The cuda kernel used in the InterOpt R package for calculating NormFinder and Genorm in parallel for lots of combinations.
This code is a modified version of the cuda kernel used in NormiRazor
The following modifications are applied to the original code:
- Aggregation weights for combinations
- Instead of all combinations, the exact combinations are defined in the input
- In each sample NormFinder only uses elements with CT<35 for calculating sample mean
- Genorm gets a list of stable genes. All genes and combinations are only comapred to those stable genes
This kernel is only needed if you want to calculate NormFinder and Genorm stability measures in the InterOpt package. You should build the InterOptCuda.cu file using nvcc and put the executable InterOptCuda into your PATH:
nvcc -O2 InterOptCuda.cu -o InterOptCuda -std=c++11
for more information please refer to: