29.01.2015 Views

Embedded Software for SoC - Grupo de Mecatrônica EESC/USP

Embedded Software for SoC - Grupo de Mecatrônica EESC/USP

Embedded Software for SoC - Grupo de Mecatrônica EESC/USP

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

130 Chapter 10<br />

The <strong>de</strong>lay of communication interface is given as a function of data<br />

size [14].<br />

The <strong>de</strong>lay of on-chip communication network is<br />

where is a constant <strong>de</strong>lay of transferring a single data item and<br />

is a <strong>de</strong>lay caused by the contention of access to the communication network<br />

(e.g. bus contention).<br />

and can be calculated during task/communication scheduling,<br />

which is explained in the next subsection.<br />

3.2.2. ILP <strong>for</strong> scheduling communication no<strong>de</strong>s and tasks<br />

In scheduling tasks and communication transactions, we need to respect data<br />

<strong>de</strong>pen<strong>de</strong>ncy between tasks and to resolve resource contention on the target<br />

architecture. In the target architecture, there are three types of resource that<br />

are shared by tasks and communications: processors, communication networks<br />

(e.g. on-chip buses), and physical buffers. There<strong>for</strong>e, we mo<strong>de</strong>l the scheduling<br />

problem with (1) data <strong>de</strong>pen<strong>de</strong>ncy constraints between tasks, (2) constraints<br />

that represent resource contention and (3) an objective function that<br />

aims to yield the minimum execution time of the task graph.<br />

3.2.2.1. Data <strong>de</strong>pen<strong>de</strong>ncy constraints<br />

In the ETG, data and communication <strong>de</strong>pen<strong>de</strong>ncy is represented by directed<br />

edges. For every edge from no<strong>de</strong> to no<strong>de</strong> the following inequalities are<br />

satisfied.<br />

3.2.2.2. Resource contention constraints<br />

Processor and on-chip communication network. Processor contention and<br />

on-chip communication network contention are similar in that the contentions<br />

occur when two or more components are trying to access the same resource<br />

(processor or on-chip communication network) at the same time. To <strong>for</strong>mulate<br />

the processor contention constraints in the ILP, first we find a set of task<br />

no<strong>de</strong>s that are mapped on the same processor. For the task set, the processor<br />

contention occurs when two tasks, and try to use the same processor<br />

simultaneously. To prevent such a processor contention, we need to satisfy<br />

or<br />

The same concept applies to the on-chip<br />

communication networks such as on-chip buses and point-to-point interconnections.<br />

Physical buffer contention constraints. Physical buffer contention constraints<br />

are different from those of processor and on-chip communication network

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!