Messages in this thread | | | Date | Tue, 28 Sep 1999 20:38:21 +0200 (CEST) | From | Andrea Arcangeli <> | Subject | Re: possible spinlock optimizations |
| |
On Tue, 28 Sep 1999, Ingo Molnar wrote:
>better kernel. (he also might thus have less social life and a tiny bit >less happy gf - which again shows that things are relative ;)
8)8)
>hiding symptoms without fixing the real cause of the symptoms is bad.
Note that if a spinlock is held for too long time the guy is going to not scale anyway. Also he should not rely on the spin_lock_irq as he may need to convert the code to spin_lock_irqsave in the future.
Basically I was only thinking to improve the kernel and I didn't thought about the possible humans-side-effects of such irq latency improvement.
My only worry was if the cli to repeat after exiting the slow path will harm more performance than the gain of having irqs enabled. I believe enabling irqs will pay the cost of the additional cli to run after exiting the slow path.
If you are _only_ worried by the human-side-effect, then I can add a #define SPIN_LOCK_SLOW on the top of spinlocks.h that will be #unset only in 2.4.0 8).
Andrea
- 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/
| |