R
Ramakrishna Upadrasta
Researcher at Indian Institute of Technology, Hyderabad
Publications - 27
Citations - 267
Ramakrishna Upadrasta is an academic researcher from Indian Institute of Technology, Hyderabad. The author has contributed to research in topics: Compiler & Computer science. The author has an hindex of 5, co-authored 22 publications receiving 154 citations. Previous affiliations of Ramakrishna Upadrasta include University of Paris-Sud & École Normale Supérieure.
Papers
More filters
GRAPHITE Two Years After First Lessons Learned From Real-World Polyhedral Compilation
Konrad Trifunovic,Albert Cohen,David Edelsohn,Feng Li,Tobias Grosser,Harsha Jagasia,Razya Ladelsky,Sebastian Pop,Jan Sjödin,Ramakrishna Upadrasta +9 more
TL;DR: This paper reports on original questions and innovative solutions that arose during the design and implementation of graphite, pioneered by the graphite branch of the GNU Compiler Collection (GCC).
Journal ArticleDOI
IR2VEC: LLVM IR Based Scalable Program Embeddings
S. VenkataKeerthy,Rohit Aggarwal,Shalini Jain,Maunendra Sankar Desarkar,Ramakrishna Upadrasta,Y. N. Srikant +5 more
TL;DR: IR2VEC as mentioned in this paper is a distributed embedding infrastructure that combines representation learning methods with flow information to capture the syntax as well as the semantics of the input programs and achieves state-of-the-art performance on heterogeneous device mapping and thread coarsening.
Proceedings ArticleDOI
Sub-polyhedral scheduling using (unit-)two-variable-per-inequality polyhedra
TL;DR: This work proposes a sub-polyhedral scheduling technique using (Unit-)Two-Variable-Per-Inequality or (U)TVPI Polyhedra, and proves that code generated by the sub- polyhedral parallelization prototype matches the performance of PLuTo-optimized code when the under-approximation preserves feasibility.
Journal ArticleDOI
IR2Vec: LLVM IR based Scalable Program Embeddings
S. VenkataKeerthy,Rohit Aggarwal,Shalini Jain,Maunendra Sankar Desarkar,Ramakrishna Upadrasta,Y. N. Srikant +5 more
TL;DR: IR2Vec as discussed by the authors is a distributed encoding infrastructure that combines representation learning methods with flow information to capture the syntax as well as the semantics of the input programs and achieves state-of-the-art performance on heterogeneous device mapping and thread coarsening.
Posted Content
PolyDL: Polyhedral Optimizations for Creation of High Performance DL primitives
Sanket Tavarageri,Alexander Heinecke,Sasikanth Avancha,Gagandeep Goyal,Ramakrishna Upadrasta,Bharat Kaul +5 more
TL;DR: Comp compiler algorithms to automatically generate high performance implementations of DL primitives that closely match the performance of hand optimized libraries and a flexible framework where it is possible to plug in library implementations of the same in lieu of a subset of the loops.