jclass unused, jfloat v)
this function works fine.
I presume that your results are gotten by calling Float.floatToIntBits() from Java code. Based on the results, you don't have any trouble passing the arguments in and computing the results. The only issue remaining is returning the result.
Have you tried the CVM_DEFINES += -DAAPCS that Chris suggested yet? If not, please try that and follow his instructions. That will give us some needed additional information.