As njn noted, trace-tests on x64 shows: > Trace stats check failed: got 1, expected 3 for sideExitIntoInterpreter This is probably because we aren't doing integer speculation on division for x64 yet.
I'll shortly be posting a bug to improve the codegen for div/mod on i386; this bug should probably wait on that one.
Created attachment 401145 [details] [diff] [review] patch I tried to duplicate the logic of the x86 backend as of bug 516909. Passes trace-tests, including testEliminatedGuardWithinAnchor.
Assignee: general → dvander
Status: NEW → ASSIGNED
Attachment #401145 - Flags: review?(nnethercote)
Comment on attachment 401145 [details] [diff] [review] patch I assume the deletion of testbug507425.js isn't meant to be in there? r=me with that removed. Also, a nit: I like 'mod' better than 'ins' in asm_div_mod() but it's not terribly important.
Attachment #401145 - Flags: review?(nnethercote) → review+
Thanks, not sure show that got in there. I kept 'ins' since both divs and mods can flow into that function. http://hg.mozilla.org/tracemonkey/rev/8b7afba2b5bc
Status: ASSIGNED → RESOLVED
Last Resolved: 9 years ago
Resolution: --- → FIXED
status1.9.2: --- → beta1-fixed
You need to log in before you can comment on or make changes to this bug.