|
Multi-Standard DSP
Acceleration Template
IP Core
Overview:
The Multi-Standard DSP Acceleration Core is a fully
customizable architecture template for acceleration of DSP algorithms.
The IP core's architecture is based on a coarse-grained reconfigurable
array and is optimized for the execution of nested program loops
containing logic and arithmetic expressions. These loops are
computationally expensive and common in many application domains.
The architectures derived from the template are able to compute one
complete group of nested loops in each runtime configuration. The core
has a functional unit array and an embedded memory array to provide
highly parallel memory accesses and data operations. This allows the
core to outperform DSPs and processors in intensive data processing.
The core is a low power solution for intensive data processing and it’s
(re)configurability allows efficient use of the available hardware
resources.

Functional Description:
Configuration Register File: This block contains the runtime
configuration for executing a group of nested loops in a given core
instance.
Configuration and Control I/F: This interface is used to
reconfigure the core and to access control and status registers. The
supported interfaces are CW-Link, AMBA™- APB and Coreconnect™ OPB.
Input and Output ports: The input interface is used to write the
data to be processed and the output interface is used to collect the
processed data. The supported interfaces are CW-Link, AMBA™-AHB and
Coreconnect™-OPB.
Address Generator: This block computes the sequences of memory
addresses for reading operands from the memories and writing the results
back.
Read Crossbar: Configures the data
path from memory unit outputs or system input ports to functional unit
inputs. The read crossbar also allows routing the output of functional
units directly into the input of other functional units.
Write Crossbar: Configures the data path from functional unit
outputs to memory unit inputs or system output ports.
Memory Units (MU) Array: RAM or ROM embedded data memories used
to store the data being processed or tables of constants, respectively.
Functional Units (FU) Array: Array of processing units chosen
from a functional unit library. These units can be of any type and as
many as required for the operations being executed.
|
Key Features:
Features available at pre-synthesis time:
• Configurable nested loop maximum depth
• Configurable number of input and output ports
• Configurable type and number of functional units
• Configurable type, number and capacity of embedded memory units
• Configurable data path width
• Configurable data path routing
• Configurable address generation capabilities
Features available at runtime:
• Reconfigurable data path
• Reconfigurable address generation
General features:
• Supports runtime partial reconfiguration using a memory mapped
configuration register file addressable word by word
• CW-Link, Coreconnect™/OPB or AMBA™/AHB interface for data and
configuration
Applications:
• Multimedia data compressing and decompressing
• Communications
• Digital signal processing in general
|
Deliverables:
- Detailed datasheet and user documentation for
system integration
- HDL testbench covering all functionalities of
the core and including automatic verification of the correctness of
the responses
- FPGA Netlist
- HDL (VHDL or VERILOG) source code
- Simulation script
Part Number: CWcomp03
Price:

How To Purchase
|
|
FPGA Implementation Results:
|
Family |
Example Device
|
Slices |
IOB |
GCLK
|
Mult18x18
|
BRAM
|
Max. Clock Frequency
|
Design Tools
|
|
Spartan-3™ |
XC3S5000-4 |
1,949 |
461 |
1 |
8 |
configurable |
66 MHz |
ISE 9.2i |
|
Virtex-II™ |
XC2V1500-4 |
1,969 |
461 |
1 |
8 |
configurable |
87 MHz |
ISE 9.2i |
|
Virtex-4™ |
XC4VLX15-10 |
1,969 |
461 |
1 |
8 |
configurable |
125 MHz |
ISE 9.2i |
|
Virtex-5™ |
XC5VLX50-1 |
1,198 |
461 |
1 |
8 |
configurable |
158 MHz |
ISE 9.2i |
ASIC Implementation Results:
|
Technology |
Gate count
|
Area |
RAM |
Max. Clock Frequency
|
|
UMC 130 nm |
58 k |
0.22 mm2 |
configurable |
>166 MHz |
|
|
|
|











|