Low-Latency Computing

 
 
 
 

Computers we've used over the years:

  • Multi-core Intel Pentium®, Xeon®
  • Reduced Instruction Set (RISC) i860, SPARC
  • DEC Alpha
  • Massively parallel (various)
  • Custom LSI Logic ASIC
  • Xilinx, Altera & QuickLogic FPGAs
  • TI TMS320® DSPs
  • Various microprocessors
 
 
 
 

AOA Xinetics started building adaptive optics (AO) systems in 1976. These measure how a flat wavefront of light gets distorted into a "potato chip shape" by the atmosphere, and corrects it by bending a deformable mirror to the same shape. When the bent optical wavefront reflects off the similarly bent mirror the result is flat, as if there was no distorting atmosphere.

The toughest computing requirement of AO systems is that for each real-time wavefront measurement, several million calculations must be performed, and the answers must be available just a few milliseconds after light reaches the camera.

Although the computer hardware we use has changed drastically over the years, we still use dataflow, multi-threaded, multi-processor software for minimum latency. Many applications benefit from getting answers faster, storing less data, and from more predictable computation times.

 
 
 
 

EXAMPLE DATAFLOW DIAGRAM

DATAFLOW DIAGRAM Operations are performed as soon as operands to each stage are available, reducing latency.
 
 
 
 

We also use software such as the Xenomai real-time kernel on the Linux OS for extremely low latencies.

Whether running on a mobile smartphone ARM processor or on a server-grade, hyperthreading, 10-core Xeon® processor, AOA Xinetics' computing approach delivers results—before you need the answer! 

If you have a latency problem, contact AOX for a solution.



SPARC® is a registered trademark of SPARC International, Inc.. Xeon® and Pentium® are registered trademarks of Intel Corporation. TMS320® is a registered trademark of Texas instruments Inc.