Reconfigurable Baseband Processing for Wireless Communications - Adaptive Signal Processing in Wireless Communications - page 467

Table 15.9 The Parameter Set for the
Parameterizable OFDM

B Bandwidth of OFDM system

N sym Number of OFDM symbols per frame

N preamble Preamble length per frame

N Number of OFDM samples per symbol

N g Number of guard samples per symbol

N data Number of useful data per symbol

mod Modulation modes

tab pilot Table for pilot information

tab format Table for format information

OFDM symbol. Modulation modes (mod) indicate the modulation type for the OFDM
samples that carry useful information. The modulation mode in one OFDM symbol can
differ on a subcarrier basis. In the cognitive radio case, the modulation mode can be set
to zero to nullify carriers.

15.4 Mapping Applications to Multicore Architectures

In the most general terms, spatial mapping is the allocation of spatial resources to
applications. As such, spatial mapping is to space what scheduling is to time. In the
context of a multiprocessor system-on-chip (MP-SoC), the spatial resources to allocate
are tiles and interconnections between tiles. These tiles may be processors or memory
(controllers). Although the allocation of columns of fabric inside an FPGA could also
be described as a spatial mapping problem, the system-on-chip context is considerably
more coarse grained.
Baseband processing applications in wireless communications are streaming applica-
tions. Such streaming applications can be decomposed into multiple communicating
computational kernels. Most streaming applications have intuitive decompositions into
chains of such kernels, which is illustrated by the fact that most standards contain block
diagrams of the components that make up the entire application [e.g., 17]. This decom-
position can be as general as a Kahn process network (KPN) [32], or if more detailed
information is available, some specialized form of a data flow graph (DFG).
A commonly used specialized DFG is the synchronous data flow graph (SDFG) [36].
The nodes of an SDFG are generally referred to as actors. The edges between actors are
annotated with tokens. Tokens are an abstract notion of data that is communicated
between two actors.
Actors are annotated for every incoming and outgoing edge. On incoming edges, the
annotation is the number of tokens required on that specific input to fire the actor once,
i.e., to execute the corresponding computational kernel. On outgoing edges, the annota-
tion is the number of tokens the actor will produce after firing. Actors in an SDFG are
self-timed, i.e., there is no global arbitration of what actor fires when, but rather a simple
local firing rule: an actor may fire when all incoming edges have at least as many tokens
on them as annotated.