At 05:57 PM 3/10/2002 -0500, Simon Glover wrote: > In Parrot's standard operational mode, the following code[1]: > > set N2, -2147483648.0 > set I2, N2 > print I2 > print "\n" > end > > produces the output: > > -2147483648 > > but in JIT mode (on an x86) it produces: > > -0 > > This smacks of some kind of overflow problem, but I haven't the faintest > idea of where in jit.c to start looking.
Using 32bit representations, you just set the int value to 0x80000000 which is all 0 with the sign bit set. For floating point the sign bit set, with all zeroes indicates -0 in IEEE floating point. In short, the output is correct for a 32bit float. Maybe the JIT can be tweaked to use double precision floats as default. -Melvin
