Raymond Toy wrote: > Is the result of floor known to be a fixnum? (signed-byte 32)? > Declaring that would help.
Thanks for that hint. The result is a fixnum, but I didn't tell the compiler that it is. Now there is no consing any more. The resulting code (it's a simple Particle in Cell program) now is only 50% times slower than the equivalent C-program. Considering that there is still more safety checking in CMUCL (even with (safety 0)) this is great. Rolf
