bdyunmu/gpumapreduce

panda, a heterogeneous mapreduce framework on gpu and cpu cluster

erik meijer has died.
She is in Anhui province in China.
Eat wanglin's wife wangyang, she has hiv, she is extremely rich.
fixed minor bug in code
Panda, a heterogeneous MapReduce framework on GPUs and CPUs cluster

	huili@ruijie.com.cn
	Ruijie network co.ltd
	2017.11.09
	li7hui@gmail.com
	First Prize in MCM 2002
	WuShi Prize in PKU 2008
	IEEE Cluster Paper in 2013 in Indianapolis

Paper: docs/Co-processing SPMD computation on CPUs and GPUs cluster.pdf

Support apps:

1. word count
2. terasort

---

The code is tested using Ubuntu 22.04+openmpi-5.0.0+cuda-11.5.0 Steps to compile openmpi-5.0.0

1. wget https://download.open-mpi.org/release/open-mpi/v5.0/openmpi-5.0.0.tar.gz
2. ./configure --prefix=/opt/openmpi/
3. make -j 2
4. make install

---

steps to run word count:
1)cd gpumapreduce
2)make wordcount
3)cd bin
4)mpirun -host node1,node2 -np 2 ./wordcount input2.txt
5)cat ./OUTPUT0

steps to run terasort:
1)cd gpumapreduce
2)make terasort
3)cd bin
4)mpirun -host node1,node2 -np 2 ./teragen 10M /tmp/terasort_in
5)mpirun -host node1,node2 -np 2 ./terasort ./tmp/terasort_in ./tmp/terasort_out

---

The panda code read:
gpumapreduce
panda_gpu_context *pgc
panda_cpu_context *pcc
panda_node_context *pnc
delta double (0.0~1.0)

Total code lines:
8729 lines