Instruction issue degree, Computer Engineering

Instruction Issue degree: The major concept in superscalar processing is how many instructions we can issue per cycle. If we issue k number of instructions per cycle in a superscalar processor, in that case the processor is called a k-degree superscalar processor. If we want to exploit the full parallelism from a superscalar processor then k instructions must be implement in parallel.

For example, we consider a 2-degree superscalar processor with 4 pipeline phases for instruction cycle, i.e. decode instruction (DI), instruction fetch (IF), fetch the operands (FO), execute the instruction (EI) as shown below in Figure. In this superscalar processor, two instructions are issued per cycle as shown below in Figure. Here, 6 instructions in 4 stage pipelines have been implemented in 6 clock cycles. Under ideal conditions, after steady state, two instructions are being implemented per cycle.

                                                 20_Instruction issue degree.png

                     Superscalar Processing of instruction cycle in 4-stage instruction pipeline

For executing superscalar processing, some important hardware must be provided which is discussed below:

  • The necessity of data path is increased with the degree of superscalar processing. Assume, one instruction size is 32 bit and we are using 2-degree superscalar processor, then 64 data path by the instruction memory is required and 2 instruction registers are also required.
  • Multiple implementation units are also required for implementing multiple instructions and to avoid resource conflicts.

Data dependency will be increased in superscalar processing if sufficient hardware is not given. The added hardware provided is known as hardware machine parallelism. Hardware parallelism makes sure that resource is available in hardware to exploit parallelism. Another alternative is to exploit the instruction level parallelism inherent in the code. This is achieved by transforming the source code by an optimizing compiler like that it reduces the dependency and resource conflicts in the resulting code.

Many popular commercial processors have been executed with superscalar architecture like IBM RS/6000, MIPS R4000, DEC 21064, Power PC, Pentium, etc.

Posted Date: 3/4/2013 5:32:31 AM | Location : United States







Related Discussions:- Instruction issue degree, Assignment Help, Ask Question on Instruction issue degree, Get Answer, Expert's Help, Instruction issue degree Discussions

Write discussion on Instruction issue degree
Your posts are moderated
Related Questions
DUPLICATING PROCESSES : As we mentioned earlier duplicating is a process whereby a master copy is prepared from which a large number of other copies are obtained with the help of

Make a page translation table the meets the requirements of the virtual memory system given below.  Suppose page (and frame) sizes of 20 with pages 0 by 3 in logical memory and fra

Q. Show the conflict in register? All micro-operations written on a line are to be executed at same time provided the statements or a group of statements to be implemented toge

What are the different storage classes in C?  Storage classes in C  There are four storage classes in C:  a. Automatic storage class:  b. Register storage class:  c

What is an on"*-input filed" statement? ON *-INPUT The ABAP/4 module is known as if the user has entered the "*" in the first character of the field, and the field has th

We might model such a scenario using three types of object: one for Customers, one for BankAccounts and another for Transactions. In terms of data required, for Customers assume we

If for a short period of time circuits goes to some dissimilar logic level then it is assumed to have then it is known as static hazard e.g. If the final logic value of output of a

Q. What is Internet Service Providers? Nobody truly owns the Internet however it is maintained by a group of volunteers interested in supporting this mode of information interc

Level of a node The root is at level zero and the level of the node is 1 more than the level of its parent

The output of a JK flipflop with asynchronous preset and clear inputs is '1'. The output can be changed to '0' with which conditions ? Ans. Through applying J = 1, K = 1 and u