"Punch It, Chewie": simulated the work of a 56-qubit quantum computer
In October, IBM scientists reported that they were able to model a 56-qubit quantum computer on a classic computer. A few months ago it was thought that it was impossible to step over the barrier into 49 qubits.
The previously unattainable result raises the topic of the imminent onset of quantum supremacy — the moment in which the capabilities of a quantum computer exceed the capabilities of any computer with our familiar architecture.
The article examines how IBM has achieved the "impossible" and what prospects it opens up.
')
/ photo by Yohanes Sanjaya CC
Modeling a quantum computer with a number of qubits larger than 49 was considered impossible on a classic machine due to memory limitations. The previous record - 45 qubits - belongs to researchers from the Swiss Higher Technical School of Zurich. To implement the project, they needed 500 terabytes of RAM. The simulation was performed using a Cori II supercomputer with 9304 nodes (a 68-core Intel Xeon Phi 7250 processor each) and petabyte memory.
The complexity of modeling a quantum computer is that when adding a qubit, the amount of memory expended increases exponentially. In classical calculations, information is encoded by bits that have two states — 1 or 0. In quantum computing, a unit of information takes the values ​​1 or 0, or is in a superposition. Qubits are connected with each other, which gives rise to new states. Thus, fifty qubits can simultaneously represent quadrillion values.
As Edwin Pednault from the IBM team says, the old methods require about one exabyte of RAM to implement a 56-qubit system — a huge amount. However, scientists from IBM had 32 terabytes. The team managed to find the original solution.
The mathematical trick that made the simulation possible came to Edwin Pednol's head while washing dishes. The researcher made an analogy of the device of the bristle brush for cleaning glasses with tensors - multidimensional tables with numbers depending on the selected coordinates.
Writing operations in a tensor form allowed us to reduce the amount of wasted memory and parallelize tasks to be performed on multiple processors simultaneously. The experiment was conducted on a Vulcan supercomputer. It is on the 25th place in the ranking of the most powerful supercomputers (according to data for November 2017).
It was believed that quantum superiority will come when the limit of 49 qubits is reached. However, this did not happen. According to Edwin, the achievement of IBM speaks of "full quantum readiness" - the ability to use quantum hardware, software and engineering tools.
So far, the resulting model is running a billion times slower than a “perfect” 56-qubit quantum computer. But the simulation of quantum computing gives researchers the opportunity to find areas of tasks that they are able to solve more effectively than ordinary computers. And imitation of errors helps to identify the causes of problems of real quantum devices.
/ photo by Richard Jones CC
The IBM team plans to continue working in this direction in order to add a few more qubits to the result already achieved. However, as Edwin Pednol points out, simulations will not replace real quantum computers. Therefore, large IT companies continue the race to create "iron" quantum solutions.
For example, Google planned to present the first 49-qubit chip by the end of 2017. In November of this year, IBM also announced the development of a 50-qubit computer, which should strengthen the company's quantum ecosystem.
The main goal of IBM and other companies experimenting with quanta is to build a quantum computer capable of finding answers to important problems that cannot be effectively solved by classical methods. For example, the tasks of quantum chemistry .
Edwin says that in the future, humanity will begin to fully use the possibilities of quantum computing to solve increasingly complex problems. At this point, modeling will be needed to support research, develop new quantum algorithms, and promote quantum technologies.
PS Here are some more materials from our corporate blog:
The previously unattainable result raises the topic of the imminent onset of quantum supremacy — the moment in which the capabilities of a quantum computer exceed the capabilities of any computer with our familiar architecture.
The article examines how IBM has achieved the "impossible" and what prospects it opens up.
')
/ photo by Yohanes Sanjaya CC
"Brush Method"
Modeling a quantum computer with a number of qubits larger than 49 was considered impossible on a classic machine due to memory limitations. The previous record - 45 qubits - belongs to researchers from the Swiss Higher Technical School of Zurich. To implement the project, they needed 500 terabytes of RAM. The simulation was performed using a Cori II supercomputer with 9304 nodes (a 68-core Intel Xeon Phi 7250 processor each) and petabyte memory.
The complexity of modeling a quantum computer is that when adding a qubit, the amount of memory expended increases exponentially. In classical calculations, information is encoded by bits that have two states — 1 or 0. In quantum computing, a unit of information takes the values ​​1 or 0, or is in a superposition. Qubits are connected with each other, which gives rise to new states. Thus, fifty qubits can simultaneously represent quadrillion values.
As Edwin Pednault from the IBM team says, the old methods require about one exabyte of RAM to implement a 56-qubit system — a huge amount. However, scientists from IBM had 32 terabytes. The team managed to find the original solution.
The mathematical trick that made the simulation possible came to Edwin Pednol's head while washing dishes. The researcher made an analogy of the device of the bristle brush for cleaning glasses with tensors - multidimensional tables with numbers depending on the selected coordinates.
Writing operations in a tensor form allowed us to reduce the amount of wasted memory and parallelize tasks to be performed on multiple processors simultaneously. The experiment was conducted on a Vulcan supercomputer. It is on the 25th place in the ranking of the most powerful supercomputers (according to data for November 2017).
Is quantum superiority already here?
It was believed that quantum superiority will come when the limit of 49 qubits is reached. However, this did not happen. According to Edwin, the achievement of IBM speaks of "full quantum readiness" - the ability to use quantum hardware, software and engineering tools.
So far, the resulting model is running a billion times slower than a “perfect” 56-qubit quantum computer. But the simulation of quantum computing gives researchers the opportunity to find areas of tasks that they are able to solve more effectively than ordinary computers. And imitation of errors helps to identify the causes of problems of real quantum devices.
/ photo by Richard Jones CC
The IBM team plans to continue working in this direction in order to add a few more qubits to the result already achieved. However, as Edwin Pednol points out, simulations will not replace real quantum computers. Therefore, large IT companies continue the race to create "iron" quantum solutions.
For example, Google planned to present the first 49-qubit chip by the end of 2017. In November of this year, IBM also announced the development of a 50-qubit computer, which should strengthen the company's quantum ecosystem.
The main goal of IBM and other companies experimenting with quanta is to build a quantum computer capable of finding answers to important problems that cannot be effectively solved by classical methods. For example, the tasks of quantum chemistry .
Edwin says that in the future, humanity will begin to fully use the possibilities of quantum computing to solve increasingly complex problems. At this point, modeling will be needed to support research, develop new quantum algorithms, and promote quantum technologies.
PS Here are some more materials from our corporate blog:
Source: https://habr.com/ru/post/343204/
All Articles