Re: Slow memory access, Not understood when using integer@(mem address)

[Benoît Minisini <benoit.minisini@xxxxxxxxxxxxxxxx>]

Le 25/09/2026 à 09:01, Benoît Minisini a écrit :
> I will see if I can optimize the signal event handler management, if 
> it's possible.
I'm afraid it's not possible to optimize that at the moment.

When you catch the illegal memory access, it's apparently not possible 
to tell Linux to handle it anyway as if the signal was not catched.

So I have to install and uninstall the signal handlers at each call of 
the 'xxx@' instructions, because I do want the illegal memory access to 
be raised in the rest of the program.

Anyway I made some test, and most of the time is spent in the 'setjmp()' 
API, so even if I don't install/uninstall the signal handlers, the 
instruction is still 30x slower than a simple variable access.

Regards,

--
Benoît Minisini.