Re: [PATCH zero-call-used-regs] Add leafy mode for zero-call-used-regs

[Alexandre Oliva <oliva@adacore.com>]

Hello, Qing,

It was a pleasure to meet you at the Cauldron.


On Oct 21, 2022, Qing Zhao <qing.zhao@oracle.com> wrote:

> Hi, Alexandre,
> Could you please explain a little bit on the motivation of this patch first?

It was a suggestion I got after the Cauldron presentation.
It made sense to me, and was easy enough to implement.

'all' for leaf functions is likely wasteful.  If no other functions are
called, one can determine exactly which registers might carry
information out and thus need zeroing, and 'used' is thus likely enough,
depending on the purpose of register scrubbing.  (In some scenarios, it
might make sense to want scrubbing of all registers, even unused ones
that carry incoming values)

Though some functions are coded as leaf functions, others may become
leaf functions because of inlining or other optimizations.  It's hard
for users to predict, so it makes sense to have a mode that tells the
compiler to figure it out.


There's room for a follow-up improvement, to save on a little more
potentially-wasteful anti-leaking scrubbing even in non-leaf functions:
for this purpose, they need not scrub registers that they don't use
themselves, if all potential callees are known to have scrubbed them.

I have not (yet?) implemented this variant; I haven't even found a name
I'm happy with for it.  (seal?  plug?  cork?  another leak antonym?)

I'm not entirely happy with leafy either, FWIW.  Bikeshedding anyone? :-)

https://gcc.gnu.org/pipermail/gcc-patches/2022-October/604083.html

-- 
Alexandre Oliva, happy hacker                https://FSFLA.org/blogs/lxo/
   Free Software Activist                       GNU Toolchain Engineer
Disinformation flourishes because many people care deeply about injustice
but very few check the facts.  Ask me about <https://stallmansupport.org>