Intel® IXP42X product line and IXC1100 control plane processors—Intel XScale® Processor
Intel® IXP42X Product Line of Network Processors and IXC1100 Control Plane Processor
DM September 2006
110 Order Number: 252480-006US
Any access to the trace buffer registers in User mode will cause an undefined
instruction exception. Specifying registers which do not exist has unpredictable results.
3.6.12.1.1 Checkpoint Registers
When the debugger reconstructs a trace history, it is required to start at the oldest
trace buffer entry and construct a trace going forward. In fill-once mode and wrap-
around mode when the buffer does not wrap around, the trace can be reconstructed by
starting from the point in the code where the trace buffer was first enabled.
The difficulty occurs in wrap-around mode when the trace buffer wraps around at least
once. In this case the debugger gets a snapshot of the last N control flow changes in
the program, where N <= size of buffer. The debugger does not know the starting
address of the oldest entry read from the trace buffer. The checkpoint registers provide
reference addresses to help reduce this problem.
The two checkpoint registers (CHKPT0, CHKPT1) on IXP42X product line and IXC1100
control plane processors provide the debugger with two reference addresses to use for
re-constructing the trace history.
When the trace buffer is enabled, reading and writing to either checkpoint register has
unpredictable results. When the trace buffer is disabled, writing to a checkpoint register
sets the register to the value written. Reading the checkpoint registers returns the
value of the register.
In normal usage, the checkpoint registers are used to hold target addresses of specific
entries in the trace buffer. Only direct and indirect entries get check-pointed. Exception
and roll-over messages are never check-pointed. When an entry is check-pointed, the
processor sets bit 6 of the message byte to indicate this (refer to Table 49., Message
Byte Formats)
When the trace buffer contains only one check-pointed entry, the corresponding
checkpoint register is CHKPT0. When the trace buffer wraps around, two entries will
typically be check-pointed, usually about half a buffers length apart. In this case, the
first (oldest) check-pointed entry read from the trace buffer corresponds to CHKPT1,
the second check-pointed entry corresponds to CHKPT0.
Although the checkpoint registers are provided for wrap-around mode, they are still
valid in fill-once mode.
Table 46. CP 14 Trace Buffer Register Summary
CP14 Register Number Register Name
11 Trace Buffer Register (TBREG)
12 Checkpoint 0 Register (CHKPT0)
13 Checkpoint 1 Register (CHKPT1)
Table 47. Checkpoint Register (CHKPTx)
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9876543210
CHKPTx
reset value: Unpredictable
Bits Access Description
31:0 Read/Write CHKPTx:
target address for corresponding entry in trace buffer