Vishal Sahu
Projects
-
Virtual Distributed File System, VMware, Palo Alto, CA
Jun 2016 - Aug 2016
- Optimized in-memory B-tree extent insertion to reduce meta data foot print.
- Implemented read interface for VDFS to connect to remote AWS S3 storage backend using AWS C++ SDKs.
- Implemented proc-like posix interface to list all the backend objects which user can see in organized directory structure.
- Added feature of VSAN volume formatting using user defined policy.
-
Design of 32-bit SPARC processor, CSE502, Computer Architecture, Stony Brook University
Jan 2016 - May 2016
- Designed 32-bit SPARC in-order Scalar processor with 5-Stage pipeline.
- Implemented write-allocate write-back Direct-mapped instruction and data caches.
- Implemented data forwarding from memory and write-back stages to resolve RAW dependencies to improve throughput of pipeline.
-
Anti-malware stackable file system(amfs), CSE506, Operating Systems, Stony Brook University
Sep 2015 - Nov 2015
- Implemented a stackable file system that efficiently quarantines the files containing malware.
- Developed mechanism to update pattern database with minimal re-scanning overhead by implementing update-on-write version keeping mechanism.
- User can define and upload forbidden words database during mount time.
-
Implementation of JOS, OSCAR Lab
Feb 2016-May 2016
- Implemented light configured 64-bit JOS from scratch, writing memory management, scheduler, process management and file system.
-
Asynchronous utility module for Linux, CSE506, Stony Brook University
Oct 2015-Dec 2015
- Developed asynchronous job queuing mechanism based on producer-consumer design paradigm.
- Implemented appropriate locking mechanisms to avoid races and deadlocks.
- Formulated fair scheduling policy to prevent starvation of low priority jobs.
-
Multimedia IP design, Samsung Research Institute, India
Jun 2013-Jul 2015
- Developed Scaler for Pinch-to-Zoom feature capable of scaling images in range of 0.25x - 4x.
- Implemented SPIHT, wavelet coefficients based image compression algorithm.
- Optimized run time of multimedia module by 60% using openGL vectorization on Qualcomm Adreno GPU.
-
Optimization of Speech Recognition System, NIT Trichy
May 2010-Jul 2010
- Problem statement: Processing delay in real time speech recognition system results in bad user experience.
- Optimized the run time of the speech-recognition system implemented in C++ by 20% using linear assembly without altering the detection accuracy.
|