[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

--- Comment #9 from Yi He  ---
Hi, Mike

I did the tests, just like you and Alex analysed: "If it runs faster than it
used to (but still slower than the MTTCG case) then it could be some of the
ancillary changes to locking un-wedged that case."

QEMU 2.9.0-rc1 with --accel tcg,thread=single: Elapsed time = 5317.460 seconds

QEMU 2.9.0-rc1 without --accel tcg,thread=single: Elapsed time = 175.690
seconds

Compared to the start: Elapsed time = 21594.914 second

Thanks and best regards, Yi

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

--- Comment #8 from Mike Holmes  ---
Yi - Alex Bennée, the author of the code pinged me and is curious if it was
locking fixes or the single threaded nature - possibly knowing that is of value
to ODP ? I am not suggesting  you need to do the test however.

His suggestion if you are interested.

It would also be an interesting experiment if the now-fast test was
re-run with --accel tcg,thread=single to force single threaded
behaviour. If it runs faster than it used to (but still slower than the
MTTCG case) then it could be some of the ancillary changes to locking
un-wedged that case.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

Yi He  changed:

   What|Removed |Added

 CC||yi...@linaro.org
 Resolution|--- |FIXED
 Status|IN_PROGRESS |RESOLVED

--- Comment #7 from Yi He  ---
Run with most recent QEMU with MTTCG support contributed by Linaro core team:
2.9.0-rc1 released Mar 21 2017

Testing ./test/common_plat/validation/api/scheduler/scheduler_main elapsed time
reduced to reasonable ~200 seconds.

QEMU release information provided by Mike:

Note my team are close to closing this out, but it might take a while for
distros to pick it up, so support will arrive eventually

https://projects.linaro.org/browse/VIRT-48?jql=text%20~%20%22MTTCG%22

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

Yi He  changed:

   What|Removed |Added

   Assignee|yi...@linaro.org|maxim.uva...@linaro.org

--- Comment #6 from Yi He  ---
Hi, Maxim

Please see the comments and if we can close this one?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

Yi He  changed:

   What|Removed |Added

 Status|CONFIRMED   |IN_PROGRESS

--- Comment #5 from Yi He  ---
Not success to run MTTCG QEMU to emulate 8-cores aarch64 upon Intel server.

Due to its QEMU issue not related ODP, suppose this is for CI to automate ODP
validation for aarch64, can we create automation tasks on aarch64 servers with
QEMU/aarch64-KVM-accelerator instead of QEMU/Intel-aarch64 emulation?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

Yi He  changed:

   What|Removed |Added

 Ever confirmed|0   |1
 Status|UNCONFIRMED |CONFIRMED

--- Comment #4 from Yi He  ---
Replicated the problem

+server with Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz, 16G RAM
+qemu-system-aach64: cortex-A75, 8 cores, 2G RAM
+Ubuntu Cloud xenial-server-cloudimg-arm64-uefi1.img

Some investigation reveals this problem may related to Multi-threaded emulation
for QEMU: https://lwn.net/Articles/697265/

Follow this article there are Linaro efforts developing Multi-Threaded TCG
(MTTCG) QEMU (not upstream yet), and I'll give a try in next days.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

--- Comment #3 from Yi He  ---
I'll work on this bug in this and next week and update once replicated.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

--- Comment #2 from Maxim Uvarov  ---
Any progress?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[lng-odp] [Bug 2797] scheduler_main runs for long time

[bugzilla-daemon]

https://bugs.linaro.org/show_bug.cgi?id=2797

Mike Holmes  changed:

   What|Removed |Added

 CC||mike.hol...@linaro.org

--- Comment #1 from Mike Holmes  ---
Yi

Can the length of the run vary with the machine type ?
Perhaps as crude as a compile flag ?

-- 
You are receiving this mail because:
You are on the CC list for the bug.