|The Eighteenth Annual Interactive Audio Conference
PROJECT BAR-B-Q 2013
|Group Report: When is Hardware Offloading Preferable, Now and in the Future?|
|Participants: A.K.A. "Outsourced! (The musical)"|
|David Berol, Audience, Inc.||Kate Werner, Microsoft|
|Andy Rumelt, Cirrus Logic, Inc.||Adam Fluckey, Dolby Laboratories|
|David Roach, Optimal Sound||Konstantin Merkher, CEVA DSP, Ltd|
|Howard Brown, IDT||Mike Spence, Texas Instruments|
|Miguel Chavez, Analog Devices||Facilitator: Linda Law, Project Bar-B-Q|
|download the PDF|
Audio is a basic feature of the most popular computing platforms today. When building a device based on these platforms the location of audio processing is always considered. To put it simply, architects have two options where the audio processing is executed: the host or the offload processor. This report explores the use cases that justify the need for the additional engineering effort, expense, and implementation of audio hardware offloading.
Summary of Findings
During the initial discussions the team quickly found that this topic can get use case and device specific very quickly. In order to up-level, the team created the following boundaries:
Generally the team found that individual feature sets make audio hardware offload less attractive. However as more and more feature sets are grouped together the benefits start to add up thus making audio hardware offload appealing. In addition, the team found some specific use cases that should always be offloaded (e.g. ANC and speaker protection) due to the latency requirements.
As day turned to night, the team discussed the impact of “walled gardens” in the industry and concluded that “walled gardens” are driven by individual business strategies. A deep dive explaining the keys to the walled gardens was mutually destructive.
Expanded solution description
We analyzed each use case to determine whether hardware acceleration provides an advantage in latency (performance), power usage, and/or cost, for both the present and a 3-to-5 year future. Here are the results:
Here are some categories of future products (edge cases for the “then” use cases) which we did not analyze.
Future Technology Breakthroughs that affect offloading
Other reference material
Cite Matrix from Doppler Chickens 2012 Project Bar-B-Q Workgroup:
select a section:
Copyright 2000-2014, Fat Labs, Inc., ALL RIGHTS RESERVED