XFS vs ReiserFS vs Ext3 comparisons abound using bonnie++, use-case analysis, etc.
The crazy performance numbers put up by XFS, like the following from Bart,
have me set to test XFS out the next time I rebuild my portage tree.
Ext3:
$ mkfs.ext3 /dev/sdb1 0.43s user 20.04s system 17% cpu 1:56.24 total
$ git clone quark.jukie.net:/scratch/linux-2.6/ 106.27s user 15.66s system 78% cpu 2:34.51 total
$ git checkout -f 0.84s user 0.19s system 91% cpu 1.136 total
$ make defconfig 2.49s user 0.86s system 33% cpu 10.053 total
$ make bzImage modules 566.28s user 55.95s system 66% cpu 15:33.44 total
$ make tags cscope 98.44s user 10.99s system 64% cpu 2:49.30 total
$ git clean -d -x -f 0.15s user 0.52s system 10% cpu 6.047 total
$ rm -i -rf linux-2.6 0.08s user 1.39s system 23% cpu 6.152 total
XFS:
$ mkfs.xfs -f /dev/sdb1 0.00s user 0.02s system 0% cpu 3.33s total
$ git clone quark.jukie.net:/scratch/linux-2.6/ 105.44s user 16.21s system 56% cpu 3:36.56 total
$ git checkout -f 0.32s user 0.14s system 55% cpu 0.826 total
$ make defconfig 2.06s user 0.73s system 25% cpu 10.755 total
$ make bzImage modules 127.72s user 31.05s system 71% cpu 3:40.74 total
$ make tags cscope 104.99s user 10.22s system 57% cpu 3:18.89 total
$ git clean -d -x -f 0.14s user 0.72s system 21% cpu 4.087 total
$ rm -i -rf linux-2.6 0.06s user 3.46s system 26% cpu 13.405 total