TC1796
System Units (Vol. 1 of 2)
CPU Subsystem
User’s Manual
2-49
V2.0, 2007-07
CPU, V2.0
2.7.2.2
Control Flow Instruction Timing
•
Each instruction is single issued.
•
All targets yield a full instruction in one access (not 16-bits of a 32-bit instruction).
•
All code fetches take a single cycle.Timing is best case; no cache misses for context
operations, no pending stores.
Table 2-15
Load/Store Control Flow Instruction Timing
Instruction
Flow
Latency
Repeat
Rate
Instruction
Flow
Latency
Repeat
Rate
Branch Instructions
J
2
2
JLI
2
2
JA
2
2
JEQ.A
1/2/3
1/2/3
JI
2
2
JNE.A
1/2/3
1/2/3
JL
2
2
JNZ.A
1/2/3
1/2/3
JLA
2
2
JZ.A
1/2/3
1/2/3
CSA Instructions
CALL
1)
1) Latency of CSA related instructions varies according to preceding instruction and status of the shadow register
file. Average latency is ~2.7 cycles
2-5
2-5
SYSCALL
–
2-5
CALLA
2-5
2-5
SVLCX
–
4-9
CALLI
2-5
2-5
RSLCX
–
4
RET
–
2-5
RFE
–
2-5
BISR
–
4-9
RFM
2)
2) Not strictly a CSA operation, but retrieves from memory a subset of context information and changes control
flow in a similar manner.
–
–
Loop Instructions
LOOP
3)
3) First time encountered executed in LS pipeline: Flow latency = 2, Repeat rate = 2
Successive time executed in Loop Pipeline: Flow latency = 1: Repeat rate = 1 (nested up to 2 deep)
Last time encountered: Flow latency = 3: Repeat rate = 3
2/1/3
2/1/3
2/1/3
2/1/3