multithreadcorner / mcbooster Goto Github PK

Hi,

let me try to explain what I mean:
So when using mcbooster to generate large amounts of random numbers you will need to do it in batches since a GPU has limited memory right? But if you use different seeds, in MCBoster it doesn't really change that much. See here: https://github.com/MultithreadCorner/MCBooster/blob/master/mcbooster/functors/DecayMother.h#L150

The number of collisions depend on the number of events generated. (They grow approx. quadratic)
I propose using an offset rather than the multiplication by "fseed", thus starting at "evt+offset" if I generate a second batch of numbers.

Hope this is somewhat understandable. If not I'll be happy to clearify this more.
Cheers

Followed the installation steps provided and compiled with:
NVCC V7.5.17
Cuda architecture 2.1(2.0)
gcc 4.8.5

All examples seem to run without a problem.
The only error I observed so far is the following:

PARSE ERROR: Argument: -e1000
Couldn't find match for argument

when trying to use the -e flag with the binary "MCBooster_Example_CUDA_GenerateSample"

If I am not mistaken, can't we get ride of these lines?

The If statement on line 322 makes sure that this only run for device -> device evaluation. So no need for the lines linked to above right?