Chapter 19 Debug (S12XDBGV2)
MC9S12XDP512 Data Sheet, Rev. 2.17
728
Freescale Semiconductor
19.4.5.3.3
Trace Buffer Reset State
The trace buffer contents are not initialized by a system reset. Thus should a system reset occur, the trace
session information from immediately before the reset occurred can be read out. The DBGCNT bits are
not cleared by a system reset. Thus should a reset occur, the number of valid lines in the trace buffer is
indicated by DBGCNT. The internal pointer to the current trace buffer address is initialized by unlocking
the trace buffer thus points to the oldest valid data even if a reset occurred during the tracing session.
Generally debugging occurrences of system resets is best handled using mid or end-trigger alignment since
the reset may occur before the trace trigger, which in the begin-trigger alignment case means no
information would be stored in the trace buffer.
19.4.6
Tagging
A tag follows program information as it advances through the instruction queue. When a tagged instruction
reaches the head of the queue a tag hit occurs and triggers the state sequencer.
Each comparator control register features a TAG bit, which controls whether the comparator match will
cause a trigger immediately or tag the opcode at the matched address. If a comparator is enabled for tagged
comparisons, the address stored in the comparator match address registers must be an opcode address for
the trigger to occur.
Both CPU and XGATE opcodes can be tagged with the comparator register TAG bits.
Using a begin-aligned trigger together with tagging, if the tagged instruction is about to be executed then
the transition to the next state sequencer state occurs. If the transition is to the nal state, tracing is started.
Only upon completion of the tracing session can a breakpoint be generated. Similarly using a mid-aligned
trigger with tagging, if the tagged instruction is about to be executed then the trace is continued for another
32 lines. Upon tracing completion the breakpoint is generated. Using an end-aligned trigger, when the
tagged instruction is about to be executed and the next transition is to nal state then a breakpoint is
generated immediately, before the tagged instruction is carried out.
R/W monitoring is not useful for tagged operations since the trigger occurs based on the tagged opcode
reaching the execution stage of the instruction queue. Similarly access size (SZ) monitoring and data bus
monitoring is not useful if tagged triggering is selected, since the tag is attached to the opcode at the
matched address and is not dependent on the data bus nor on the size of access. Thus these bits are ignored
if tagged triggering is selected.
When congured for range comparisons and tagging, the ranges are accurate only to word boundaries.
CPU tagging is disabled when the BDM becomes active. Conversely, BDM rmware commands are not
processed while tagging is active. XGATE tagging is possible when the BDM is active.