Hi Thomas, Jingjing
I guess Jingjing intended to ack another patch, which is for XL710 performance
enhancement. Thanks for the good catch!
Regards,
Helin
> -Original Message-
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Friday, January 9, 2015 11:48 PM
> To: Wu,
Thanks, Thomas.
I supposed to review another patch for XL710 performance. Due to the similar
patch subject, I reviewed this applied patch.
> -Original Message-
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Friday, January 09, 2015 11:48 PM
> To: Wu, Jingjing
> Cc:
> -Original Message-
> From: Wu, Jingjing
> Sent: Thursday, December 25, 2014 11:14 AM
> To: dev at dpdk.org
> Cc: Wu, Jingjing; Zhang, Helin; Cao, Waterman
> Subject: [PATCH 2/4] e1000: new functions replace old ones for ethertype
> filters
>
> This patch removes old functions which de
From: Vlad Zolotarov [mailto:vl...@cloudius-systems.com]
Sent: Friday, January 09, 2015 9:50 PM
To: Ouyang, Changchun; dev at dpdk.org
Subject: Re: [dpdk-dev] [PATCH v5 4/6] ether: Check VMDq RSS mode
On 01/09/15 07:54, Ouyang, Changchun wrote:
-Original Message-
From: Vlad Zolota
Hi,
> -Original Message-
> From: Olivier MATZ [mailto:olivier.matz at 6wind.com]
> Sent: Friday, January 9, 2015 6:45 PM
> To: Ananyev, Konstantin; Liu, Jijiang
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v3 0/3] enhance TX checksum command and
> csum forwarding engine
>
> Hi,
> -Original Message-
> From: Vlad Zolotarov [mailto:vladz at cloudius-systems.com]
> Sent: Friday, January 9, 2015 10:02 PM
> To: Ouyang, Changchun; dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v5 5/6] ixgbe: Config VF RSS
>
>
> On 01/09/15 08:07, Ouyang, Changchun wrote:
> >
> >> -
This patch enables VF RSS for Niantic, which allow each VF having at most 4
queues.
The actual queue number per VF depends on the total number of pool, which is
determined by the max number of VF at PF initialization stage and the number of
queue specified in config:
1) If the max number of VF is
Put global register configuring out of loop for queue; also fix typo and indent.
Signed-off-by: Changchun Ouyang
---
lib/librte_pmd_ixgbe/ixgbe_rxtx.c | 35 ++-
1 file changed, 18 insertions(+), 17 deletions(-)
diff --git a/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
b/li
Negotiate API version with VF when receiving the IXGBE_VF_API_NEGOTIATE message.
Signed-off-by: Changchun Ouyang
---
lib/librte_pmd_ixgbe/ixgbe_ethdev.h | 1 +
lib/librte_pmd_ixgbe/ixgbe_pf.c | 25 +
2 files changed, 26 insertions(+)
diff --git a/lib/librte_pmd_ixgb
Get the available Rx and Tx queue number when receiving IXGBE_VF_GET_QUEUES
message from VF.
Signed-off-by: Changchun Ouyang
changes in v5
- Add some 'FIX ME' comments for IXGBE_VF_TRANS_VLAN.
---
lib/librte_pmd_ixgbe/ixgbe_pf.c | 40 +++-
1 file changed,
It needs config RSS and IXGBE_MRQC and IXGBE_VFPSRTYPE to enable VF RSS.
The psrtype will determine how many queues the received packets will distribute
to,
and the value of psrtype should depends on both facet: max VF rxq number which
has been negotiated with PF, and the number of rxq specified
Check mq mode for VMDq RSS, handle it correctly instead of returning an error;
Also remove the limitation of per pool queue number has max value of 1, because
the per pool queue number could be 2 or 4 if it is VMDq RSS mode;
The number of rxq specified in config will determine the mq mode for VMDq
Set VMDq RSS mode if it has VF(VF number is more than 1) and has RSS
information.
Signed-off-by: Changchun Ouyang
changes in v6
- Put common statement outside the if branch.
changes in v5
- Assign txmode.mq_mode with ETH_MQ_TX_NONE explicitly;
- Remove one line wrong comment.
---
app/t
Hello
Does anybody has any more review comments or ideas for this? I will send out
the formal patch soon. Thanks a lot!
Regards,
Helin
> -Original Message-
> From: Zhang, Helin
> Sent: Friday, December 19, 2014 3:27 PM
> To: dev at dpdk.org
> Cc: Wu, Jingjing; Liu, Jijiang; Chen, Jing D
v2 changes:
change the return value if adding an existing filter from the filter's index
to negative value.
The patch set uses new filter_ctrl API to replace old ethertype filter APIs.
It uses new functions and structure to replace old ones in igb/ixgbe driver,
new commands to replace old on
This patch removes old functions which deal with ethertype filter in ixgbe
driver.
It also defines ixgbe_dev_filter_ctrl which is binding to filter_ctrl API,
and ethertype filter can be dealt with through this new entrance.
Signed-off-by: Jingjing Wu
---
lib/librte_pmd_ixgbe/ixgbe_ethdev.c | 35
Following commands of ethertype filter are removed:
- add_ethertype_filter (port_id) ethertype (eth_value)
- remove_ethertype_filter (port_id) index (idx)
- get_ethertype_filter (port_id) index (idx)
New command is added for ethertype filter by using filter_ctrl API and new
ethertype filter s
This patch removes old functions which deal with ethertype filter in igb driver.
It also defines eth_igb_filter_ctrl which is binding to filter_ctrl API,
and ethertype filter can be dealt with through this new entrance.
Signed-off-by: Jingjing Wu
---
lib/librte_pmd_e1000/e1000_ethdev.h | 13 ++
Structure rte_ethertype_filter is removed.
Following APIs are removed:
- rte_eth_dev_add_ethertype_filter
- rte_eth_dev_remove_ethertype_filter
- rte_eth_dev_get_ethertype_filter
Signed-off-by: Jingjing Wu
---
lib/librte_ether/rte_ethdev.c | 57
lib/librte_ethe
Acked-by: Helin Zhang
> -Original Message-
> From: Wu, Jingjing
> Sent: Monday, January 12, 2015 3:16 PM
> To: dev at dpdk.org
> Cc: Wu, Jingjing; Zhang, Helin; Cao, Min
> Subject: [PATCH v2 0/4] Integrate ethertype filter in igb/ixgbe driver to new
> API
>
> v2 changes:
> change the
> -Original Message-
> From: Zhang, Helin
> Sent: Monday, December 29, 2014 9:41 AM
> To: dev at dpdk.org
> Cc: nhorman at tuxdriver.com; Xu, Qian Q; Cao, Waterman; Lu, Patrick; Liu,
> Jijiang; Wu, Jingjing; Zhang, Helin
> Subject: [PATCH v2] i40e: workaround for XL710 performance
>
> On
Thank you, Jingjing!
Regards,
Helin
> -Original Message-
> From: Wu, Jingjing
> Sent: Monday, January 12, 2015 3:34 PM
> To: Zhang, Helin; dev at dpdk.org
> Cc: nhorman at tuxdriver.com; Xu, Qian Q; Cao, Waterman; Lu, Patrick; Liu,
> Jijiang
> Subject: RE: [PATCH v2] i40e: workaround for
On Fri, Jan 09, 2015 at 09:14:16AM -0800, Ravi Kerur wrote:
> Hi,
>
> Has anyone successfully cross compiled bsdapp on Ubuntu or other linux
> flavor? From the Linux documentation I see
>
> "To compile all 64-bit targets using gcc, use:
>
> make install T=x86_64*gcc"
>
> which makes me believe
On Fri, Jan 09, 2015 at 05:20:26PM -0800, Aziz Hajee wrote:
> I am using the dpdk1.6.0r1
> The rte_kni.lo is loaded:
> lsmod | grep kni
> rte_kni 279134 1
>
> however, the ifconfig vEth0, and vEth1 does not show link up ?
> How do i get the operational state up for these interfaces.
On Sun, Jan 11, 2015 at 07:10:30PM +0200, Vlad Zolotarov wrote:
> Hi,
> guys could you share form your experience what is the best way to distribute
> the DPDK libraries with the DPDK-based app:
>
> * Is there any significant benefit in compiling the libraries on a
>target machine?
> * Is th
Hi Jijiang,
Please find some comments below.
On 01/12/2015 04:41 AM, Liu, Jijiang wrote:
> There are some examples for the different packet types:
>
> 1. For L2 Packet types:
> MAC, ARP
> MAC, PAY2
> ...
> They are forwarded without beeing modified no matter if these above commands
> are set.
On Mon, Jan 12, 2015 at 11:21:32AM +, Bruce Richardson wrote:
> On Fri, Jan 09, 2015 at 09:14:16AM -0800, Ravi Kerur wrote:
> > Hi,
> >
> > Has anyone successfully cross compiled bsdapp on Ubuntu or other linux
> > flavor? From the Linux documentation I see
> >
> > "To compile all 64-bit targ
On 01/12/15 05:41, Ouyang, Changchun wrote:
>
> *From:*Vlad Zolotarov [mailto:vladz at cloudius-systems.com]
> *Sent:* Friday, January 09, 2015 9:50 PM
> *To:* Ouyang, Changchun; dev at dpdk.org
> *Subject:* Re: [dpdk-dev] [PATCH v5 4/6] ether: Check VMDq RSS mode
>
> On 01/09/15 07:54, Ouyang, Ch
On 01/12/15 07:59, Ouyang Changchun wrote:
> It needs config RSS and IXGBE_MRQC and IXGBE_VFPSRTYPE to enable VF RSS.
>
> The psrtype will determine how many queues the received packets will
> distribute to,
> and the value of psrtype should depends on both facet: max VF rxq number which
> has be
On 01/12/15 07:59, Ouyang Changchun wrote:
> Set VMDq RSS mode if it has VF(VF number is more than 1) and has RSS
> information.
>
> Signed-off-by: Changchun Ouyang
Reviewed-by: Vlad Zolotarov
>
> changes in v6
>- Put common statement outside the if branch.
>
> changes in v5
>- Assig
On 01/12/15 07:59, Ouyang Changchun wrote:
> Check mq mode for VMDq RSS, handle it correctly instead of returning an error;
> Also remove the limitation of per pool queue number has max value of 1,
> because
> the per pool queue number could be 2 or 4 if it is VMDq RSS mode;
>
> The number of rxq
Hello:
I have basic questions related to dpdk and trying to find help.
I am about to create a daemon process, is there a way for other process to
know whether the daemon is already created? I doesn't mean to get the pid,
because it changes every time.
If the daemon is created, how do
Date: Mon, 12 Jan 2015 15:39:39 +0100
Message-Id: <1421073581-6644-1-git-send-email-michalx.k.jastrzebski at
intel.com>
X-Mailer: git-send-email 2.1.1
From: Pawel Wodkowski
Hi,
this patchset enables DCB in SRIOV (ETH_MQ_RX_VMDQ_DCB and ETH_MQ_TX_VMDQ_DCB)
for each VF and PF for ixgbe driver.
Date: Mon, 12 Jan 2015 15:39:40 +0100
Message-Id: <1421073581-6644-2-git-send-email-michalx.k.jastrzebski at
intel.com>
X-Mailer: git-send-email 2.1.1
In-Reply-To: <1421073581-6644-1-git-send-email-michalx.k.jastrzebski at
intel.com>
References: <1421073581-6644-1-git-send-email-michalx.k.jastrze
Date: Mon, 12 Jan 2015 15:39:41 +0100
Message-Id: <1421073581-6644-3-git-send-email-michalx.k.jastrzebski at
intel.com>
X-Mailer: git-send-email 2.1.1
In-Reply-To: <1421073581-6644-1-git-send-email-michalx.k.jastrzebski at
intel.com>
References: <1421073581-6644-1-git-send-email-michalx.k.jastrze
On Mon, Jan 12, 2015 at 11:30:26AM +, Bruce Richardson wrote:
> On Sun, Jan 11, 2015 at 07:10:30PM +0200, Vlad Zolotarov wrote:
> > Hi,
> > guys could you share form your experience what is the best way to distribute
> > the DPDK libraries with the DPDK-based app:
> >
> > * Is there any signi
On Mon, Jan 12, 2015 at 02:28:20PM +, Ni, Xun wrote:
> Hello:
>
>I have basic questions related to dpdk and trying to find help.
>
>I am about to create a daemon process, is there a way for other process to
> know whether the daemon is already created? I doesn't mean to get the pid,
On 01/12/15 16:50, Neil Horman wrote:
> On Mon, Jan 12, 2015 at 11:30:26AM +, Bruce Richardson wrote:
>> On Sun, Jan 11, 2015 at 07:10:30PM +0200, Vlad Zolotarov wrote:
>>> Hi,
>>> guys could you share form your experience what is the best way to distribute
>>> the DPDK libraries with the DPDK
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Michal Jastrzebski
> Sent: Monday, January 12, 2015 3:41 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH 0/2] Enable DCB in SRIOV mode for ixgbe driver
>
> Date: Mon, 12 Jan 2015 15:39:39 +0100
> Message-I
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Michal Jastrzebski
> Sent: Monday, January 12, 2015 3:43 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH 1/2] pmd: add DCB for VF for ixgbe
>
> Date: Mon, 12 Jan 2015 15:39:40 +0100
> Message-Id: <14210735
> -Original Message-
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Michal Jastrzebski
> Sent: Monday, January 12, 2015 3:45 PM
> To: dev at dpdk.org
> Subject: [dpdk-dev] [PATCH 2/2] testpmd: fix dcb in vt mode
>
> Date: Mon, 12 Jan 2015 15:39:41 +0100
> Message-Id: <1421073581
From: Pawel Wodkowski
Hi,
this patchset enables DCB in SRIOV (ETH_MQ_RX_VMDQ_DCB and ETH_MQ_TX_VMDQ_DCB)
for each VF and PF for ixgbe driver.
As a side effect this allow to use multiple queues for TX in VF (8 if there is
16 or less VFs or 4 if there is 32 or less VFs) when PFC is not enabled.
From: Pawel Wodkowski
This patch add support for DCB in SRIOV mode. When no PFC
is enabled this feature might be used as multiple queues
(up to 8 or 4) for VF.
It incorporate following modifications:
- Allow zero rx/tx queues to be passed to rte_eth_dev_configure().
Rationale:
in SRIOV mo
From: Pawel Wodkowski
This patch incorporate fixes to support DCB in SRIOV mode for testpmd.
It also clean up some old code that is not needed or wrong.
Signed-off-by: Pawel Wodkowski
---
app/test-pmd/cmdline.c |4 ++--
app/test-pmd/testpmd.c | 39 +--
Remove CONFIG_RTE_BUILD_COMBINE_LIBS and CONFIG_RTE_LIBNAME.
Signed-off-by: Sergio Gonzalez Monroy
---
config/common_bsdapp| 6 --
config/common_linuxapp | 6 --
config/defconfig_ppc_64-power8-linuxapp-gcc | 2 -
lib/Makefile
This is equivalent to:
git mv lib/librte_eal lib/core
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_eal/Makefile |39 +
lib/core/librte_eal/bsdapp/Makefile|38 +
lib/core/librte_eal/bsdapp/contigmem/BSDmakefile |36 +
lib/core/libr
This patch series updates the DPDK build system.
Following are the goals it tries to accomplish:
- Create a library containing core DPDK libraries (librte_eal,
librte_malloc, librte_mempool, librte_mbuf and librte_ring).
The idea of core libraries is to group those libraries that are
alw
Update path to libraries inside core subdirectory.
Signed-off-by: Sergio Gonzalez Monroy
---
app/test/test_eal_fs.c | 2 +-
lib/Makefile | 6 +-
lib/core/librte_eal/bsdapp/eal/Makefile| 14 +++---
lib/core/librte_eal
This is equivalent to:
git mv lib/librte_malloc lib/core
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_malloc/Makefile | 48 +
lib/core/librte_malloc/malloc_elem.c | 321
lib/core/librte_malloc/malloc_elem.h | 190 +++
lib/c
This is equivalent to:
git mv lib/librte_ring lib/core
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_ring/Makefile | 48 ++
lib/core/librte_ring/rte_ring.c | 338 +++
lib/core/librte_ring/rte_ring.h | 1214 +++
lib/librte_ring/Makefil
This patch does:
- Update the app building command to link against librte_core.
- Set --start-group/--end-group and --whole-archive/--no-whole-archive
flags only when linking against static DPDK libs.
- Set --as--need/--no-as-needed when linknig against shared DPDK libs.
- Link against EXECENV
This patch creates a new rte.corelib.mk file and updates core libraries
to use it.
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_eal/bsdapp/eal/Makefile | 2 +-
lib/core/librte_eal/linuxapp/eal/Makefile | 3 +-
lib/core/librte_malloc/Makefile | 2 +-
lib/core/librte_mb
This patch creates a new subdirectory 'core' which contains DPDK core
libraries.
The goal is to generate a librte_core library that contains all
libraries under the core subdirectory. For that purpose, a synthetic
library librte_core is created.
When building the DPDK, all object files from core
We need to add -lpthread to EXECENV_LDLIBS because we are not passing
-pthread flags in EXECENV_CFLAGS to GCC when linking apps/
Signed-off-by: Sergio Gonzalez Monroy
---
mk/exec-env/linuxapp/rte.vars.mk | 2 ++
1 file changed, 2 insertions(+)
diff --git a/mk/exec-env/linuxapp/rte.vars.mk b/mk/
This is equivalent to:
git mv lib/librte_mempool lib/core
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_mempool/Makefile | 51 +
lib/core/librte_mempool/rte_dom0_mempool.c | 134 +++
lib/core/librte_mempool/rte_mempool.c | 901 ++
lib/core/librte_m
This patch set LDLIBS for each library.
When creating shared libraries, each library will be linked against
their dependant libraries - LDLIBS.
Signed-off-by: Sergio Gonzalez Monroy
---
lib/librte_acl/Makefile | 1 +
lib/librte_cfgfile/Makefile | 1 +
lib/librte_cmdline/Makefile
This is equivalent to:
git mv lib/librte_mbuf lib/core
Signed-off-by: Sergio Gonzalez Monroy
---
lib/core/librte_mbuf/Makefile | 48 ++
lib/core/librte_mbuf/rte_mbuf.c | 252 +
lib/core/librte_mbuf/rte_mbuf.h | 1133 +++
lib/librte_mbuf/Makefile
This patch mainly makes use of the LDLIBS variable when linking shared
libraries, setting proper DT_NEEDED entries.
This patch also fix a few nits like syntax highlighting, the command
string (O_TO_S_STR) used for linking shared libraries and the displayed
of dependencies when debugging is enable (
Hi Sergio,
2015-01-12 16:33, Sergio Gonzalez Monroy:
> This patch series updates the DPDK build system.
Thanks for proposing such rework.
We need discussions on that topic. So I ask some questions below.
> Following are the goals it tries to accomplish:
> - Create a library containing core DPDK
Hi Thomas,
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Monday, January 12, 2015 4:52 PM
>
> Hi Sergio,
>
> 2015-01-12 16:33, Sergio Gonzalez Monroy:
> > This patch series updates the DPDK build system.
>
> Thanks for proposing such rework.
> We need discussions on that
On Mon, Jan 12, 2015 at 05:21:48PM +, Gonzalez Monroy, Sergio wrote:
> Hi Thomas,
>
> > From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> > Sent: Monday, January 12, 2015 4:52 PM
> >
> > Hi Sergio,
> >
> > 2015-01-12 16:33, Sergio Gonzalez Monroy:
> > > This patch series updates
On Mon, 12 Jan 2015 09:52:10 -0500
Neil Horman wrote:
> On Mon, Jan 12, 2015 at 02:28:20PM +, Ni, Xun wrote:
> > Hello:
> >
> >I have basic questions related to dpdk and trying to find help.
> >
> >I am about to create a daemon process, is there a way for other process
> > to know
v2 changes:
- When build with the compilers that don't support AVX2 instructions,
make rte_acl_classify_avx2() do nothing and return an error.
- Remove unneeded 'ifdef __AVX2__' in acl_run_avx2.*.
- Reorder order of patches in the set, to keep RTE_LIBRTE_ACL_STANDALONE=y
always buildable.
This pat
Make data_indexes long enough to survive idle transitions.
That allows to simplify match processing code.
Also fix incorrect size calculations for data indexes.
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_bld.c | 5 +++--
lib/librte_acl/acl_run.h | 4
2 files changed, 3 inserti
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/rte_acl_osdep_alone.h | 12
1 file changed, 12 insertions(+)
diff --git a/lib/librte_acl/rte_acl_osdep_alone.h
b/lib/librte_acl/rte_acl_osdep_alone.h
index a84b6f9..2a99860 100644
--- a/lib/librte_acl/rte_acl_osdep_alone.h
+++ b/
Reorganise SSE code-path a bit by moving lo/hi dwords shuffle
out from calc_addr().
That allows to make calc_addr() for SSE and AVX2 practically identical
and opens opportunity for further code deduplication.
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_run_sse.h | 38 +++
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_bld.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/librte_acl/acl_bld.c b/lib/librte_acl/acl_bld.c
index 8bf4a54..22f7934 100644
--- a/lib/librte_acl/acl_bld.c
+++ b/lib/librte_acl/acl_bld.c
@@ -1907,7 +1907,7 @@
Introduced division of whole 256 child transition enties
into 4 sub-groups (64 kids per group).
So 2 groups within the same node with identical children,
can use one set of transition entries.
That allows to compact some DFA nodes and get space savings in the RT table,
without any negative performa
Current rule-wildness based heuristsics can cause unnecessary splits of
the ruleset.
That might have negative perfomance effect:
more tries to traverse, bigger RT tables.
After removing it, on some test-cases with big rulesets (~10K)
observed ~50% speedup.
No difference for smaller rulesets.
Signe
Right now we allocate indexes for all types of nodes, except MATCH,
at 'gen final RT table' stage.
For MATCH type nodes we are doing it at building temporary tree stage.
This is totally unnecessary and makes code more complex and error prone.
Rework the code and make MATCH indexes being allocated a
Make classify_scalar to behave in the same way as it's vector counterpart:
move match check out of the inner loop, etc.
That makes scalar and vector code look more identical.
Plus it improves scalar code performance.
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_run_scalar.c | 23
Make sure that test_acl would not ignore error conditions.
Run classify() with all possible values.
Signed-off-by: Konstantin Ananyev
---
app/test/test_acl.c | 8 ++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/app/test/test_acl.c b/app/test/test_acl.c
index 356d620..7119a
In test-acl replace command-line option "--scalar" with new one:
"--alg=scalar|sse|avx2".
Allows user manually select preferred classify() method.
Signed-off-by: Konstantin Ananyev
---
app/test-acl/main.c | 93 ++---
1 file changed, 75 insertions(+
Previous improvements made scalar method the fastest one
for tiny bunch of packets (< 4).
That allows us to remove specific vector code-path for small number of packets
(search_sse_2)
and always use scalar method for such cases.
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_run_avx2.c
Move common check for input parameters up into rte_acl_classify_alg().
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl_run_scalar.c | 4
lib/librte_acl/acl_run_sse.c| 4
lib/librte_acl/rte_acl.c| 19 ---
3 files changed, 12 insertions(+), 15 delet
If at build phase we don't make any trie splitting,
then temporary build structures and resulting RT structure might be
much bigger than current.
>From other side - having just one trie instead of multiple can speedup
search quite significantly.
>From my measurements on rule-sets with ~10K rules:
R
New data type to manipulate 256 bit AVX values.
Rename field in the rte_xmm to keep common naming accross SSE/AVX fields.
Signed-off-by: Konstantin Ananyev
---
examples/l3fwd/main.c | 2 +-
lib/librte_acl/acl_run_sse.c| 88 -
Signed-off-by: Konstantin Ananyev
---
lib/librte_acl/acl.h | 39 ++-
lib/librte_acl/acl_run.h | 1 -
2 files changed, 38 insertions(+), 2 deletions(-)
diff --git a/lib/librte_acl/acl.h b/lib/librte_acl/acl.h
index 61b849a..e65e079 100644
--- a/lib/librte_
v2 changes:
When build with the compilers that don't support AVX2 instructions,
make rte_acl_classify_avx2() do nothing and return an error.
Remove unneeded 'ifdef __AVX2__' in acl_run_avx2.*.
Introduce new classify() method that uses AVX2 instructions.
>From my measurements:
On HSW boards when pr
Vector code reorganisation/deduplication:
To avoid maintaining two nearly identical implementations of calc_addr()
(one for SSE, another for AVX2), replace it with a new macro that suits
both SSE and AVX2 code-paths.
Also remove no needed any more MM_* macros.
Signed-off-by: Konstantin Ananyev
-
80 matches
Mail list logo
