diff --git a/tools/git2graph/README.md b/tools/git2graph/README.md
index 7a7caa6..2bc1360 100644
--- a/tools/git2graph/README.md
+++ b/tools/git2graph/README.md
@@ -1,67 +1,67 @@
 git2graph
 =========
 
 `git2graph` crawls a Git repository and outputs it as a graph, i.e., as a pair
 of textual files <nodes, edges>. The nodes file will contain a list of graph
 nodes as Software Heritage (SWH) Persistent Identifiers (PIDs); the edges file
 a list of graph edges as <from, to> PID pairs.
 
 
 Dependencies
 ------------
 
 Build time dependencies:
 
 - [libgit2](https://libgit2.org/)
 
 Test dependencies:
 
 - [bats](https://github.com/bats-core/bats-core)
 
 
 Micro benchmark
 ---------------
 
-    $ time ./git2graph -n >(pigz -c > nodes.csv.gz) -e >(pigz -c > edges.csv.gz) /srv/src/linux
-    232,06s user 16,24s system 90% cpu 4:40,35 total
+    $ time ./git2graph -n >(zstdmt > nodes.csv.zst) -e >(zstdmt -c > edges.csv.zst) /srv/src/linux
+    160,38s user 12,72s system 98% cpu 2:55,02 total
     
-    $ zcat nodes.csv.gz | wc -l
+    $ zstdcat nodes.csv.zst | wc -l
     6503402
-    $ zcat edges.csv.gz | wc -l
+    $ zstdcat edges.csv.zst | wc -l
     305095437
 
 
 Parallel use
 ------------
 
 `git2graph` writes fixed-length lines, long either 51 bytes (nodes) or 102
 bytes (edges). When writing to a FIFO less than `PIPE_BUF` bytes (which is 4096
 bytes on Linux, and guaranteed to be at least 512 bytes by POSIX), writes are
 atomic. Hence it is possible to mass analyze many repositories in parallel with
 something like:
 
     $ mkfifo nodes.fifo edges.fifo
-    $ sort -u < nodes.fifo | pigz -c > nodes.csv.gz &
-    $ sort -u < edges.fifo | pigz -c > edges.csv.gz &
+    $ sort -u < nodes.fifo | zstdmt > nodes.csv.zst &
+    $ sort -u < edges.fifo | zstdmt > edges.csv.zst &
     $ parallel git2graph -n nodes.fifo -e edges.fifo -- repo_dir_1 repo_dir_2 ...
     $ rm nodes.fifo edges.fifo
 
 Note that you most likely want to tune `sort` in order to be parallel
 (`--parallel`), use a large buffer size (`-S`), and use a temporary directory
 with enough available space (`-T`).  (The above example uses `parallel`
 from [moreutils](https://joeyh.name/code/moreutils/), but it could trivially be
 adapted to use [GNU parallel](https://www.gnu.org/software/parallel/) or
 similar parallelization tools.)
 
 
 Limitations
 -----------
 
 Snapshot PID calculation is not fully compatible with
 the
 [spec](https://docs.softwareheritage.org/devel/apidoc/swh.model.html#swh.model.identifiers.snapshot_identifier),
 because currently only HEAD is considered as a symbolic reference. Other
 symbolic refs, if present, will be ignored, ultimately leading to a different
 snapshot PID than what Software Heritage will obtain. The graph structure is
 not affected, but looking up obtained snapshots by PID on the main Software
 Heritage archive might fail.