Function that adds recommended efficiency speedup lines and points to benchmarks

Description

This function adds to the timing results (or the benchmarked data). It computes the recommended efficiency speedup line and the point which denotes the recommended thread count, both being based on the specified efficiency value.

Usage

addRecommendedEfficiency(benchmarkData, recommendedEfficiency = 0.5)

Arguments

Details

This function allows users to add a "Recommended" efficiency line to previously computed benchmark data (without needing to recompute the timings). The recommended speedup is based on the provided efficiency value, which adjusts the slope of the speedup curve and correspondingly helps in the computation of the closest point of measured speedup to the "Recommended" speedup curve.

Value

The input data.table with the recommended efficiency added to the plot data (attributes).

See Also

findOptimalThreadCount for computing the benchmark data with measured and ideal speedup data.

Examples

# Finding the best performing thread count for each benchmarked data.table function
# with a data size of 1000 rows and 10 columns:
benchmarks <- data.table.threads::findOptimalThreadCount(1e3, 10)
# Adding recommended efficiency to the plot data:
addRecommendedEfficiency(benchmarks, recommendedEfficiency = 0.6)

Function that finds the optimal (fastest) thread count for different data.table functions

Description

This function finds the optimal thread count for running data.table functions with maximum efficiency.

Usage

findOptimalThreadCount(rowCount, colCount, times = 10, verbose = FALSE)

Arguments

Details

Iteratively runs benchmarks with increasing thread counts and determines the optimal number of threads for each data.table function.

Value

A data.table of class data_table_threads_benchmark containing the optimal thread count for each data.table function.

Examples

# Finding the best performing thread count for each benchmarked data.table function
# with a data size of 1000 rows and 10 columns:
(optimalThreads <- data.table.threads::findOptimalThreadCount(1e3, 10))

Function to make speedup plots for the benchmarked data.table functions

Description

Function to make speedup plots for the benchmarked data.table functions

Usage

## S3 method for class 'data_table_threads_benchmark'
plot(x, ...)

Arguments

Details

Creates a comprehensive ggplot showing the ideal, sub-optimal, and measured speedup trends for the data.table functions benchmarked with varying thread counts.

Value

A ggplot object containing a speedup plot for each benchmarked data.table function.

Examples

# Finding the best performing thread count for each benchmarked data.table function
# with a data size of 1000 rows and 10 columns:
benchmarkData <- data.table.threads::findOptimalThreadCount(1e3, 10)
# Generating speedup plots based on the data collected above:
plot(benchmarkData)

Function to concisely display the results returned by findOptimalThreadCount() in an organized table

Description

Function to concisely display the results returned by findOptimalThreadCount() in an organized table

Usage

## S3 method for class 'data_table_threads_benchmark'
print(x, ...)

Arguments

Details

Prints a table enlisting the best performing thread count along with the runtime (median value) for each benchmarked data.table function.

Value

NULL.

Examples

# Finding the best performing thread count for each benchmarked data.table function
# with a data size of 1000 rows and 10 columns:
(benchmarkData <- data.table.threads::findOptimalThreadCount(1e3, 10))

Function to run a set of predefined benchmarks for different data.table functions with varying thread counts

Description

Function to run a set of predefined benchmarks for different data.table functions with varying thread counts

Usage

runBenchmarks(rowCount, colCount, threadCount, times = 10, verbose = TRUE)

Arguments

Details

Benchmarks various data.table functions that are parallelizable (setorder, GForce_sum, subsetting, frollmean, fcoalesce, between, fifelse, nafill, and CJ) with varying thread counts.

Value

A data.table containing benchmarked timings for each data.table function with different thread counts.

Function to set the thread count for a specific data.table function

Description

Function to set the thread count for a specific data.table function

Usage

setThreadCount(
  benchmarkData,
  functionName,
  efficiencyFactor = 0.5,
  verbose = FALSE
)

Arguments

Details

Sets the thread count to either the optimal (fastest median runtime) or recommended value (default) based on the chosen type argument for the specified data.table function based on the results obtained from findOptimalThreadCount().

Value

NULL.

Examples

# Finding the best performing thread count for each benchmarked data.table function
# with a data size of 1000 rows and 10 columns:
benchmarkData <- data.table.threads::findOptimalThreadCount(1e3, 10)
# Setting the optimal thread count for the 'forder' function:
setThreadCount(benchmarkData, "forder", efficiencyFactor = 1)
# Can verify by checking benchmarkData and getDTthreads():
data.table::getDTthreads()