RFC: Polishing Up The Result Graphs

[Michael]

in the SVG graph images, the PTS logo and the base.png is a data uri for a PNG - surely this would be better as SVG vectors

Haven't found a better way of handling it. The SVG of the PTS logo created in Inkscape is quite large. I haven't experimented to see how easy it is to embed an SVG inside an SVG and work reliably cross-browser, etc. Otherwise if just embedding the SVG tags inside the graphs, it's a lot of wasted bandwidth, etc.

[geearf]

There is summary graphs on OpenBenchmarking.org, but won't be in the articles, since I need as everyone as possible clicking through the articles to actually break even on all the testing that goes into each article....

Oh I never realized you have summaries on openbenchmarking, cool!

Now I looked at one article and could not find the link to openbenchmarking, how do I get that?


As a side thing, not about graphs, but I personally feel that justified text is better than simply aligned-left

[monraaf]

Another benchmarking site that I regularly visit (computerbase.de) has a cool feature on their graphs, which is very useful for graphs with lots of lines.

The individual lines respond to mouse-over and clicks:
- Mouse-over will highlight (make it bold) the line and at the same time highlight the corresponding entry in the legend (especially useful if some lines have similar colors).
- Click will hide that line from the graph, useful for when you don't care about the performance of one uut (to make the others more visible).

It also has another useful features:
In bar graphs, mouse-over additionally replaces the absolute performance numbers by relative performance to the mouse-overed bar (in %, i.e. the mouse-overed bar becomes 100% and others become 90%, 110% etc).

[kaprikawn]

For the bar graphs, I'd prefer it if they were ordered by performance based on whatever test you're running, best performer at the top then cascading down.

[curaga]

libframetime in all my trials still doesn't work always well.

Patches welcome, or at least reports to the github tracker so people who have specific game Z can help. I just pushed a flush after every line, which should help with those that don't exit gracefully.

[microcode]

Legible colours would be nice, there's a decent formula from W3C which you can use to calculate if your colour contrasts enough to be legibile against a specific background.

[microcode]

Haven't found a better way of handling it. The SVG of the PTS logo created in Inkscape is quite large. I haven't experimented to see how easy it is to embed an SVG inside an SVG and work reliably cross-browser, etc. Otherwise if just embedding the SVG tags inside the graphs, it's a lot of wasted bandwidth, etc.

For kicks I created a reasonably minimal SVG of the pts logo, and it came out to 469 bytes raw, 298 bytes gzipped. (https://gist.github.com/xorgy/65c6d0e87757dbb56a75)

With regard to embedding, I think it should be fine to embed it as an image the same way you do the current png one.

Bonus: The SVG is pixel-aligned at natural resolution, and looks a heck of a lot better as a result.

Cheers.

[Michael]

For the bar graphs, I'd prefer it if they were ordered by performance based on whatever test you're running, best performer at the top then cascading down.

I do it when feel it's more beneficial, but regardless of how I choose to show the graphs, on OpenBenchmarking.org results page is an option to "sort by performance"

[Michael]

Legible colours would be nice, there's a decent formula from W3C which you can use to calculate if your colour contrasts enough to be legibile against a specific background.

But what about an algorithm for calculating random, legible colors? That's what I had been trying on before but to no luck.

[Michael]

For kicks I created a reasonably minimal SVG of the pts logo, and it came out to 469 bytes raw, 298 bytes gzipped. (https://gist.github.com/xorgy/65c6d0e87757dbb56a75)

With regard to embedding, I think it should be fine to embed it as an image the same way you do the current png one.

Bonus: The SVG is pixel-aligned at natural resolution, and looks a heck of a lot better as a result.

Cheers.

Will take a look, thanks!