henman at it.to-be.co.jp
Tue Feb 20 06:15:14 CET 2007
>...snipped... "Parallel Architecture & Programming".
> We're considering to offer GnuPG as a parallelization project to the students
> this semester.
I don't think its a good idea. Technically gpg is by nature a sequential process.
Using pthreads isn't going to turn a sequential process into a parallel one.
Actually using multiple threads has probably made it less robust, less efficient, and more likely to be breached as has been pointed out earlier.
Though why don't you do a neuron network or some such or the traveling salesman's shortest distance.
Take a look at ibm's new cell chip see what you can do with the eight identical synergistic (vector) processing elements (SPEs). They run independently of each other and can be used for real parallel processing, still synchronization is necessary if you
want to use them all in coneert, like using a mailbox type communications between them. There are 128 registers, 128 bits wide and use simd. If you have unix/linux you can program them.
Or consider about plugable authorization modules.
Just my 2 cents.
More information about the Gnupg-devel