For a while now I’ve had a task on my plate to enable jemalloc for Thunderbird on Windows builds, after confirming it doesn’t regress performance.
A few weeks ago, I managed to finish writing a start-up performance test (although still run manually) for Thunderbird, and with thanks to a couple of volunteers, start-up speed showed a reduction of about 2-3% for a warm start.
Our bloat tests (which are currently quite simple) showed slightly reduced allocations. Although not as definitive as I would have liked, knowing the improvements that jemalloc helped make with memory management on Firefox it should be a great bonus to Thunderbird. I think the main reason it didn’t show much of an improvement is that our bloat tests are currently quite limited and short-running; Over a longer session a bigger improvement should be seen.
We finally pushed the patch to enable jemalloc last Friday, so Windows nightly users should already be running with it, and it will be included in Thunderbird 3 beta 2. This should be (at least) a small performance bonus for Windows users.