afonso360 opened PR #5201 from fuzz-more-ops-2
to main
:
:wave: Hey,
This adds
bitselect
,select
andselect_spectre_guard
to the fuzzgen list of ops.Fuzzing on x86 & AArch64 for the past couple of hours has not raised any issues.
jameysharp submitted PR review.
jameysharp created PR review comment:
Could you add links to #5199 and #5200 here?
jameysharp created PR review comment:
I'm a little hesitant to add opcodes to this list where we have to disable all versions of the opcode on all of the architectures that we're actually fuzzing on. I suppose we could add an
insert_select_spectre_guard
function that generates anicmp
first. It's not great in the long run, since I like your idea of eventually generating this table from the meta instruction definitions. But it would get us some more fuzz test coverage quickly, at least.
jameysharp created PR review comment:
Based on your report in #5200, shouldn't these
select.i128
opcodes exclude aarch64 as well?
jameysharp submitted PR review.
jameysharp created PR review comment:
Could you add links to #5197 and #5198 here?
afonso360 updated PR #5201 from fuzz-more-ops-2
to main
.
afonso360 submitted PR review.
afonso360 created PR review comment:
Yeah, I don't mind adding that right now.
I didn't open issues for
select_spectre_guard
since I wasn't sure, but are we planning on implementing that? Or does that opcode only make sense when paired with something else?
afonso360 edited PR review comment.
jameysharp submitted PR review.
jameysharp created PR review comment:
That's a good question! I've opened #5206 for it.
afonso360 updated PR #5201 from fuzz-more-ops-2
to main
.
afonso360 submitted PR review.
afonso360 created PR review comment:
:+1: Thanks!
I've added the special
insert_select_spectre_guard
, and surprisingly this works for all variations ofselect_spectre_guard
.I'll wait for a decision to be reached on #5206 before opening the "missing lowerings" issues.
jameysharp submitted PR review.
jameysharp merged PR #5201.
Last updated: Dec 23 2024 at 13:07 UTC