Intel NETWORK PROCESSOR IXP2800 - Manual (Page 109)

Hardware Reference Manual 109
Intel
®
IXP2800 Network Processor
Intel XScale ® Core
3.8.1.2 Data Cache Efficiency Mode
PMN0 totals the number of data cache accesses, which includes cacheable and non-cacheable
accesses, mini-data cache access and accesses made to locations configured as data RAM.
Note that STM and LDM will each count as several accesses to the data cache depending on the
number of registers specified in the register list. LDRD will register two accesses.
PMN1 counts the number of data cache and mini-data cache misses. Cache operations do not
contribute to this count.
The statistic derived from these two events is:
• Data cache miss-rate. This is derived by dividing PMN1 by PMN0.
3.8.1.3 Instruction Fetch Latency Mode
PMN0 accumulates the number of cycles when the instruction-cache is not able to deliver an
instruction to the Intel XScale
®
core due to an instruction-cache miss or instruction-TLB miss.
This event means that the processor core is stalled.
PMN1 counts the number of instruction fetch requests to external memory. Each of these requests
loads 32 bytes at a time. This is the same event as measured in instruction cache efficiency mode
and is included in this mode for convenience so that only one performance monitoring run is need.
Statistics derived from these two events:
• The average number of cycles the processor stalled waiting for an instruction fetch from
external memory to return. This is calculated by dividing PMN0 by PMN1. If the average is
high then the Intel XScale
®
core may be starved of the bus external to the Intel XScale
®
core.
• The percentage of total execution cycles the processor stalled waiting on an instruction fetch
from external memory to return. This is calculated by dividing PMN0 by CCNT, which was
used to measure total execution time.
3.8.1.4 Data/Bus Request Buffer Full Mode
The Data Cache has buffers available to service cache misses or uncacheable accesses. For every
memory request that the Data Cache receives from the processor core, a buffer is speculatively
allocated in case an external memory request is required or temporary storage is needed for an
unaligned access. If no buffers are available, the Data Cache will stall the processor core.
The frequency of Data Cache stalls depends on the performance of the bus external to the Intel
XScale ® core and what the memory access latency is for Data Cache miss requests to external
memory. If the Intel XScale
®
core memory access latency is high (possibly due to starvation) these
Data Cache buffers will become full. This performance monitoring mode is provided to determine
whether the Intel XScale ® core is being starved of the bus external to the Intel XScale ® core —
which affects the performance of the application running on the Intel XScale
®
core.
PMN0 accumulates the number of clock cycles by which the processor is stalled due to this
condition and PMN1 monitors the number of times this condition occurs.