Happy to see what's going on here!
I've set pogs to no new tasks and hope to get one of the arm7 wu's next time a wu is requested.
For now my Nexus (running in Dock, set to use 2 CPU's) crunches a 'single DM'. Now it has reached 32% after 24 hours.
Sounds about right. The NEON version should be almost 4 times faster than this one. Even the ARMv6 version (after the latest update) should be almost twice as fast as the old ARMv6 version you are using now....hopefully.
Cheers
HB
Hi HB, Any idea when us Raspi users will see this new version? I currently get about 48 hours a work unit using the old app with a medium overclock.
That looks like 5 times faster then my armv6 device @600mhz and your has 1300
both with app 1.03 not bad this v7.
Your device is a dualcore.
I don´t know if there are existing armv6 with multicore.
Ok if i watch after the Samsung Galaxy mini 2 for example it is to have for 99 Euro and uses an armv7 @800mhz so my estimation over the sell price was wrong.
I went out from prices of actual popular phones like nexus4 s3 or note2
I´m not disappointed if the v6 support for NEON will be dropped.
Hi HB, Any idea when us Raspi users will see this new version? I currently get about 48 hours a work unit using the old app with a medium overclock.
There can't be a NEON version for Raspi (it has an older ARMv6 w/o NEON), and the new version for the Android ARMv6 was all about compiler settings. I played around with those settings for the Raspi compile but didn't see any significant performance increase so far. I'm looking into other ways of (moderately) speeding it up, tho, so there might be a new version "soon". ;-)
But what i can see now in crunching time is that with the 1.03 app it takes a little less then 1h to crunch 1% so i have done with that phone in 20h 20 ore more then 20 percentage of the task so it´s how you said Bikeman.
The armv6 runs with doubled speed with this NEON app instead of app 1.02
Are you saying you are running the 1.03 app on your ARMv6 device??? That shouldn't work.... the 1.03 app version is compiled for ARMv7 CPUs with NEON, and ARMv6 CPUs don't come with NEON. Sure enough, your CPU doesn't report the NEON feature in the scheduler request
CPU features: swp half thumb fastmult vfp edsp java
Theoretically, Android could emulate NEON (and ARMv7) instructions by catching the "illegal instruction" interrupt that would occur in such a situation and then emulate the instruction in software, but that should be sooooooo sloooooow.
I played around with those settings for the Raspi compile but didn't see any significant performance increase so far.
You know playing around with settings (like i did too with my phone temperature settings)is not the best solution but sometimes it can help :D
I didn´t want to encourage you with that. :)
My device runs on armv6-compatible processor rev 5(v6l)@600mhz
processor features are swp half thumb fastmult vfp edsp java
OS: Adroid: 2.6.35.7-perf-CL1178095
no usual gpu´s found
last firmware update(FOTA firmware over the air) i did was for a week or two
I´m running following task
p2030.20120226.G194.26-02.01.S.b6s0g0.00000-792-4
using einsteinbinary_BRP4U version 103 (VFP) in slot 0
Yes it runs brp single dm 1.03 but it looks like it´s how you said with neon doubled as fast as with 1.02 sorry that i aborted the 1.02 so you have no comparison
now after 9:48:06 it has crunched 11,273 percentage of the task isn´t that good for this machine?
p2030.20120226.G194.26-02.01.S.b6s0g0.00000-792-4
using einsteinbinary_BRP4U version 103 (VFP) in slot 0
Yes it runs brp single dm 1.03 but it looks like it´s how you said with neon doubled as fast as with 1.02 sorry that i aborted the 1.02 so you have no comparison
I see. No, you are not running a NEON version, you are running the new 1.03 VFP version. The Neon version would show up as "einsteinbinary_BRP4U version 103 (NEON)"
Quote:
now after 9:48:06 it has crunched 11,273 percentage of the task isn´t that good for this machine?
Should be normal for this type of hardware. 600 Mhz isn't that fast, and ARMv6 is basically a 10 year old design. More modern ARMv7 with NEON and higher clock rates can be up to ca 9 times as fast (!).
Ok thanks for that info
I think then it´s not especially good for live crunching but for testing it´s ok.
I´ve send you my log of nativeboinc to MPI... i thought there´s something wrong with my device or firmware :)
delete it whatever
RE: RE: Happy to see
)
Hi HB, Any idea when us Raspi users will see this new version? I currently get about 48 hours a work unit using the old app with a medium overclock.
That looks like 5 times
)
That looks like 5 times faster then my armv6 device @600mhz and your has 1300
both with app 1.03 not bad this v7.
Your device is a dualcore.
I don´t know if there are existing armv6 with multicore.
Ok if i watch after the Samsung Galaxy mini 2 for example it is to have for 99 Euro and uses an armv7 @800mhz so my estimation over the sell price was wrong.
I went out from prices of actual popular phones like nexus4 s3 or note2
I´m not disappointed if the v6 support for NEON will be dropped.
RE: Hi HB, Any idea when
)
There can't be a NEON version for Raspi (it has an older ARMv6 w/o NEON), and the new version for the Android ARMv6 was all about compiler settings. I played around with those settings for the Raspi compile but didn't see any significant performance increase so far. I'm looking into other ways of (moderately) speeding it up, tho, so there might be a new version "soon". ;-)
Cheers
HB
Hi I'm
)
Hi
I'm confused...
Are you saying you are running the 1.03 app on your ARMv6 device??? That shouldn't work.... the 1.03 app version is compiled for ARMv7 CPUs with NEON, and ARMv6 CPUs don't come with NEON. Sure enough, your CPU doesn't report the NEON feature in the scheduler request
Theoretically, Android could emulate NEON (and ARMv7) instructions by catching the "illegal instruction" interrupt that would occur in such a situation and then emulate the instruction in software, but that should be sooooooo sloooooow.
Cheers
HB
I played around with those
)
I played around with those settings for the Raspi compile but didn't see any significant performance increase so far.
You know playing around with settings (like i did too with my phone temperature settings)is not the best solution but sometimes it can help :D
I didn´t want to encourage you with that. :)
My device runs on
)
My device runs on armv6-compatible processor rev 5(v6l)@600mhz
processor features are swp half thumb fastmult vfp edsp java
OS: Adroid: 2.6.35.7-perf-CL1178095
no usual gpu´s found
last firmware update(FOTA firmware over the air) i did was for a week or two
I´m running following task
p2030.20120226.G194.26-02.01.S.b6s0g0.00000-792-4
using einsteinbinary_BRP4U version 103 (VFP) in slot 0
Yes it runs brp single dm 1.03 but it looks like it´s how you said with neon doubled as fast as with 1.02 sorry that i aborted the 1.02 so you have no comparison
now after 9:48:06 it has crunched 11,273 percentage of the task isn´t that good for this machine?
i didn´t rooted my device or
)
i didn´t rooted my device or something like that and the firmware cames from the manufacturer.
That´s all i can say.
That´s my Scheduler log here
)
That´s my Scheduler log here from albert if that helps
2013-04-24 13:40:52.0424 [PID=9073] Request: [USER#xxxxx] [HOST#6955] [IP xxx.xxx.xxx.58] client 7.0.36
2013-04-24 13:40:52.0435 [PID=9073 ] [send] Not using matchmaker scheduling; Not using EDF sim
2013-04-24 13:40:52.0436 [PID=9073 ] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2013-04-24 13:40:52.0436 [PID=9073 ] [send] work_req_seconds: 0.00 secs
2013-04-24 13:40:52.0436 [PID=9073 ] [send] available disk 3.60 GB, work_buf_min 8640
2013-04-24 13:40:52.0436 [PID=9073 ] [send] active_frac 0.997705 on_frac 0.928794
2013-04-24 13:40:52.0436 [PID=9073 ] [send] p_vm_extensions_disabled: no
2013-04-24 13:40:52.0436 [PID=9073 ] [send] CPU features: swp half thumb fastmult vfp edsp java
2013-04-24 13:40:52.4394 [PID=9073 ] Sending reply to [HOST#6955]: 0 results, delay req 60.00
2013-04-24 13:40:52.4397 [PID=9073 ] Scheduler ran 0.403 seconds
RE: p2030.20120226.G194.2
)
I see. No, you are not running a NEON version, you are running the new 1.03 VFP version. The Neon version would show up as "einsteinbinary_BRP4U version 103 (NEON)"
Should be normal for this type of hardware. 600 Mhz isn't that fast, and ARMv6 is basically a 10 year old design. More modern ARMv7 with NEON and higher clock rates can be up to ca 9 times as fast (!).
Thanks for testing
HBE
Ok thanks for that info I
)
Ok thanks for that info
I think then it´s not especially good for live crunching but for testing it´s ok.
I´ve send you my log of nativeboinc to MPI... i thought there´s something wrong with my device or firmware :)
delete it whatever