Re: [lng-odp] [API-NEXT PATCH 4/5] linux-generic: cpu: implementation for cycle count API

> -Original Message- > From: ext Ivan Khoronzhuk [mailto:ivan.khoronz...@linaro.org] > Sent: Monday, September 07, 2015 5:45 PM > To: Savolainen, Petri (Nokia - FI/Espoo); lng-odp@lists.linaro.org > Subject: Re: [lng-odp] [API-NEXT PATCH 4/5] linux-generic: cpu: > implementation for cycle

[lng-odp] [Bug 1615] odp_timer fails in CI with Segmentation fault

https://bugs.linaro.org/show_bug.cgi?id=1615 --- Comment #12 from Ola Liljedahl --- Perhaps we should add an explicit check that the sigval pointer is correct. Add a field with some magic value to the timer pool and check that tp->magic has the expected value. -- You are receiving this mail bec

Re: [lng-odp] [API-NEXT PATCH 5/5] performance: sched: update scheduling test to use cycle counts

sorry one typo . -printf("\nTime accuracy test (%i sec)\n", TEST_SEC); +printf("\nCPU cycle count accuracy test (%i sec)\n", TEST_SEC); Maybe beter to put this line under clock_gettime? In order to not miss *under -> above t1 == t2? Maybe here it's not so important but if it's

Re: [lng-odp] [API-NEXT PATCH 5/5] performance: sched: update scheduling test to use cycle counts

On 07.09.15 15:42, Petri Savolainen wrote: Use CPU cycle count API instead of time API to measure CPU cycles. Signed-off-by: Petri Savolainen --- test/performance/odp_scheduling.c | 132 -- 1 file changed, 57 insertions(+), 75 deletions(-) diff --git a/

Re: [lng-odp] [Patch] validation: scheduler: increase time check

We should have an application design "deep dive" discussion during SFO15. Let's get that on the schedule. Timely since the theme as we move into 2016 needs to be ODP Applications. On Mon, Sep 7, 2015 at 10:53 AM, Mike Holmes wrote: > Can any of this discussion be translated into a paragraph in

Re: [lng-odp] [Patch] validation: scheduler: increase time check

Can any of this discussion be translated into a paragraph in the users guide, explaining the philosophy described here ? The difference between control and worker tasks, how to achieve background processing, how to integrate with non event driven SW etc. On 7 September 2015 at 10:22, Bill Fischof

Re: [lng-odp] [PATCHv2 7/7] validation: pktio: test for transmit error handling

On Mon, Sep 07, 2015 at 02:20:46PM +0300, Maxim Uvarov wrote: > On 09/04/15 12:55, Stuart Haslam wrote: > >Test that transmit errors are handled correctly by attempting to send a > >packet larger than the MTU of the interface. > > > >Signed-off-by: Stuart Haslam > >--- > > test/validation/pktio/p

Re: [lng-odp] [API-NEXT PATCH 1/5] api: cpu: added cpu cycle count API

+ On 07.09.15 17:38, Ivan Khoronzhuk wrote: Petri, On 07.09.15 15:41, Petri Savolainen wrote: Raw CPU cycle counts can be used to measure performance in CPU cycles. These functions will replace some usage of odp_time_cycles() of odp_time_diff_cycles(). Signed-off-by: Petri Savolainen --- P

Re: [lng-odp] [PATCHv2 6/7] validation: pktio: add support for direct receive

On Mon, Sep 07, 2015 at 02:14:50PM +0300, Maxim Uvarov wrote: > On 09/04/15 12:55, Stuart Haslam wrote: > >Add a couple of tests for receiving packets directly via > >odp_pktio_recv(). > > > >Signed-off-by: Stuart Haslam > >--- > > test/validation/pktio/pktio.c | 103 > >

Re: [lng-odp] [API-NEXT PATCH 4/5] linux-generic: cpu: implementation for cycle count API

On 07.09.15 15:41, Petri Savolainen wrote: Added implementation for CPU cycle diff, max and resolution. Signed-off-by: Petri Savolainen --- platform/linux-generic/Makefile.am | 1 + platform/linux-generic/arch/linux/odp_cpu_cycles.c | 10 ++ platform/linux-gener

Re: [lng-odp] [API-NEXT PATCH 1/5] api: cpu: added cpu cycle count API

Petri, On 07.09.15 15:41, Petri Savolainen wrote: Raw CPU cycle counts can be used to measure performance in CPU cycles. These functions will replace some usage of odp_time_cycles() of odp_time_diff_cycles(). Signed-off-by: Petri Savolainen --- Probably proposed API covers all existent cases

Re: [lng-odp] [PATCHv2 4/7] validation: pktio: don't print to stdout

On Mon, Sep 07, 2015 at 02:05:29PM +0300, Maxim Uvarov wrote: > When I added that print I expected that person who executed test for > specific pktio can verify that > printed mac address is the same as read pktio (device). Hrm, I've never done that. > With removing that print there will be no >

Re: [lng-odp] [Patch] validation: scheduler: increase time check

Isn't that why we have the distinction between control and worker threads now? It sounds like a job for a control thread to handle the "non-ODP" stuff while the worker threads should just be heads-down doing the main application processing as directed by the scheduler. On Mon, Sep 7, 2015 at 9:19

Re: [lng-odp] [Patch] validation: scheduler: increase time check

From: ext Ola Liljedahl [mailto:ola.liljed...@linaro.org] Sent: Monday, September 07, 2015 4:51 PM To: Savolainen, Petri (Nokia - FI/Espoo) Cc: Bill Fischofer; LNG ODP Mailman List Subject: Re: [lng-odp] [Patch] validation: scheduler: increase time check On 7 September 2015 at 14:59, Savolainen,

Re: [lng-odp] [Patch] validation: scheduler: increase time check

On 7 September 2015 at 14:59, Savolainen, Petri (Nokia - FI/Espoo) < petri.savolai...@nokia.com> wrote: > Scheduler timeout (wait time) enables application to wait on a schedule > call, for an finite time. Otherwise application would need to wait > infinitely (SCHED_WAIT) or poll full scheduler on

Re: [lng-odp] [API-NEXT PATCHv4 1/4] api: classification: move odp_pmr_match_t definition

Hello Benoît, did you manually edit patch before sending? For some reason I can not apply it: 16:02 /opt/Linaro/odp2.git (api-next)$git am -s -3 /tmp/z9/\[lng-odp\]\ \[API-NEXT\ PATCHv4\ 1_4\]\ api\:\ classification\:\ move\ odp_pmr_match_t\ definition.eml Applying: api: classification: mov

Re: [lng-odp] [Patch] validation: scheduler: increase time check

Scheduler timeout (wait time) enables application to wait on a schedule call, for an finite time. Otherwise application would need to wait infinitely (SCHED_WAIT) or poll full scheduler on full CPU speed (SCHED_NO_WAIT). Periodic return from scheduler can be implemented also using a timer, but

Re: [lng-odp] [API-NEXT PATCH 2/5] linux-generic: cpu: created arch depedent cpu_cycles files

Petri, looks like you did not add -M option to format-patch. Maxim. On 09/07/15 15:41, Petri Savolainen wrote: Renamed time_cycles files to cpu_cycles, since those files implement cpu cycle counter read needed by cpu API but not needed by new time API. Signed-off-by: Petri Savolainen --- pl

[lng-odp] [NEXT PATCHv3] api: define pktio statistics api

Signed-off-by: Maxim Uvarov --- v3: - reduce number of counters; v2: - add function to check supported cnts; - add optional functio to reset cnts; include/odp/api/packet_io_stats.h | 94 ++ platform/linux-generic/include/odp/packet_io.h | 1 + 2 fil

[lng-odp] [API-NEXT PATCH 4/5] linux-generic: cpu: implementation for cycle count API

Added implementation for CPU cycle diff, max and resolution. Signed-off-by: Petri Savolainen --- platform/linux-generic/Makefile.am | 1 + platform/linux-generic/arch/linux/odp_cpu_cycles.c | 10 ++ platform/linux-generic/arch/mips64/odp_cpu_cycles.c | 10 ++ p

[lng-odp] [API-NEXT PATCH 5/5] performance: sched: update scheduling test to use cycle counts

Use CPU cycle count API instead of time API to measure CPU cycles. Signed-off-by: Petri Savolainen --- test/performance/odp_scheduling.c | 132 -- 1 file changed, 57 insertions(+), 75 deletions(-) diff --git a/test/performance/odp_scheduling.c b/test/perform

[lng-odp] [API-NEXT PATCH 2/5] linux-generic: cpu: created arch depedent cpu_cycles files

Renamed time_cycles files to cpu_cycles, since those files implement cpu cycle counter read needed by cpu API but not needed by new time API. Signed-off-by: Petri Savolainen --- platform/linux-generic/Makefile.am | 8 ++--- platform/linux-generic/arch/linux/odp_cpu_cycles.c | 38

[lng-odp] [API-NEXT PATCH 3/5] linux-generic: cpu: rename time_cycles to cpu_cycles

Implemented odp_cpu_cycles() be renaming odp_time_cycles() implementation. Time implementation uses odp_cpu_cycles temporarely. Signed-off-by: Petri Savolainen --- platform/linux-generic/arch/linux/odp_cpu_cycles.c | 4 ++-- platform/linux-generic/arch/mips64/odp_cpu_cycles.c | 4 ++-- platform

[lng-odp] [API-NEXT PATCH 0/5] CPU cycle count API

Raw CPU cycle count API is needed for measuring CPU cycle consumption. This can replace many of the odp_time_cycles() when time API moves to real time from CPU cycles. Petri Savolainen (5): api: cpu: added cpu cycle count API linux-generic: cpu: created arch depedent cpu_cycles files linux

[lng-odp] [API-NEXT PATCH 1/5] api: cpu: added cpu cycle count API

Raw CPU cycle counts can be used to measure performance in CPU cycles. These functions will replace some usage of odp_time_cycles() of odp_time_diff_cycles(). Signed-off-by: Petri Savolainen --- include/odp/api/cpu.h | 52 +++ 1 file changed, 52 in

Re: [lng-odp] [Patch] validation: scheduler: increase time check

On 4 September 2015 at 15:36, Bill Fischofer wrote: > That's saying that a specific worker has knowledge of what events it > should be receiving and when. The whole point of event scheduling is that > an individual worker thread is not aware of such considerations, but simply > process events th

Re: [lng-odp] [API-NEXT PATCHv2 0/4] classification API name change

Merged, Maxim. On 09/03/15 23:06, Bill Fischofer wrote: I thought I had responded already. But just in case: Reviewed-by: Bill Fischofer > On Thu, Sep 3, 2015 at 10:58 AM, Bala Manoharan mailto:bala.manoha...@linaro.org>> wrote: Ping. On 25 Augus

[lng-odp] [Bug 1711] Rename APIs for consistency with ODP naming standard

https://bugs.linaro.org/show_bug.cgi?id=1711 Maxim Uvarov changed: What|Removed |Added CC||maxim.uva...@linaro.org --- Comment #6 from Max

Re: [lng-odp] [RFC Patch] example: timer: move cycles/ns table to validation tests

Ola, On 07.09.15 14:45, Ola Liljedahl wrote: On 7 September 2015 at 11:45, Ivan Khoronzhuk mailto:ivan.khoronz...@linaro.org>> wrote: Petri, Maybe this peace of code should migrate to some other place, but on my opinion it has more common with time API then cpu API. The table

[lng-odp] [PATCHv2] validation: renaming module libs

Renaming module libs in /test/validation/. Module test libs are renamed from lib.la to libtestmodule.la. This is done to avoid libname clashes, such as libcrypto, which seems to confuse libtool, and should therefore fix bug ID 1787. Signed-off-by: Christophe Milard --- since v1: libodp changed t

Re: [lng-odp] [RFC Patch] example: timer: move cycles/ns table to validation tests

On 7 September 2015 at 11:45, Ivan Khoronzhuk wrote: > Petri, > > Maybe this peace of code should migrate to some other place, > but on my opinion it has more common with time API then cpu API. > > The table to check time table mixes time/cycles with timer API example, > and it > is confusing a l

Re: [lng-odp] [PATCH] validation: renaming module libs

I don't really have a strong opinion about this. AS long as I can rebuild my code :) Nicolas On 09/07/2015 12:29 PM, Christophe Milard wrote: > I'd prefer libtest as "odp" does not give any extra info, and, as the > prefix would be the same for all, the underscore could be removed. > But we nee

Re: [lng-odp] [API-NEXT PATCH v2 2/2] api: rwlock_recursive: added recursive rwlock

On 09/03/15 15:01, Petri Savolainen wrote: Added recursive read-write lock. It allows threads to read lock or write lock the rwlock multiple times without deadlocking. Mixing read and write lock operations is not supported. ODP version of recursive rwlock enables porting legacy applications, whi

Re: [lng-odp] [PATCHv2 7/7] validation: pktio: test for transmit error handling

On 09/04/15 12:55, Stuart Haslam wrote: Test that transmit errors are handled correctly by attempting to send a packet larger than the MTU of the interface. Signed-off-by: Stuart Haslam --- test/validation/pktio/pktio.c | 124 +- 1 file changed, 121 in

Re: [lng-odp] [API-NEXT PATCH v4 05/10] linux-generic: sysinfo: clarify the API for max CPU frequency

On 07.09.15 14:08, Hongbo Zhang wrote: On 3 September 2015 at 22:47, Ivan Khoronzhuk wrote: Hi, Hongbo On 11.08.15 10:33, hongbo.zh...@freescale.com wrote: From: Hongbo Zhang Currently the API to get CPU frequency is vague, it needs to be clarified whether max or current frequency is ret

Re: [lng-odp] [PATCHv2 6/7] validation: pktio: add support for direct receive

On 09/04/15 12:55, Stuart Haslam wrote: Add a couple of tests for receiving packets directly via odp_pktio_recv(). Signed-off-by: Stuart Haslam --- test/validation/pktio/pktio.c | 103 ++ test/validation/pktio/pktio.h | 2 + 2 files changed, 66 inse

Re: [lng-odp] [API-NEXT PATCH v4 05/10] linux-generic: sysinfo: clarify the API for max CPU frequency

On 3 September 2015 at 22:47, Ivan Khoronzhuk wrote: > Hi, Hongbo > > On 11.08.15 10:33, hongbo.zh...@freescale.com wrote: >> >> From: Hongbo Zhang >> >> Currently the API to get CPU frequency is vague, it needs to be clarified >> whether max or current frequency is returned, and now most use cas

Re: [lng-odp] [PATCHv2 4/7] validation: pktio: don't print to stdout

When I added that print I expected that person who executed test for specific pktio can verify that printed mac address is the same as read pktio (device). With removing that print there will be no way to prove that right mac addr was returned. Maxim. On 09/04/15 12:55, Stuart Haslam wrote: T

Re: [lng-odp] [API-NEXT PATCH 1/2] api: pktio: specify that pool type must be ODP_POOL_PACKET

Merged to api-next. Maxim. On 09/04/15 13:24, Savolainen, Petri (Nokia - FI/Espoo) wrote: For both patches: Reviewed-by: Petri Savolainen -Original Message- From: lng-odp [mailto:lng-odp-boun...@lists.linaro.org] On Behalf Of ext Stuart Haslam Sent: Friday, September 04, 2015 1:16

Re: [lng-odp] [PATCH] validation: renaming module libs

On 09/07/15 13:29, Christophe Milard wrote: I'd prefer libtest as "odp" does not give any extra info, and, as the prefix would be the same for all, the underscore could be removed. But we need this to be merged in some way, so if you all agree on libodptest_, I can do that. Anders? Nicolas? l

Re: [lng-odp] [API-NEXT PATCH 1/2] api: pktio: specify that pool type must be ODP_POOL_PACKET

merged to api-next. Maxim. On 09/04/15 13:24, Savolainen, Petri (Nokia - FI/Espoo) wrote: For both patches: Reviewed-by: Petri Savolainen -Original Message- From: lng-odp [mailto:lng-odp-boun...@lists.linaro.org] On Behalf Of ext Stuart Haslam Sent: Friday, September 04, 2015 1:16

Re: [lng-odp] [API-NEXT PATCH v4 05/10] linux-generic: sysinfo: clarify the API for max CPU frequency

On 3 September 2015 at 21:47, Savolainen, Petri (Nokia - FI/Espoo) wrote: > > >> -Original Message- >> From: ext hongbo.zh...@freescale.com >> [mailto:hongbo.zh...@freescale.com] >> Sent: Tuesday, August 11, 2015 10:34 AM >> To: lng-odp@lists.linaro.org >> Cc: mike.hol...@linaro.org; stuar

Re: [lng-odp] [PATCH] validation: renaming module libs

I'd prefer libtest as "odp" does not give any extra info, and, as the prefix would be the same for all, the underscore could be removed. But we need this to be merged in some way, so if you all agree on libodptest_, I can do that. Anders? Nicolas? On 7 September 2015 at 11:59, Maxim Uvarov wrote:

Re: [lng-odp] [PATCH] doc: implementers add include structure

Merged, Maxim. On 09/04/15 22:56, Bill Fischofer wrote: On Fri, Sep 4, 2015 at 1:39 PM, Mike Holmes > wrote: Signed-off-by: Mike Holmes mailto:mike.hol...@linaro.org>> Reviewed-by: Bill Fischofer > --- doc/impl

Re: [lng-odp] [API-NEXT PATCH v4 04/10] linux-generic: sysinfo: move CPU model API to cpu.h

On 3 September 2015 at 21:35, Savolainen, Petri (Nokia - FI/Espoo) wrote: > >> --- a/include/odp/api/cpu.h >> +++ b/include/odp/api/cpu.h >> @@ -51,6 +51,13 @@ int odp_cpu_count(void); >> uint64_t odp_cpu_hz(void); >> >> /** >> + * CPU model name >> + * >> + * @return Pointer to CPU model name s

Re: [lng-odp] [PATCH] test: api_test: remove un-used directory

On 09/07/15 12:18, Christophe Milard wrote: I agree with this Patch: I don't see any reason to keep non working code in the repos'head. This code is still in git, if we want to get it later. /Christophe. my point is: or remove test and odp ring. Or keep odp ring and move test to linux-generic

Re: [lng-odp] [API-NEXT PATCH v4 03/10] linux-generic: sysinfo: move CPU HZ API to cpu.h

On 3 September 2015 at 21:29, Savolainen, Petri (Nokia - FI/Espoo) wrote: > > >> -Original Message- >> From: ext hongbo.zh...@freescale.com >> [mailto:hongbo.zh...@freescale.com] >> Sent: Tuesday, August 11, 2015 10:34 AM >> To: lng-odp@lists.linaro.org >> Cc: mike.hol...@linaro.org; stuar

Re: [lng-odp] [PATCH] validation: update README files

Merged, Maxim. On 08/05/15 16:20, Stuart Haslam wrote: On Fri, Jul 31, 2015 at 09:56:44PM +0200, Christophe Milard wrote: Signed-off-by: Christophe Milard Reviewed-by: Stuart Haslam --- test/README| 8 +--- test/validation/README | 50 ++

Re: [lng-odp] [PATCHv2 0/4] linux-generic: add pktio pcap type

On Mon, Sep 07, 2015 at 10:42:13AM +0300, Maxim Uvarov wrote: > On 09/04/15 20:37, Bala Manoharan wrote: > >This method of using pcap file to generate packets is fine. > >But why should we use a dedicated interface with "pcap" as the name? > > > >I was imagining something like an ODP application wh

Re: [lng-odp] [PATCH] validation: renaming module libs

On 09/07/15 12:10, Christophe Milard wrote: These libs are test only libs. And hopefully they will be merged in one global test superlib, so they will probably not be referenced much. In my eyes, we are just creating a prefix to avoid a name clash with the standart lib name. The prefix has not m

Re: [lng-odp] [API-NEXT PATCH v4 00/10] linux-generic: sysinfo: CPU frequency API clean up

On 3 September 2015 at 21:25, Savolainen, Petri (Nokia - FI/Espoo) wrote: > Is this the latest version of the patch set ? > Yes, it is the latest although sent some time ago. > Generic comments: > 1) All patches which touch api files (03/10 and others) should be labeled > with "api:", instead of

Re: [lng-odp] [PATCHv2 0/4] linux-generic: add pktio pcap type

On Fri, Sep 04, 2015 at 02:50:47PM -0500, Bill Fischofer wrote: > On Fri, Sep 4, 2015 at 8:20 AM, Stuart Haslam > wrote: > > > This is pretty handy for testing, for example to test classifier rules > > using packets from a pcap; > > > > odp_classifier -ipcap:in=test.pcap -p -m 0 > > "ODP_PMR_SIP

Re: [lng-odp] [RFC Patch] example: timer: move cycles/ns table to validation tests

Petri, Maybe this peace of code should migrate to some other place, but on my opinion it has more common with time API then cpu API. The table to check time table mixes time/cycles with timer API example, and it is confusing a little. Here it doesn't bear any sense except time API has something

Re: [lng-odp] [PATCH] test: api_test: remove un-used directory

I agree with this Patch: I don't see any reason to keep non working code in the repos'head. This code is still in git, if we want to get it later. /Christophe. On 7 September 2015 at 09:52, Maxim Uvarov wrote: > Did you review [PATCHv8 00/12] IPC (pktio) patches as we discussed before? > > I do

Re: [lng-odp] [PATCH] validation: renaming module libs

These libs are test only libs. And hopefully they will be merged in one global test superlib, so they will probably not be referenced much. In my eyes, we are just creating a prefix to avoid a name clash with the standart lib name. The prefix has not much greater function. If it says "odp", it is r

[lng-odp] [Patch] linux-generic: odp_schedule: fix odp_schdule_wait_time

The resolution of schedule time can be more than 1ns. As result can happen that 1ns corresponds 0 ticks of timer counter, but if passed 1ns it's obvious that user wanted to schedule at least once. Currently it can lead to wait forever, as 0 corresponds to ODP_SCHED_WAIT, it can block program flow a

Re: [lng-odp] [PATCH] validation: renaming module libs

On 09/04/15 16:59, Nicolas Morey-Chaisemartin wrote: It actually works in the end. We link with the right lib. It's only an issue of rebuilding... As long as the implementation library does not link with the one from the test it should work alright. You can never make sure that you won't have n

Re: [lng-odp] [PATCH] test: api_test: remove un-used directory

Did you review [PATCHv8 00/12] IPC (pktio) patches as we discussed before? I don't want to kill test for odp_ring. Patch: [PATCHv8 06/12] linux-generic: move odp_ring from helper to linux-generic platfrom moves it to platform internal test. Event if we don't agree to include pktio ipc yet,

Re: [lng-odp] [PATCHv2 0/4] linux-generic: add pktio pcap type

On 09/04/15 20:37, Bala Manoharan wrote: This method of using pcap file to generate packets is fine. But why should we use a dedicated interface with "pcap" as the name? I was imagining something like an ODP application which reads from a given pcap file constructs the packet and sends the packe

Re: [lng-odp] odp v1.3.0.0

Hi, True, that's now the latest commit: commit 3afd410eaa2e55f47b42508ac0b86390a7b4c711 Author: Bill Fischofer Date: Thu Sep 3 13:59:22 2015 -0500 Thanks! José. On Monday 07 September 2015 09:54:57 ext Maxim Uvarov wrote: > Please delete local tag and do g