I saw a failure last night. I don't think it's related to these new changes to the miner, but:
2014-Jun-18 22:49:36.476494 Connected bbr.extremepool.org:7777 OK
2014-Jun-18 22:49:37.349759 Getting scratchpad...
2014-Jun-18 22:49:37.468708 Scratchpad received ok, size: 0Kb, heigh=0
This caused the miner to fail in processing its scratchpad. I'm guessing this was a pool crash and the resulting invalid data sent back to the miner caused the miner to crash.
Did others running the normal simpleminer also see this problem?
I'll get a fix to this crash bug ASAP. In the meantime, I've submitted a pull request for efficiency improvements, because I think they're safe.
Expect an invalid scratchpad size robustness patch within two hours.
I've put in a quick fix to avoid this specific problem and included it in the pull request for efficiency.
update: This is now merged. Update your tree to get the now more-robust simpleminer code.In the meantime:
(a) I still recommend that everyone run their miner in a shell script that auto-restarts. For example, on linux, the following will work nicely:
#!/bin/sh
ADDRESS=YOURADDR
POOL=YOURPOOL
THREADS=NUMTHREADS
while true
do
./build/release/src/simpleminer --pool-addr $POOL --login $ADDRESS --pass x --mining-threads $THREADS
done
(b) I'm going to do a bit of refactoring of simpleminer focusing on overall robustness. This shouldn't be too bad, but it's not 5 minutes of hacking and deserves to be done with a little care, so I'll hop on it tonight/tomorrow.
. If porting a GPU miner is not hard as you said, there is high chance that you're not the only one who have optimized miner for this coin. There are many skilled programmers in this board, you know. 
