Week	What We Plan To Do	What We Actually Did
Apr 11-14	Get familiar with Suffix Array construction algorithms and code	We did this!
Apr 15-20	Go deep into the Skew parallel code in Guy's benchmark	We did this!
Apr 21-26	Implement Skew algorithms in CUDA	We read the "Parallel suffix array and least common prefix for the GPU" paper and discuss the pseudocode and implementing schedule. With pseudocode ready we are implementing the algorithm.
Apr 27-30	Prepare final exam	Done
May 1-3	Finish implementing Skew algorithms in CUDA	Tried but not done
May 4-7	Identify bottlenecks and make optimizations	Finished skew algorithm in CUDA correctly with small input, have bug with large input
May 8-9	Continue optimizations and run benchmarks	Debugging....
May 10-13	Final report and demo	We catched up with the scheduler, now it is working and performance is better than PBBS parallelKS, demo finished. Report done.

• [April 14] Finish reading the Skew paper and understand the serial start code
• [April 20] Study the PBBS parallel Skew code and indentify which parts to parallelize in GPU
• [April 23] Read the "Parallel suffix array and least common prefix for the GPU" paper and discuss the pseudocode and implementing schedule
• [April 26] Checkpoint report
• [May 1] Get PBBS benchmark result of serialSA and parallelSA
• [May 2] Implementing Skew Algorithm in CUDA using radixSort and mergeSort library
• [May 10] Skew Algorithm in CUDA using Thrust library implemented working correctly on small size inputs, but have some memory bugs for large inputs
• [May 11] Bug fixed about in mergesort, now can run with large data input, but very slow, still optimizing.......
• [May 11] Performance improved for large data sets, now performance is better than PBBS serialKS and parallelKS.
• [May 12] Fixed some bugs about out of memory.
• [May 12] merge on CUDA sucessfully implemented, overcome the bottlneck! Final results got and ppt ready! Let's rock!