Added seq_time to analysis.csv. Joined the big analysis data csv's. C…

…reated some R scripts to generate graphs. Got details from nsight.

big_data_analsis.R

@@ -0,0 +1,26 @@
+# Plot the data
+
+# Imports
+library(reshape2)
+library(ggplot2)
+
+# Load the data
+# Change path to file
+# data_path <- '../parallel/analysis.csv'  # You can use this instead
+data_path <- '/home/reynaldo/Documents/School/Spring2017/HPC/project/parallel_mcmc/parallel/complete_big_analysis.csv'
+perf_data <- read.csv(file=data_path, header=TRUE, sep=",")
+melted_data <- melt(perf_data, id=c("trial", "block", "thread"))
+five_hundred_trials <- melted_data[which(melted_data$trial == 500), ]
+five_hundred_trials$thread = as.factor(five_hundred_trials$thread)
+
+
+# Find which combination of blocks and threads works best
+ours_five_hundred_trials <- five_hundred_trials[which(five_hundred_trials$variable == "our_time"), ]
+ours_five_hundred_trials <- ours_five_hundred_trials[which(!is.na(ours_five_hundred_trials$thread)), ]
+
+line_base <- ggplot(ours_five_hundred_trials, aes(x=block, y=value, group=thread, shape=thread, color=thread)) +
+  geom_line() +
+  geom_point()
+line_base <- line_base +  xlab("Blocks") + ylab("Time (seconds)") + labs(colour = "Num Threads", shape="Num Threads")
+line_base <- line_base + ggtitle("Blocks vs Time vs Threads") + theme(plot.title = element_text(hjust = 0.5))
+line_base

data_analysis.R

@@ -0,0 +1,43 @@
+# Plot the data
+
+# Imports
+library(reshape2)
+library(ggplot2)
+
+# Load the data
+# Change path to file
+# data_path <- '../parallel/analysis.csv'  # You can use this instead
+data_path <- '/home/reynaldo/Documents/School/Spring2017/HPC/project/parallel_mcmc/parallel/analysis.csv'
+perf_data <- read.csv(file=data_path, header=TRUE, sep=",")
+melted_data <- melt(perf_data, id=c("trial", "block", "thread"))
+five_hundred_trials <- melted_data[which(melted_data$trial == 500), ]
+
+# Generate Plots
+
+# Create Scatter plot (Blocks vs Threads)
+base_plot <- ggplot(five_hundred_trials, aes(x=block, y=time, color=variable))
+base_plot <- base_plot  +  geom_point(aes(x=block, y=thread, size=value, color=variable)) 
+base_plot <- base_plot + scale_size(range = c(0, 10))
+base_plot <- base_plot + xlab("Threads") + ylab("Time (seconds)")
+base_plot <- base_plot + ggtitle("Blocks vs Time") + theme(plot.title = element_text(hjust = 0.5))
+base_plot
+
+# Find which combination of blocks and threads works best
+five_hundred_trials$thread = as.factor(five_hundred_trials$thread)
+ours_five_hundred_trials <- five_hundred_trials[which(five_hundred_trials$variable == "our_time"), ]
+ours_five_hundred_trials <- ours_five_hundred_trials[which(!is.na(ours_five_hundred_trials$thread)), ]
+
+line_base <- ggplot(ours_five_hundred_trials, aes(x=block, y=value, group=thread, shape=thread, color=thread)) +
+  geom_line() +
+  geom_point()
+line_base <- line_base +  xlab("Blocks") + ylab("Time (seconds)") + labs(colour = "Num Threads", shape="Num Threads")
+line_base <- line_base + ggtitle("Blocks vs Time vs Threads") + theme(plot.title = element_text(hjust = 0.5))
+line_base
+
+# Line Plot
+line_base <- ggplot(five_hundred_trials, aes(x=block, y=value, group=variable, shape=variable, color=variable)) +
+  geom_line() +
+  geom_point()
+line_base <- line_base +  xlab("Blocks") + ylab("Time (seconds)")
+line_base <- line_base + ggtitle("Blocks vs Time") + theme(plot.title = element_text(hjust = 0.5))
+line_base

parallel/analysis.csv

@@ -1,4 +1,4 @@
-trial,block,thread,our_time,their_time
+trial,block,thread,seq_time,our_time,their_time
 500,1,1,1.32,11.04,5.19
 500,1,2,1.25,7.83,5.12
 500,1,4,1.23,6.29,5.26

parallel/complete_big_analysis.csv

@@ -0,0 +1,22 @@
+trial,block,thread,seq_time,our_time,their_time
+500,1,1,5519.37,3321.13,18.31
+500,1,2,5519.37,3379.05,18.31
+500,1,4,5519.37,3533.75,18.31
+500,1,8,5519.37,3803.11,18.31
+500,1,16,5519.37,4390.35,18.31
+500,1,32,5519.37,5553.23,18.31
+500,1,64,5519.37,16.49,18.31
+500,1,128,5519.37,16.45,18.31
+500,1,256,5519.37,16.52,18.31
+500,1,512,5519.37,16.45,18.31
+500,1,1024,5519.37,16.34,18.31
+500,2,64,5519.37,16.40,18.42
+500,2,128,5519.37,16.47,18.42
+500,2,256,5519.37,16.40,18.42
+500,2,512,5519.37,17.03,18.42
+500,2,1024,5519.37,16.63,18.42
+1000,2,64,20407.01,16.68,18.31
+1000,2,128,20407.01,16.35,18.31
+1000,2,256,20407.01,16.49,18.31
+1000,2,512,20407.01,16.42,18.31
+1000,2,1024,20407.01,16.39,18.31

parallel/nsight_summary_details.csv

@@ -0,0 +1,7 @@
+Name,Start Time(ns),Duration(ns),GridX,GridY,GridZ,BlockX,BlockY,BlockZ,Registers/Thread,Static Shared Memory,Dynamic Shared Memory,Size(bytes),Throughput(bytes/sec),Achieved Occupancy
+"Memcpy HtoD [sync]",222838415,3008,,,,,,,,,,12220,4062500000,
+"Memcpy HtoD [sync]",222854191,2912,,,,,,,,,,12220,4196428571,
+"setup_kernel(curandStateXORWOW*, unsigned int, int)",224295348,3893552,1,1,1,64,1,1,33,0,0,,,0.027
+"seqMetroProcess(int, int, int*, float*, curandStateXORWOW*, float*, float*, float, float, float*, float*, int)",228192356,5093374624,1,1,1,8,1,1,43,0,0,,,0.016
+"sampleTdot(int, int*, curandStateXORWOW*)",5321570148,6976,1,1,1,8,1,1,13,0,0,,,0.016
+"mergePosterior(int, float*, float*, int*, curandStateXORWOW*, int, int, float*, float*)",5321583364,818372125,1,1,1,1,1,1,40,0,0,,,0.016

parallel/nsight_summary_details_theirs.csv

@@ -0,0 +1,5 @@
+Name,Invocations,Avg. Duration(ns),Registers/Thread,Static Shared Memory,Avg. Dynamic Shared Memory
+"sample_theta(curandStateXORWOW*, float*, float*, int*, float*, float, float, int)",500,73326,28,0,0
+"void thrust::system::cuda::detail::bulk_::detail::launch_by_value<unsigned int=128, thrust::system::cuda::detail::bulk_::detail::cuda_task<thrust::system::cuda::detail::bulk_::parallel_group<thrust::system::cuda::detail::bulk_::concurrent_group<thrust::system::cuda::detail::bulk_::agent<unsigned long=7>, unsigned long=128>, unsigned long=0>, thrust::system::cuda::detail::bulk_::detail::closure<thrust::system::cuda::detail::reduce_detail::reduce_partitions, thrust::tuple<thrust::system::cuda::detail::bulk_::detail::cursor<unsigned int=1>, thrust::detail::normal_iterator<thrust::pointer<float, thrust::system::cuda::detail::tag, thrust::use_default, thrust::use_default>>, thrust::detail::normal_iterator<thrust::pointer<float, thrust::system::cuda::detail::tag, thrust::use_default, thrust::use_default>>, thrust::detail::normal_iterator<thrust::pointer<float, thrust::sy",1000,548364,58,32,6112
+"void thrust::system::cuda::detail::bulk_::detail::launch_by_value<unsigned int=128, thrust::system::cuda::detail::bulk_::detail::cuda_task<thrust::system::cuda::detail::bulk_::parallel_group<thrust::system::cuda::detail::bulk_::concurrent_group<thrust::system::cuda::detail::bulk_::agent<unsigned long=7>, unsigned long=128>, unsigned long=0>, thrust::system::cuda::detail::bulk_::detail::closure<thrust::system::cuda::detail::reduce_detail::reduce_partitions, thrust::tuple<thrust::system::cuda::detail::bulk_::detail::cursor<unsigned int=1>, thrust::device_ptr<float>, thrust::system::cuda::detail::aligned_decomposition<long>, thrust::detail::normal_iterator<thrust::pointer<float, thrust::system::cuda::detail::tag, thrust::use_default, thrust::use_default>>, float, thrust::plus<float>, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type>>>>(unsigned long=7)",1000,713547,74,32,8160
+"setup_kernel(curandStateXORWOW*, unsigned int, int)",1,7552158,33,0,0