Re: [PATCH] x86emul: Cache the amd_like() output on x86_emulate() entry

[Jan Beulich <jbeulich@suse.com>]

On 21.09.2026 11:25, Alejandro Vallejo wrote:
> The current code instantiates amd_like() way too many times, leading to
> codegen explosion. Unconditionally call it early on, and use that
> variable everywhere. This shrinks the emulator by ~3KiB.
> 
> No a functional change.
> 
> Signed-off-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
> ---
> pipeline: https://gitlab.com/xen-project/people/agvallejo/xen/-/pipelines/2860748404
>           (it's red because of an caching issue and personal branch name
>            conventions it's just arm. x86 passes in full)
> 
> bloat-o-meter before-after patch.
> 
> add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-3366 (-3366)
> Function                                     old     new   delta
> x86_emulate                               209351  205985   -3366
> Total: Before=3617694, After=3614328, chg -0.09%

That's surprisingly big a difference, considering how simple amd_like() is.
I'm counting 24 instances, so the savings would be well over 100 bytes per
instance.

That said - there are more uses, and I wonder why you ...

> @@ -1310,6 +1310,7 @@ x86_emulate(
>      /* Shadow copy of register state. Committed on successful emulation. */
>      struct cpu_user_regs _regs = *ctxt->regs;
>      const struct cpu_policy *__maybe_unused cp = ctxt->cpu_policy;
> +    bool is_amd_like = amd_like(ctxt);
>      struct x86_emulate_state state;

... make a standalone local variable here when you could introduce a new
field in struct x86_emulate_state, which would then filled early in
x86emul_decode(), and which could then be used elsewhere as well. With
that amd_like() could then likely also move from private.h to decode.c;
better yet - perhaps the function then wouldn't be needed at all anymore.

Another way to reduce the overhead at least some would be to make more use
of _amd_like(), now that x86_emulate() has a "cp" local variable.

Jan