seL4 - getting the execution time of different threads
Hello together, I was wondering how you do get the execution time in seL4 if you are using multiple threads. In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on. But how do I use the timer to measure the execution time of the other threads without stopping their execution to "pull" the time from the initial thread? Is it possible to share the timer or create new ones in the other threads? Or is there another way to measure the execution time? Some background information: The threads have different VSpace's and CSpace's. Used tutorials: seL4-tutorial 4, seL4-tutorial timer. The other threads do not have a bootinfo in their environment. Kind regards, Chris
Hi Chris, It depends on your use case. If you are looking to monitor the execution times of threads for debugging and/or benchmarking purposes, you can use our kernel benchmarking support, detailed here [1]. Otherwise, you could consider using the mcs branch of the kernel, which allows you to obtain the execution time of a thread via the seL4_SchedContext_Consumed invocation [2]. If you need to use the master branch of the kernel, then you would need to intercept the scheduler on thread switch, by running a high priority thread which enables other threads. Cheers Anna. [1] https://docs.sel4.systems/BenchmarkingGuide.html#cpu-utilisation [2] https://github.com/seL4/seL4/releases/tag/9.0.0-mcs? ________________________________ From: Devel <devel-bounces@sel4.systems> on behalf of Mr. Peace <chrisdenneberg7@gmail.com> Sent: Monday, 9 July 2018 3:34 AM To: devel@sel4.systems Subject: [seL4] seL4 - getting the execution time of different threads Hello together, I was wondering how you do get the execution time in seL4 if you are using multiple threads. In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on. But how do I use the timer to measure the execution time of the other threads without stopping their execution to "pull" the time from the initial thread? Is it possible to share the timer or create new ones in the other threads? Or is there another way to measure the execution time? Some background information: The threads have different VSpace's and CSpace's. Used tutorials: seL4-tutorial 4, seL4-tutorial timer. The other threads do not have a bootinfo in their environment. Kind regards, Chris
Hi Anna, Thanks for your reply. I was going through this email chain. I understand the 3 techniques that you told. I was wondering about this particular question of Chris: "In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on. Is it possible to share the timer or create new ones in the other threads?" Can you please suggest is it possible to share or create a new timer in the other thread? Creating a timer in the other thread will require vka, allocman etc. This boils down to, if it is possible to share vka, allocman etc. with the other thread or create new ones in the other thread. If yes, how can we do that? -- Thanks and Regards, Amit Goyal On 2018-07-10 06:55, Anna.Lyons@data61.csiro.au wrote:
Hi Chris,
It depends on your use case. If you are looking to monitor the execution times of threads for debugging and/or benchmarking purposes, you can use our kernel benchmarking support, detailed here [1].
Otherwise, you could consider using the mcs branch of the kernel, which allows you to obtain the execution time of a thread via the seL4_SchedContext_Consumed invocation [2].
If you need to use the master branch of the kernel, then you would need to intercept the scheduler on thread switch, by running a high priority thread which enables other threads.
Cheers
Anna.
[1] https://docs.sel4.systems/BenchmarkingGuide.html#cpu-utilisation
[2] https://github.com/seL4/seL4/releases/tag/9.0.0-mcs?
________________________________ From: Devel <devel-bounces@sel4.systems> on behalf of Mr. Peace <chrisdenneberg7@gmail.com> Sent: Monday, 9 July 2018 3:34 AM To: devel@sel4.systems Subject: [seL4] seL4 - getting the execution time of different threads
Hello together,
I was wondering how you do get the execution time in seL4 if you are using multiple threads. In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on.
But how do I use the timer to measure the execution time of the other threads without stopping their execution to "pull" the time from the initial thread? Is it possible to share the timer or create new ones in the other threads? Or is there another way to measure the execution time?
Some background information: The threads have different VSpace's and CSpace's. Used tutorials: seL4-tutorial 4, seL4-tutorial timer. The other threads do not have a bootinfo in their environment.
Kind regards, Chris
_______________________________________________ Devel mailing list Devel@sel4.systems https://sel4.systems/lists/listinfo/devel
Hi Amit, Yes, it's possible to share vka, allocman etc (if they are in the same address space you can make them global and use them - although beware, these libraries are not thread safe). Between processes you can set up new allocators, I recommend you take a look at the sel4bench project [1] which passes the timer to each new benchmark process. However, it depends on what you want. If you just want to be able to read the time and get timeouts, you don't want to share the timer itself (its a device) but expose a timer service either through a library, or RPC if it is between processes. I'll note this down as an example we should build into future tutorials, but you can start by using the timer manager [2]. Good luck! Anna. [1] https://github.com/seL4/sel4bench [2]https://github.com/seL4/util_libs/blob/9784983ec4038236ba53dd090c57e27ee499d... ________________________________________ From: Amit Goyal <amitgoyal@cse.iitb.ac.in> Sent: Wednesday, 11 July 2018 12:31 AM To: Lyons, Anna (Data61, Kensington NSW) Cc: devel@sel4.systems; chrisdenneberg7@gmail.com Subject: Re: [seL4] seL4 - getting the execution time of different threads Hi Anna, Thanks for your reply. I was going through this email chain. I understand the 3 techniques that you told. I was wondering about this particular question of Chris: "In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on. Is it possible to share the timer or create new ones in the other threads?" Can you please suggest is it possible to share or create a new timer in the other thread? Creating a timer in the other thread will require vka, allocman etc. This boils down to, if it is possible to share vka, allocman etc. with the other thread or create new ones in the other thread. If yes, how can we do that? -- Thanks and Regards, Amit Goyal On 2018-07-10 06:55, Anna.Lyons@data61.csiro.au wrote:
Hi Chris,
It depends on your use case. If you are looking to monitor the execution times of threads for debugging and/or benchmarking purposes, you can use our kernel benchmarking support, detailed here [1].
Otherwise, you could consider using the mcs branch of the kernel, which allows you to obtain the execution time of a thread via the seL4_SchedContext_Consumed invocation [2].
If you need to use the master branch of the kernel, then you would need to intercept the scheduler on thread switch, by running a high priority thread which enables other threads.
Cheers
Anna.
[1] https://docs.sel4.systems/BenchmarkingGuide.html#cpu-utilisation
[2] https://github.com/seL4/seL4/releases/tag/9.0.0-mcs?
________________________________ From: Devel <devel-bounces@sel4.systems> on behalf of Mr. Peace <chrisdenneberg7@gmail.com> Sent: Monday, 9 July 2018 3:34 AM To: devel@sel4.systems Subject: [seL4] seL4 - getting the execution time of different threads
Hello together,
I was wondering how you do get the execution time in seL4 if you are using multiple threads. In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on.
But how do I use the timer to measure the execution time of the other threads without stopping their execution to "pull" the time from the initial thread? Is it possible to share the timer or create new ones in the other threads? Or is there another way to measure the execution time?
Some background information: The threads have different VSpace's and CSpace's. Used tutorials: seL4-tutorial 4, seL4-tutorial timer. The other threads do not have a bootinfo in their environment.
Kind regards, Chris
_______________________________________________ Devel mailing list Devel@sel4.systems https://sel4.systems/lists/listinfo/devel
Hi All, As per the suggestion, I tried running the seL4bench (https://github.com/seL4/sel4bench) but it did not work. I tried to run it as a separate project using the following commands on x86 system, ubuntu 16.04 on qemu: mkdir sel4bench cd sel4bench repo init -u https://github.com/seL4/sel4bench-manifest.git repo sync mkdir build cd build Cases are i) ../init-build.sh -DPLATFORM=sabre -DHARDWARE=TRUE -DAARCH32=TRUE -DRELEASE=TRUE -DFAULT=TRUE -DFASTPATH=TRUE It produces Error1.txt ii) ../init-build.sh -DPLATFORM=sabre -DHARDWARE=TRUE -DRELEASE=TRUE -DFAULT=TRUE -DFASTPATH=TRUE (*successful*) ninja It produces Error2.txt iii) ../init-build.sh -DPLATFORM=x86_64 -DSIMULATION=TRUE (*successful*) ninja It produces the following error. cpio: unrecognized option '--reproducible' Try `cpio --help' or `cpio --usage' for more information. ninja: build stopped: subcommand failed. Can you please suggest a possible way out for running it? -- Thanks and Regards, Amit Goyal On 2018-07-11 06:06, Anna.Lyons@data61.csiro.au wrote:
Hi Amit,
Yes, it's possible to share vka, allocman etc (if they are in the same address space you can make them global and use them - although beware, these libraries are not thread safe). Between processes you can set up new allocators, I recommend you take a look at the sel4bench project [1] which passes the timer to each new benchmark process.
However, it depends on what you want. If you just want to be able to read the time and get timeouts, you don't want to share the timer itself (its a device) but expose a timer service either through a library, or RPC if it is between processes. I'll note this down as an example we should build into future tutorials, but you can start by using the timer manager [2].
Good luck!
Anna.
[1] https://github.com/seL4/sel4bench
[2]https://github.com/seL4/util_libs/blob/9784983ec4038236ba53dd090c57e27ee499d... ________________________________________ From: Amit Goyal <amitgoyal@cse.iitb.ac.in> Sent: Wednesday, 11 July 2018 12:31 AM To: Lyons, Anna (Data61, Kensington NSW) Cc: devel@sel4.systems; chrisdenneberg7@gmail.com Subject: Re: [seL4] seL4 - getting the execution time of different threads
Hi Anna,
Thanks for your reply. I was going through this email chain. I understand the 3 techniques that you told. I was wondering about this particular question of Chris: "In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on. Is it possible to share the timer or create new ones in the other threads?"
Can you please suggest is it possible to share or create a new timer in the other thread? Creating a timer in the other thread will require vka, allocman etc. This boils down to, if it is possible to share vka, allocman etc. with the other thread or create new ones in the other thread. If yes, how can we do that?
-- Thanks and Regards, Amit Goyal
On 2018-07-10 06:55, Anna.Lyons@data61.csiro.au wrote:
Hi Chris,
It depends on your use case. If you are looking to monitor the execution times of threads for debugging and/or benchmarking purposes, you can use our kernel benchmarking support, detailed here [1].
Otherwise, you could consider using the mcs branch of the kernel, which allows you to obtain the execution time of a thread via the seL4_SchedContext_Consumed invocation [2].
If you need to use the master branch of the kernel, then you would need to intercept the scheduler on thread switch, by running a high priority thread which enables other threads.
Cheers
Anna.
[1] https://docs.sel4.systems/BenchmarkingGuide.html#cpu-utilisation
[2] https://github.com/seL4/seL4/releases/tag/9.0.0-mcs?
________________________________ From: Devel <devel-bounces@sel4.systems> on behalf of Mr. Peace <chrisdenneberg7@gmail.com> Sent: Monday, 9 July 2018 3:34 AM To: devel@sel4.systems Subject: [seL4] seL4 - getting the execution time of different threads
Hello together,
I was wondering how you do get the execution time in seL4 if you are using multiple threads. In the initial thread I can use the timer as it is shown in the tutorial about timers since I do have the bootinfo, vka, allocman and so on.
But how do I use the timer to measure the execution time of the other threads without stopping their execution to "pull" the time from the initial thread? Is it possible to share the timer or create new ones in the other threads? Or is there another way to measure the execution time?
Some background information: The threads have different VSpace's and CSpace's. Used tutorials: seL4-tutorial 4, seL4-tutorial timer. The other threads do not have a bootinfo in their environment.
Kind regards, Chris
_______________________________________________ Devel mailing list Devel@sel4.systems https://sel4.systems/lists/listinfo/devel
"Amit" == Amit Goyal <amitgoyal@cse.iitb.ac.in> writes:
Hi Amit, Please don't top-post --- it makes the email thread hard to read, and rapidly increases the size of messages to where they won;t be accepted by the list processor. Amit> As per the suggestion, I tried running the seL4bench Amit> (https://github.com/seL4/sel4bench) but it did not work. I tried Amit> to run it as a separate project using the following commands on Amit> x86 system, ubuntu 16.04 on qemu: Did you install an ARM cross compiler first? The error message arm-linux-gnueabi-gcc is not a full path and was not found in the PATH. suggests that you haven't got the 32-bit ARM-el compiler installed. You can either override the compiler by setting export CXX=arm-linux-gnueabihf-gcc (if you have the hardfloat compiler installed) or install the softfloat compiler with apt-get install gcc-arm-linux-gnueabi I'd strongly suggest also installing ccache to speed subsequent compilations. Amit> cpio: unrecognized option '--reproducible' Try `cpio --help' or Amit> `cpio --usage' for more information. ninja: build stopped: Amit> subcommand failed. --reproducible is in upstream cpio. You can just edit the makefile to remove the flag for now, until Ubuntu packages the latest version. Peter C -- Dr Peter Chubb Tel: +61 2 9490 5852 http://ts.data61.csiro.au/ Trustworthy Systems Group Data61 (formerly NICTA)
participants (4)
-
Amit Goyal -
Anna.Lyons＠data61.csiro.au -
Mr. Peace -
Peter.Chubb＠data61.csiro.au