Ckelty at Savage Minds has some pointers (and, also refers to this very resourceful seven page guide (pdf)). As is noted by ckelty, however, this method works only for humanity texts without much of mathematical details. A text in materials science and engineering may probably be read in a few hours if you are very familiar with the material, and if most of the text is not full of mathematical details and derivations. If not, of course, as Euclid is supposed to have said, there is no royal road.
Having made that comment, a mathematically dense text can still be read faster than is usual; I follow several strategies (and, though the following discussion is explicitly in terms of texts in materials science and engineering, most of what I say might also work for texts in physics and chemistry; however, they might not work for a mathematics or programming text).
Approach the text with a particular problem in mind (and, skipping everything that is not relevant to the problem at hand). If you are lucky, four or five problems should see you through a text with ease and give you considerable proficiency. If you cannot come up with any problem of your own, choose those texts which are nicely structured in that each chapter does start with an intriguing problem and at the end of the chapter one would have solved the problem; or, browse through the problems at the end of the chapter, choose one or two from the list, and keep them at the back of your mind when you start reading the text.
The other method is to choose an important result/theorem/explanation, and begin from there and backtrack to the earlier sections as and when they are needed. This works for reading papers too.
Sometimes, I also google stuff, and sample lots of peripheral information at a brisk pace before getting down to the text. This works wonderfully if you are completely new to a field.
I have also found that while reading texts, if I devote myself exclusively to one book (and area) at a time, I can read more than if I scatter my efforts (and, I have heard that Knuth follows the same strategy to master different areas when he writes his TAoCP series texts:
The only way to gain enough efficiency to complete The Art of Computer Programming is to operate in batch mode, concentrating intensively and uninterruptedly on one subject at a time, rather than swapping a number of topics in and out of my head.
However, this strategy backfires (as I learnt from my own experience), if you are doing coursework, when you are supposed to read more than one text and from more than one area at the same time.)
One should also note that even though reading the first few texts are excruciatingly slow, the effort pays; later texts can be mastered with lesser and lesser effort.
Before I end this post, some of the most intense and least painful reading of texts happened to me when I read them in a group; if you are lucky (as I was) and can find a couple of deep readers, and, if you can read the same sections from the same text, and try to explain the stuff to each other, you will be surprised by the amount you can learn in a couple of days.