Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

Options
Go to last post Go to first unread
Offline alyles  
#1 Posted : 03 December 2016 18:17:25(UTC)
alyles


Rank: Advanced Member

Groups: Registered
Joined: 23/03/2016(UTC)
Posts: 125
United States

Was thanked: 40 time(s) in 26 post(s)
EDIT: I should note that the following may void warranty on chrombebook machines. I'm not advocating anyone do this on their device as it's not officially supported by ChromeOS nor by SMathStudio. I was performing my own experiment and simply publishing the results. If you attempt to perform the same actions do so at your own risk.

I recently purchased a an Acer Chromebook 14 and was curious if I could somehow get SMath installed on the device.
I really like the ChromeOS experience. It's pretty fast, light, easy to use. For my personal computing device, I don't really need any offline applications. However, I am taking a couple Coursera classes on the side and SMath is a great tool to use in performing some of the homeworks and keeping notes.

I've used SMath on other Linux devices using mono so I figured the trick would be to get mono installed on the Chromebook. Turns out there's a great tool called Crouton that will allow me to run a linux distro alongside ChromeOS. The down side is that I had to put my Chromebook into a developer mode which is less secure and also that the linux distro takes up vital space. My Chromebook only has a 32GB hard drive. I don't anticipate needing a lot of space, but I didn't want to take up 1/8th or more of the drive drive space.

To solve the second problem, I decided to install crouton on a 32 GB USB 3.0 thumb drive using steps 1-3 on this wiki.

Once that's installed I have a linux distro running alongside ChromeOS and I simply followed standard instructions for installing Mono and SMath Studio. I haven't used it a great amount yet, but it seems to work just fine. I'll keep you updated on how well it works for me.

Edited by user 03 December 2016 18:37:11(UTC)  | Reason: Added warning statement.

thanks 2 users thanked alyles for this useful post.
on 03/12/2016(UTC),  on 03/12/2016(UTC)
Offline Efried  
#2 Posted : 08 January 2018 18:46:42(UTC)
Efried

Rank: Newbie

Groups: Registered
Joined: 08/01/2018(UTC)
Posts: 5
Austria

very interesting. BTW do you have compared performance on different OS?
How does Mono support multiprocessing?
I might want migrating from Mathcad 14/15 since it does not support multiprocessing.
The open questions is whether more cores as 16 with AMDs Ryzen or much mor as 500 cores with Intels XEON PHI (used) are benefitial?
And if yes what OS would deleiver the best performance for number crunching - using monte Carlo-approaches.
May be there is need for a function - calculate the program in parallel using different parameters.
thanks a lot
Offline Jean Giraud  
#3 Posted : 09 January 2018 03:37:59(UTC)
Jean Giraud


Rank: Advanced Member

Groups: Registered
Joined: 04/07/2015(UTC)
Posts: 3,154
Canada

Was thanked: 610 time(s) in 473 post(s)
Originally Posted by: Efried Go to Quoted Post
I might want migrating from Mathcad 14/15 since it does not support multiprocessing.
The open questions is whether more cores as 16 with AMDs Ryzen or much mor as500 cores with Intels XEON PHI (used) are benefitial?
And if yes what OS would deleiver the best performance for number crunching - using monte Carlo-approaches.
May be there is need for a function - calculate the program in parallel using different parameters.
thanks a lot


1. Mathcad/PTC does not read back lower Mathcad/Mathsoft
2. All legal owers of Mathcad 11 Single User Edition got flushed om May 30 2015 15:30

Good idea to migrate to Smath. Not as powerful as MCD, but what it does, it does it well
and in many instances construct a document much faster. Absolute top for publishing !

Your question reminds me a Mathsoft Collab got pest off. He was asking to time his work
sheet that toke minutes on his brand new 4 quad ... on my e-machine < 1 sec.
Advice: return, get the $ back.
Useful stuff run on Win. Can you tell a single Engineering/Scientific that could run
on separate processor [parallel]. Eng/Scien. execute top down, no need for parallel.
Hyper computers are Automata machines for which they run stuff specifically designed for.
The Mathsoft big switch between 8 => 2001i was like a semi-compile technology.
My beta testing of 2001i concluded 1/1 speed from their own designed stuff to
approximately 3/1 for math objects from foreign.
In Mathcad, you can open as many instances as you wish, in Smath as well.
The only one that computes is the one shown on screen. All other ones are silent.
As you click on some of the silent, this one now on screen runs from where it was.
All those gadgets are piano keyboad ... maths instead of music.

Jean
Offline Jean Giraud  
#4 Posted : 09 January 2018 07:33:05(UTC)
Jean Giraud


Rank: Advanced Member

Groups: Registered
Joined: 04/07/2015(UTC)
Posts: 3,154
Canada

Was thanked: 610 time(s) in 473 post(s)
... this document crunches lot of maths 10 sec on my slow laptop.
Collabs reported running in ½ time in their PC. By itself, the most
demanding is the scalar plot integrator [ it runs 164 times].
Examine more closely: you could defend that the last Σ(data) & Deriv(data)
could run in parallel to gain ~ ½ sec. On the other hand, for project
convenience, many modules can be compacted in program.
Thus, excluding // computing ... Project(f,L,H,N).

By itself, Mathcad 11 was such a bomb !
The only maths MCD 11 let you breath are the Fourier Quantum.
It runs double integrator zillion times. Here again // computing
could not be coded by any Bible. Alternately, Automata machines
do run the arithmetic [+,-,*,/] simultaneously.
In few seconds, all the world fingers prints data base can be examined
and reported for match [not from Windows !].

Jean

Finite Differences [Fnct, Deriv, Cumul].sm (74kb) downloaded 16 time(s).
Offline Jean Giraud  
#5 Posted : 09 January 2018 08:49:51(UTC)
Jean Giraud


Rank: Advanced Member

Groups: Registered
Joined: 04/07/2015(UTC)
Posts: 3,154
Canada

Was thanked: 610 time(s) in 473 post(s)
... recomposed this original document to expose what
could be split // ... assuming Smath could manage
a Win version suitable for two processors, including
the graphic sub-microprocessor.

Jean

Integrate_00000 [ RECAST Scalar].sm (30kb) downloaded 19 time(s).
Offline Efried  
#6 Posted : 09 January 2018 11:37:40(UTC)
Efried

Rank: Newbie

Groups: Registered
Joined: 08/01/2018(UTC)
Posts: 5
Austria

My computing times for the last block in your sheet are
2 core i5 6 GB Version 0.98.6484 i5 0.08 sec. 0.15 sec.
4 core i7 2700K 4GB Version 0.98.6580 0.08 sec. 0.19 sec.
4 core i7 Notebook 0.03 sec. 0.09 sec but only after storing and reloading
all on Win10 64 bit using SSD for programs

Seems that the newest beta version has introduced some speed improvements? Appearently no multiprocessing takes place. May be it needs a function parallelprocessing to be used in Monte Carlo Variation as example.
cheers

Edited by user 09 January 2018 15:06:05(UTC)  | Reason: improving with details

Offline Jean Giraud  
#7 Posted : 09 January 2018 18:19:23(UTC)
Jean Giraud


Rank: Advanced Member

Groups: Registered
Joined: 04/07/2015(UTC)
Posts: 3,154
Canada

Was thanked: 610 time(s) in 473 post(s)
Originally Posted by: Efried Go to Quoted Post
My computing times for the last block in your sheet are
2 core i5 6 GBVersion 0.98.6484i5 0.08 sec.0.15 sec.
4 core i7 2700K4GB Version 0.98.6580 0.08 sec.0.19 sec.
4 core i7 Notebook 0.03 sec. 0.09 sec but only after storing and reloading
all on Win10 64 bit using SSD for programs

Seems that the newest beta version has introduced some speed improvements? Appearently no multiprocessing takes place. May be it needs a function parallelprocessing to be used in Monte Carlo Variation as example.


Not clear about your timing. I understand that you mean
2 core i5 6 GBVersion 0.98.6484i5 0.08 sec ... vs 1 core 0.15 sec
Calculations aren't linear wrt GHz
As it looks the 4 core is not much faster

True: Andrey mentioned some speed improvement [from recollection].
If true that 2 core i5 6 GBVersion runs 0.08 sec compared to 1 core 0.15 sec
that's most interesting [assuming Win 10] for an eventual purchase.
I downloaded 6484 on my old XP Pro ... opened 2 work sheets = 2 scrapped *.sm
Could not recuperate/rescue.

Jean

Timing.PNG
Offline Efried  
#8 Posted : 09 January 2018 18:40:08(UTC)
Efried

Rank: Newbie

Groups: Registered
Joined: 08/01/2018(UTC)
Posts: 5
Austria

Thanks,
I upgraded to 8 GB with the old 4 core i7 on my PC and got
0.08 for Denom (x)*c(x) and 0.12 sec for Num(x)*c(x).
This does fit more into the expectations.
However as you stated multiple cores are propably not exploited, the difference in time may be attributed to the CPU frequency (but not completely since the newer i7 on the notebook is still faster having less GHz).

Best Regards
Offline alyles  
#9 Posted : 06 December 2018 05:29:28(UTC)
alyles


Rank: Advanced Member

Groups: Registered
Joined: 23/03/2016(UTC)
Posts: 125
United States

Was thanked: 40 time(s) in 26 post(s)
Originally Posted by: Efried Go to Quoted Post
very interesting. BTW do you have compared performance on different OS?
How does Mono support multiprocessing?
I might want migrating from Mathcad 14/15 since it does not support multiprocessing.
The open questions is whether more cores as 16 with AMDs Ryzen or much mor as 500 cores with Intels XEON PHI (used) are benefitial?
And if yes what OS would deleiver the best performance for number crunching - using monte Carlo-approaches.
May be there is need for a function - calculate the program in parallel using different parameters.
thanks a lot


Hi Efried.

Sorry I completely missed your reply. I initially started this thread as a simple experiment when I got my first Chromebook just to see what I would need to do to run SMath. It wasn't very user friendly. I had to put my device into a developer mode therefore bypassing much of the security features of Chromebooks and my Chromebook didn't have a lot of processing power. The Acer Chromebook 14 has an Intel Celeron N3160 processor Quad-core 1.60 GHz and 4GB of RAM. That being said, it seemed to run ok on the device. I don't recall it being laggy, but I also didn't do much that required lots of processing.


[UPDATE]
It's been two years since my initial post and I haven't done much with SMath on my Chromebook since that initial experiment. However, ChromeOS has come a long way since then and I just received a new Pixelbook. I thought I would revisit this thread in case there may be others with Chromebooks looking to use SMath.

There are a couple more methods of installation available now that the OS development has progressed. My initial method is still valid however there are a couple ways of doing this now without having to use developer mode.

SMath Android App
ChromeOS now supports the installation of native Android Apps on most devices. Therefore, the simplest way to use SMath is to install the SMath Studio. Visit this help page to install the android app store on the Chromebook if not already.

The downside to this approach is the android application is not as feature rich as the desktop application. For example, no plugins are available for the android application. However, if basically functionality is all your looking for then this is sufficient.

Linux Installation
ChromeOS has started rolling out Linux Application support (Crostini) on some devices. Note this is only in Beta. This allows a user to run a Linux VM without the need to put the device in Developer mode. This method still requires a little bit of Linux terminal knowledge for the installation and execution of Mono. Info on which chromebooks are supported and how to enable Crostini can be found here.

Once Crostini is installed installation of SMath can continue via the normal Linux Methods.

I did run into some scaling issues when launching the program. This has been a common issue with some linux apps running within Crostini. However, I was able to address these issues by following info on the wiki here.

In the end, the scaling issue was fixed using the following:

Code:
sommelier -X --scale=0.5 --dpi=160 mono "~/SMath/SMathStudio_Desktop.exe"


I wanted to perform some sort of benchmark testing in regards to Efried's initial comment. Recently, Jean posted an example that had some wide range of execution times.

On my work PC (HP ZBook 15, Gen 8 Intel i7, 16GB RAM) the execution time was around 45 seconds. Others noted between 3 and 12 minutes. On my Pixelbook (Gen 7 Intel i5, 8GB ram) running SMath within Crostini, the execution took 1 minute 28 seconds! Not bad!

Edited by user 06 December 2018 05:32:12(UTC)  | Reason: Not specified

thanks 1 user thanked alyles for this useful post.
on 06/12/2018(UTC)
Offline Jean Giraud  
#10 Posted : 06 December 2018 21:37:18(UTC)
Jean Giraud


Rank: Advanced Member

Groups: Registered
Joined: 04/07/2015(UTC)
Posts: 3,154
Canada

Was thanked: 610 time(s) in 473 post(s)
Originally Posted by: alyles Go to Quoted Post
On my work PC (HP ZBook 15, Gen 8 Intel i7, 16GB RAM) the execution time was around 45 seconds. Others noted between 3 and 12 minutes. On my Pixelbook (Gen 7 Intel i5, 8GB ram) running SMath within Crostini, the execution took 1 minute 28 seconds! Not bad!


Thanks alyles for your speedy Gonzalez:
From fresh start, timing dropped from 12 min down to 5 min [1.66GHz].
Users browsing this topic
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.