • Reference hourly costs are expressed in $, and are taken from AWS pricing.

Ksw2 Aligner

Comparison between Huxelerate Optimized version of ksw2_extd_sse on FPGA and non-optimized version on CPU. Baseline run performed with ksw2_extd_sse parallelized with OpenMP on multiple threads.

Experimental Settings

  • Extensions alignment of multiple pairs of sequences (query/target length: 32000)
  • Target Architectures:
    • Intel Xeon E5-2666 v3 (36 threads)
    • Intel Xeon Platinum (72 threads)
    • AMD EPYC 7002 (96 threads)
    • Xilinx VU9P+ Intel Xeon E5-2686 v4 (8 threads)
  • Reference hourly costs are expressed in $, and are taken from AWS pricing (hourly cost for AMD Ryzen 3700x not available).

You can directly test them

Sequence Aligner

Huxelerate sequence aligner computes a similarity measure of two text sequences or subsequences and identifies the best way to align them according to a user-defined scoring scheme. The function perform an accelerated version of Smith-Waterman and Needleman–Wunsch alignment algorithms, computational bottlenecks in bioinformatic which find applications also in computer security and text analysis.

Signal Aligner

Huxelerate signal aligner computes a similarity measure across two time series which may vary in speed. The function implements an accelerated version of the Dynamic Time Warping algorithm, a computational bottleneck for signal classification and recognition tasks. Example applications involve genome sequencing and analysis with nanopore technology, connected health with ECG-based diagnostic and humanized machines with speech recognition.

Ksw2c Aligner

Huxelerate ksw2c aligner computes the optimal scores and the CIGAR for global and extension alignment of two sequences using dual gap affine cost function and a custom scoring matrix. The core also supports banded alignment and the z-drop heuristics - info here The function implements an accelerated version of ksw2_extd software

ksw2s Aligner

Huxelerate ksw2s provides the same functionality as ksw2c aligner but the CIGAR. This FPGA implementation is in fact optimized in performance to calculate only the optimal scores.

Want to know more?
Talk to an expert