Hi Ravi,
1. Does csmith has any standard recommendation for range of seed to used for effective test case generation . i.e Min range =< seed <= Max range
Csmith uses the rand48 library, I would imagine because its algorithm is well-defined, making seeds portable across platforms. 32 bits of any seed you pass should be useful.
2. How effective does checksum helps to keep a tab on runtime errors , can you explain how checksum calculation is happening, so that I can understand its effectiveness?
It's just a CRC32. It should be very close to 100% effective in detecting errors in global variables.
http://en.wikipedia.org/wiki/Cyclic_redundancy_checkThe only purpose of the checksum is to create a concise summary of a program's execution. If you're at all worried about it being less than perfectly effective, the --check-global option causes the generated code to just print the values of all globals before the program terminates.
John