Verify and load_vec

[Kyle Klein]

Hi all,

Correct me if I am wrong, but it seems to me that load_vec and verify both assume that you are running on a single processor. My question is do we need to support loading a vector and distributing it across our machines and similarly do you want us to verify correctness when the vector is distributed? The latter only requires a simple modification of the harness, but I assume such modification would be disregarded.

Thanks,

Kyle

--
Kyle Klein 

Ph.D. Student 

Department of Computer Science UC Santa Barbara

http://www.cs.ucsb.edu/~kyleklein

[Lijie Ren]

I think we need to parallelize both load_vec and save_vec to make them work on multiple processors... I've done nothing to the harness file and used diff to verify.

-Lijie

[Matt Weiden (Reader)]

Hello Kyle and Lije,

Yes, if you want load_vec to work in parallel you must parallelize it
yourself. I included this extra method just in case you wanted to try
out your own problem sets (yeah!). The important thing to have working
is the model problem given by cs240_getB.

You can use Verify on problem sets small enough that you can collect
the whole vector on one machine and run it serially. Once you have
verified that your calculations are correct, I don't believe its
necessary to keep it in your program. Then you can run problem sets as
big as you want without having to parallelize verify or collect the
vector on one machine.

Let me know if you have any other questions.

Matt

[Kyle Klein]

Hi Matt,

I would just like to comment that simple dividing n*n by the number of processors will allow each processor to check its own piece of the vector itself.

This allows checking large problem sizes, and doing small ones more quickly...

I am confident our code is correct already because of this, do I need to implement a check without modifying the harness for you to use when grading? 

Thanks,

Kyle