Large Bitcoin Collider (Collision Finders Pool)

[rico666]

It only happens on the cpan install of JSON.
I always manually run cpan install JSON prior to running LBC for the first time.

I'll try to improve that situation. This "return pressing" should work though.

Then after JSON installs I run the LBC client and everything except xdelta3 will install.
I then install xdelta3 using apt-get.

LBC cannot install distribution packages, because the way how this is done varies between distributions and also their package names (and some distributions do not even have xdelta3).

So this is something that has to be done always manually, but prior to calling LBC. When LBC is called, "make, gcc, openssl-devel, libgmp-devel" must be installed, else some Perl package installations will fail.

openssl-devel may be called libssl-dev on some distributions.


Rico

[Jude Austin]

Is the latest appliance broke?

Downloaded it last night and it won't boot, goes to recovery, then tells me to get bent, haha.

Thanks,
Jude

[rico666]

Is the latest appliance broke?

Downloaded it last night and it won't boot, goes to recovery, then tells me to get bent, haha.

Hm. Latest appliance is 1st appliance. Pretty old stuff.

prior to anything else, do a

Code:

./LBC -u

then, you might want to update the system a little bit:

Code:

pacman -Syu

Might take a while until I come to update this (if ever) as I'm 100% on EC arithmetics and how to get it
efficiently on GPU. A nice optimization today gave me 50000 keys/s per CPU core.
Also, the GPU client now needs only 100 MB RAM instead of 600 MB (as the bloom filter goes to GPU memory)


Rico

[Jude Austin]

Is the latest appliance broke?

Downloaded it last night and it won't boot, goes to recovery, then tells me to get bent, haha.

Hm. Latest appliance is 1st appliance. Pretty old stuff.

prior to anything else, do a

Code:

./LBC -u

then, you might want to update the system a little bit:

Code:

pacman -Syu

Might take a while until I come to update this (if ever) as I'm 100% on EC arithmetics and how to get it
efficiently on GPU. A nice optimization today gave me 50000 keys/s per CPU core.
Also, the GPU client now needs only 100 MB RAM instead of 600 MB (as the bloom filter goes to GPU memory)


Rico

Hmm, really weird about the appliance, won't boot.

Did you fix LBC for my dedicated?

I assume you are moving away from CPU and focusing on GPU.

Thanks,
Jude

[rico666]

Did you fix LBC for my dedicated?

I assume you are moving away from CPU and focusing on GPU.

Not moving away from CPU, but focussing on GPU right now. Older architectures have low priority at the moment.

For the newest gossip, see https://twitter.com/LBC_collider (it has already 2 Followers! :-)

My notebook does now 9.05 Mkeys/s - with 38% GPU load. My goal is to reach 100% GPU load with 2 CPU cores firing at it - at most.


Rico

[Jude Austin]

Did you fix LBC for my dedicated?

I assume you are moving away from CPU and focusing on GPU.

Not moving away from CPU, but focussing on GPU right now. Older architectures have low priority at the moment.

For the newest gossip, see https://twitter.com/LBC_collider (it has already 2 Followers! :-)

My notebook does now 9.05 Mkeys/s - with 38% GPU load. My goal is to reach 100% GPU load with 2 CPU cores firing at it - at most.


Rico

Nice, I completely understand and agree that you should prioritize GPU.

I will check out the Twitter also.

And nice gain on your notebook, looking forward to running it.

Thanks,
Jude

[rico666]

Nice, I completely understand and agree that you should prioritize GPU.

It's kind of Sisyphean work, but fun. Here's a small part of field element multiplication and how I convert
my CPU code to the GPU primitives.

Code:

  // ------5--------
  // c = (c >> 52)
  //   + (uint128_t)a0 * b2
  //   + (uint128_t)a1 * b1
  //   + (uint128_t)a2 * b0
  //   + (d & M) * R;
  uint128_shiftright(&c, c, 52);
  uint128_mul(&d1, a0, b2);
  uint128_mul(&d2, a1, b1);
  uint128_mul(&d3, a2, b0);
  uint128_mul(&d4, d.Lo & M, R);
  uint128_add(&c, c, d1);
  uint128_add(&c, c, d2);
  uint128_add(&c, c, d3);
  uint128_add(&c, c, d4);
  // ------5--------

I have no clue what will happen when the porting is done and I press the start button... 


Rico

[TooDumbForBitcoin]

How many days?

[rico666]

How many days?

Until what? #50? (0 - 23 days at current speed) #51? (23-81 days at current speed)

@clients: Please have a look if you haven't already a FOUND.txt in your LBC directory. #48 did rot there for about a week until the operator of the succesful client found out...

Me having a new GPU client? Probably March. I'm hitting some nasty mysterious SEGV at the moment.
Other than that it's already munging Jacobi coordinates @ GPU, so we're not far away from a x2 Speedup again.


RIco

[rico666]

Schrödinger's Cat is both alive and dead until someone has a look.

Turns out, the same is true for the private keys your LBC client has found. There may be private keys to thousands of bitcoins on your hard disk, and the wave function will only collapse if you have a look. So until you do - you may be a billionaire. 

It also turns out, the LBC pool has found #50 about a day ago and today someone - going by the id HodlAllTheCoins - had a look.

0x22bd43c2e9354 - time of find: "Not earlier than Monday 13th March 2017 10:27:20 AM UTC"


Rico

[unknownhostname]

I never find anything

[becoin]

I never find anything

Can't you search for your own pubkeys and find the corresponding privkeys? Do what others do on this funny project!

[Ollebolle22]

I never find anything

Wow, that's a lot of GKeys you have delivered
What's your setup and total speed? Botnet, or gpu?

I'm just waiting for a multi gpu client that can use at least 90% of the gpu cores. Then I will switch my mining rigs of 20 gpus at this project.

[rico666]

I never find anything

With your current Mkeys/s power, you have around 50% of the pool capacity (may be even more).
So actually the probability for you to find something is that high compared to the rest of the pool.

But luck's a bitch... My funny 0.2 Mkeys/s 32bit client could find something sooner than your keygenerator army.
It's just not very probable.

@Ollebolle22:

I'm just waiting for a multi gpu client that can use at least 90% of the gpu cores. Then I will switch my mining rigs of 20 gpus at this project.

For the past two days I've been fighting with a segmentation fault on my new GPU implementation and I will have to take a few steps back and build up a complete forensic test scenario to figure out what's happening.

LBC is multi-GPU capable already, but you need still quite a lot CPU power to be able to put some load on the GPUs.
The new client is way better, but unfortunately unstable. As soon as this is sorted out, I believe we can start to measure the pool performance in GKeys/s!

Oh - BTW - You actually can use LBC with a GPU mining software on the same card. LBC just takes some 20-30% of the GPU capacity, so you can have a co-existant operation.


Rico

[SlarkBoy]

Rico can I request change client ID?

[rico666]

Rico can I request change client ID?

Send me a PM or mail to bots@cryptoguru.org

Rico

[TooDumbForBitcoin]

Schrödinger's Cat is both alive and dead until someone has a look.

Turns out, the same is true for the private keys your LBC client has found. There may be private keys to thousands of bitcoins on your hard disk, and the wave function will only collapse if you have a look. So until you do - you may be a billionaire. 

It also turns out, the LBC pool has found #50 about a day ago and today someone - going by the id HodlAllTheCoins - had a look.

0x22bd43c2e9354 - time of find: "Not earlier than Monday 13th March 2017 10:27:20 AM UTC"


Rico

I should get a reward for asking how many days, which made you urge colliders to look.  BTW, I asked on the same day it was found - coincidence, or am I so dumb I'm smart?

[rico666]

I should get a reward for asking how many days, which made you urge colliders to look.  BTW, I asked on the same day it was found - coincidence, or am I so dumb I'm smart?

Becoin has an explanation: It's all staged and you are just my alter ego.
Also, your question was way too unspecific - like numerology. So your reward is a smiley: 

As for "urging colliders to look" - In the german thread it became clear that for people who cannot set up a correct hook another way of signalling finds should be provided. I try to think of something where the pool could signal them (by mail etc.) that one of their clients has found something, without having the information about the PK itself. Tricky.

Sidenote:

With the current All-Time-Highs of search speed and in a sudden rush of nostalgia I looked up the pool history

2016-09-14: 500 bn keys (1 tn addresses) searched
2016-09-10: New client available 3x speedup
2016-09-07: Windows clients - although quite bad - available
2016-08-29: 1st "real" pool bounty found
2016-08-10: pool inception - roughly 0.15 MKeys/s
16 Jul/Aug: stand-alone experiments, then client and pool development
2016-07-28: standalone client: 36bits searched

36bits does my Notebook today (same machine as in July 2016) in 2 hours. And the 1st 500bn keys (something that took the project like two months) are searched today within 30 minutes.


Rico

[becoin]

And the 1st 500bn keys (something that took the project like two months) are searched today within 30 minutes.

And we all are going to be dead when 2^256 keys are searched in this project within next couple of centuries.

[rico666]

And the 1st 500bn keys (something that took the project like two months) are searched today within 30 minutes.

And we all are going to be dead when 2^256 keys are searched in this project within next couple of centuries.

Get your basics right. This project is not searching 2^256 keys.

Rico