On Friday 09 February 2007 10:43, David Miller wrote:
> Current gcc does the right thing, even for weird sizes like 56 and 52
> which expands to many IALU operations.
... except if you use -Os, which at least on x86* is default and is what
distros ship with.
-Andi
-
To unsubscribe from this lis
From: Eric Dumazet <[EMAIL PROTECTED]>
Date: Fri, 9 Feb 2007 10:36:58 +0100
> Thats strange, because pointer arithmetic is unsigned...
> I dont know when gcc started to use reciprocal division, maybe your gcc was
> very old ?
Yep, it was only on older gcc's.
And as the sparc gcc backend co-main
On Friday 09 February 2007 10:15, David Miller wrote:
> From: Eric Dumazet <[EMAIL PROTECTED]>
> Date: Fri, 9 Feb 2007 10:06:24 +0100
>
> > Yes, but a decent C compiler for such targets should not use a
> > multiply instruction to perform a (idx * 12) operation... :)
>
> Good point.
>
> Actually, I
From: Eric Dumazet <[EMAIL PROTECTED]>
Date: Fri, 9 Feb 2007 10:06:24 +0100
> Yes, but a decent C compiler for such targets should not use a
> multiply instruction to perform a (idx * 12) operation... :)
Good point.
Actually, I could never get GCC to avoid a divide on sparc64 for
certain kinds o
On Friday 09 February 2007 09:40, David Miller wrote:
> From: Andi Kleen <[EMAIL PROTECTED]>
> Date: 09 Feb 2007 10:18:03 +0100
>
> > David Miller <[EMAIL PROTECTED]> writes:
> > > I've applied this, but I _REALLY_ don't like the new multiply
> > > instructions that are used now in the hash indexin
On Friday 09 February 2007 09:40, David Miller wrote:
> From: Andi Kleen <[EMAIL PROTECTED]>
> Date: 09 Feb 2007 10:18:03 +0100
>
> > David Miller <[EMAIL PROTECTED]> writes:
> > >
> > > I've applied this, but I _REALLY_ don't like the new multiply
> > > instructions that are used now in the hash
From: Andi Kleen <[EMAIL PROTECTED]>
Date: 09 Feb 2007 10:18:03 +0100
> David Miller <[EMAIL PROTECTED]> writes:
> >
> > I've applied this, but I _REALLY_ don't like the new multiply
> > instructions that are used now in the hash indexing paths when
> > CONFIG_SMP is set.
> >
> > I think that's
David Miller <[EMAIL PROTECTED]> writes:
>
> I've applied this, but I _REALLY_ don't like the new multiply
> instructions that are used now in the hash indexing paths when
> CONFIG_SMP is set.
>
> I think that's a higher cost than the memory waste.
You're serious? multiply on a modern CPU is _mu
From: Eric Dumazet <[EMAIL PROTECTED]>
Date: Wed, 7 Feb 2007 11:59:34 +0100
> ehash table layout is currently this one :
>
> First half of this table is used by sockets not in TIME_WAIT state
> Second half of it is used by sockets in TIME_WAIT state.
>
> This is non optimal because of for a give
ehash table layout is currently this one :
First half of this table is used by sockets not in TIME_WAIT state
Second half of it is used by sockets in TIME_WAIT state.
This is non optimal because of for a given hash or socket, the two chain heads
are located in separate cache lines.
Moreover the
10 matches
Mail list logo
