>>1
Lisp is usually slower than C, but it can be as fast as C if you push it enough. Usually most people that use Lisp tend to just write everything as expressive/readable as possible, which can make things slow, then they optimize the parts that are too slow. Of course, Lisp offers a wide variety of optimization strategies, some of which are not offered by other languages (full control over compile-time calculations for one, caching and various code generation/manipulation tricks are also very easy to do). Lisp can be as fast or as slow as you want it to be, providing you have a decent implementation.
If you want speed, I'd say go for ASM and C. Lisp is a good choice too, but it won't be fast by default, it's usually slow because it lets you do whatever you want without questions asked, so you have to work to achieve speed, but its advantage is that it makes hard things simple, and it gives you a lot of flexibility on how you rewrite/redesign your code (usually with very little effort required), so you can test many algos easily. If you have a O(logn) algo in your Lisp program, and a O(n*n) algo in you C program, the Lisp program would be faster because of better choice of code, which tends to be easier to do in Lisp (altough, such choices are language independent, I usually end up writing a lot more code in C than in CL, so I find rewriting code much easier in CL). As for taking the same algo you wrote in Lisp and porting it to C, you'll only get linear performance gains from that if lucky, but sometimes it may be worth it if the CL implementation isn't good enough (it usually is just fine, but if you're writing C/Fortran in CL, you might as well just do it in C/Fortran).