which architecture are you talking about? it depends on the individual programs: if the kernel supports emulation you can still build programs with -msoft-float. to find out what the kernel supports look at the usual places like /proc or the boot messages, or write a little test program and look if it catches an exception.

If -msoft-float is not used, you will find, of course, floating point instructions in the assembler dump (objdump -DS myprogram or object file).
Now if the kernel has floating point emulation disabled and you do not have an FPU, your program gets signalled and usually terminates.
Whether you have an FPU can be seen in /proc/cpuinfo, whether it is used too (and /proc/cmdline in case you used nofpu or similar)

This method does not answer all questions, but quite a lot :-)