Posts by cncr04s
log in
1) Message boards : News : Back to Normal (Message 8910)
Posted 2743 days ago by cncr04s
You could always offload non project communication related web traffic to a separate vmserver located elsewhere, essentially what's listed as data-driven web pages on the server status, you don't have to have everything in one location (your home line)
I know i'm not the first person to offer, but I wish this project well and offer my assistance in server admin exp, as well as programming (C/asm) if any of it is requested.
2) Message boards : Number crunching : Multi-GPU Problem (Message 8909)
Posted 2743 days ago by cncr04s
Are these 3 cards in sli mode? If not then not having a monitor, dummy or otherwise is a more likley cause. Other causes may be the drivers them selves, I have problems with newer versions of nvidia drivers on win7 64bit, namely the driver reset and has recovered issue. I have 190.38 drivers on here and have upped my ram's voltage, this fixed my issues.
3) Message boards : Number crunching : ATI and Nvidia on Same Machine? (Message 8808)
Posted 2749 days ago by cncr04s
I have a GTX 275 that I got last year for about 250$.
AT the time that I built my system I wasn't worried about having more then one card or knew that you could number crunch with it.
Now that I crunch with my card, I realize that nvidia is behind the game in raw compute power.

I went and got a ATI 5770 from best buy, Before I say any thing else, I was able to return it for 100% refund,
Now I put it in along side my nvidia. I'm using windows 7 and was able to install the ati drivers along side my main display's nvidia drivers.
At first the catalyst control center told me I needed to have the drivers loaded, so I plugged in my other monitor into it, and I had a desktop, and was able to move windows to and from both displays.
I was confident that I had every thing set right, I mean it was working, so I load up BOINC(6.10.58) and it wasn't able to detect the ati card, I was disappointed.
I've heard of people using ati and nvidia cards in the same machine, but they were using nvidia for physx, so I'm wondering if I needed to have the ati card as the MAIN display and the nvidia as the secondary display to have both cuda and cal processing ability?
I didn't have time to try it before I took it back, I was wondering if any one can confirm or deny that having an ati main and a nvidia second would allow both processing technologies to be detected in BOINC.
I don't really want to have to get a second nvidia card to up my processing ability, not that I can sli on this board ither, it can crossfire though if I decide to go all ati
4) Message boards : Number crunching : CUDA 2.04 Application Hints, Tips, Performance, and Issues (Message 8794)
Posted 2750 days ago by cncr04s
I tried the 2.04 app, but all the workunits I get error out immediately.
Did I install it wrong, or is there something else going on.
5) Message boards : Science : collatz parity help (Message 8729)
Posted 2755 days ago by cncr04s
Alright, I've fixed up my code, it's running better now, k=19 is what i'm using now, it seems to be the top now. as for k=20, I get an annoying infinite loop involving 2 ending up as 2, and then 2 again.. and so on, I have to add a second check to the code to test for this, it's probably not worth it. Although I don't get 3 billion steps per second with my implementation, it's due to the fact it's designed for really large int's that only fit in memory, instead of a 256bit int which can fit easy into 4 registers. I get about 150 million steps per second on 6400 bit int's.
Next step is to write a verification function.
6) Message boards : Science : collatz parity help (Message 8727)
Posted 2755 days ago by cncr04s
Edit: I've noticed my blunder of a mistake, i need to apply the F function K times, not just 5 times, that was a relic from when my table was k= 5 =P I'll edit this after i fix it up
7) Message boards : Science : collatz parity help (Message 8725)
Posted 2755 days ago by cncr04s
all my memory is aligned at 8 bytes at least, this is all that is needed for a 64bit cpu, but the function I am using aligns at 16 and 64 as well.

I've tried 17,18,19,20 and they are all between 50% and 10% slower then using k = 16. I have no idea why, other than, k = 16 results in about 512k of memory, most of which can fit into my cpu's L2 cache.
Even just 17 causes a 10% loss in speed.

That packing of 3 values into 8 byte only works up to k=17 and breaks down at k=18, isn't it?

actually no, //{c,p,zz,dddd} - This is how my data is stored, d is 32bits, c is 8 bits,p is 8 bits. p is c that has been pre calculated 3 to the power of c. I only access c to add to the step counter. I need to actually make it so it c will subtract from the step count if necessary, do to the fact that applying the F function 5 times to lets say the number 1 for example would result in a overflow into the repeating sequence, when we calculate the real steps as ending at one.
8) Message boards : Science : collatz parity help (Message 8717)
Posted 2755 days ago by cncr04s
I've tried k=20, but I notice a slowdown compared to k=16, as it doesn't fit into my cpu's cache(AMD phenom), k=16 on my cpu has the fastest speeds. Keep in mind AMD cpu's arent the same as Intels. If the stock cpu apps are using k=20, you might consider testing AMD versions with k=16.
I've also implemented a method of storing all 3 data requirements in a single 8 byte location, as the first access loads at least 8 bytes into the cache, subsequent access bypass memory because it's already loaded, this resulted in a 5% speedup. I'd also take any suggestions you might have on optimizations.
9) Message boards : Science : collatz parity help (Message 8715)
Posted 2755 days ago by cncr04s
If any one is interested, I have it up and running with k=16, which seems to fit nicely in my cpu's cache. It's processing at about ~1.3 billion steps per second on 320bit starting integers, which can expand up to as big as memory allows (8 gig in my case);
10) Message boards : Science : collatz parity help (Message 8700)
Posted 2756 days ago by cncr04s

I completely understand this now, the wiki doesn't explain this to thoroughly, at least not to a regular old C++/ assembly coder.
I was very close to your solution, after spending hours reading the sentences over and over, as well as win calc in the bit setting mode, to look at patterns.
The 2 steps I had not yet thought of was, it wasn't 5 steps, but k+c[b], as well as multiplying a (shifted over k bits) by 3^c[b] + d[b].

so far I've tested these new bits of information added to my code, and the verifications of the numbers obtained at k+c[b] steps compared to a regular step by step, seem to be true.

I appreciate your help in this, this will substantially speed up my code, all I need to do are code up some very fast assembly routines to deal with the new operations being preformed.

The basis behind my bigint lib(designed for 3x+1 only) is based off what is stated in this section
It is also known that {4,2,1} is the only repeating cycle possible with fewer than 35400 terms.

I aim to test numbers that have more terms then this, I assume terms means illiterations, or the number size.
11) Message boards : Science : collatz parity help (Message 8697)
Posted 2756 days ago by cncr04s
Allot of neat information on that page, but I can't find any thing specific to the speedup of sequence simulation.
12) Message boards : Number crunching : Strange messages from server (Message 8696)
Posted 2756 days ago by cncr04s
that's strange, I can't get any new workunits
7/4/2010 10:34:32 PM Collatz Conjecture Message from server: No work sent
7/4/2010 10:34:32 PM Collatz Conjecture Message from server: Your computer has no ATI GPU

... I have a Nvidia gpu, and have old workunits running right now so, i dunno.
13) Message boards : Science : collatz parity help (Message 8678)
Posted 2758 days ago by cncr04s
I'm having a lot of trouble understanding the step ahead k steps using a parity sequence.

The "parity" section above gives a way to speed up simulation of the sequence. To jump ahead k steps on each iteration
(using the f function from that section), break up the current number into two parts, b (the k least significant bits, interpreted as an integer),
and a (the rest of the bits as an integer). The result of jumping ahead k steps can be found as:

f k(a 2k+b) = a 3^c[b]+d[b].
The c and d arrays are precalculated for all possible k-bit numbers b, where d [b] is the result of applying the f function k times to b,
and c [b] is the number of odd numbers encountered on the way. For example, if k=5, you can jump ahead 5 steps on each iteration by separating out the 5 least significant bits of a number and using:

c [0...31] = {0,3,2,2,2,2,2,4,1,4,1,3,2,2,3,4,1,2,3,3,1,1,3,3,2,3,2,4,3,3,4,5}
d [0...31] = {0,2,1,1,2,2,2,20,1,26,1,10,4,4,13,40,2,5,17,17,2,2,20,20,8,22,8,71,26,26,80,242}.

Using the above example and k = 5
Take for example the number 25, in 5 steps I know my result should be 44(76 38 19 29 44). I know that 3^c[b]+d[b] ends up as 49, being c=3 and d=22.
I've thought every way of applying the number 49 to come up with the result of 44, and I can not find it.( nor can I use it to come up with any number in the sequence)

I don't completely understand this math function as a C++ programmer. I've spent hours and hours on this and I can't figure out this math function, and apply it to some C++ code in order to help my big number tester for collatz sequences.
If any one could help me clarify what exactly this function is doing in some C or some psuedo code, I would greatly appreciate it.
14) Message boards : News : Back to Comcast... (Message 8674)
Posted 2758 days ago by cncr04s
He does not need a host, just the ISP connection. He has all the hardware he needs in his house.

I checked once about getting a T1 line into a house to set up my own ISP and it was over $1000.00 a month!

A T1 around here goes for about $250/mo plus installation w/ a 3yr contract. A single T1 would handle Collatz, but there wouldn't be any bandwidth left for other things such as working from home. So, that means 2 bonded T1s. We don't have DSL or a home phone because the AT&T service is not good in the area (20% wiring, 80% either lack of training, politics and unions, and/or just plain low IQs). We had both DSL and phone for a while, but when the DSL was down 50% of the time and the phone was down 10% of the time, we canceled both. Think a T1 line on that same last mile of wire would really work? I have my doubts.

That hosting provider I told you about in the private message also does co location, that would be the cheapest option if you have your own hardware, and not willing to use hardware they provide like a dedicated server.
15) Message boards : News : Back to Comcast... (Message 8649)
Posted 2760 days ago by cncr04s
I'm not sure if this is the proper place to suggest low cost hosting providers, but I have one that I've used twice in the recent past. They are fairly cheap, depending on the type of hardware you are needing. they offer various bandwidth options, a shared 10mbit is the cheapest for dedicated servers(at least 1 TB of transfer), they have Virtual machines as well,
send me a pm if you want to know who specifically I am talking about, I don't want to "advertise" them unwantingly

Main page · Your account · Message boards

Copyright © 2018 Jon Sonntag; All rights reserved.