Section A Flashcards
Define ‘synthesis’
“Generation of a circuit model starting from a less abstract one: a transformation between two model views (e.g. Behavioural –> Structural)”
What are the three Model Abstraction Levels?
- Architectural (performs operations)
- Logic (evaluates logic functions)
- Geometrical (set of geometrical entities)
Define ‘model’
“Representation that shows relevant features without associated details”
What is the ‘RTL’?
“Register Transfer Level” : design level of abstraction below behavioural but above structural, using a subset of the HDL.
Events are defined in terms of clocks.
How many bits are in a single precision floating point number?
32 bits
List the bit numbers and what they relate to in a single precision floating point number:
23 bits: 0 - 22: Mantissa/Fraction
8 bits: 23 - 30: Exponent
1 bit: 31: Sign
What is the need for Wireload Models (WLMs)?
To model and investigate the interconnect delays at higher levels of design.
What is responsible for interconnect delay?
Interconnect capacitance (wiring+routing), and interconnect resistance.
Which factors, in combination, produce the Total Delay for DSM circuits?
- Delay due to wires
2. Delay due to cells
What causes wire delays in DSM circuits?
The Resistance.Capacitance (RC) Coefficient (assume => 8ns)
ANSI/IEEE Floating Point:
- Biased Exponent
- No leading zeros
- Implied most significant 1 in the mantissa
- Binary point of mantissa is immediately after most significant 1
- Stored part of mantissa is only fractional part
What is the motivation for using Floating Point notation?
Allows representation of very small and very large numbers, which is otherwise restricted with fixed point notation (only integers).
How can Floating Point be compared to scientific notation?
e.g. 10011101 = 1.0011101 x 2^7
first 1 not stored as assumed present, mantissa = 0011101, exponent = 111
What is the exponent biases in floating point?
Single: +127 (2^(8-1) - 1)
Double: +1023 (2^(11-1) -1)
How is each decimal digit represented in BCD?
Each digit is represented by a four bit binary code that varies from 0000 - 1001 (0 to 9)
If correction is required in BCD, which number do we add and why?
- Add 0110 (decimal 6)
- This produces a carry to by-pass the invalid codes (i.e. 15 - 9 = 6)
What is the defining feature of Gray Code?
Only one bit changes state when transitioning between adjacent codewords in the sequence
What is the purpose of a parity bit?
Error detection: A parity bit, or check bit is a bit added to the end of a string of binary code that indicates whether the number of bits in the string with the value ‘1’ is even or odd.
N.B. When added to given set of bits:
‘Even Parity Bits’ ensure that total number of 1s in result is even.
‘Odd Parity bits’ ensure that total number of 1s in result is odd.
What is the Datapath?
The Datapath is the core of the processor, where computations are performed.
Consists of interconnection of basic combinational functions (e.g. Arithmetic Operations/Logic operators)
What makes up a Generic Digital Processor?
Datapath, Memory, Control, Input/Output
What is the purpose of Memory in a Generic Digital Processor?
To store results or fetch data
What is the purpose of Control in a Generic Digital Processor?
To determine what will happen next in the cycle
How to convert a positive binary number to a negative 2’s C?
- Invert all bits of positive number
- Add 1
e.g. +5
0101 -> 1010 -> 1011 = (-5)
Why is 2’s C used almost universally?
The same hardware (invert, add) can be used for addition and subtraction
What is an FSM?
Finite State Machine: Any device/computational model that can be in one of a number of finite states at any time.
