How do I implement signed int64 multiplication with overflow checking in 32-bit mode?
Thanks.

You'll have far fewer problems if you first convert this to unsigned multiplication. Check the sign of both numbers -- if just one of the numbers is signed, remember to make the answer signed; otherwise the answer will be positive.

Then use the classic method you learned in elementary school (you'll do four unsigned multiply operations), and add up the results keeping the components in the proper columns. There will be twice as many bits in the result as in the multiplier and multiplicand combined, so you may want to make 128 bits available to hold the answer.

Treat each 32-bit component as a column; so the answer has potentially four columns. If any part of the result spills over past the first 64 bits (into the last two columns), the number has overflowed. And if bit 63 is set, the number won't fit into a signed 64-bit number (there could be a boundary condition you have to test). And remember to make the answer negative if just one of the numbers was originally negative.

This should at least get you started... good luck!