FEATURES OF THE ELLIOTT 502 COMPUTER

INSTRUCTION TIMING

Single Order

The existence of two stores with different timing characteristics allows some simple optimisation to be carried out when evaluating coding methods. The essential data are:
Access Cycle
Fast Store 20 bits .625 microseconds 1.25 microseconds  
Main Store 40 bits 1.5 microseconds 3 microseconds

The time to obey a particular order may be assessed by considering the sequence of store cycles required.

Example 1

Fast Store Order 54 : 16 Add 1 to main store location 16

Starting at time zero, the order itself is accessed from the fast store. Since only the 'real' half of the cycle is relevant, the Central Processor can start the main store cycle immediately. Since only the 'write' half of the main store cycle is relevant, the operation is concluded after 3.[265]625 microseconds. A new fast store access then begins.

Example 2

Fast Store order 33 : 700 Put the contents of the main store, specified by the contents of fast store location 700, into the accumulator.

The order is accessed with the first fast store cycle. Note that since the second fast store access is required immediately, the 'write' half of the cycle is wasted. The indirect address is accessed with the 'read' half of the next cycle, and the main store cycle, commences at the end of it.

Sequences of Orders

The main store has a 40-bit structure, and an access (orders) results in an order-pair being extracted to the order registers. Thus, two can be had for the time-price of one, provided that the first order does not itself refer to the main store.

Example 3

07 : 819 /2 Write b-line to main store location 819 /2
Main Store 54 17 Add one to content of fast store location 17
fig3

Example 4

Main Store 07 : 819 /2 Write b-line to main store location 819 /2
  54 17 Add one to content of fast store location 17
fig4

NOTE: The two main store cycles of the first order add 3 microseconds to the total time.

Example 5

Main Store 33 3 Accumulator = content of fast store location 3
  55 18 Subtract one from fast store location 18
fig5

NOTE: The corresponding time for fast store operation of the above orders is 5 microseconds. Thus it is advantageous to place programs with few or no "jumps" in the main store, with their data in the fast store.

The above examples show how overlapping store cycles can save time; they also indicate that the main store program and data combination is time-consuming, but that main store program/fast store data can be the fastest.

Back - Contents - Next
Page created by Bill Purvis, last update 20th December 2003
Contact me at: mailto:bill 'at' beeb.net