different total currents between CPU and GPU for the branch "improved_gpu_support_dftu"

I have compiled the branch "improved_gpu_support_dftu" using GPU on Ada. However, I noticed that the total current at t = 0 for a td calculation does not vanish for the GPU, but it vanishes if I used the option "DisableAccel = yes" in the input file.

The total current at t=0 for the GPU:

################################################################################
# HEADER
# Iter               t                      I(1)                    I(2)                    I(3) 
################################################################################
       0  0.0000000000000000e+00  8.0776266996049981e-03  2.3341125218222691e-02 -7.5193841120253010e-04  

The total current at t=0 when the input variable DisableAccel is set to yes:

################################################################################
# HEADER
# Iter               t                      I(1)                    I(2)                    I(3)  
################################################################################
       0  0.0000000000000000e+00 -4.2085775729522693e-09  3.3200316141268904e-10 -3.1976306691647950e-09  

Here I also uploaded the input, output, and jobscript files.

inp

jobscript.slurm

output-disableAccel

output-gpu