Having a drop queue per drop flow consumes a lot of memory and reduce the
speed capabilities of the NIC to handle such cases.
To avoid this and reduce memory consumption, an RSS drop queue is created
for all drop flows.
In mlx5 PMD handling a single queue of several destination queues ends in
creating the same Verbs attribute, the main difference resides in the
indirection table and the RSS hash key.
This helps to prepare the supports to the RSS queues by first handling the
queue action has being an RSS queue with a single queue. No RSS hash key
will be provided to the Verbs flow.
Creating a drop queue in mlx5 ends by creating a non polled queue, but if
the associated work queue could not be created the error was not handled
ending in a undefined situation.
Fixes: 2097d0d1e2cc ("net/mlx5: support basic flow items and actions") Cc: stable@dpdk.org Signed-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Wenzhuo Lu [Wed, 1 Mar 2017 06:04:51 +0000 (14:04 +0800)]
net/ixgbe/base: fix build error
Fix ICC build error by removing the EWARN third parameter.
Build error:
.../drivers/net/ixgbe/base/ixgbe_phy.c(1543):
error #268: the format string ends before this argument
EWARN(hw, "WARNING: Intel (R) Network "
^
.../drivers/net/ixgbe/base/ixgbe_phy.c(1805):
error #268: the format string ends before this argument
EWARN(hw, "WARNING: Intel (R) Network "
^
Fixes: aa4fc14d2cee ("ixgbe: update base driver") Fixes: b94a06c1b451 ("ixgbe/base: support qsfp and lco") Cc: stable@dpdk.org Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Wenzhuo Lu [Wed, 1 Mar 2017 06:04:49 +0000 (14:04 +0800)]
net/ixgbe/base: disable FC for 15B0
Disable Ethernet Flow Control (FC) for device 15B0.
Make sure that ixgbe_device_supports_autoneg_fc()
returns false and hw->fc.disable_fc_autoneg is set
to true to avoid running the fc_autoneg function
for the device 15B0, as this device doesn't support
this function.
Wenzhuo Lu [Wed, 1 Mar 2017 06:04:48 +0000 (14:04 +0800)]
net/ixgbe/base: complete HW init when SFP not present
If SFP module is not present, reset_hw doesn't return success.
SW should complete the initialization, or with specific module
it resulted in no link when the module was later inserted.
Yong Wang [Tue, 21 Feb 2017 09:33:23 +0000 (04:33 -0500)]
net/e1000/base: fix multicast setting in VF
In function e1000_update_mc_addr_list_vf(), "msgbuf[0]" is used prior
to initialization at "msgbuf[0] |= E1000_VF_SET_MULTICAST_OVERFLOW".
And "msgbuf[0]" is overwritten at "msgbuf[0] = E1000_VF_SET_MULTICAST".
Fix it by moving the second line prior to the first one that mentioned
above.
Fixes: dffbaf7880a8 ("e1000: revert fix for multicast in VF") Cc: stable@dpdk.org Signed-off-by: Yong Wang <wang.yong19@zte.com.cn> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Chas Williams [Fri, 10 Feb 2017 20:12:06 +0000 (15:12 -0500)]
net/bnx2x: fix transmit queue free threshold
The default tx_free_thresh is potentially larger than the allocated queue
which will result in TX queue cleanup never happening. To fix this,
lower the default free threshold and ensure that the free threshold is
never greater than the maximum outstanding transmit buffers.
Shahaf Shuler [Tue, 21 Feb 2017 14:37:24 +0000 (16:37 +0200)]
net/mlx5: fix extended statistics
The number of extended statistics counters is queried through ETHTOOL.
ETHTOOL provides a different number when the link is up or down.
Since extended statistics query occurs at device start,
segmentation fault might happen when changing the link state before and
after the device start.
this commit address this issue, and query the number of statistics
before every call to ETHTOOL.
Qi Zhang [Mon, 20 Feb 2017 18:11:56 +0000 (13:11 -0500)]
net/i40e: fix compile error
Fix the compile error when RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC
is disabled.
Also fake_mbuf is required to be initialized and assigned to
additional sw_ring entries for vector PMD independent from
RTE_LIBRTE_I40E_RX_ALLOW_BULK_ALLOC config option.
Shahaf Shuler [Tue, 14 Feb 2017 14:31:06 +0000 (16:31 +0200)]
net/mlx5: add out of buffer counter to extended statistic
This commit adds RX out of buffer counter to xstats report.
The counter counts the number of dropped occurred due to lack of buffers
on device RX queues.
Keith Wiles [Fri, 17 Feb 2017 15:43:04 +0000 (09:43 -0600)]
net/tap: fix possibly unterminated string
Calling strncpy with a maximum size argument of 16 bytes on destination
array "ifr.ifr_ifrn.ifrn_name" of size 16 bytes might leave the
destination string unterminated.
Allain Legacy [Fri, 31 Mar 2017 13:52:03 +0000 (09:52 -0400)]
cfgfile: support empty value
This commit adds support to the cfgfile library for parsing a key=value
line that has no value string specified (e.g., "key="). This can be used
to override a configuration attribute that has a default value or default
list of values to set it back to an undefined value to disable
functionality.
Joseph Richard [Fri, 31 Mar 2017 13:52:02 +0000 (09:52 -0400)]
cfgfile: fix parsing of long fields
When parsing a ini file with a "key = value" line that has both "key" and
"value" sized to the maximum allowed length causes a parsing failure. The
internal "buffer" variable should be sized at least as large as the maximum
for both fields. This commit updates the local array to be sized to hold
the max name, max value, " = ", and the nul terminator.
Allain Legacy [Fri, 31 Mar 2017 13:52:01 +0000 (09:52 -0400)]
cfgfile: constrain string search
The call to memchr() uses the absolute length of the string buffer instead
of the actual length of the string returned by fgets(). This causes the
search to go beyond the '\n' character and find ';' characters in random
garbage on the stack. This then causes the 'len' variable to be updated
and the subsequent search for the '=' character to potentially find one
beyond the first newline character.
Since this bug relies on ';' and '=' characters appearing in random places
in the 'buffer' variable it is intermittently reproducible at best.
Allain Legacy [Fri, 31 Mar 2017 13:52:00 +0000 (09:52 -0400)]
cfgfile: support configurable comment character
The current cfgfile comment character is hardcoded to ';'. This commit a
new API to allow the user to specify which comment character to use while
parsing the file.
This is to ease adoption by applications that have an existing
configuration file which may use a different comment character. For
instance, an application may already have a configuration file that uses
the '#' as the comment character.
The approach of using a new API with an extensible parameters structure was
used rather than simply adding a new argument to the existing API to allow
for additional arguments to be introduced in the future.
Allain Legacy [Fri, 31 Mar 2017 13:51:59 +0000 (09:51 -0400)]
cfgfile: support global properties section
The current implementation of the cfgfile library requires that all
key=value pairs be within [SECTION] definitions. The ini file standard
allows for key=value pairs in an unnamed section.
This commit adds the capability of parsing key=value pairs from such an
unnamed section. The CFG_FLAG_GLOBAL_SECTION flag must be passed to the
rte_cfgfile_load() API to enable this functionality. Any key=value pairs
found before the first section can be accessed in the section named
"GLOBAL".
Allain Legacy [Fri, 31 Mar 2017 13:51:58 +0000 (09:51 -0400)]
test/cfgfile: add basic unit tests
This commit adds the basic infrastructure for the cfgfile library unit
tests. It includes success path tests for the most commonly used APIs.
More unit tests will be added later.
Yong Wang [Wed, 29 Mar 2017 08:27:35 +0000 (04:27 -0400)]
doc: fix a typo in howto guide
Fixes: 0ba3870e7559 ("doc: add guide to use virtio-user as exceptional path") Cc: stable@dpdk.org Signed-off-by: Yong Wang <wang.yong19@zte.com.cn> Acked-by: John McNamara <john.mcnamara@intel.com>
Jerin Jacob [Mon, 3 Apr 2017 08:35:14 +0000 (14:05 +0530)]
eal/linux: fix build with glibc 2.25
glibc 2.25 is warning about if applications depend on
sys/types.h for makedev macro, it expects to be included
from <sys/sysmacros.h>
Found this error while testing with GCC 6.3.1 on archlinux.
lib/librte_eal/linuxapp/eal/eal_pci_uio.c: In function ‘pci_mknod_uio_dev’:
lib/librte_eal/linuxapp/eal/eal_pci_uio.c:134:13:
error: In the GNU C Library, "makedev" is defined
by <sys/sysmacros.h>. For historical compatibility, it is
currently defined by <sys/types.h> as well, but we plan to
remove this soon. To use "makedev", include <sys/sysmacros.h>
directly. If you did not intend to use a system-defined macro
"makedev", you should undefine it after including <sys/types.h>. [-Werror]
dev = makedev(major, minor);
^~~~~~~~~~~~~~~~~
Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Bruce Richardson [Fri, 24 Mar 2017 14:30:11 +0000 (14:30 +0000)]
nic_uio: fix device binding at boot
When loading nic_uio from /boot/loader.conf as specified in the Getting
Started Guide doc, the NIC devices were not bound at boot. Unloading the
nic_uio driver and reloading it would cause them to be bound, however.
The root cause appears to be the fact that when the module is loaded at
boot, the call to find the pci device when parsing the b:d:f parameter
fails to return the device. That means that later on when the device
is probed as part of a PCI scan, no action is taken as it's not recorded
as a device to be used.
We fix this by having the b:d:f string parsed again on probe if the
initial check to see if it's an already-known device fails. In my tests,
this causes the NIC devices to be successfully bound at boot time, as
well as leaving things working as before in the case the module is loaded
post-boot.
Jianfeng Tan [Thu, 16 Mar 2017 16:28:44 +0000 (16:28 +0000)]
vfio: fix secondary process start
When binding with vfio-pci, secondary process cannot be started with
an error message:
cannot find TAILQ entry for PCI device.
It's due to: struct rte_pci_addr is padded with 1 byte for alignment
by compiler. Then below comparison in commit 2f4adfad0a69
("vfio: add multiprocess support") will fail if the last byte is not
initialized.
Some compilers require definition of vfio_iommu_spapr_tce_ddw_info
before its use in vfio_iommu_spapr_tce_info, so move tce_info
definition below tce_ddw_info.
Fixes: 468f42cc2645 ("vfio: fix build on old kernel") Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Recently added "dma_zalloc_coherent()" call is causing build error
for Linux kernels < 3.2.
compile error:
lib/librte_eal/linuxapp/igb_uio/igb_uio.c:
In function ‘igbuio_pci_probe’:
lib/librte_eal/linuxapp/igb_uio/igb_uio.c:434:2:
error: implicit declaration of function ‘dma_zalloc_coherent’
[-Werror=implicit-function-declaration]
map_addr = dma_zalloc_coherent(&dev->dev, 1024,
^
dma_zalloc_coherent() introduced with Linux kernel 3.2, with commit
Linux: 842fa69f3e0c ("include/linux/dma-mapping.h: add dma_zalloc_coherent()")
Since it does not exist for older kernels, causing a build error.
Switched to dma_alloc_coherent() API to prevent build error.
Shreyansh Jain [Fri, 31 Mar 2017 05:35:37 +0000 (11:05 +0530)]
mempool: move stack handler as a driver
Moved from lib/librte_mempool, stack mempool handler is an independent
driver.
Shared builds would now require to link in librte_mempool_stack for
"stack" mempool handler.
Shreyansh Jain [Fri, 31 Mar 2017 05:35:36 +0000 (11:05 +0530)]
mempool: move ring handler as a driver
Moved from lib/librte_mempool, ring mempool is now an independent
driver.
Shared builds would now need to add librte_mempool_ring for:
* ring_mp_mc
* ring_sp_sc
* ring_sp_mc
* ring_mp_sc
Shreyansh Jain [Fri, 31 Mar 2017 05:35:35 +0000 (11:05 +0530)]
mempool: fix crash when handler not found
In case the stack or ring mempool handler are compiled as shared
library and not linked in with test binary, segfault is reported.
This is because return value of rte_mempool_set_ops_byname is not
being checked in rte_mempool_ops_alloc.
This patch handles error returned from rte_mempool_set_ops_byname
when a mempool is not found.
Gage Eads [Thu, 30 Mar 2017 23:02:00 +0000 (18:02 -0500)]
mempool: update non-EAL thread note
Commit 30e6399892276 ("mempool: support non-EAL thread") added the
capability for non-EAL threads to use the mempool library. This commit
removes the note indicating that the mempool library cannot be used safely
by non-EAL threads, and replaces it with a more up-to-date note.
For using a DPDK app when iommu is enabled, it requires to
add iommu=pt to the kernel command line. But using igb_uio driver
makes DMAR errors because the device has not an IOMMU domain.
Since kernel 3.15, iommu=pt requires to use the internal kernel
DMA API for attaching the device to the IOMMU 1:1 mapping, aka
si_domain. Previous versions did attach the device to that
domain when intel iommu notifier was called.
This is not a problem if the driver does later some call to the
DMA API because the mapping can be done then. But DPDK apps do
not use that DMA API at all.
Doing this dma map and unmap is harmless even when iommu is not
enabled at all.
Ferruh Yigit [Tue, 21 Mar 2017 09:54:48 +0000 (09:54 +0000)]
kni: fix build with kernel 4.11
compile error:
.../build/build/lib/librte_eal/linuxapp/kni/kni_net.c:124:6:
error: implicit declaration of function ‘signal_pending’
[-Werror=implicit-function-declaration]
if (signal_pending(current) || ret_val <= 0) {
^~~~~~~~~~~~~~
Linux 4.11 moves signal function declarations to its own header file:
Linux: 174cd4b1e5fb ("sched/headers: Prepare to move signal wakeup &
sigpending methods from <linux/sched.h> into <linux/sched/signal.h>")
Use new header file "linux/sched/signal.h" to fix the build error.
Cc: stable@dpdk.org Reported-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com> Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com> Tested-by: Pankaj Gupta <pagupta@redhat.com>
Olivier Matz [Thu, 30 Mar 2017 12:32:28 +0000 (14:32 +0200)]
mk: fix dependencies to optional configs
In rte.lib.mk, the list of libraries passed to the link
command (LDLIBS) is generated from the DEPDIRS-xxx variables.
If a library is not compiled because it is disabled in
configuration, it should not appear in DEPDIRS-xxx.
- librte_port depends on librte_kni only if it is enabled.
- librte_table depends on librte_acl only if it is enabled.
Olivier Matz [Wed, 29 Mar 2017 08:36:32 +0000 (10:36 +0200)]
net/mlx5: implement descriptor status API
Since there is no "descriptor done" flag like on Intel drivers, the
approach is different on mlx5 driver.
- for Tx, we call txq_complete() to free descriptors processed by
the hw, then we check if the descriptor is between tail and head
- for Rx, we need to browse the cqes, managing compressed ones,
to get the number of used descriptors.
Olivier Matz [Wed, 29 Mar 2017 08:36:28 +0000 (10:36 +0200)]
ethdev: add descriptor status API
Introduce a new API to get the status of a descriptor.
For Rx, it is almost similar to rx_descriptor_done API, except it
differentiates "used" descriptors (which are hold by the driver and not
returned to the hardware).
For Tx, it is a new API.
The descriptor_done() API, and probably the rx_queue_count() API could
be replaced by this new API as soon as it is implemented on all PMDs.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com> Reviewed-by: Andrew Rybchenko <arybchenko@solarflare.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:30 +0000 (16:21 +0100)]
ring: make struct and macros type agnostic
Modify the enqueue and dequeue macros to support copying any type of
object by passing in the exact object type. Rather than using the "ring"
structure member of rte_ring, which is of type "array of void *", instead
have the macros take the start of the ring a a pointer value, thereby
leaving the rte_ring structure as purely a header value. This allows it
to be reused by other future ring types which can add on extra fields if
they want, or even to have the actual ring elements, of whatever type
stored separate from the ring header.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:28 +0000 (16:21 +0100)]
ring: separate out head index manipulation
We can write a single common function for head manipulation for enq
and a common one for deq, allowing us to have a single worker function
for enq and deq, rather than two of each. Update all other inline
functions to use the new functions.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:26 +0000 (16:21 +0100)]
examples/quota_watermark: use ring space for watermarks
Now that the enqueue function returns the amount of space in the ring,
we can use that to replace the old watermark functionality. Update the
example app to do so, and re-enable it in the examples Makefile.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:25 +0000 (16:21 +0100)]
ring: return remaining entry count when dequeuing
Add an extra parameter to the ring dequeue burst/bulk functions so that
those functions can optionally return the amount of remaining objs in the
ring. This information can be used by applications in a number of ways,
for instance, with single-consumer queues, it provides a max
dequeue size which is guaranteed to work.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:24 +0000 (16:21 +0100)]
ring: return free space when enqueuing
Add an extra parameter to the ring enqueue burst/bulk functions so that
those functions can optionally return the amount of free space in the
ring. This information can be used by applications in a number of ways,
for instance, with single-producer queues, it provides a max
enqueue size which is guaranteed to work. It can also be used to
implement watermark functionality in apps, replacing the older
functionality with a more flexible version, which enables apps to
implement multiple watermark thresholds, rather than just one.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:23 +0000 (16:21 +0100)]
ring: make bulk and burst return values consistent
The bulk fns for rings returns 0 for all elements enqueued and negative
for no space. Change that to make them consistent with the burst functions
in returning the number of elements enqueued/dequeued, i.e. 0 or N.
This change also allows the return value from enq/deq to be used directly
without a branch for error checking.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:22 +0000 (16:21 +0100)]
ring: remove watermark support
Remove the watermark support. A future commit will add support for having
enqueue functions return the amount of free space in the ring, which will
allow applications to implement their own watermark checks, while also
being more useful to the app.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:21 +0000 (16:21 +0100)]
ring: remove the yield when waiting for tail update
There was a compile time setting to enable a ring to yield when
it entered a loop in mp or mc rings waiting for the tail pointer update.
Build time settings are not recommended for enabling/disabling features,
and since this was off by default, remove it completely. If needed, a
runtime enabled equivalent can be used.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:20 +0000 (16:21 +0100)]
ring: remove debug setting
The debug option only provided statistics to the user, most of
which could be tracked by the application itself. Remove this as a
compile time option, and feature, simplifying the code.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:19 +0000 (16:21 +0100)]
ring: eliminate duplication of size and mask fields
The size and mask fields are duplicated in both the producer and
consumer data structures. Move them out of that into the top level
structure so they are not duplicated.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:18 +0000 (16:21 +0100)]
ring: create common structure for prod and cons metadata
create a common structure to hold the metadata for the producer and
the consumer, since both need essentially the same information - the
head and tail values, the ring size and mask.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Bruce Richardson [Wed, 29 Mar 2017 15:21:17 +0000 (16:21 +0100)]
ring: remove split cacheline build setting
Users compiling DPDK should not need to know or care about the arrangement
of cachelines in the rte_ring structure. Therefore just remove the build
option and set the structures to be always split. On platforms with 64B
cachelines, for improved performance use 128B rather than 64B alignment
since it stops the producer and consumer data being on adjacent cachelines.
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com> Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com> Acked-by: Olivier Matz <olivier.matz@6wind.com>
David Hunt [Mon, 20 Mar 2017 10:08:40 +0000 (10:08 +0000)]
doc: update distributor lib guide for new burst API
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: John McNamara <john.mcnamara@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Hunt [Mon, 20 Mar 2017 10:08:39 +0000 (10:08 +0000)]
examples/distributor: give Rx thread a core
Now that we're printing out a page of stats every second to the console,
we should give the stats it's own core so that we don't interfere with
the performance of the Rx core.
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Hunt [Mon, 20 Mar 2017 10:08:38 +0000 (10:08 +0000)]
examples/distributor: tweak for performance
This patch tunes Rx, Tx, and rte_distributor_process() burst sizes to
maximize performance.
It also addresses some checkpatch issues.
The result is approximately 10% performance increase.
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Hunt [Mon, 20 Mar 2017 10:08:30 +0000 (10:08 +0000)]
distributor: switch over to new API
This is the main switch over between the legacy API and the new
burst API. We rename all the functions in rte_distributor.c to remove
the _v1705, and we add in _v20 in the rte_distributor_v20.c
We also rename the rte_distributor_next.h as rte_distributor.h, as
this is now the public header.
At the same time, we need the autotests and sample app to compile
properly, hence those changes are in this patch also.
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Hunt [Mon, 20 Mar 2017 10:08:29 +0000 (10:08 +0000)]
test/distributor: add extra parameters
In the next few patches, we'll want to test old and new API,
so here we're allowing different parameters to be passed to
the tests, instead of just a distributor struct.
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
David Hunt [Mon, 20 Mar 2017 10:08:28 +0000 (10:08 +0000)]
distributor: add SIMD flow matching
Add an optimised version of the in-flight flow matching algorithm
using SIMD instructions. This should give up to 1.5x over the scalar
versions performance.
Falls back to scalar version if SSE4.2 not available
Signed-off-by: David Hunt <david.hunt@intel.com> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
git.droids-corp.org / dpdk.git / log