I finally got around making the palindrome generator. that was easy.
Stats :
The first five (2, 3, 5, 7, 11) are hardcoded because my palindrome gen wont work with less than 3 digits.
42nd palprime : 18181, 164/512 = 0.32 seconds 155/512 = 0.3 seconds.
70th palprime : 39293, 428/512 = 0.84 seconds 381/512 = 0.74 seconds. Unfortunately, that's the furthest we can go with 16 bits.
This is an Axe program that works with 16 bits integers at full speed.

I measure the speed with the max speed interrupt running theta++ every time. I display the value of theta at the end. Since this is a 512Hz interrupt, I just have to divide by 512 to know the time it took.

Edit : Dropped in a crazy optimization that brought my times down by a fair amount.

I totally agree with you Runer, checkpoints are unfair and even though they decrease max computing time, they make searching time random depending on wether you're just before a checkpoint or just after.

I really don't think such checkpoints should be allowed. It wouldn't contribute anything to the algorithmic thinking the competition is supposed to promote. Suppose checkpoints were allowed. Competitors would need to use them to keep up with other programs that use checkpoints, so all the good entries have them.

If, like  Dapianokid said, people could only store the 10th, 50th, and 100th palindromic primes, then all the programs would still start testing with the same number (say if we wanted the 74th prime palindrome, everyone would start with the 50th one), so the winner would be the person with the most efficient algorithm. That person still would have won if they started testing numbers from the beginning, so why not just start testing numbers from the beginning? It makes the programs simpler, smaller, and more "pure" anyways.

Meanwhile, if people could store whatever palindromic primes they wanted, the winner would be either:
a) lucky that he/she happened to store the palindromic primes just below the numbers wanted
b) someone who precomputed a huge number of palindromic primes
And both a) and b) imply that the efficiency of the actual algorithm used to search for palindromic primes is not important, which is bad.

Okay, I agree with having official numbers to get timings.

I have a question: What program do we submit if we're doing Axe? The compiled executable, or the source?

Okay, thanks.

Edit: Runer suggested that I don't post timings so as not to discourage anyone, so I'll just say that my program can get the 5953rd palprime in under an hour.

Lol, that is too late, I got discouraged when I learnt that Runer and you were participating, no need to post your timings