4. Pipelining

Rev.1.00 Jan. 10, 2008 Page 67 of 1658

REJ09B0261-0100

I1 I2 I3
(I1) (ID)
ID E1/S1 E2/s2 E3/s3 WB
I3
I3
I3
(I2) (I3)
I1 I2 ID E1/S1 E2/S2 E3/S3 WB
I1 I2 ID E1/S1 E2/S2 E3/S3 WB
I1 I2 ID s1 s2 s3 WB
E2s2
ID E3s3
ID
WB
ID
I1 I2 I3 ID S1 S2 S3 WB
E1s1 E3s3
E2s2
E1s1
E1s1
E1s1
E1s1
E2s2
E2s2
E2s2
E2s2
E3s3
E3s3
E3s3
E3s3
WB
WB
WB
WB
E2s2 E3s3 WB
E2s2E3s3 WB
E1s1
E1s1
I1 I2 I3 ID S1 S2 S3 WB
E1s1
(I1) (ID)(I2) (I3)
(I1) (ID)(I2) (I3)
(I1) (ID)(I2) (I3)
(I1) (ID)(I2) (I3)
E2s2 E3s3 WB
E1s1
ID
ID
ID
ID
ID
ID ID
WB
ID
(1-1) BF, BF/S, BT, BT/S, BRA, BSR:1 issue cycle + 0 to 3 branch cycles
(1-2) JSR, JMP, BRAF, BSRF: 1 issue cycle + 4 branch cycles
(Branch destination instruction)
(1-3) RTS: 1 issue cycle + 0 to 4 branch cycles
(1-4) RTE: 4 issue cycles + 2 branch cycles
It is 15 cycles to the ID stage
in the first instruction of exception handler
(1-5) TRAPA: 8 issue cycles + 5 cycles + 2 branch cycle
It is not constant cycles to
the clock halted period.
(1-6) SLEEP: 2 issue cycles
(Branch destination instruction)
(Branch destination instruction)
(Branch destination instruction)
In branch instructions that are categorized
as (1-1), the number of branch cycles
may be reduced by prefetching.
Note:
Note:
Note:
Note: The number of branch cycles may be
0 by prefetching instruction.
Figure 4.2 Instruction Execution Patterns (1)