blank-canvas-mark

benchmarks for blank canvas

To run Haskell

$ make build
$ make test SYS=OSX-Chrome

To access blank-canvas, use the URL:

http://localhost:3000/?width=800&height=600

Other ways of running tests:

$ ./dist/build/blank-canvas-mark/blank-canvas-mark -o foo.html -u foo.csv

OR $ ./dist/build/blank-canvas-mark/blank-canvas-mark -o foo.html -u foo.csv Bezier

Environment

To use a weak remote monad

$ BLANK_WEAK=1 ./dist/build/blank-canvas-mark/blank-canvas-mark 

To create a version of the benchmark that be be used across networks

$ BLANK_REMOTE=1 ./dist/build/blank-canvas-mark/blank-canvas-mark

To run JavaScript

JavaScript automatically defaults to a canvas of 800x600

OS X

$ cd js
$ open benchmark.html

To run a single test, use URL with dut

file:///.../benchmark.html?dut=Bezier

Linux/BSD

$ cd js
$ xdg-open benchmark.html

To look at results

$ python -m SimpleHTTPServer
$ open http://localhost:8000/summary/results.html

Results format

We store our results internally as json records. For each point, we use the following format.

Required

Field	Values	Notes
test	"blank-canvas-mark"	Name of test set; required.
os	"OSX" or "Linux"	OS test run on
browser	"Chrome" or "Firefox"
platform	"Blank Canvas" or "JavaScript"
options	"weak" and/or "web-sockets"	Optional
name	"Bezier", ...	Name of specific benchmark
time	(clock) runtime	The data point, a mean
ub / lb	upper and lower bound on mean
confidence	0.95	confidence in the mean
hash	"ABC123"	Hash of specific device under test
raw	{...}	Raw results to reproduce quality of mean
outliers	{...}	Quality of test data

The hash it there so we can compare like with like.

Optional

• _id: optional unique id for each line
• report: (individual report from criterion)