ISCA-2002 Paper Submission Site
	See Outcome, Reviews and Comments For Paper #119
Wednesday 30th of January 2002 11:47:52 AM

This paper was not selected for the conference

EDUCATING REVIEWERS: Reviewing papers is a difficult process, and it is sometimes difficult for reviewers to determine if their review was appropriate or to understand why a paper was rejected or accepted. If you would like to allow the reviewers for your specific paper to see either the author response or the other reviews, select the appropriate options below. This information is only available to reviewers of your paper, and unless you or reviewers have revealed any information, the information is anonymous.

Allow reviewers to see all reviews

Paper #119	(Download paper of type application/pdf )
Title:	Architectural Implications of Quantum Data Transport in Silicon
Abstract:	Computing with quantum states has become an increasingly intriguing reality. Prototype quantum computers of 5 to 7 bits have begun to appear using molecules in solution and photon traps \cite{Vandersypen00b,Laflamme99a}. For true scalability and to exploit our tremendous historical investment in silicon, however, solid-state silicon quantum implementations are needed. We focus on a fundamental component of any quantum architecture on silicon: quantum wires to transport quantum data. We propose a novel approach to low-latency, reliable communication through teleportation of error-coded quantum bits. Our study then examines three technologies to implement these quantum wires. We discover three fundamental constraints common to all our technologies that will shape the design of future architectures. First, quantum wires are very wide, making it impossible to build a fine grid of wires for the "sea-of-gates" implementations implied by most quantum studies. Larger computational components, perhaps von Neumann architectures, are inevitable. Second, the decoherence of quantum data over time implies several basic circular structures. Third, the bandwidth and reliability of quantum wires will constrain wire length, and consequently, the size of computational components.
Authors Response:	The pitch-matching effect described in this paper is fundamental to the interface of quantum and classical physics, regardless of technology. Fabrication research is underway, but architectural study now can identify key challenges for both fabrication and device development. To wait five years without looking forward would be to miss a valuable opportunity to guide current research towards scalable systems. The primary contributions of the paper are teleportation as a data transport primitive and identification of pitch-matching and bandwidth constraints. As to the discussion of pre-existing quantum techniques, the contribution here is the identification of the right building blocks for a scalable system. We apologize for the lack of clarity in controlled quantum operations and other minor errors. The correction of these errors will be straightforward in the final version of the paper. --- Note, at this time (1/15/2001, 4:30pm PST) only two reviews (114, 174) are available. This response is based upon those only.

Review #114 For Paper #119

Attribute	Value
Provide a short summary of the paper	The paper describes a method for producing "wires" for interconnection in a quantum processor. This will be a required fundamental technology for quantum processors to be practical.
Your qualifications to review this paper	I know nothing about this area
Overall paper merit	I can't make up my mind
Novelty of paper	This is very novel
Was the paper technically sound?	Average
Do you think the paper will be of interest to the ISCA community?	I do not think people would read the paper nor attend the session
Will this paper be important over time?	Average
Spelling and Grammar?	Poor
Could this be a short paper?	No, this can not be a short paper
Provide detailed comments to the author	I am not very familiar with this area, but I suspect that will be true of most readers. There were several confusing issues in the basic descriptions. 1) In Figure 2, there are C parameters in the equations. The text mentions probabilities as c (lower case) but nothing about C (upper case). Are these the same? 2) Figure 2 describes the H, X and Z operators as unary. However, Figure 3 shows X and Z operators with a control signal. This is inconsistent and confusing. 3) There is a detailed description of Figure 10, but the Figure itself has little information. It would really help if this figure contained the actual circuit which is described in the text. 4) Spell check should be run. There are several run together word pairs, and "attiqeuetly" (adequately?) is in section 6.1. Teleportation is misspelled several times.

Review #174 For Paper #119

Attribute	Value
Provide a short summary of the paper	This paper attempts to provide the reader with the background neccesary to understand issues involved in building quantum computers. It discusses decoherence and what can be done to overcome decoherence (error correction). It introduces some basic quantum building blocks and the design constraints introduced by the building blocks, the manufacturing techniques and the need to overcome errors.
Your qualifications to review this paper	I know the material, but am not an expert
Overall paper merit	Reject, but there is some value
Novelty of paper	Incremental improvement
Was the paper technically sound?	Good
Do you think the paper will be of interest to the ISCA community?	People would read it before confernece and attend session
Will this paper be important over time?	Average
Spelling and Grammar?	Poor
Could this be a short paper?	No, this can not be a short paper
Provide detailed comments to the author	The paper is an excellent attempt at describing the design contraints for a class of quantum computers given that the audience is unfamiliar with quantum computing. However, the background information falls just short of making me understand what I think I need to understand to follow your arguments. It also fails to differentiate between what you are introducing and what other people have already introduced. In addition, I was never convinced that the underlying manufacturing technology described is either plausible or general enough to draw conclusions about a target quantum computing architecture. In the introduction there are 3 contributions listed: end-to-end study: How precise do the phosphorus atoms need to be placed? How precise do the control lines need to be placed? What if there are errors in manufacturing? Is all of the precision implied required? In the end, I did not believe the that sections 5.1, 5.2, 5.3 described a feasible or even reasonable underlying mass-production technology. (BTW: SETs still have not shown much gain and are unlikely to do so. Is there any other way to replace the transisotr?) Since I didn't believe the basics I couldn't really buy the rest of the quantum wire architecture. Identification of key components: This appeared to be a summary of other peoples work. Design constraints: section 6 was the most interesting section of the paper. I would like to have been given the information to understand why p ~ 10^{-8} is reasonable. Why is D in qubits? are wires all pipelines D deep? Is this just a process independent measure (i.e., like lambda for VLSI)? If so, it seems to me that 1000 qubits is 20microns, not 500nm. Isn't it 20nm per qubit? ----- choose a common scheme for your figures. (Figs 1,2, and 3 use a different key) in 5.4.2 why do you choose to make "measure" a 2 input gate. Why not make the 0-qubits internal? in 5.4.3 and other places you say "quantum bit" why not "qubit" What is a CAT? 6.1: Do we need to know what an A or J gate is? 6.1: I don't understand the 3rd paragraph and it is crucial. 2N conrol points (i.e., A and J control, right?) for N qubits, implies that each qubit requires 200nm sq [which is a loose lower bound, but I can deal with that]. Somehow that translates into 10qubits. I didn't follow that, unless this means that 10 qubits are all controlled identically?

Review #854 For Paper #119

Attribute	Value
Provide a short summary of the paper	This paper begins by giving an overview to quantum computing, some issues that have arisen and some issues that have already been solved. Based on best-known solutions, the second half of the paper works to describe likely computer architecture modules for building QC compute devices and likely wiring issues at the classical/quantum boundary.
Your qualifications to review this paper	I have passing familiarity
Overall paper merit	I can't make up my mind
Novelty of paper	Incremental improvement
Was the paper technically sound?	Average
Do you think the paper will be of interest to the ISCA community?	People would attend session, but not read it beforehand
Will this paper be important over time?	Average
Spelling and Grammar?	Average
Could this be a short paper?	No, this can not be a short paper
Provide detailed comments to the author	This paper is hard to review using the boxed ratings above. On the one hand, I read it eagerly because it is forward-looking and because the area clearly will become more important over time. On the other hand, I am unsure as to whether this paper, as written, itself makes a contribution that moves forward either the general field of computer architecture or the more specific area of quantum architecture. It's a good paper tackling a tough problem and I wanted to like it. I didn't feel like it quite made a publishable contribution as it currently stands, but could easily see that some significant but not too onerous revisions would turn this paper into a great one. The first half of the paper reads a bit like a tutorial on the state of the art in quantum computing. You were probably correct in assuming that the average reviewer/reader would not have complete QC background and would need some tutorial. Unfortunately, I felt that the style of this first half left the reader in an awkward position. You name a problem and then cite someone who's fixed it. Then name another issue and then cite the solution for that one. (Eg Steane codes, Kane's device technology...) Don't get me wrong: the citing is good and I understand teh interdisciplinary nature of a paper that has to explain everything from device technology to info theory. The problem is that it was hard to discern how much research went into selecting these solutions... Why a 7-bit Steane code? Why Kane's device technology? How much would your issues/answers change if you used a different technology starting point? The second half of the paper is the part that is (1) hardest to write, (2) most important, and (3) hardest for a reviewer to evaluate without a crystal ball! You are working to explore what will be the building blocks of QC for architects to deal with and you are trying to express design constraints for them. Whether or not you predict correctly, the interesting stickiness is how to justify these choices. Could it be that by the time we are able to build large quantities of qubits, a new style of EPR generator or purification unit will change the technical/fabrication issues of these sections? I was left wondering if it is too soon to try to settle on a fixed bag of tricks. You might have changed my mind here (ie "it's not too soon!") by saying what sort of timescale you contemplated for when you or others would be working with fab people to build this. Nonetheless, Sections 5 and 6 left me with the feeling that you were trying to make everything more cut-and-dry, more done that it really is. For me, a preferred paper approach would have been to talk about a couple different possible starting technologies, a couple different options of building blocks, and then to show either that very similar or very different design constraints (Section 6) would arrive depending on how you set things up. How to conclude?.... There are interesting ideas in this paper and I really wanted to like it. I didn't hope for or expect you to graph sim-quantum results for SPEC2000 running on a quantum computer. But I did hope to get more of a feel for how stable your conclusions are or whether different technology trends at the quantum fab level will change the whole path of the QC field. If edited to be a bit more verbose (gasp!) and to let us readers in on a few more of the tradeoffs, this could be one of the seminar papers in the field.

Review #817 For Paper #119

Attribute	Value
Provide a short summary of the paper	This paper analyzes a Si-based quantum computing proposal from an architecture perspective, looking in particular at using EPR-pairs to transport qubits, and at quantum effects in the control lines.
Your qualifications to review this paper	I know the material, but am not an expert
Overall paper merit	Reject, but there is some value
Novelty of paper	This is very novel
Was the paper technically sound?	Poor
Do you think the paper will be of interest to the ISCA community?	People would attend session
Will this paper be important over time?	Average
Spelling and Grammar?	Good
Could this be a short paper?	Hard to say
Provide detailed comments to the author	I find the analysis of using EPR-pairs in quantum "wires" to be quite interesting. However, the discussion of the constraints required to keep the control gates "classical" appears to be somewhat confused. In particular, I do not see that using 100nm metallic wires will avoid quantum effects that would appear in 5nm wires. What quantum effects? The coherence time of an electron in ultra-clean Au at very low temperature (mK, or lower) is a few ns. The gates in Bruce Kane's proposal would switch at the rate of the quantum operations, or a couple orders of magnitude slower than this time. The one problem I might forsee is one of localization and relatively high resistance in the wires, but that becomes an RC problem, and again the times are relatively long. The applied magnetic fields will need to be considered here, as well, since they affect the localization. Also, directly transporting spin information with electrons is probably not as difficult as one might think - ballistic transport is not necessary. In Si, especially, phonon scattering does not decohere the electron spins too rapidly. However, values for coherence times in this case (free electrons) are not yet well established.

Paper #119

Conference Review Package -- Copyright © 2001 Univ. of Colorado. All rights reserved.

Conference Review Package -- Copyright © 2001 Univ. of Colorado.
All rights reserved.