[QUOTE=chalsall;507985]Take it amusingly.

Sometimes people don't have a clue about magnitude. What is a Watt per Hour vs. a Pound per square inch?[/QUOTE]

Way back when in one of my freshman (American college 1st-year) engineering classes, taught by an expatriate Scotsman, one of my fellow students just wasn't getting the distinction between force and pressure. Finally, Prof. Scottie, in exasperation, asks student to hold out hand as though to receive change from a storekeeper, student does so. Professor takes a heavy book and lays it flat on student's outstretched palm, says "that's forrrce." Lifts book, inserts sharp pencil between it and student's hand, pencil point in middle of palm, starts to transfer weight of book onto eraser side of pencil. Student winces, reflexively pulls hand back, professor says "...and [i]that's[/i] prrressure."

The connection trouble with the Helio X25 device may be down to the wifi antenna being nerfed when the cores were loaded and it having a poor signal to begin with, with 10 loaded cores ssh is responsive now that the signal situation has been improved.

Now that I can see what's going on during load the /proc/cpuinfo oddity seems to be down to the way big.LITTLE has been implemented. /proc/cpuinfo only shows which cores are active at any given time. When starting workers manually "sched_setaffinity: Invalid argument" occurs on some workers not others so mlucas only seems to be able to set affinities for active cores. Not being able to properly set affinities doesn't seem to matter, the workers get moved to idle cores regardless and the timings aren't off AFAIK. Still no joy probing /sys/devices/system/cpu/cpu*/...

The cores appear to be numbered from weakest to strongest as 4x4x2:
[code]Processor : AArch64 Processor rev 1 (aarch64)
processor : 0
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 1
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 2
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 3
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 4
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 5
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 6
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 7
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd03
CPU revision : 4

processor : 8
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd08
CPU revision : 1

processor : 9
model name : AArch64 Processor rev 1 (aarch64)
BogoMIPS : 26.00
BogoMIPS : 26.00
Features : fp asimd evtstrm aes pmull sha1 sha2 crc32
CPU implementer : 0x41
CPU architecture: 8
CPU variant : 0x0
CPU part : 0xd08
CPU revision : 1

Hardware : MT6797T [/code]
Still working out some issues with sim testing, namely twice now the 18432K test has depleted the battery. The other tests show fully charged battery and pretty steady 5W power draw (which sounds reasonable and well under the 3A the cable and hub can deliver). Need to figure out whether it's lying about being fully charged during smaller tests, the phone stops charging overnight for some reason, there's an issue with charging or the battery on this device, there's an issue with the model in general or if there's a weird edge case for 18432K in particular.

[QUOTE=M344587487;508181]Still working out some issues with sim testing, namely twice now the 18432K test has depleted the battery. The other tests show fully charged battery and pretty steady 5W power draw (which sounds reasonable and well under the 3A the cable and hub can deliver). Need to figure out whether it's lying about being fully charged during smaller tests, the phone stops charging overnight for some reason, there's an issue with charging or the battery on this device, there's an issue with the model in general or if there's a weird edge case for 18432K in particular.[/QUOTE]If you can, then I suggest to remove the battery from the system. Long term it will die anyway. So unless you have a source of new batteries every few years then you will at some point have to deal with having no available battery. Might as well deal with it now IMO.

Got this query re. Android-app-versus-rooted-phone-build from a prospective user:
[quote]I was wondering whether mlucas would be a suitable background task for
a cluster of arm64 servers, but perhaps server hardware isn't much
faster than phone hardware for running mlucas, so the real opportunity
would be to make an Android or iPhone app, perhaps? Do you really need
a "rooted" phone to run this code?[/quote]

Without rooting I haven't found a way to get the mlucas ELF binary to run but I could just be ignorant. With an APK you definitely don't need a rooted phone and the overhead may be minimal. The gist is you need to compile mlucas into a library function using the NDK then call it using the JNI from a barebones java app. Every app has a writable directory which the user can access which can contain the working directories.



Had a quick peek at the NDK example in android-studio weeks ago but attempting to modify it into something more complicated defeated me due to inexperience with CMakeLists among other things. It doesn't look too complicated for someone competent, the next time I'm feeling masochistic I might give it another go. If going the APK route there is a small complication in that without jumping through some annoying hoops you can only run one instance of an app, so the java wrapper will need basic worker management built in.

Final Exynos 8890 and Helio X25 benchmarks
 

Found the issue with the Helio X25 testing, the cable was limiting draw to 1A for unknown reasons, with a different cable it's now drawing what it needs which is ~7.69W under load. If it actually had ~5W draw it looked to be trading blows with the S7 efficiency wise (but not throughput) which I put down to less overhead per core making up for the process, but now that the proper power draw has been measured we see it has lower throughput than an S7 at higher power draw. The Exynos 8890 is ~60% more efficient at 2560K and ~50% more efficient at 4608K. The Helio X25 seems to scale better with FFT but that's irrelevant for the useful ranges.

[QUOTE=retina;508184]If you can, then I suggest to remove the battery from the system. Long term it will die anyway. So unless you have a source of new batteries every few years then you will at some point have to deal with having no available battery. Might as well deal with it now IMO.[/QUOTE]
The back glass is smashed on this device so I can't remove the battery without destroying the back. Given how poorly it compares to the S7's I'm just going to sell it but I take your point for the future.

[code] Combined Synthetic* Timings (msec/iter)
SoC Model Power (W) 1024K 2560K 4608K 7680K 18432K
Exynos 8890 Samsung S7 6.82 13.71 38.26 72.54 128.82 394.55
Helio X25 Elephone S8 7.69 21.55 54.62 96.17 163.34 477.80[/code]Redid the Exynos 8890 tests with 8 single core workers and used 5 dual core workers on the Helio X25 as it was faster in that config than 10 single core workers. It looks like the 20nm of the Helio X25 doesn't do it any favours compared to the Exynos's 14nm. Makes me hopeful for the performance of the 10nm and 7nm models released this year and last year.

Was on eBay looking for something unrelated just now, figured I'd have a gander at the current S7 "for parts" offerings:

[1] [url=https://www.ebay.com/itm/Samsung-Galaxy-S7-Edge-SM-G935P-Sprint-Parts-Read/123656500101?hash=item1cca807385:g:4JYAAOSwaRBcbc1m:rk:7:pf:0]Samsung Galaxy S7 Edge SM-G935P Sprint *Parts Read[/url]:
This is a parts phone only!!
Has a cracked front screen.
Is Google locked. <*** Is this a dealbreaker as far as rooting/running-Mlucas? ***
Battery doesn't stay charged long.

[2] [url=https://www.ebay.com/itm/Samsung-Galaxy-S7-Edge-For-Parts/292972691831?epid=1078096493&hash=item4436888177:g:FCoAAOSwSjpcbXW-:rk:9:pf:0]Samsung Galaxy S7 Edge For Parts 607376307780[/url]
Samsung Galaxy S7 Edge For Parts. Condition is Used. Screen right half went blank suddenly about two weeks ago. Unit had no damage before then. Took to CPR shop and they diagnosed as motherboard failure (and no charge for diagnosing!). They tried a new screen, which accounts for the marks on the back of the unit. Called Samsung and was told repair would be $70.31, so I sent it in. Less than a week later an email says repair will be $502.23!!! Ridiculous! The phone was working fine before this failure. Motherboards are available here and there is a good youtube on the repair but I am not that good. This will ship in the original box with charger and cable and the protector I had on it. Purchased directly from Samsung as unlocked. Model #SMG930UZKAX. I'm waiting on a response from Samsung re refunding the repair charge but not holding my breath.

[3] [url=https://www.ebay.com/itm/Samsung-Galaxy-S7-edge-SM-G935F-lcd-screen-for-parts-cracked-READ-1201/192824410791?epid=679668194&hash=item2ce53b02a7:g:FZgAAOSwq4xcZCPm:rk:33:pf:0]Samsung Galaxy S7 edge SM-G935F lcd screen for parts cracked READ#1201[/url]
For parts. Glass broken . Digitizer working.Lcd has black stamp.

[QUOTE=ewmayer;509299]Was on eBay looking for something unrelated just now, figured I'd have a gander at the current S7 "for parts" offerings:

[1] [URL="https://www.ebay.com/itm/Samsung-Galaxy-S7-Edge-SM-G935P-Sprint-Parts-Read/123656500101?hash=item1cca807385:g:4JYAAOSwaRBcbc1m:rk:7:pf:0"]Samsung Galaxy S7 Edge SM-G935P Sprint *Parts Read[/URL]:
This is a parts phone only!!
Has a cracked front screen.
Is Google locked. <*** Is this a dealbreaker as far as rooting/running-Mlucas? ***
Battery doesn't stay charged long.[/QUOTE]If it's FRP it's a dealbreaker as it means you can't get into the OS without logging in with a prior connected account. It's probably FRP but is worded vaguely enough that you should ask them what they mean.

[QUOTE=ewmayer;509299] [2] [URL="https://www.ebay.com/itm/Samsung-Galaxy-S7-Edge-For-Parts/292972691831?epid=1078096493&hash=item4436888177:g:FCoAAOSwSjpcbXW-:rk:9:pf:0"]Samsung Galaxy S7 Edge For Parts 607376307780[/URL]
Samsung Galaxy S7 Edge For Parts. Condition is Used. Screen right half went blank suddenly about two weeks ago. Unit had no damage before then. Took to CPR shop and they diagnosed as motherboard failure (and no charge for diagnosing!). They tried a new screen, which accounts for the marks on the back of the unit. Called Samsung and was told repair would be $70.31, so I sent it in. Less than a week later an email says repair will be $502.23!!! Ridiculous! The phone was working fine before this failure. Motherboards are available here and there is a good youtube on the repair but I am not that good. This will ship in the original box with charger and cable and the protector I had on it. Purchased directly from Samsung as unlocked. Model #SMG930UZKAX. I'm waiting on a response from Samsung re refunding the repair charge but not holding my breath.[/QUOTE]If it can get into the OS and the mobo damage is limited to the screen you may be able to use it, a working touch would make things easier. I have a device with a half broken screen but the touch still works across all of it. By orientating the screen various ways you could muddle your way through setting it up, a USB to HDMI could help: [URL]https://www.ebay.co.uk/itm/1080P-MHL-Micro-USB-MHL-to-HDMI-Cable-Adapter-for-Samsung-Galaxy-S6-S7-S8-Plus/152706907230?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649[/URL]


If touch doesn't work you can plug in a mouse or keyboard. I doubt you can plug a screen and mouse at the same time but don't have the wires to try.


[QUOTE=ewmayer;509299][3] [URL="https://www.ebay.com/itm/Samsung-Galaxy-S7-edge-SM-G935F-lcd-screen-for-parts-cracked-READ-1201/192824410791?epid=679668194&hash=item2ce53b02a7:g:FZgAAOSwq4xcZCPm:rk:33:pf:0"]Samsung Galaxy S7 edge SM-G935F lcd screen for parts cracked READ#1201[/URL]
For parts. Glass broken . Digitizer working.Lcd has black stamp.[/QUOTE]
Picture shows they're in android which is a good sign, don't know what they mean by black stamp. There is a second screen for notifications or a clock or something while it's locked which may be what's broken, that would display as a rectangle possibly black and is a non-issue AFAIK.

Thanks - I asked the followup questions you suggested (FRP? for [1], 'OS ok and touch yes-or-no' for [2]) for the first 2 items ... what would you suggest by way of make Offer for [3]? The ask of $40 (+ free ship) seems not too bad to begin with.

[QUOTE=M344587487;509303]If it's FRP it's a dealbreaker as it means you can't get into the OS without logging in with a prior connected account.[/QUOTE][url]https://www.samsung.com/us/support/frp/[/url][quote=https://www.samsung.com/us/support/frp/]If you can't remember any of your Google Account information, you can send your device, along with proof of purchase, to an authorized Samsung Service Center. Samsung has special tools to reset the phone to factory defaults with no protection enabled.[/quote]I guess the "special tool" is just a re-programming of the flash. So a JTAG programmer is all that is needed. Or become friends with someone who works at a Samsung Service Center.

[size=1]Of course the "special tools" override makes a mockery of the "prevents use of a device" claim. Standard tech marketing: pretend it is safe and secure when really it isn't.[/size]

The lack of a back makes the third one less appealing for normal buyers, I'd try $25 as a somewhat fair bargain. You get three attempts to make offers and how it normally works is you meet in the middle or the seller refuses the offer and you creep up. The seller can also set a minimum offer that isn't automatically rejected if they're not interested in playing the game. I wouldn't feel guilty about lowballing, as a seller I find it amusing when someone does it to me.


The S10 is coming out next week which is interesting. Most people are on contract so it won't be instant but I'm hoping in the coming months that the market will be flooded with cheap S8 and S9. The S10 is quite a big upgrade over the S8/S9, unlike the S9 to the S8, so with a bit of luck there's going to be big demand for it. The S10 and a few of the other top end recent models have the successor to big.LITTLE called DynamIQ, which does away with the idea of clusters. With big.LITTLE different core types have to be in a different cluster, with DynamIQ they are in the same cluster and allows for more flexibility. For example the Snapdragon 855 has 1xA76, 3xA76, 4xA55, with the lone core being clocked higher. For a normal user this should feel like a big upgrade.