v2.9.1

A small performance patch, released so that piscem-infer 0.8.0 can build on
it. Results are bit-for-bit identical to 2.9.0, and no index rebuild is
needed.

Speed

The EM and VBEM M-step that salmon-infer runs for each connected group of
transcripts is a little cheaper:

• Plain EM (--useEM, and piscem-infer's default): the step no longer
  bounds-checks every transcript it reads and writes, or tests every product
  for NaN.
  • Every transcript id in a component is checked once, when the components
    are laid out.
  • Because a NaN product always leaves a NaN in the step's output, that
    output is scanned once per step. In the (so far never seen) case that it
    holds one, the step is recomputed exactly as before.
• VBEM (the default): with a prior of at least 0.002 (the default is
  0.01), the step uses the same unchecked loop.

On the GENCODE v50 muscle sample (SRR13638690), time spent in the EM only:

	2.9.0	2.9.1
plain EM, 1 thread (1420 iterations)	6.62 s	6.18 s
plain EM, 8 threads	0.98 s	0.93 s
VBEM, 1 thread	1.013 s	0.968 s

In a full salmon quant run, the EM is a small share of the time.

Checked

• quant.sf is byte-identical to 2.9.0 on that sample with --sketch -p 8,
  both with the default VBEM and with --useEM, and is identical across
  thread counts.
• cargo test --release --workspace --locked: 383 passed. Clippy
  (-D warnings) and fmt are clean.

For library users

salmon-infer's public API is unchanged.

Full details: docs/release-notes-2.9.1.md.

---

Install salmon-cli 2.9.1

Install prebuilt binaries via shell script

curl --proto '=https' --tlsv1.2 -LsSf https://github.com/COMBINE-lab/salmon/releases/download/v2.9.1/salmon-cli-installer.sh | sh

Download salmon-cli 2.9.1

File	Platform	Checksum
salmon-cli-aarch64-apple-darwin.tar.xz	Apple Silicon macOS	checksum
salmon-cli-x86_64-apple-darwin.tar.xz	Intel macOS	checksum
salmon-cli-aarch64-unknown-linux-gnu.tar.xz	ARM64 Linux	checksum
salmon-cli-x86_64-unknown-linux-gnu.tar.xz	x64 Linux	checksum