Changes between Version 15 and Version 16 of cypress/XeonPhi


Ignore:
Timestamp:
Aug 21, 2015 10:01:24 AM (6 years ago)
Author:
cmaggio
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • cypress/XeonPhi

    v15 v16  
    486486=== Offloading ===
    487487
     488Offloading allows us to designate specific sections of our code that we wish to have executed on the MIC. Unlike the first two methods, this requires (minimal) alteration of the source code. Lets take our helloflop3.c example and modify it so that it offloads the area of heavy computation
    488489
    489490
     
    494495== Programming Considerations ==
    495496
    496 The number one thing to keep in mind is that all data traffic to and from the coprocessors must travel over PCIE. This is a relatively slow connection when compared to memory and the more you can minimize this communication, the faster you code will run.
     497As we've seen, getting code to execute on the Xeon Phi Coprocessor can be accomplished in a manner of minutes or hours. However, getting production code to run optimally on the MICs is often an effort spanning weeks or months. A few of the key considerations that can speed up your code are:
     498* Cache alignment
     499* Vectorization
     500* Blocking
     501* Minimal data transfer
    497502
    498503