WG211/M15Grebhahn

Conversation opened. 17 messages. 1 message unread.

Skip to content Using University of Minnesota Mail with screen readers Eric You are invisible. Go visible Gopalan Nadathur Gopalan Nadathur Gopalan Nadathur Marvin Van Wyk Peggy Van Wyk Ted Kaminski Ted Kaminski Tony Sloane Ulrik Pagh Schultz Xinyan Li

More 1 of 42

Invitation to the next WG 2.11 meeting Inbox x WG 2.11/London x Ulrik Pagh Schultz Dear Alexander: It is my pleasure to invite you, on behalf of the members of ... Sep 24 Ulrik Pagh Schultz Dear Bruce: Sep 24 Eric Van Wyk Dear Alexander, We hope that you've had a chance to consider our invitation t... Oct 8 Eric Van Wyk Dear Bruce, Oct 8 Alexander Grebhahn Hallo, sorry for my late reply. It is an honor for me being invited to the me... Oct 12 Bruce Watson Good morning gentlemen, Ulrik+Eric, sorry for my radio silence…wasn’t intende... Oct 15 (13 days ago) Eric Van Wyk Hi Bruce, I'm sorry to hear about your wife's mother's illness and hope she r... Oct 15 (13 days ago) Bruce Watson Good morning gentlemen, Whew on the snub and :-) on some time in the future, ... Oct 16 (12 days ago) Eric Van Wyk Hi Bruce, I will be a Parsing@SLE and will be there for the rest of the week ... Oct 16 (12 days ago) Ulrik Pagh Schultz hi Alexander, It seems that you requested a registration refund, is there a p... Oct 20 (8 days ago) Alexander Grebhahn Hi, sorry, It seams that I accidentally clicked this button. There was no pro... Oct 21 (7 days ago) Ulrik Pagh Schultz Expect a new registration from Alexander :-) /Ulrik From: Alexander Grebhahn ... Oct 21 (7 days ago) Ulrik Pagh Schultz hi Alexander, We'd like to encourage you to give a talk about your current wo... Oct 25 (3 days ago) Bruce Watson Hi Eric and Ulrik, I’d love to be there, but the current situation (and our s... Oct 25 (3 days ago) Alexander Grebhahn <grebhahn@fim.uni-passau.de>

Attachments7:36 AM (3 hours ago)

to Ulrik, me Hi,

sorry for the delay. Thank you for the encouragement. I attached the abstract and the title of my talk is "Performance-Influence Models: Prediction, Optimization, Debugging".

Best regards, Alex Attachments area Preview attachment abstract.txt [Text] Eric Van Wyk Thanks, Alex. Can you also send me an abstract and I will put that on the WG2... 8:52 AM (2 hours ago) Alexander Grebhahn

10:15 AM (46 minutes ago)

to me, Ulrik Sure. Sorry I thought I have attached the abstract.

Here the abstract:

Configuration options allow users to optimize the performance of configurable software systems for a specific use case. However, the flexibility gained from configuration gives rise to new challenges for optimization. For example, finding the performance-optimal configuration is a complex task, because users typically do not know the influence of all individual configuration options on performance and when and how options interact with each other. So, understanding the performance characteristics of a complex configurable system is often infeasible. To address this problem, we use a unique combination of machine-learning and sampling techniques, in particular, forward feature selection and linear regression, to learn models describing the influence of configuration options and their interactions on performance. Based on these models, we can predict performance of all variants of a configurable software system with a high accuracy and improve understanding of the system, for example, by using performance-influence models in interviews to validate the expectations of domain experts with the actual influences of the options. In a series of experiments, we analyzed the prediction accuracy of our approach for a number of subject systems from different domains, including configurable multigrid solvers (e.g., the Dune framework), the x264 video encoder, and the LLVM compiler. We are able to predict performance of all variants with an accuracy of more than 80% after measuring only a small number of all variants. Furthermore, we conducted interviews with domain experts on the usefulness and desired properties of performance-influence models.

Best regards, Alex

Click here to Reply, Reply to all, or Forward Using 5.4 GB Manage Program Policies Powered by Google Last account activity: 49 minutes ago Details

Compose: The GCD program I'd like to prove in Whiley MinimizePop-outClose

Performance-Influence Models: Prediction, Optimization, Debugging

Configuration options allow users to optimize the performance of configurable software systems for a specific use case. However, the flexibility gained from configuration gives rise to new challenges for optimization. For example, finding the performance-optimal configuration is a complex task, because users typically do not know the influence of all individual configuration options on performance and when and how options interact with each other. So, understanding the performance characteristics of a complex configurable system is often infeasible. To address this problem, we use a unique combination of machine-learning and sampling techniques, in particular, forward feature selection and linear regression, to learn models describing the influence of configuration options and their interactions on performance. Based on these models, we can predict performance of all variants of a configurable software system with a high accuracy and improve understanding of the system, for example, by using performance-influence models in interviews to validate the expectations of domain experts with the actual influences of the options. In a series of experiments, we analyzed the prediction accuracy of our approach for a number of subject systems from different domains, including configurable multigrid solvers (e.g., the Dune framework), the x264 video encoder, and the LLVM compiler. We are able to predict performance of all variants with an accuracy of more than 80% after measuring only a small number of all variants. Furthermore, we conducted interviews with domain experts on the usefulness and desired properties of performance-influence models.