Surpass specialized hardware in high-speed processing

Surpass specialized hardware in high-speed processing in the field requires excellent computational performance such as latest scientific calculations and finance simulation.

Reduction of product development costs

GPU makes it possible to solve complicated problems with fewer resources and energy by high performance and high density computing.

Excellent scalability

The many-core-architecture realizes¡¡applications which require the high performance computing that cannot be achieved by other CPUs.

Ported and optimized PIV algorithm to CUDA. PIV algorithm is used to calculate velocity vector from particle displacement. More than 60 fold improvement in processing performance compared to the original PIV processing code ran on AMD Opteron 2220 2.8GHz.

Consulting and Profiling

Decided porting line and hardware by analysis of original source code.

• What we need from you:
  
  • Your existing code, preferably not optimized for any processor; or
  • Pseudo-code, code that is available to the public such as an open source which equivalent, a similar algorithm or program to your own code.
• Procedure:
  
  1. Agree upon target performance.
  2. Profile your code.
  3. Porting and optimize a portion of your code.
  4. Report the results of the initial optimization.

Porting and Optimization

Fixstars's engineers, with our expertise in the software development for multi-core processor, port and optimize applications on CUDA when it reached the target performance through profiling.

• Procedure:
  
  1. Submit the quotation and schedule coordination
  2. Contract
  3. Porting and optimization
  4. Delivery
  5. Report the results of the initial optimization

Software Development Support

We support after shipping such as install support and technical transfer.

Training

We are providing a practical training for CUDA programming to make it possible that clients develop and maintain software by themselves. Also, we are organizing CUDA programming seminar constantly.