Blogs

05-06 summarization

Still researching on how to choose a basic model for classify. Preparing for Statistics exams on 5.8. Received welcome package today,thank you all.

Arindam Banerjee's picture

Finalizing algorithm

Still working on shaping and finalizing the algorithm and getting familiarized with the input file formats. Implementing efficient algorithm is very crucial here considering circuit complexity and program run time. And yes, received GSoC welcome kit yesterday!!

Collecting data and read some documentation of Stanford parser, etc

Continuously collect some weibo posts.

After reading a paper about syntactic sentence structure, trying to find useful tools from Stanford parser.

Doing some CUDA homework.

Arindam Banerjee's picture

Working with file formats

The netlist parser of the PCB autoplacer needs to read the PCB design from some universal file format. Now I am working with different file formats that hold the netlist and design information. The information generally includes PCB layers, components, pins, locations, netlists etc. Also the output file from the autoplacer should contain the “placed” circuit design information that can be fed to other design tools or routers. Some particular input and output file formats need to be finalized which are neat and convenient to extract information.

Read paper

  1. opinion mining and sentiment analysis:www.cs.cornell.edu/home/llee/omsa/omsa.pdf‎

Care package from GSOC

Has anyone else received a package from Google with a GSOC notebook and pen? Alas ... no t-shirt yet...

Arindam Banerjee's picture

PCB Autoplacer - why do we need it?

The main aim of the placement is to make the PCB design compact with less wire length and without affecting its functionalities. A manual placement may result into a substandard design that affects circuit's performance and makes the circuit nonmanufacturable by producing excessive wirelength beyond available resources. Placement is a crucial stage in VLSI design where millions of gates need to be placed before routing. In PCB design, number of components is less but they vary a lot in shape and size. Once I developed a basic analytical, quadratic, global placer for VLSI circuits that was tested with basic MCNC benchmark circuits. But there are considerable amount of differences between the placement in VLSI and PCB. VLSI placement algorithm cannot be directly applied for PCB but need to be modified significantly. Now I am working on finalizing the algorithm that need to be implemented for auto placement in PCB design. In PCB design, pins of a single component cannot be changed but only the component can be rotated. A netlist parser would read the initial design from netlist file. Now using the information available from netlist, the placement should be done optimally.

The first steps

Attended our second IRC session this morning. Getting familiar with the blog provided by Bart. It seems to hang and lose my changes more often than not when I try to save them. Perhaps this is because I am using Internet Explorer ...

Arindam Banerjee's picture

First blog on GSoC 2014

It’s thrilling to be selected as one of the student participants of GSoC 2014 under the mentorship of Portland State University. Today I got my first access to post blogs here on my project, my progress and overall experience of working in GSoC. My topic is “A routability-driven placer for Printed Circuit Board design” where an open source auto placer would be developed to be integrated with existing PCB designer tools. Considering the community bonding period as the first phase of SDLC, right now I am aiming to shape the proposed timeline in more practical and logical way so that defined milestones can be achieved smoothly. Also the requirement analysis and necessary planning part are going on with my mentor Bart Massey. Hope the overall development experience in GSoC period would be enjoyable and memorable.

jfischy's picture

Initial draft of proposal done

Well, today got initial proposal done and sent to Jamey Sharp and Mick Thomure for some feedback. It appears that they have given feedback, but since there is work going on with the PSU system, I can't get access to my mail. I guess I will have to wait until tommorrow to check and work on it some more.

Tonight I took the children to Cinco de Mayo festival by the waterfront. Boy was it crowded but it was fun to be near the rides and see the kids get excited about the some of them. It has become quite spendy but we conserved some by bringing our own drinks and being selective about what we wanted to do. The fireworks were really nice too and gots lots of exercise walking thats for sure!

Syndicate content