Transcript QUANTUM COMPUTING:

QUANTUM COMPUTING:
• Quantum computing is an attempt to unite
Quantum mechanics and information science
together to achieve next generation computation.
• A Quantum computer is a machine that performs
calculations based on the laws of quantum
mechanics which is the behaviour of particles at the
sub-atomic level.
• Quantum computers have simultaneity and
parallelism built inherently.
• Moore’s law states that transistors doubles every 18
months in a microprocessor.
• Transistor size should reduce proportionally.
• CMOS-size-5nm
• In other few years the transistor size reaches subatomic scale i.e in the range of 0.1A
Classical Computers:
• Use bits which contain either zero or a one.
• Operate on these bits using a series of binary logic
gates.
• Components have been decreasing in size.
• Classical designs are reaching the theoretical limit of
miniaturization.(only a few atoms)
• On the atomic scale matter obeys the rules of quantum
not classical physics.
• Quantum technology could not only further reduce the
size of components , but could allow for development
of new algorithms based on quantum concepts.
Qubit(Quantum bit)
• A bit of data is represented by a single atom that is
in one of two states is known as qubit.
• Physical implementation of a qubit uses the two
energy levels of an atom.
• Excited state representing |1> and a ground state
representing |0>.
• Spin up-state represents a 1,spin-down a 0.
• A single bit can be forced into a superposition of the
two states.
So What’s the Point?
• While a single classical bit can store either 0 or 1,a
single qubit can simultaneously store both 0 and 1.
• Two qubits can store four states simultaneously
while two classical bits can store one of four bits.
• In general if L is the number of qubits in a quantum
register, that register can store 2^L different states
simultaneously.
• Classical registers store only one state.
• The speed of classical computers can be improved
by using parallelism.
• In contrasted with quantum systems, parallelism is
exponentially increased with the linear increase in
the size of system.
• Because of its inheritance. Parallelism is inbuilt in
quantum systems.
Quantum error detection:
• . The qubits are highly unstable and they keep their
state which is termed as ‘decoherence’. This
requires constant error correction for building a
fault tolerant system.
• Quantum error correction is very expensive
because arbitrary reliability is achieved by
recursively encoding physical qubits numerous
times and is achieved at the expense of speed.
• It is the most basic operation of a quantum
computer.
Parallelism
• Exploitable parallelism is limited by resource and
application structure.
• Now specialize into memory and computing
blocks.
• Encode them differently.
• High processing speed and slow memory.
• The problem of stalls.
• Now,
Memory hierarchy
• Reliability can be increased by recursive
encoding.
• When level 1 encoding creates N bits,
• Level 2 encoding creates N^2 logical bits.
• Qubits in ion trap quantum processor have
large life times when left idle.
• Volatility increases with interactions.
•
•
•
•
Error correction procedure for every gate.
Processor spends most of its time on ECP.
So design should enable fast error correction.
Increase the number of the cloned/ancillary
qubits.
• For each level of concatenation error
correction time and error increase
exponentially.
• But reliability increases double exponentially 
Revised architecture
• Data locality is a common phenomenon.
• The logical qubit can start at level 2, go to level
1 in peak and return to level 2 when idle.
• Memory at level 2 is for area and reliability.
• It is slower than level 1 structure designed for
gate execution.
• So what do we do for optimisation?
Cache
•
•
•
•
Alleviate the need for constant communication.
Memory at level 2 encoding (slow and reliable).
Cache at level 1 (faster and less reliable).
Compute region is fastest and as reliable as
cache.
• But they differ in speed due to the number of
ancilla bits in compute region.
• (a) Memory is denser. The figure shows 3 data
qubits in the compute block which take the same
area as 8 data qubits in memory.
• (b) Memory is at level 2 encoding, while the
compute and cache are at level 1 encoding.
• The revised architecture consists of memory at level
2, compute regions at level 2 and also a cache and
compute region at level 1.
•
•
•
•
•
Changed the ratio of logical to ancillary bits.
Earlier it was 1:2 entirely. Now 8:1 and 1:2.
Eg: adder
The cache hit rate was around 20%.
Static scheduling is done and the dependency is
calculated.
• The optimised approach fetches hit rate up to 85%
irrespective of the cache and adder size.
• Now, we have seen that balanced design with
architectural techniques shows 13X improvement in
speed and 8X in performance.
THANK YOU