delay computation is performed on 64 bits to avoid overflows (replaced the solution provided earler by Tim Rentsch, both approach have the same performance hit but using long long is just simpler)

git-svn-id: https://svn.code.sf.net/p/dosmid/code/trunk@336 850bf26b-01f9-4b63-b2ae-8b26f6eb5263
2 files changed