Add a parser benchmark based on the PLT
Created attachment 62380 [details]
Comment on attachment 62380 [details]
Unfortunately, this benchmark has a much larger divergence problem than the original html-parser.html benchmark. I think the problem is that the parser is kicking off subresource loads that somehow accumulate and slow things down. This dataset probably has many more subresources than the HTML5 spec.
I'm not sure how to do this in a reasonable way. :(