papiplot: Cache Miss Benchmark Plotter

In RBigData/hpcvis: Plot Utilities for MPI Codes and Hardware Counters

Description

Cache Miss Benchmark Plotter

Plots Cacahe Misses

FLOPs Benchmark Plotter

Create plots from PAPI performance counter data.

Usage

## S3 method for class 'cachebench'
papiplot(x, ..., title, opnames, facet.by = "operation",
  label.angle = 0, levels = 1:3)

## S3 method for class 'papi_cache'
papiplot(x, ..., title, opnames, color = FALSE,
  facet.by = "operation", bar.label = FALSE, label.angle = 0)

## S3 method for class 'flopsbench'
papiplot(x, ..., title, opnames, groupby = "color")

papiplot(x, ..., title, facet.by = "operation", label.angle = 0)

Arguments

x	PAPI object.
...	Additional objects.
title	The label for the plot title. Should be a character string of your choice, NULL for no label, or left blank for the default plot label. In the latter case, this is chosen based on the input data.
opnames	An optional argument to specify different names for the expressions/operations used to generate the profiler data.
facet.by	Choice to facet cache plots by the different expressions/operations (facet.by="operation"), or by the cache level (facet.by="level").
label.angle	The angle of x-axis labels.
levels	Which cache levels to display.
color	Logical; should different groups be colored?
bar.label	Logical; should numeric values of heights of bars be shown?
groupby	Should operations be grouped by "color" or by "shape"? The latter is useful when black/white plots are needed.

Details

One may wonder why we do not simply overload plot(). In fact that was the original incarnation of this package. However, overloading plot() leads to several issues. Simply, the basic plot command is *too* overloaded. This makes it difficult to find documentation, view function arguments, etc. Using papiplot(), finding help is easier, and one can also enjoy function argument autocompletion via the tab key.

Value

A ggplot2 object.

Examples

## Not run: 
library(pbdPAPI)
x <- cachebench(rnorm(1e4), rnorm(2e4), rnorm(3e4))

library(hpcvis)
papiplot(x)
papiplot(x, label.angle=15)

## End(Not run)

## Not run: 
library(pbdPAPI)
x <- system.cache(rnorm(1e4))
y <- system.cache(rnorm(4e4))
z <- system.cache(rnorm(8e4))

library(hpcvis)
papiplot(x)
papiplot(x, opnames=NULL)

opnames <- c("small", "medium", "large")
papiplot(x, y, z, opnames=opnames)
papiplot(x, y, z, opnames=opnames, color=TRUE, facet.by="level")

## End(Not run)

## Not run: 
library(pbdPAPI)
data <- list(rnorm(1e4), rnorm(2e4), rnorm(3e4))
x <- flopsbench(exp, sqrt, sum, data=data)

library(hpcvis)
papiplot(x)
papiplot(x, groupby="shape")

## End(Not run)

RBigData/hpcvis documentation built on May 8, 2019, 4:54 a.m.