Hi Anton,
I am curious of what is the concurrency use case for the ByteQ lib. Because
currently we found some UT cases failure on some weak memory model CPU, like
arm and ppc.
Besides this question, I have a few comments inline.
Best Regards,
Wei Yanqin
> -Original Message-
> From: d
Hi Haifeng,
It looks indeed a risk for using ovs-rcu. A few comments inline.
Best Regards,
Wei Yanqin
> -Original Message-
> From: dev On Behalf Of Linhaifeng
> Sent: Monday, June 1, 2020 11:13 AM
> To: d...@openvswitch.org
> Subject: [ovs-dev] [PATCH] ovs rcu: update rcu pointer first
rformance for all kinds of traffic.
Reviewed-by: Lijian Zhang
Reviewed-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 14 +++---
lib/flow.c| 2 +-
lib/packets.h | 46 --
3 files changed, 52 insertions(+),
flows 40%
EMC disable 10%
Malvika Gupta (1):
[ovs-dev] dpif-netdev: Modify dfc_processing function to void function
Yanqin Wei (5):
netdev: avoid unnecessary packet batch refilling in netdev feature
check
dpif-netdev: add tunnel_valid flag to skip ip/ipv6 address comparison
(flow->ufid) in smc disabled case.
Reviewed-by: Lijian Zhang
Reviewed-by: Malvika Gupta
Reviewed-by: Lance Yang
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 23 +--
1 file changed, 13 insertions(+), 10 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netde
. Moreover, dp_packet_batch_is_empty() called after function
return from dfc_processing itself calls dp_packet_batch_size to check if
'packets' array is empty. This patch modifies the dfc_processing function to a
void function to remove the above code redundancy and cleans the code.
Reviewed-by: Yanqi
Zhang
Reviewed-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
lib/dp-packet.h | 12
lib/netdev.c| 13 ++---
2 files changed, 18 insertions(+), 7 deletions(-)
diff --git a/lib/dp-packet.h b/lib/dp-packet.h
index 0430cca8e..1345f46e7 100644
--- a/lib/dp-packet.h
+++ b/lib
: Malvika Gupta
Reviewed-by: Ruifeng Wang
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 55 +++
1 file changed, 32 insertions(+), 23 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index c94d5e8c7..3994f41e4 100644
--- a/lib/dpif
Key length is only useful for emc table. This patch moves the key length
calculation into emc_change_entry for fast path performance.
Reviewed-by: Lijian Zhang
Reviewed-by: Malvika Gupta
Reviewed-by: Ruifeng Wang
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 10 ++
1 file changed
mat:"CC: %an <%ae>%nFixes: %h (\"%s\")" \
--abbrev=12 COMMIT_REF"
Best Regards,
Wei Yanqin
> -Original Message-
> From: Linhaifeng
> Sent: Tuesday, June 2, 2020 3:13 PM
> To: Yanqin Wei ; d...@openvswitch.org
> Cc: nd
> Subject: RE: [P
> Sent: Wednesday, June 3, 2020 1:28 AM
> To: Linhaifeng
> Cc: Yanqin Wei ; d...@openvswitch.org; nd
> ; Lilijun (Jerry) ; chenchanghu
> ; Lichunhe
> Subject: Re: [ovs-dev] [PATCH v2] ovs rcu: update rcu pointer first
>
> On Tue, Jun 02, 2020 at 07:27:59AM +, Linhaifeng w
Hi Ben,
This patch is from Linhai, but I have the same concern about this. I will read
ovs-rcu comments and feedback.
Thanks for your time.
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ben Pfaff
> Sent: Wednesday, June 3, 2020 8:35 AM
> To: Yanqin Wei
> Cc:
original patch. Please help to correct.
https://patchwork.ozlabs.org/project/openvswitch/patch/4099de2e54afad489356c6c9161d53339735a...@dggeml522-mbs.china.huawei.com/
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ben Pfaff
> Sent: Wednesday, June 3, 2020 8:48 AM
> To: Y
] dpif-netdev: add tunnel_valid flag to skip ip/ipv6
address comparison
[ovs-dev][PATCH v1 3/6] dpif-netdev: improve emc lookup performance by
contiguous storage of hash value.
Any comments from anyone are appreciated.
Best Regards,
Wei Yanqin
> -Original Message-
> From: Yanq
Hi Shahaji,
Adding isb instruction can help rdtsc precise, which sync system counter to
cntvct_el0. There is a patch in DPDK. https://patchwork.dpdk.org/patch/66561/
So it may be not related with intermittent drops you observed.
Best Regards,
Wei Yanqin
> -Original Message-
> From: dev
:11 AM
To: Yanqin Wei
Cc: Flavio Leitner ; ovs-dev@openvswitch.org; nd
; Ilya Maximets ; Lee Reed
; Vinay Gupta ; Alex Barba
Subject: Re: [ovs-dev] 10-25 packet drops every few (10-50) seconds TCP (iperf3)
Hi Yanqin,
I added the patch you gave me to my script which runs a do nothing for loop
degradation of
multi-rings.
Best Regards,
Wei Yanqin
From: Shahaji Bhosle
Sent: Thursday, July 2, 2020 9:27 PM
To: Yanqin Wei
Cc: Flavio Leitner ; ovs-dev@openvswitch.org; nd
; Ilya Maximets ; Lee Reed
; Vinay Gupta ; Alex Barba
Subject: Re: [ovs-dev] 10-25 packet drops every few (10-50
Regards,
Wei Yanqin
>
> On Tue, Jun 30, 2020 at 2:26 AM Yanqin Wei wrote:
> >
> > Hi, every contributor
> >
> > These patches could significantly improve multi-flow throughput of
> userspace datapath. If you feel it will take too much time to review all
> patche
Hi William,
>
> On Tue, Jun 2, 2020 at 12:10 AM Yanqin Wei wrote:
> >
> > OVS userspace datapath is a program with heavy memory access. It needs
> > to load/store a large number of memory, including packet header,
> > metadata, EMC/SMC/DPCLS tables and so on. It
Hi Ilya,
> > ---
>
> Hi.
> First of all, thanks for working on performance improvements!
Thanks, I saw some slides where OVS was used to compare flow scalability with
other projects. It inspired me to optimize this code.
>
> However, this doesn't look as a clean patch.
There are some trade-off
1000LL
Would you like to disable it if it is not necessary?
Best Regards,
Wei Yanqin
From: Shahaji Bhosle
Sent: Monday, July 6, 2020 8:24 PM
To: Yanqin Wei
Cc: Flavio Leitner ; ovs-dev@openvswitch.org; nd
; Ilya Maximets ; Lee Reed
; Vinay Gupta ; Alex Barba
Subject: Re: [ovs-dev] 10-25 packet
, 2020 12:23 PM
To: Yanqin Wei
Cc: Flavio Leitner ; ovs-dev@openvswitch.org; nd
; Ilya Maximets ; Lee Reed
; Vinay Gupta ; Alex Barba
Subject: Re: [ovs-dev] 10-25 packet drops every few (10-50) seconds TCP (iperf3)
Thanks Yangin,
What does this define mean? Every 10 second some kind of book
The 2nd one is another periodic task for dpcls ranking.
From: Yanqin Wei
Sent: Tuesday, July 7, 2020 1:19 PM
To: Shahaji Bhosle
Cc: Flavio Leitner ; ovs-dev@openvswitch.org; nd
; Ilya Maximets ; Lee Reed
; Vinay Gupta ; Alex Barba
Subject: RE: [ovs-dev] 10-25 packet drops every few (10-50
Hi Harry,
> >
> > OVS userspace datapath is a program with heavy memory access. It needs
> > to load/store a large number of memory, including packet header,
> > metadata, EMC/SMC/DPCLS tables and so on. It causes a lot of cache
> > line missing and refilling, which has a great impact on flow
> >
Hi Ilya,
> >
> >>> ---
> >>
> >> Hi.
> >> First of all, thanks for working on performance improvements!
> > Thanks, I saw some slides where OVS was used to compare flow scalability
> with other projects. It inspired me to optimize this code.
> >
> >>
> >> However, this doesn't look as a clean patc
rm. It was
observed that 50% performance improvement in thunderX2 and 40% improvement
in TaiShan(Cortex-A72).
Signed-off-by: Yanqin Wei
---
lib/hash.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/hash.c b/lib/hash.c
index c64f25e..06f8339 100644
--- a/lib/hash.c
+++ b/
need 2 cache lines in most modern CPU.
Performance test was run in some arm platform. 1000/1 flows NIC2NIC
test achieved around 10% throughput improvement in thunderX2(aarch64
platform).
Signed-off-by: Yanqin Wei
Reviewed-by: Gavin Hu
---
lib/dpif-netdev.c | 81
need 2 cache lines in most modern CPU.
Performance test was run in some arm platform. 1000/1 flows NIC2NIC
test achieved around 10% throughput improvement in thunderX2(aarch64
platform).
Signed-off-by: Yanqin Wei
Reviewed-by: Gavin Hu
---
lib/dpif-netdev.c | 80
need 2 cache lines in most modern CPU.
Performance test was run in some arm platform. 1000/1 flows NIC2NIC
test achieved around 10% throughput improvement in thunderX2(aarch64
platform).
Signed-off-by: Yanqin Wei
Reviewed-by: Gavin Hu
---
lib/dpif-netdev.c | 80
Yanqin
> -Original Message-
> From: Lance Yang (Arm Technology China)
> Sent: Tuesday, December 17, 2019 9:54 AM
> To: Ilya Maximets ; dwilder
> Cc: ovs-dev@openvswitch.org; b...@ovn.org; Yanqin Wei (Arm Technology
> China) ; Gavin Hu (Arm Technology China)
>
Yanqin
> -Original Message-
> From: Lance Yang
> Sent: Tuesday, January 21, 2020 9:06 AM
> To: Ilya Maximets ; ovs-dev@openvswitch.org
> Cc: b...@ovn.org; Yanqin Wei ; dwil...@us.ibm.com;
> Gavin Hu ; Ruifeng Wang ;
> Jieqiang Wang ; Malvika Gupta
> ; nd
> Subje
Hi Ben,
There is a patch to enable travis CI on Arm. Is it possible to be put into 2.13?
https://patchwork.ozlabs.org/patch/1204923/
Best Regards,
Wei Yanqin
> -Original Message-
> From: dev On Behalf Of Ben Pfaff
> Sent: Tuesday, February 11, 2020 1:44 AM
> To: d...@openvswitch.org
>
, we hope it can be enabled soon to prevent new issue introduced during
development.
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ben Pfaff
> Sent: Thursday, February 13, 2020 12:23 AM
> To: Ilya Maximets
> Cc: Yanqin Wei ; d...@openvswitch.org; nd
> ; Malvika
: Lijian Zhang
Signed-off-by: Yanqin Wei
---
lib/netdev-dummy.c| 5 -
lib/netdev-offload.c | 11 +++
lib/netdev-provider.h | 1 +
lib/netdev.c | 1 +
4 files changed, 17 insertions(+), 1 deletion(-)
diff --git a/lib/netdev-dummy.c b/lib/netdev-dummy.c
index 71df29184
this is
implemented by one-way barrier instead of two-way memory fence.
Reviewed-by: Gavin Hu
Reviewed-by: Lijian Zhang
Signed-off-by: Yanqin Wei
---
lib/pvector.c | 14 +++---
lib/pvector.h | 12 +++-
2 files changed, 14 insertions(+), 12 deletions(-)
diff --git a/lib/pvector.c b
Hi Ilya,
> -Original Message-
> From: Ilya Maximets
> Sent: Tuesday, February 25, 2020 5:25 PM
> To: Yanqin Wei ; d...@openvswitch.org
> Cc: nd ; Lijian Zhang ; Gavin Hu
> ; i.maxim...@ovn.org
> Subject: Re: [ovs-dev] [PATCH v1] netdev: fix partial offloading
Hi Ilya,
> -Original Message-
> From: Ilya Maximets
> Sent: Tuesday, February 25, 2020 5:54 PM
> To: Yanqin Wei ; d...@openvswitch.org
> Cc: Lijian Zhang ; Gavin Hu ; nd
> ; i.maxim...@ovn.org; Ben Pfaff
> Subject: Re: [ovs-dev] [PATCH v1] lib: use acquire-
> -Original Message-
> From: Ilya Maximets
> Sent: Tuesday, February 25, 2020 6:31 PM
> To: Yanqin Wei ; Ilya Maximets
> ; d...@openvswitch.org
> Cc: nd ; Lijian Zhang ; Gavin Hu
>
> Subject: Re: [ovs-dev] [PATCH v1] netdev: fix partial offloading test cases
&
this is
implemented by one-way barrier instead of two-way memory fence.
Reviewed-by: Gavin Hu
Reviewed-by: Lijian Zhang
Signed-off-by: Yanqin Wei
---
lib/pvector.c | 16 +---
lib/pvector.h | 10 ++
2 files changed, 15 insertions(+), 11 deletions(-)
diff --git a/lib/pvector.c b
his patch.
Suggested-by: Ilya Maximets
Reviewed-by: Gavin Hu
Reviewed-by: Lijian Zhang
Signed-off-by: Yanqin Wei
---
tests/dpif-netdev.at | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tests/dpif-netdev.at b/tests/dpif-netdev.at
index 0aeb4e788..12e468744 100644
--- a/te
this is
implemented by one-way barrier instead of two-way memory fence.
Reviewed-by: Gavin Hu
Reviewed-by: Lijian Zhang
Signed-off-by: Yanqin Wei
---
lib/pvector.c | 18 +++---
lib/pvector.h | 12
2 files changed, 19 insertions(+), 11 deletions(-)
diff --git a/lib
Hi Ilya,
V3 has been updated based on your comments. Thanks.
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ilya Maximets
> Sent: Wednesday, February 26, 2020 9:41 PM
> To: Yanqin Wei ; d...@openvswitch.org
> Cc: nd ; Lijian Zhang ; Gavin Hu
> ; i.maxim...@ovn
this is
implemented by one-way barrier instead of two-way memory fence.
Fixes: fe7cfa5c3f19 ("lib/pvector: Non-intrusive RCU priority vector.")
Reviewed-by: Gavin Hu
Reviewed-by: Lijian Zhang
Signed-off-by: Yanqin Wei
---
lib/pvector.c | 18 +++---
lib/pvector.h | 13 ++
Hi Ilya,
Thanks for your comments. V4 has been updated, could you please check it again?
https://patchwork.ozlabs.org/patch/1248440/
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ilya Maximets
> Sent: Friday, February 28, 2020 9:40 PM
> To: Yanqin Wei ; d...@ope
; From: Lance Yang
> Sent: Tuesday, January 21, 2020 9:06 AM
> To: Ilya Maximets ; ovs-dev@openvswitch.org
> Cc: b...@ovn.org; Yanqin Wei ; dwil...@us.ibm.com;
> Gavin Hu ; Ruifeng Wang ;
> Jieqiang Wang ; Malvika Gupta
> ; nd
> Subject: RE: [ovs-dev] [PATCH v2 3/3] travis: E
Cmap_find is a hot spot in partial offloading datapath. This patch applies
sda-table for mark2flow lookup. The throughput uplift is more than 10% in
case of single flow and 20% with 1000 mega flows.
Reviewed-by: Gavin Hu
Reviewed-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
lib/dpif
eviewed-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
lib/automake.mk | 2 +
lib/sda-table.c | 166
lib/sda-table.h | 127
3 files changed, 295 insertions(+)
create mode 100644 lib/sda-table.c
create mode 1006
Add check test for sda-table lookup, insertion and deletion.
Reviewed-by: Gavin Hu
Reviewed-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
tests/automake.mk | 3 +-
tests/test-sda-table.c | 197 +
2 files changed, 199 insertions(+), 1 deletion
ookup. The
throughput uplift is more than 10% in case of single flow and 20% with
>1000 mega flow.
Yanqin Wei (3):
lib: implement scalable direct address table for fast lookup
dpif-netdev: improve partial offloding datapath by sda-table lookup
tests/test-sda-table: add check test for sda
Thanks for the feedback. Replied in line.
> -Original Message-
> From: Ilya Maximets
> Sent: Thursday, March 12, 2020 12:20 AM
> To: Yanqin Wei ; Ilya Maximets
> ; ovs-dev@openvswitch.org; b...@ovn.org; Lance Yang
>
> Cc: dwil...@us.ibm.com; Gavin Hu ; Ruifeng W
Hi Xiangxia,
The meter id is allocated by id_pool, which can always return the lowest
available id.
There is a light scalable data struct which supports direct address lookup. It
can achieve several times performance improvement than cmap_find. And it has
not copy memory when expanding.
https:
Hi Ilya,
The patch is updated for reducing Arm CI running time. Do you think if the
time increment is acceptable?
https://patchwork.ozlabs.org/patch/1258100/
Best Regards,
Wei Yanqin
> -Original Message-
> From: Yanqin Wei
> Sent: Thursday, March 12, 2020 5:39 PM
> To: I
1. add thread fence to ensure hash setting safe.
2. make reader not blocked by writer.
Yanqin Wei (2):
[ovs-dev] [PATCH 1/2] cmap: add thread fence for slot update
[ovs-dev] [PATCH 2/2] cmap: non-blocking cmap lookup
lib/cmap.c | 159 ++---
1
bucket update in the cmap lib is protected by a counter. But hash setting
is possible to be moved before counter update. This patch fix this issue.
Reviewed-by: Ola Liljedahl
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/cmap.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion
se counter and bitmap
share one 32 bits variable. No additional load/store for reader and
writer.
Reviewed-by: Ola Liljedahl
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/cmap.c | 157 ++---
1 file changed, 101 insertions(+), 56 deletion
"rxq_enabled" of netdev is writen in the vhost thread and read by pmd
thread once it observes 'change_seq' is updated. This patch is to keep
order on aarch64 or other weak memory model CPU to ensure 'rxq_enabled' is
observed before 'change_seq'.
Reviewed
"rxq_enabled" of netdev is writen in the vhost thread and read by pmd
thread once it observes 'change_seq' is updated. This patch is to keep
order on aarch64 or other weak memory model CPU to ensure 'rxq_enabled' is
observed before 'change_seq'.
Reviewed
only under the medium number of flows scenario.
Yanqin Wei (2):
dpif-netdev: add EMC entries counter
dpif-netdev: dfc_process performance optimization by prefetching EMC
entry.
lib/dpif-netdev.c | 88 +++
1 file changed, 57 insertions
Implement entries counter for EMC. It could be used for improvement for
EMC lookup.
Signed-off-by: Yanqin Wei
Reviewed-by: Gavin Hu
---
lib/dpif-netdev.c | 22 ++
1 file changed, 14 insertions(+), 8 deletions(-)
diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
index
ent in 100k flows RFC2544 test. And low number
of flows has almost no performance impact.
Signed-off-by: Yanqin Wei
Reviewed-by: Gavin Hu
---
lib/dpif-netdev.c | 66 ---
1 file changed, 43 insertions(+), 23 deletions(-)
mode change 100644 =>
s run in two aarch64 machines. In the NIC2NIC test, one
tuple dpcls lookup case achieves around 4% throughput improvement and
10(average) tuples case achieves around 5% improvement.
Tested-by: Malvika Gupta
Signed-off-by: Yanqin Wei
---
lib/util.h | 7 ++-
1 file changed, 6 insertions(+)
P2P forwarding cases.
Change-Id: I5d606afb52bfa68e8afa6f886d69b9665cdad51a
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 13 +++---
lib/flow.c| 116 --
lib/flow.h| 2 +
3 files changed, 79 inserti
The padding length is (packet size - ipv6 header length - ipv6 plen). This
patch fixes incorrect ipv6 size checking and improves it by combining branch.
Change-Id: I7a6212872f89ef9c11e3fbcec4dbecbcc6e89c47
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/flow.c | 10 --
1 file
: I6d503c904d0354c94882196d7720a574b2d07e44
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/flow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/flow.c b/lib/flow.c
index 1b21f51..4d895e5 100644
--- a/lib/flow.c
+++ b/lib/flow.c
@@ -343,7 +343,6 @@ parse_vlan
P2P forwarding cases.
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/dpif-netdev.c | 13 +++---
lib/flow.c| 116 --
lib/flow.h| 2 +
3 files changed, 79 insertions(+), 52 deletions(-)
diff --git a/lib/dpif-netdev.c
The padding length is (packet size - ipv6 header length - ipv6 plen). This
patch fixes incorrect ipv6 size checking and improves it by combining branch.
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/flow.c | 10 --
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git
-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/flow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/flow.c b/lib/flow.c
index 1b21f51..4d895e5 100644
--- a/lib/flow.c
+++ b/lib/flow.c
@@ -343,7 +343,6 @@ parse_vlan(const void **datap, size_t *sizep, union
flow_vlan_hdr
The padding length is (packet size - ipv6 header length - ipv6 plen). This
patch fixes incorrect padding size checking in ipv6_sanity_check.
Reviewed-by: Gavin Hu
Signed-off-by: Yanqin Wei
---
lib/flow.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/flow.c b/lib
.
Test bed: aarch64(Centriq 2400) platform.
Test case: DPCLS forwarding(disable EMC + avg 10 subtable lookups)
Test result: improve around 10%.
Signed-off-by: Yanqin Wei
---
lib/automake.mk| 1 +
lib/hash-aarch64.h | 151 +
m. It was
observed that 50% performance improvement in thunderX2 and 40%
improvement in TaiShan(Cortex-A72).
Signed-off-by: Yanqin Wei
---
lib/hash.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/lib/hash.c b/lib/hash.c
index c64f25e..06f8339 100644
--- a/lib/hash.c
+++ b/
Hi Ben,
Sorry for the format issue. I will resubmit it by "git send-email".
Best Regards,
Wei Yanqin
-Original Message-
From: Ben Pfaff
Sent: Saturday, February 23, 2019 5:03 AM
To: Yanqin Wei (Arm Technology China)
Cc: d...@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v1
Thanks for the comments. It will be fixed in the following patch v3.
-Original Message-
From: Ilya Maximets
Sent: Tuesday, March 12, 2019 9:31 PM
To: Yanqin Wei (Arm Technology China) ; d...@openvswitch.org
Cc: nd ; Gavin Hu (Arm Technology China)
Subject: Re: [ovs-dev,v2] netdev-dpdk
Hi , OVS Maintainers,
Could you help to have a look at this patch? Thanks a lot.
Best Regards,
Wei Yanqin
-Original Message-
From: Yanqin Wei
Sent: Wednesday, March 13, 2019 1:28 PM
To: d...@openvswitch.org
Cc: nd ; Gavin Hu (Arm Technology China) ;
Yanqin Wei (Arm Technology China
From: Ian Stokes
Sent: Monday, March 25, 2019 6:16 AM
To: Yanqin Wei (Arm Technology China) ; d...@openvswitch.org
Cc: nd ; Gavin Hu (Arm Technology China) ; Ilya
Maximets
Subject: Re: [ovs-dev] [PATCH v3] dpif-netdev: dfc_process optimization by
prefetching EMC entry.
On 3/13/2019 5:27 AM, Y
even in worst case(EMC/SMC disable & no hash-feature/load
balance enable).
Best Regards,
Wei Yanqin
-Original Message-
From: Ilya Maximets
Sent: Friday, March 22, 2019 9:12 PM
To: Yanqin Wei (Arm Technology China) ;
d...@openvswitch.org; ian.sto...@intel.com
Cc: nd
Subject: Re:
will try to improve it . For 1-8 flows
cases, it should be not very important because of the lack of actual deployment.
Best Regards,
Wei Yanqin
-Original Message-
From: Ilya Maximets
Sent: Monday, April 1, 2019 2:16 PM
To: Yanqin Wei (Arm Technology China) ;
d...@openvswitch.org
avoided by configuration.
Is my understanding correct?
Best Regards,
Wei Yanqin
-Original Message-
From: Ilya Maximets
Sent: Monday, April 1, 2019 3:15 PM
To: Yanqin Wei (Arm Technology China) ;
d...@openvswitch.org; ian.sto...@intel.com
Cc: nd ; Gavin Hu (Arm Technology China)
Subject
Hi,
We are working to support arm64 build for ovs travis CI. It is indeed to use
arch: arm64 to choose cpu architecture, because travis has provided native
arm64 option now.
But in this patch it seems ppc64 builds run on the ppc-VM + x86 native machine.
Currently arm only select a part of jo
Yanqin
-Original Message-
From: dwilder
Sent: Saturday, November 2, 2019 1:09 AM
To: Yanqin Wei (Arm Technology China)
Cc: Ilya Maximets ; ovs-dev@openvswitch.org;
wil...@us.ibm.com; nd
Subject: Re: RE: [ovs-dev] [PATCH ] travis: support ppc64le builds
On 2019-10-30 19:04, Yanqin Wei
Hi David,
I am sorry to reply this late. Yes, if travis support this configuration, it
should be a good solution for me.
Best regards,
Wei Yanqin
-Original Message-
From: dwilder
Sent: Wednesday, November 6, 2019 4:32 AM
To: Yanqin Wei (Arm Technology China)
Cc: Ilya Maximets ; ovs
Hi David
> -Original Message-
> From: David Wilder
> Sent: Thursday, November 7, 2019 3:21 AM
> To: ovs-dev@openvswitch.org
> Cc: i.maxim...@ovn.org; b...@ovn.org; Yanqin Wei (Arm Technology China)
> ; wil...@us.ibm.com
> Subject: [PATCH v3] travis: support ppc64le b
e found in anther candidate bucket.
Best Regards,
Wei Yanqin
> -Original Message-
> From: Ilya Maximets
> Sent: Tuesday, November 19, 2019 12:06 AM
> To: Ola Liljedahl ; i.maxim...@ovn.org; Yanqin Wei
> (Arm Technology China) ; ovs-dev@openvswitch.org
> Cc: Gavin Hu (Arm
Hi Ben,
This issue is difficult to reproduce in local machine. We will try to do it.
Could you suggest the log we need to collect or enable?
Best Regards,
Wei Yanqin
> -Original Message-
> From: dev On Behalf Of Ben Pfaff
> Sent: Friday, November 22, 2019 1:34 AM
> To: Lance Yang (Arm
Hi Ilya,
Reply inline.
Best Regards,
Wei Yanqin
> -Original Message-
> From: dev On Behalf Of Ilya Maximets
> Sent: Friday, November 22, 2019 3:30 AM
> To: Lance Yang (Arm Technology China) ;
> d...@openvswitch.org; ovs-dev@openvswitch.org
> Cc: Jieqiang Wang (Arm Technology China) ;
>
age-
> From: Ben Pfaff
> Sent: Tuesday, November 26, 2019 5:54 AM
> To: Yanqin Wei (Arm Technology China)
> Cc: d...@openvswitch.org; ovs-dev@openvswitch.org; nd ;
> Gavin Hu (Arm Technology China)
> Subject: Re: [ovs-dev] [PATCH v1] netdev: use acquire-release semantic
is portable to all aarch64
> > architectures and has been verified on an 64-bit arm server.
> >
> > Suggested-by: Yanqin Wei
> > Signed-off-by: Malvika Gupta
> > ---
>
> Thanks for the patch!
>
> Are you trying to use AF_XDP on aarch64? Asking becaus
Hi William,
I think the main objective of this patch is to introduce AF_XDP socket for OVS
userspace datapath. This is an alternative to DPDK PMD library. Is it
possible to continue using other DPDK libraries before XDP provides the
corresponding functionality?
Best Regards,
Wei Yanqin
Hi,
I am trying to compile OVS(master branch) via gcc 6.5.0, but there is gcc
option error.
gcc: error: unrecognized command line option '-Wmultistatement-macros'; did you
mean '-Wunused-macros'?
My question is which versions of GCC are supported by OVS. Is it a bug of OVS
compiling? The compi
your
support.
Best Regards,
Wei Yanqin
-Original Message-
From: Ben Pfaff
Sent: Tuesday, June 4, 2019 7:11 AM
To: Yanqin Wei (Arm Technology China)
Cc: ovs-disc...@openvswitch.org; d...@openvswitch.org
Subject: Re: [ovs-dev] OVS compiling issue(gcc version 6.5.0 )
On Mon, Jun 03, 2019
on in this patch. What do you think of
it?
Best Regards,
Wei Yanqin
-Original Message-
From: Ben Pfaff
Sent: Thursday, August 29, 2019 5:58 AM
To: Yanqin Wei (Arm Technology China)
Cc: d...@openvswitch.org; nd ; Gavin Hu (Arm Technology China)
Subject: Re: [ovs-dev] [PATCH v2] flow:
data-oriented and well-defined flag to bypass some
time-consuming handling, it can improve performance a lot.
Best Regards,
Wei Yanqin
-Original Message-
From: Ben Pfaff
Sent: Thursday, August 29, 2019 6:11 AM
To: Yanqin Wei (Arm Technology China) ; Ilya Maximets
Cc: d
Yanqin
-Original Message-
From: Ilya Maximets
Sent: Thursday, August 29, 2019 9:46 PM
To: Yanqin Wei (Arm Technology China) ; Ben Pfaff
Cc: d...@openvswitch.org; nd ; Gavin Hu (Arm Technology China)
Subject: Re: [ovs-dev] [PATCH v2] flow: miniflow_extract metadata branchless
Hi James,
CRC32 is not a mandatory feature for Armv8.0. What is the arm64 CPU used in
your platform?
I'm not sure if it's necessary to use mandatory feature (neon) to optimize hash
libraries, because I thought most arm64 CPUs supported CRC32.
Best Regards,
Wei Yanqin
-Original Message-
Understood. Thanks for the information.
Best Regards,
Wei Yanqin
From: James Page
Sent: Wednesday, September 4, 2019 5:45 PM
To: Yanqin Wei (Arm Technology China)
Cc: Ilya Maximets ; ovs-dev@openvswitch.org
Subject: Re: [ovs-dev] SIGILL ovs branch-2.12/arm64/DPDK
Hi
On Wed, Sep 4, 2019 at 10
-
From: Ilya Maximets
Sent: Wednesday, September 4, 2019 8:46 PM
To: Yanqin Wei (Arm Technology China) ; James Page
Cc: ovs-dev@openvswitch.org
Subject: Re: [ovs-dev] SIGILL ovs branch-2.12/arm64/DPDK
BTW, I submitted a bug to DPDK:
https://bugs.dpdk.org/show_bug.cgi?id=344
Best regards, Ilya
Hi Ilya,
Thanks for the clarification. It looks application is strong couple with DPDK.
I will follow your discussion under this bug.
Best Regards,
Wei Yanqin
-Original Message-
From: Ilya Maximets
Sent: Thursday, September 5, 2019 3:41 PM
To: Yanqin Wei (Arm Technology China) ; James
Thanks James,
I can understand the root cause now.
Best Regards,
Wei Yanqin
From: James Page
Sent: Thursday, September 5, 2019 3:42 PM
To: Yanqin Wei (Arm Technology China)
Cc: Ilya Maximets ; ovs-dev@openvswitch.org
Subject: Re: [ovs-dev] SIGILL ovs branch-2.12/arm64/DPDK
On Thu, Sep 5, 2019
97 matches
Mail list logo