PBKDF2 and bcrypt are configured with an "iteration count", which means that they can be made as slow as you want. Therefore, there cannot be a table which shows how fast they go. What you need to do, instead, is to decide how much time you allocate to the function (e.g. you want it to take 0.05 seconds on your server) and then set the iteration count accordingly. This very much depends on the computing abilities of your server, so this is a matter of measure on your target hardware.
For the other functions, you can find an awful lot of data on this site for performance measures of many hash functions on a lot of platforms (although all these platforms tend to be servers, desktop systems, or at worst smartphones -- smaller embedded systems are not covered). The high-level picture is that:
- Performance depends a lot on the hardware.
- Performance depends a lot on the implementation.
- All currently defined hash functions are way too fast to be adequate for password hashing by themselves; you need some kind of slowdown through thousands or millions of iterations.
While PBKDF2 adds iterations properly, it can still be made much faster in the context of a parallel attack, by using a GPU; therefore, you may prefer to use a function which is GPU-unfriendly, and, right now, this means bcrypt (scrypt being a nice candidate, when it becomes old enough to be trusted). See this answer for details.