untar-in-parallel.sh

#!/bin/bash

# function to extract each tar file
extract_tar() {
  local file="$1"
  local folder="$(basename "$file" .tar.bz2)"
  mkdir "$folder"
  tar -xjf "$file" -C "$folder"
}

# get the number of CPUs
num_cpu=$(grep -c ^processor /proc/cpuinfo)

# create a semaphore to limit the number of parallel processes
semaphore=$(mktemp -u)
mkfifo "$semaphore"
exec 3<>"$semaphore"
for ((i=0;i<num_cpu;i++)); do
  echo >&3
done

# loop through all .tar.bz2 files in the current directory
for file in *.tar.bz2; do
  read -u 3
  (
    extract_tar "$file"
    echo >&3
  ) &
done

# wait for all background jobs to finish
wait

# clean up the semaphore
exec 3>&-
rm "$semaphore"

I really like what how you implemented the max thread count. That's clever.

BTW If you have gnu parallel installed, I think this makes basically the same results as your script:

#make 1 directory per archive
for tarfile in *.tar *.tar.*; do mkdir ${tarfile/.tar*/}; done

#extract archives to their own directory, 1 tar process per CPU core
parallel -P $(nproc) bsdtar -xf {} --directory {=s/.tar.*//=} ::: *.tar*

both the for loop to make the directories and the extract command should work on any *.tar* archives, compressed or not, but will not work on archives named like .tgz instead of .tar.gz. I use bsdtar because it does single threaded decompression (gnu tar may use multiple threads for xz and zstd and other compressiosn, which could get excessive as we're already making 1 process per cpu core)

Thanks for your feedback @bobpaul! I really appreciate it :)

Let me introduce my untar version