The ChipList, by Adrian Offerman; The Processor Portal

new
Processor Selector

Platform:
Segment:
View: show / edit

bookmark bookmark site
bookmark permalink
Fri 8 Jun 2012, 9:00

Performance Interactions of OpenCL* Code and Intel® Quick Sync Video on Intel® HD Graphics 4000


Performance Interactions of OpenCL* Code and Intel® Quick Sync Video on Intel® HD Graphics 4000

Download Article

Download Performance Interactions of OpenCL* Code and Intel® Quick Sync Video on Intel® HD Graphics 4000 [PDF 488KB]

Introduction

Developers of video editing and other applications that generate or process frames of video, and subsequently encode them using Intel® Quick Sync Video, may find it challenging to gain performance advantages using OpenCL* to shift frame processing from the CPU to Intel® HD Graphics. This paper explains why this is and how to identify workloads that can show positive improvement, enabling developers to make the most productive use of their OpenCL coding efforts.

 

An Unexpected Problem

In support of preparing OpenCL on Intel processor graphics for the 3rd generation Intel® Core™ processor launch, I optimized a number of OpenCL kernels for a video editing application. The expectation was that by moving some processing of video image effects from the CPU to processor graphics, application performance would be substantially improved.

The OpenCL kernels did indeed show good unit test performance gains as compared to a single CPU core running the same effect on well-optimized code—on average about three times (3x) faster, with some kernels as much as 20x faster than the equivalent CPU code! ...

Filed under: Visual Computing Source
Also published by:

ChipList news channel Last Months News


ChipList developers news channel Last Months Developers News


Page viewed 2043 times since Fri 8 Jun 2012, 23:42.