I compared the speed (plain text extraction) of xpdf, poppler and mupdf on 100k PDFs. mupdf is in 95% cases the fastest, then comes xpdf and then poppler (the latter two crashed on a few files). SumatraPDF viewer went from poppler-only to two engines (poppler & mupdf) to mupdf+patches. At the moment from my experience SumatraPDF has the fastest and most reliable PDF engine, that is open source. So it will be interesting how this Chrome PDF open source engine (based on Foxit?) performs outside of Chrome as standalone library/commandline tool.