sure, let me first clarify that the "Usage:" section of a help page from a function in R normally shows the function call with all the arguments and their default values in the form f(argument=value) but this does not mean that you should call the function f() exactly in that way, it means that if you do not set that argument, then the function will use that default value. in fact, i see that in your case, it may suffice to set the argument parallel.sz=3, in addition to the input expression data matrix, the input list of gene sets and your desired cutoffs on the minimum and maximum gene set size.

the following simulation of an input expression matrix with 10k genes and 1k samples, and 1000 gene sets, runs in about three minutes using 4 cores in my computer:

p <- 10000
n <- 1000
sizeGeneSets <- sample(5:500, size=1000, replace=TRUE)
geneSets <- lapply(as.list(sizeGeneSets), sample,
                   x=paste0("g", 1:p), replace=FALSE)
names(geneSets) <- paste0("gs", 1:length(geneSets))
length(geneSets)
[1] 1000
y <- matrix(rnorm(n*p), nrow=p, ncol=n,
            dimnames=list(paste("g", 1:p, sep="") , paste("s", 1:n, sep="")))
dim(y)
[1] 10000  1000

library(GSVA)

es <- gsva(y, geneSets, parallel.sz=4))
Setting parallel calculations through a MulticoreParam back-end
with workers=4 and tasks=100.
Estimating GSVA scores for 1000 gene sets.
Estimating ECDFs with Gaussian kernels
Estimating ECDFs in parallel
iteration: 100
  |======================================================================| 100%
dim(y)
[1] 10000  1000

by setting the argument parallel.sz the gsva() function will try to run the calculations in parallel using a multicore backend with the number of cores specified in parallel.sz. if you need additional fine tuning of the parallelization then you need to set the argument BPPARAM. to learn how to set that argument you should read sections 1 to 3 of the vignette "Introduction to BiocParallel` from the BiocParallel package.