A Debug Probe for Concurrently Debugging Multiple Embedded
Download
Report
Transcript A Debug Probe for Concurrently Debugging Multiple Embedded
Presenter : Shao-Cheih Hou
Sight count : 11
ASPDAC ‘08
Existing SoC debug techniques mainly target busbased systems. They are not readily applicable to the
emerging system that use Networkon-Chip (NoC) as
on-chip communication scheme. In this paper, we
present the detailed design of a novel debug probe
(DP) inserted between the core under debug (CUD)
and the NoC. With embedded configurable triggers,
delay control and timestamping mechanism, the
proposed DP is very effective for inter-core
transaction analysis as well as controlling embedded
cores’ debug processes. Experimental results show
the functionalities of the proposed DP and its area
overhead1.
More and more IP connect on bus in SoC
Single bus is not efficiency
Full connect is too complex and high cost
Network is a good way for chip(NoC)
In NoC, too many IPs need to verification
Design for debug => debug platform
Debug need more flexible
。Programmable trigger
。Trace (timestamping)
。Delay control
。Run stop
。Debug prob
Tracing[1,3,
12,13,22]
Coresight[4]
JTAG[21]
Internal
nodes[21]
Monitor for
NoC[7,8]
DfD for
embedded
cores[4]
Silicon
debug
platform
Open Core
peotocol[15]
Debug
method
This paper
interface
Debug software
Software control
Each core structure config
Driver match
On-chip debug architecture
DA(debug agent)
DP(debug probe)
Off-chip debug controller
OCP and JTAG interface
Command I/O
Data output
DA (Debug Agent)
Receive/send debug command/data from/to Off-chip
Debug Controller by using system JTAG port
• Change debug command to package form, and passing to
specific CUD or DP
• Receive package form network and change to debug data
6
Connect to Network interface(NI)
OCP protocol (slave interface)
Memory-map for each DP
Control the delay of the debug probe
For data synchronize
JTAG Interface
Another way to configure the delay control
ARM ICE JTAG protocol use in this paper
Dump trace data from core by bus
Solve “many-to-one”
Time slot for each core under debug
Trace bandwidth : TD(i)~Bchip(TS((𝑁 − 1)/𝑁)
Trigger Unit
Control for config
Detector as comparator
MUX for output record
Record format
Trace Unit
Trace after trigger
Shadow buffer for start cycle
Record package
The method for DP setup and run
Environment setup
ARM EICE JTAG
OCP protocol
Function Test – delay R/W
Function Test – Transaction operation
Area of DP
The paper propose a debug platform
For complex NoC base
Bus config and JTAG parallel
Trace for both core information and protocol information
Reconfigurable HW
Similar as problem in my target
NoC(not bus system) debug
Memory interface signal trace
JTAG and interconnect debug parallel
Problem for this paper
Only the simulation, no real case
Area analyze only for DP, not all the platform
No detail for SW and off-chip debug controller