I'm trying to mine a genesis block for an experimental Peercoin clone.

I've already written one using boost threads, and it's been really fun since this is the most processor intensive program I've ever written, but I'm only getting 1M hashes per second, so I'm guessing it's going to take about 72 minutes based assuming initial difficulty 1 and this equation is correct.

Is there a relatively simple way to send the functions to the GPU?

All I'm doing so far is using a "silent" (returns false instead of writing the error to debug.log) CheckBlock() to verify if the nonce is valid or not.

Please show me a simple library that can use the GPU to verify the nonces.