From: Markus Elfring Sent: Tuesday, December 26, 2023
11:09 AM
>
> The kfree() function was called in two cases by
> the create_gpadl_header() function during error handling
> even if the passed variable contained a null pointer.
> This issue was detected by using the Coccinelle software.
>
> T
From: Saurabh Sengar Sent: Friday, January 5,
2024 2:29 AM
>
> The current method for signaling the compatibility of a Hyper-V host
> with MSIs featuring 15-bit APIC IDs relies on a synthetic cpuid leaf.
> However, for higher VTLs, this leaf is not reported, due to the absence
> of an IO-APIC.
>
From: Yury Norov Sent: Tuesday, January 9, 2024 3:29 PM
>
> Hi Michael,
>
> So, I'm just a guy who helped to formulate the heuristics in an
> itemized form, and implement them using the existing kernel API.
> I have no access to MANA machines and I ran no performance tests
> myself.
Agreed. :-)
Hi Michael,
So, I'm just a guy who helped to formulate the heuristics in an
itemized form, and implement them using the existing kernel API.
I have no access to MANA machines and I ran no performance tests
myself.
On Tue, Jan 09, 2024 at 07:22:38PM +, Michael Kelley wrote:
> From: Souradeep C
> -Original Message-
> From: Michael Kelley
> Sent: Tuesday, January 9, 2024 2:23 PM
> To: Souradeep Chakrabarti ; KY Srinivasan
> ; Haiyang Zhang ;
> wei@kernel.org; Dexuan Cui ;
> da...@davemloft.net; eduma...@google.com; k...@kernel.org;
> pab...@redhat.com; Long Li ; yury.no...@
From: Souradeep Chakrabarti Sent: Tuesday,
January 9, 2024 2:51 AM
>
> From: Yury Norov
>
> Souradeep investigated that the driver performs faster if IRQs are
> spread on CPUs with the following heuristics:
>
> 1. No more than one IRQ per CPU, if possible;
> 2. NUMA locality is the second pri
On Tue, 2024-01-09 at 02:51 -0800, Souradeep Chakrabarti wrote:
> This patch set introduces a new helper function irq_setup(),
> to optimize IRQ distribution for MANA network devices.
> The patch set makes the driver working 15% faster than
> with cpumask_local_spread().
>
> Souradeep Chakrabarti
Existing MANA design assigns IRQ to every CPU, including sibling
hyper-threads. This may cause multiple IRQs to be active simultaneously
in the same core and may reduce the network performance.
Improve the performance by assigning IRQ to non sibling CPUs in local
NUMA node. The performance improve
From: Yury Norov
Souradeep investigated that the driver performs faster if IRQs are
spread on CPUs with the following heuristics:
1. No more than one IRQ per CPU, if possible;
2. NUMA locality is the second priority;
3. Sibling dislocality is the last priority.
Let's consider this topology:
No
From: Yury Norov
Now we can simplify code that allocates cpumasks for local needs.
Signed-off-by: Yury Norov
---
include/linux/cpumask.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
index 228c23eb36d2..1c29947db848 100644
--- a/includ
From: Yury Norov
Similarly to cpumask_weight_and(), cpumask_weight_andnot() is a handy
helper that may help to avoid creating an intermediate mask just to
calculate number of bits that set in a 1st given mask, and clear in 2nd
one.
Signed-off-by: Yury Norov
Reviewed-by: Jacob Keller
---
inclu
This patch set introduces a new helper function irq_setup(),
to optimize IRQ distribution for MANA network devices.
The patch set makes the driver working 15% faster than
with cpumask_local_spread().
Souradeep Chakrabarti (1):
net: mana: Assigning IRQ affinity on HT cores
Yury Norov (3):
cpum
12 matches
Mail list logo
