Pmc2 Selectable Events - IBM PowerPC 604 User Manual

Risc
Table of Contents

Advertisement

9.1.1.1.2 PMC2 Selectable Events
The events counted by PMC2 follow the same groupings explained in the previous section.
The differences between PMCl and PMC2 event selection are as follows:
• Different bits of the MMCRO register are decoded in order to select events.
• PMC2 has fewer events.
Events selectable for counting by
PMC2
are listed. along with their
MMCR0[26-31]
encodings,
are listed in Table 9-3.
Table 9-3. PMC2 Events-MMCRO [26-31) Select Encoding
Encoding
Description
000000
Nothing. Register counter holds current value
000001
Processor cycles
00 0010
Numberof instructions completed. Legal values are 000, 001, 010, 011, and 100.
00 0011
RTCSELECT
bit
transition. O
=
47, 1
=
51, 2
=
55, 3
=
63 bits from the time base lower register.
000100
Number of instructions dispatched
00 0101
Number of cycles a load miss takes
00 0110
Data cache misses (data cache line fill)
00 0111
Number of lttb misses
001000
Number of branches completed. Indicates the number ol branch instructions completed every cycle.
00
None
01
Illegal value
10
One
11
Two
001001
Number of reservations successfully obtained
001010
Number of mfspr instructions dispatched (speculative)
00 1011
Number of lcbl instructions. The lcbl instruction may not hit in the cache.
00 1100
Number of pipeline-flushing operations (sc, lsync, mtspr[xer], floating-point operations with divide
by O or invalid operand when the 604 is in precise mode, branch when MSR[BE] is set, lswx with
XER
=
0 and SO set).
001101
Branch unit produced result (branch or CR-logical instruction finished)
00 1110
SCIUO unit produced result (add, subtract, compare, rotate, shift, or logical Instruction)
00 1111
MCIU unit produced result (multiply/divide or SPR instruction)
010000
Number of instructions dispatched to the branc;h unit
01 0001
Number of instructions dispatched to the SCIUO unit
01 0010
Number of loads completed. From Oto 4 instructions per cycle. Indicates the number of load
instructions being completed every cycle. These include all cache operations, tlble, tlbsync, sync,
elelo, and lcbl.
01 0011
Number of instructions dispatched to the MCIU
01 0100
Number of snoop hits occurred
Chapter 9. Performance Monitor
9-5

Advertisement

Table of Contents
loading

Table of Contents