## Paper Writing
## Introduction
## Background
1. Graph Computing (Where paralleism is coming in graph from) CSR potential problem
-- Discuss fundamentals of graph and its application algorithm like page rank, single source shortest path and etc briefly
--Highlight non-amortization of memory access time in graph computation
3. GPU Architecture (Describe stream multiprocessor in detail)
4. GPU Execution Model
5. Microarchitectural Limitation of GPU for Graph Processing Load
--Highlighting memory coalescing, uncoalescing and load imbalance.
--Enlist inefficiencies of vertex-centric graph parallel computing in GPU
--Underutilization of SIMD (Load Imbalance and branch divergence issues)
--Discuss "Scalarized vertex-centric parallel graph computing" method from "Graph waving architecture paper" in detail
7. Enlist few bottlenecks in performance imporvement
-- Amdhal's law (Law of diminishing return): Illustrate it in current context of graph computation on GPU. And explain upper performance improvement limit from proposed method
## Related Work (Literature Survey)
1. Discuss utilization of SIMD unit reported by other authors for Graph computation

## Charaacterization section
## Proposed Architecture
## Table , PTS assembly code,
DID library