Research
  GPU Software Development
    21st Century Super Computing
    Real World Example – Fast K-Nearest Neighbor Search
  Hyperspectral Image Processing
  Precision Time Synchronization

 

Inc. 500

GPU Software Development 


Celestech continues to develop software application expertise and world-class algorithms for high performance parallel processors.  This includes adapting Graphics Processing Units (GPUs) to perform signal and hyperspectral processing applications with amazing performance.  Specific platforms include the NVIDIA GeForce and Tesla using the Compute Unified Device Architecture (CUDA) application program interface. 

 

This software interface allows an off the shelf Graphics Processing Unit to serve as a high performance floating point compute engine for a wide range of applications.  These Graphics Processing Units deliver extraordinary performance (Teraflop+) in a small, inexpensive package with unequaled FLOPS per dollar, pound, watt and cubic foot ratios.

 

Our technology research applies to a wide range of computationally intensive problems including:

  • Image Processing (Hyperspectral, SAR, etc.)
  • Signal Processing
  • Fast Search Algorithms (K-Nearest Neighbor)
  • Artificial Intelligence
  • Array Processing
  • Matrix Operations
  • Encryption/Decryption

 

Celestech can provide:

  • Proof of Concept Demonstrations
  • Parallel GPU Algorithm Development
  • Desktop and Server Applications and Subsystems

 

While GPU technology offers significant advantages, development has a steep learning curve.  Celestech’s GPU Software developers and computer scientists have mastered these learning curve challenges and have developed the skills and experience to rapidly apply their knowledge and expertise to resolving your most challenging computational and development problems.

 


Related Links

Research
21st Century Super Computing
Real World Example – Fast K-Nearest Neighbor Search