Freescale Semiconductor, Inc.
four. If there is no time in the head to perform a prefetch due to a previous trailing write,
then additional time to perform the prefetches must be allotted in the middle of the
instruction or after the tail.
8 (2 /1 /0)
TOTAL NUMBER OF CLOCKS
NUMBER OF READ CYCLES
NUMBER OF INSTRUCTION ACCESS CYCLES
NUMBER OF WRITE CYCLES
The total number of clocks for bus activity is as follows:
(2 Reads × 2 Clocks/Read) + (1 Instruction Access × 2 Clocks/Access) +
(0 Writes × 2 Clocks/Write) = 6 Clocks of Bus Activity
The number of internal clocks (not overlapped by bus activity) is as follows:
10 Clocks Total − 6 Clocks Bus Activity = 4 Internal Clocks
Memory read requires two bus cycles at two clocks each. This read time, implied in the tail
figure for the EA, cannot be overlapped with the instruction because the instruction has a
head of zero. An additional two clocks are required for the ADD instruction itself. The total
is 6 + 4 + 2 = 12 clocks. If bus cycles take more time (i.e., the memory is off-chip), add an
appropriate number of clocks to each memory access.
The instruction sequence MOVE.L D0, (A0) followed by LSL.L #7, D2 provides an
example of overlapped execution. The MOVE instruction has a head of zero and a tail of
four because it is a long write. The LSL instruction has a head of four. The trailing write
from the MOVE overlaps the LSL head completely. Thus, the two-instruction sequence
has a head of zero and a tail of zero, and a total execution of 8 rather than 12 clocks.
General observations regarding calculation of execution time are as follows:
• Any time the number of bus cycles is listed as "X", substitute a value of one for byte
and word cycles and a value of two for long cycles. For long bus cycles, usually add a
value of two to the tail.
• The time calculated for an instruction on a three-clock (or longer) bus is usually longer
than the actual execution time. All times shown are for two-clock bus cycles.
• If the previous instruction has a negative tail, then a prefetch for the current
instruction can begin during the execution of that previous instruction.
• Certain instructions requiring an immediate extension word (immediate word EA,
absolute word EA, address register indirect with displacement EA, conditional
branches with word offsets, bit operations, LPSTOP, TBL, MOVEM, MOVEC,
MOVES, MOVEP, MUL.L, DIV.L, CHK2, CMP2, and DBcc) are not permitted to begin
until the extension word has been in the instruction pipeline for at least one cycle.
This does not apply to long offsets or displacements.
5- 98
MC68340 USER’S MANUAL
MOTOROLA
For More Information On This Product,
Go to: www.freescale.com
FREESCALE [ Freescale ]
