BUILD: Current tracemonkey tip STEPS TO REPRODUCE: 1) Save https://bugzilla.mozilla.org/attachment.cgi?id=336658 2) Run it in the JS shell EXPECTED RESULTS: C-like speeds, or at least as fast as pure-interpreted ACTUAL RESULTS: 4x slower than pure-interpreted SEE ALSO: Bug 451673 comment 16. Once this is fixed, the testcase in bug 451673 comment 2 should be retested.
9 years ago
time for a retest?
Apparently. On current TM (both shell and browser) the jit version takes 0.8 seconds instead of the 15 it used to take. The non-jit version takes a bit over 3 seconds. So now we have a 4x speedup instead of a 4x slowdown. Is there any way we could create regression tests for this sort of thing?