Projects

Team Parkas, INRIA, Paris-Rocquencourt

Aimed at hunting concurrency bugs in GCC and CLANG to improve compiler optimizations for multi-processor applications
Added global memory trace to study effect of compiler optimizations on global memory accesses, and replay instrumentation to study the manner in which a load instruction affects subsequent instructions
Added support for control dependency analysis to study effects of conditional statements, and MMX/SSE (128 bit SIMD) instructions
Project Link: (http://raweb.inria.fr/rapportsactivite/RA2013/parkas/uid30.html)

Approximation Algorithms for common subtree & related problems (Dec ‘13 – now)

Investigating approximation algorithms and parameterization techniques for the common sub-tree problem and trying to apply them to specialized classes of graphs and related problems like tree edit distance

Anaphora Without Syntax (Aug ‘13 - now)

Designed a language-independent system for high-school geometry construction problems
Achieved an accuracy of more than 90% for English and Hindi using cross lingual mapping (probabilistically mapping constructs/words in different languages), heuristic based parsing and context based semantic analysis to handle anaphora
The corpus, code and report are available at http://cse.iitk.ac.in/users/pratikkr/btp/index.html

Multi-Lingual word learning for containment situations (Mar ‘13 – Apr ‘13)

Attempted to learn synonymous words in multiple languages for a given context using common ground semantics & label association
Learned design specifications in the given context (peg-in-a-hole) and then tried to learn linguistic semantics for them
The code and report is available at http://home.iitk.ac.in/~pratikkr/se367/project/

Using Progressive Stochastic Search to solve Sudoku CSP (Jan ‘12 – Apr ‘12)

Modelled Sudoku as a constraint satisfaction problem and implemented PSS and iterative PSS to solve a given Sudoku puzzle
Observed that PSS and IPSS are better than other stochastic algorithms like Simulated Annealing and Cultural Genetic Algorithm
The code and report is available at http://home.iitk.ac.in/~pratikkr/cs365/projects/

C++ Compiler Design

Mentor: Prof. Sanjeev Kumar Aggarwal

Implemented a working compiler for a subset of C++ language features using C++, Lex/Yacc for MIPS architecture
Built all compiler modules - lexer, parser (with error checking grammar), symbol table and code generator - from scratch

Extensions to PintOS System

Mentor: Prof. Subhajit Roy

Extended barebones operating system to add features for system calls, inter-process communication using shared memory model, virtual memory with pure demand paging and buffer caching for efficient I/O
Programmed PintOS file system on the lines of ext2 file system to support sub-directories and 3-level indexed inodes