It depends on how far you unroll the SHA256 algorithm. The open source code
https://bitcointalk.org/index.php?topic=9047.0 fits two fully unrolled SHA256 engines (for the two stages of SHA256D) into a 115kLE cyclone IV (I think the utilization is around 90kLE). However if you just want the minimum then a single fully rolled engine comes in at around 1kLE (and runs 128 times slower). As for the conversion between Logic Elements (CLBs) and gates, I've seen estimates of around 12 gates per LE. Anyway have a look through the thread I linked above, lots of info there and the linked github has the verilog code.