On Fri, 5 Jan 2007, Andrew Morton wrote:
> Does this actually clean things up, or does it randomly move things around
> while carefully retaining existing obscurity? Not sure..
Looks like a good cleanup to me.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body
On Fri, 5 Jan 2007 13:46:45 +0200 (EET)
Pekka J Enberg <[EMAIL PROTECTED]> wrote:
> Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
> longer need to do NUMA_BUILD tricks and the UMA allocation path is much
> simpler. No functional changes in this patch.
>
> Note: saves few
On Fri, 5 Jan 2007, Pekka J Enberg wrote:
> From: Pekka Enberg <[EMAIL PROTECTED]>
>
> Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
> longer need to do NUMA_BUILD tricks and the UMA allocation path is much
> simpler. No functional changes in this patch.
Looks good.
-
From: Pekka Enberg <[EMAIL PROTECTED]>
Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
longer need to do NUMA_BUILD tricks and the UMA allocation path is much
simpler. No functional changes in this patch.
Note: saves few kernel text bytes on x86 NUMA build due to using
From: Pekka Enberg [EMAIL PROTECTED]
Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
longer need to do NUMA_BUILD tricks and the UMA allocation path is much
simpler. No functional changes in this patch.
Note: saves few kernel text bytes on x86 NUMA build due to using gotos
On Fri, 5 Jan 2007, Pekka J Enberg wrote:
From: Pekka Enberg [EMAIL PROTECTED]
Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
longer need to do NUMA_BUILD tricks and the UMA allocation path is much
simpler. No functional changes in this patch.
Looks good.
-
To
On Fri, 5 Jan 2007 13:46:45 +0200 (EET)
Pekka J Enberg [EMAIL PROTECTED] wrote:
Clean up __cache_alloc and __cache_alloc_node functions a bit. We no
longer need to do NUMA_BUILD tricks and the UMA allocation path is much
simpler. No functional changes in this patch.
Note: saves few kernel
On Fri, 5 Jan 2007, Andrew Morton wrote:
Does this actually clean things up, or does it randomly move things around
while carefully retaining existing obscurity? Not sure..
Looks like a good cleanup to me.
-
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of
On 1/4/07, Christoph Hellwig <[EMAIL PROTECTED]> wrote:
Seems to work nicely on my 2node cell blade.
Thanks for testing. Unfortunately as the other Christoph pointed out,
my patch reintroduces a bug that was fixed a while ago. kmalloc_node
should not be using mempolicies...
-
To unsubscribe
On Tue, Jan 02, 2007 at 03:47:06PM +0200, Pekka J Enberg wrote:
> [Andrew, I have been unable to find a NUMA-capable tester for this patch,
> so can we please put this in to -mm for some exposure?]
>
> From: Pekka Enberg <[EMAIL PROTECTED]>
>
> This patch cleans up __cache_alloc and
On Tue, Jan 02, 2007 at 03:47:06PM +0200, Pekka J Enberg wrote:
[Andrew, I have been unable to find a NUMA-capable tester for this patch,
so can we please put this in to -mm for some exposure?]
From: Pekka Enberg [EMAIL PROTECTED]
This patch cleans up __cache_alloc and __cache_alloc_node
On 1/4/07, Christoph Hellwig [EMAIL PROTECTED] wrote:
Seems to work nicely on my 2node cell blade.
Thanks for testing. Unfortunately as the other Christoph pointed out,
my patch reintroduces a bug that was fixed a while ago. kmalloc_node
should not be using mempolicies...
-
To unsubscribe from
On Tue, 2 Jan 2007, Pekka J Enberg wrote:
> +
> + if (nodeid == -1 || nodeid == numa_node_id()) {
> + if (unlikely(current->flags & (PF_SPREAD_SLAB | PF_MEMPOLICY)))
{
> + obj = alternate_node_alloc(cache, flags);
> + if (obj)
> +
On Tue, 2 Jan 2007 15:47:06 +0200 (EET)
Pekka J Enberg <[EMAIL PROTECTED]> wrote:
> I have been unable to find a NUMA-capable tester for this patch,
Any x86_64 box can be used to test NUMA code via the numa=fake=N boot option.
fake-numa is somewhat sick in mainline and you might find that it
On Tue, 2 Jan 2007, Pekka J Enberg wrote:
> +
> + if (nodeid == -1 || nodeid == numa_node_id()) {
> + if (unlikely(current->flags & (PF_SPREAD_SLAB | PF_MEMPOLICY)))
> {
> + obj = alternate_node_alloc(cache, flags);
> + if (obj)
> +
Pekka J Enberg wrote:
> [Andrew, I have been unable to find a NUMA-capable tester for this patch,
> so can we please put this in to -mm for some exposure?]
>
> From: Pekka Enberg <[EMAIL PROTECTED]>
>
> This patch cleans up __cache_alloc and __cache_alloc_node functions. We no
> longer need
[Andrew, I have been unable to find a NUMA-capable tester for this patch,
so can we please put this in to -mm for some exposure?]
From: Pekka Enberg <[EMAIL PROTECTED]>
This patch cleans up __cache_alloc and __cache_alloc_node functions. We no
longer need to do NUMA_BUILD tricks and the UMA
[Andrew, I have been unable to find a NUMA-capable tester for this patch,
so can we please put this in to -mm for some exposure?]
From: Pekka Enberg [EMAIL PROTECTED]
This patch cleans up __cache_alloc and __cache_alloc_node functions. We no
longer need to do NUMA_BUILD tricks and the UMA
Pekka J Enberg wrote:
[Andrew, I have been unable to find a NUMA-capable tester for this patch,
so can we please put this in to -mm for some exposure?]
From: Pekka Enberg [EMAIL PROTECTED]
This patch cleans up __cache_alloc and __cache_alloc_node functions. We no
longer need to do
On Tue, 2 Jan 2007, Pekka J Enberg wrote:
+
+ if (nodeid == -1 || nodeid == numa_node_id()) {
+ if (unlikely(current-flags (PF_SPREAD_SLAB | PF_MEMPOLICY)))
{
+ obj = alternate_node_alloc(cache, flags);
+ if (obj)
+
On Tue, 2 Jan 2007 15:47:06 +0200 (EET)
Pekka J Enberg [EMAIL PROTECTED] wrote:
I have been unable to find a NUMA-capable tester for this patch,
Any x86_64 box can be used to test NUMA code via the numa=fake=N boot option.
fake-numa is somewhat sick in mainline and you might find that it
On Tue, 2 Jan 2007, Pekka J Enberg wrote:
+
+ if (nodeid == -1 || nodeid == numa_node_id()) {
+ if (unlikely(current-flags (PF_SPREAD_SLAB | PF_MEMPOLICY)))
{
+ obj = alternate_node_alloc(cache, flags);
+ if (obj)
+
22 matches
Mail list logo