[Bug translator/23634] WARNING: Can't parse SDT_V3 operand for x86_64 qemu tapset probe points

["fche at redhat dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=23634

Frank Ch. Eigler <fche at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fche at redhat dot com

--- Comment #2 from Frank Ch. Eigler <fche at redhat dot com> ---
> Would there be a problem changing the default STAP_SDT_ARG_CONTRAINT from "nor" to "nr"?

Not so easy, unfortunately, or we would've done it already.  "nr" forces the
compiler to push nonliteral arguments into registers.  There may not be many
registers free at that point in the code, or in the processor generally, so
there are performance costs.

If we can satisfy a systemtap.apps testsuite run with "nr" on x86-64 (not so
register-starved an architecture), then for that one perhaps we can switch the
default, and force people to accept the cost.

The underlying problem here is that in gcc, there seems to be no code that
corresponds to "public symbol relative offset" or "register-relative offset",
both of which we could or do easily support in stap or other consumers.  It's
only this "local scratch symbol relative offset" case that's a loser.

-- 
You are receiving this mail because:
You are the assignee for the bug.