I tried really hard to find the point to all this, but to say the author is vague would be a huge understatement.
The claims are that Perl is slow. Slow how, exactly? The only reference given for this observation is a heavy floating point matrix math algorithm. That's hardly a job that would prompt many to reach for the Perl hammer, but if you're making that choice we have PDL.
The complaints are about Perl's "magic" and how that's the core of its perceived performance problem. RPerl promises a "low-magic" Perl...what parts are you planning on cutting out? I hazard to guess from the marketing that the list includes anything that isn't well suited for raw number crunching. That would basically make it not-Perl.
While Perl is a full feature language well capable of performing across a very wide range of domains, it really feels like the author has decided he only works in one of Perl's least natural settings. Having still chosen Perl as his tool of choice, he declares the entire thing bunk because it under-performs in his one niche domain. Brilliant!
The first thing any decent Perl Hacker will tell you is use the right tool for the job... Clearly, for his job, Perl isn't it. There's precisely zero chance of RPerl returning Perl to "its former glory" (whatever the fuck that even means) because it's increasingly clear that RPerl wants to blow up Perl to save Perl. And that's just stupid.
Either download PDL and move on with your life, or pick another language that performs in your niche domain: Ripping out everything that makes Perl be Perl under some strange and highly selective notion of "performance", is a waste of time. Of course, it's pretty clear the author is smart...but a mathematician, which might explain some of this strange obsession and odd way of approaching the "problem": Mathematicians as a generalization just aren't good software developers (or much of anything else that has to do with the daily grind of the computing world), and Perl really wasn't built for them. Perl is a practical tool for solving practical problems, which doesn't typically include plotting the paths of planets in the universe. That's just not its optimized use case.
All the reasons Perl is fantastic are extremely likely to drive mathematician types absolutely bonkers, the same way it drives (real) Computer Science types bonkers. And they can curse it as they always have, while they tinker away at their elegant minimalistic research toy langauges, and yet Perl is still the duct tape that continues to hold the Internet together.