R terra:: cumulative sum reset function across layers of a stack

I have a stack with large number of layers, and I want to run a pixel-wise cumulative sum function across the layers with reset when sum becomes <0. Ive compiled the following code:

# given that "A" is an input stack with lengthy number of layers

B<-A 

for(i in 2:nlyr(B)){
  B[[i]]<-ifel((A[[i]]+B[[i-1]])>0, A[i]]+B[[i-1]], 0 )
  
}

, which works well but I have a large stack of rasters (in terms of nlyrs). Im wondering if there is a faster way to implement this task.

UPDATED. Apparently the function I posted before was incorrect. Below is a solution which benefits from parallelization of the operation via cores argument in the terra::app function. Though I assume this is not the most optimal suggestion.

library(terra)
v <-  c(1, 2, -40, 35, 10, 10)
r <- rast(ncol=1, nrow=1, nlyr=6, vals=v)
r

cumsumreset <- function(x) {
  for (i in 2:length(x)) {
    if (x[i-1] < 0) x[i-1] <- 0
    x[i] <- x[i] + x[i-1]
    if (x[i] < 0) x[i] <- 0
  }
  x
}


a <- app(r, cumsumreset, cores=5)

a

P.S. My appologies and thanks to @RobertHijmans