Friday, 25 August 2017

Pipelining Concepts with Overlapped mechanisms

Pipelining – An overlapped parallelism

vTo achieve pipelining one must subdivide the input into a sequence of subtask, each of that executed concurrently with other stages in the pipeline
Principles of Linear Pipelining

qAssembly lines used in automated industries to increase productivity

qAll assembly lines in a pipelined processors must have same processing speed
qElse congestion problem may occur.
qThe precedence relation of a set of subtasks says that one state Tj cannot start until Ti finishes.
qLinear Pipelining process.
Clock Period:
     -time delay
Tk=k+(n-1) denotes the output /result at each clock cycles
Speed Up:
Sk=T1/Tk=n.k/k+(n-1)

Classification of pipelined processors
Handler has proposed the following 6 schemas:
ÒArithmetic pipelining
ÒInstruction pipelining
ÒProcessor pipelining
ÒUnifunction vs multifunction pipelines
ÒStatic vs dynamic pipelines
ÒScalar vs vector pipelines
GENERAL PIPELINES AND RESERVATION TABLES

ÒLinear Pipeline – depends on the output to process the next instruction

ÒPipelines with feedback may have a non linear data flow
ÒGeneral Pipelines with feedback connections and feed forward
ÒNeed to use the reservation table
ÒForward connection –> j>=i+2
Feedforward connection j<=i
Interleaved memory organisation

ÒMemory bandwidth – avg no of words accessed per second

ÒFactors affecting the bandwidth are:
   Processor architecture
   Memory configuration
   Memory module characteristics(Clock cycle, memory org)
Eg: four pipeline vector processors
ÒS- access memory organization
ÒC- access memory organization
ÒC/S access memory organization
S- access memory organization
ÒIt uses simply low order interleaving and applies (n-m) bits of the address to all M=2 memory modules in one access
ÒAll the modules gets accessed simultaneously
ÒHere, high order bits selects modules
ÒWords from modules are latched at the same time
ÒLower order bits selects word from latches
ÒThis is done through MUX with higher speed
C-access memory organisation
ÒIt allows m memory  words to be accessed concurrently
ÒMore than one bank can be accessed so that it increases the bank utilisation and reducing the bank cost.
ÒThis is C-access m/y organisation
C/S access memory organisation

ÒN access busses with m interleaved memory modules

Ò n busess operates in parallel to allow S access
ÒCombination of both S access and C access

Multiprocessing Control and Algorithms

Inter Process Communication A process can be of two type: Independent process. Co-operating process. An independent process is no...