Re: -fno-strength-reduce

Richard Henderson (rth@twiddle.net)
Fri, 22 Jan 1999 19:38:25 -0800

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: Richard B. Johnson: "Re: Setting terminal baud-rate."
Previous message: David: "Re: Scheduler set to 1000Hz"
In reply to: Gerard Roudier: "Re: Should raw I/O be added to the kernel?"

On Sat, Jan 23, 1999 at 01:33:05AM +0100, Marc Espie wrote:
> I've been happily surprised to see that egcs-current actually outputs
> better code than 2.8.1 on x86 in many situations, so maybe it also does
> a better job of deciding when to apply strength reduction on
> SMALL REGISTER CLASSES ?

EGCS 1.1.1 does a better job of strength reduction in general;
previously we would create many artificially independant induction
variables. It has nothing to do with S_R_C except that they are
most likely to lose.

However: on x86 scaled index addressing is nearly free, and if use
can be made of it, we may be able to save registers.

-
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/

Next message: Richard B. Johnson: "Re: Setting terminal baud-rate."
Previous message: David: "Re: Scheduler set to 1000Hz"
In reply to: Gerard Roudier: "Re: Should raw I/O be added to the kernel?"