Isn't it possible the processor may be idle for a second or two BEFORE
you start your compile? (While you're typing on the command line?) And
that during that time, it can precharge a zero-page cache which,
admittedly, gets used up pretty quickly at the start of the compile.
BUT is likely to completely overwhelm the extra couple of assembly
instructions to check if the cache has anything in it during a normal
alloc.
And I'd think the argument about using up in-cpu instruction cache with
those extra instructions would be lost in compiler optimizations and
page boundary issues, but I bow easily before anybody else's knowledge
there...
Personally, I think you're nit-picking... Now if you brought up memory
bus bandwidth issues on SMP machines, then maybe you'd have a point I
could follow. But since we have to do page zeroing ANYWAY, and are just
moving it to a place where the resources consumed are cheaper... And if
the idle task continues for a while the page cache gets as big as we
want to make it and the idle task goes back to twiddling its thumbs...
Rob
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/