VanitySearch (Yet another address prefix finder)

[WanderingPhilospher]

Can I specify the desired search range for private keys in "VanitySearch"? I've seen BitCrack has --keyspace but I'm most interested in "VanitySearch" from @Jean_Luc.

Not without tweaking the existing code

Please tell me how to make "VanitySearch" generate random private keys in the range of no more than 2 ^ 180 keys. I'm not good at C ++ programming, but what needs to be changed in the "Random.cpp" code to get this result?

It requires more than Random.cpp changes.  Especially if you are wanting to remove the endo and symmetry keys.

[MisterCooper]

Can I specify the desired search range for private keys in "VanitySearch"? I've seen BitCrack has --keyspace but I'm most interested in "VanitySearch" from @Jean_Luc.

Not without tweaking the existing code

Please tell me how to make "VanitySearch" generate random private keys in the range of no more than 2 ^ 180 keys. I'm not good at C ++ programming, but what needs to be changed in the "Random.cpp" code to get this result?

It requires more than Random.cpp changes.  Especially if you are wanting to remove the endo and symmetry keys.

Of all the "VanitySearch" GitHub has very high quality and works well only "@Jean_Luc". How to make it so that "VanitySearch" generates random private keys in the range of no more than 2 ^ 180 keys?

[WhyFhy]

There is a way to get custom middle? (or end / but its cheksum)

1*custom*(middle) or 1*custom(end)

[escobol]

There is a way to get custom middle? (or end / but its cheksum)

1*custom*(middle) or 1*custom(end)

Hm, not working.

[xhomerx10]

There is a way to get custom middle? (or end / but its cheksum)

1*custom*(middle) or 1*custom(end)

Hm, not working.

That's too vague.  No one can help you if you don't describe the problem.

[WanderingPhilospher]

There is a way to get custom middle? (or end / but its cheksum)

1*custom*(middle) or 1*custom(end)

Hm, not working.

Works just fine; your input should be like this:

1*23*Bama
Would produce 1xxxxxx23xxxxBama

1*23
Would produce 1xxxxxxxxxxxxxxx23

1*Bama*
Would produce 1xxxxBamaxxxxxxxx

[escobol]

now clear
Thnx

There is a way to get custom middle? (or end / but its cheksum)

1*custom*(middle) or 1*custom(end)

Hm, not working.

Works just fine; your input should be like this:

1*23*Bama
Would produce 1xxxxxx23xxxxBama

1*23
Would produce 1xxxxxxxxxxxxxxx23

1*Bama*
Would produce 1xxxxBamaxxxxxxxx

[MisterCooper]

I run VanitySearch on Ubuntu using several files to check:

Code:

@root> ./VanitySearch -u -i List01.txt -u -i List02.txt -u -i List03.txt -u -i List04.txt -u -i List05.txt -u -i List06.txt -u -i List07.txt -o result.txt -r 1

But unfortunately I cannot see in the terminal how many keys are generated and checked during the process. Is there a way to check this process through logs?

[NotATether]

I run VanitySearch on Ubuntu using several files to check:

Code:

@root> ./VanitySearch -u -i List01.txt -u -i List02.txt -u -i List03.txt -u -i List04.txt -u -i List05.txt -u -i List06.txt -u -i List07.txt -o result.txt -r 1

But unfortunately I cannot see in the terminal how many keys are generated and checked during the process. Is there a way to check this process through logs?

It prints the same text in the file for each private key found:

Pub Addr:
Priv (WIF):
Priv (HEX):

So you can try searching for all the lines that have Pub Addr in them and then tally them up using:

grep "Pub Addr" result.txt | wc -l

[MisterCooper]

I run VanitySearch on Ubuntu using several files to check:

Code:

@root> ./VanitySearch -u -i List01.txt -u -i List02.txt -u -i List03.txt -u -i List04.txt -u -i List05.txt -u -i List06.txt -u -i List07.txt -o result.txt -r 1

But unfortunately I cannot see in the terminal how many keys are generated and checked during the process. Is there a way to check this process through logs?

It prints the same text in the file for each private key found:

Pub Addr:
Priv (WIF):
Priv (HEX):

So you can try searching for all the lines that have Pub Addr in them and then tally them up using:

grep "Pub Addr" result.txt | wc -l

@NotATether I want to see how many keys are generated when I run from the CPU and when I run from the GPU? I would like to see their difference on the speed of key generation?

@NotATether and another question, does it affect the key generation speed if I add a large list   (70 million Bitcoin Addresses (P2SH) in uncompressed format)?

[NotATether]

@NotATether I want to see how many keys are generated when I run from the CPU and when I run from the GPU? I would like to see their difference on the speed of key generation?

In the interactive display it shows you the number of MKeys/s it is searching at for GPU and combined CPU/GPU use, like this:

Code:

$ ./VanitySearch -gpu 1notatether
VanitySearch v1.19
Difficulty: 10054102514374868992
Search: 1notatether [Compressed]
Start Mon Mar  1 06:53:26 2021
Base Key: 2A1422BAED92CCEAC4B23F7DF4CB207FAA88DC1D550D86EA3C6FE7D0881F08E5
Number of CPU thread: 1
GPU: GPU #0 GRID K520 (8x192 cores) Grid(64x128)
[123.15 Mkey/s][GPU 121.60 Mkey/s][Total 2^30.46][Prob 0.0%][50% in 1794.37y][Found 0]

The GPU speed is listed in the [GPU] bracket and the combined GPU+CPU speed is in the first bracket. There is no separate listing for CPU only.

(Yes, I mistakenly got a K520 so I'm stuck with compute cap 30 and CUDA Toolkit 10.2. I don't know the status of my other Tesla server yet. I guess on the flip side this makes it easier for me to run Bitcrack through a debugger even though it's non-reproducible there? )

@NotATether and another question, does it affect the key generation speed if I add a large list   (70 million Bitcoin Addresses (P2SH) in uncompressed format)?

What do you mean? VanitySearch can't take a list of addresses, only a list of prefixes, and adding more of those will slow the program down.

[MisterCooper]

What do you mean? VanitySearch can't take a list of addresses, only a list of prefixes, and adding more of those will slow the program down.

I have a large list of uncompressed addresses. Total number of 70 million addresses.
I have divided this list into 7 files. I expect VanitySearch to find a key to at least one address from the list. Of course, this is a cosmic sum in terms of chance! But theoretical is it possible? Or won't it work? That is, in terms of selecting a perfix for the full value consisting of 33-34 characters "Base58 (P2SH)".

List: "10 000 addr (P2SH)":

Code:

1LAqmtxdjfsM2xsVVthsMUPsZ4Lw2X2sen
1JaS1LCj7m8BfJyexEqMVmSYs8RJhcRasy
1PTyMf9Z2TagnkMJVFNBEzR95QFqH99NQf
1NGL7KxXiYDcxxxhapukprgm2xm3D4MSL4
...
...
...
e.t.c

To increase the speed, maybe I need to cut addresses, leaving, for example, 10-12 characters for search?

Something like this:

Code:

1LAqmtxdjf
1JaS1LCj7m
1PTyMf9Z2T
1NGL7KxXiY
...
...
...
e.t.c

[WanderingPhilospher]

What do you mean? VanitySearch can't take a list of addresses, only a list of prefixes, and adding more of those will slow the program down.

It does and will take a list of full addresses. I think Jean Luc said it was better to put in full, if you were looking for a full address.

[NotATether]

I have a large list of uncompressed addresses. Total number of 70 million addresses.
I have divided this list into 7 files. I expect VanitySearch to find a key to at least one address from the list. Of course, this is a cosmic sum in terms of chance! But theoretical is it possible? Or won't it work? That is, in terms of selecting a perfix for the full value consisting of 33-34 characters "Base58 (P2SH)".

List: "10 000 addr (P2SH)":

Code:

1LAqmtxdjfsM2xsVVthsMUPsZ4Lw2X2sen
1JaS1LCj7m8BfJyexEqMVmSYs8RJhcRasy
1PTyMf9Z2TagnkMJVFNBEzR95QFqH99NQf
1NGL7KxXiYDcxxxhapukprgm2xm3D4MSL4
...
...
...
e.t.c

To increase the speed, maybe I need to cut addresses, leaving, for example, 10-12 characters for search?

Something like this:

Code:

1LAqmtxdjf
1JaS1LCj7m
1PTyMf9Z2T
1NGL7KxXiY
...
...
...
e.t.c

It's not as much as the type of address you're searching for (bech32 does slightly more operations than the other two) or the length of your prefixes, the program stores all of the prefixes in an std::vector, so ignoring memory requirements (70,000,000*34 characters makes 238 megabytes to store the data itself so it's doable) it iterates over all of those addresses, vector is not exactly a fast container type for iteration so you'll be spending too much time waiting for a key to be tried for all 70 million addresses to have a realistic chance of finding a match.

This can be fixed by storing all the prefixes in a bloom filter instead of a vector, but then again, that's not exactly the purpose of VanitySearch.

What do you mean? VanitySearch can't take a list of addresses, only a list of prefixes, and adding more of those will slow the program down.

It does and will take a list of full addresses. I think Jean Luc said it was better to put in full, if you were looking for a full address.

There are special codepaths taken when VanitySearch detects a full address that check the ripemd160 hash for a match instead of a lengthy strncmp.

[NotATether]

Also some benchmark results for my K520 in lieu of DaveF's list:

Number of CPU thread: 8 <--- Xeon 2650 8 threads
GPU: GPU #0 GRID K520 (8x192 cores) Grid(64x128) 
128.80 Mkey/s (GPU 117.41 Mkey/s)

So Volta is already 10x faster than equivalent Kepler datacenter GPUs 

I got a good deal for a Tesla T4 that's better than a V100 so I'm eagerly awaiting for that to come and I'll post benchmarks of that too.

[WanderingPhilospher]

There are special codepaths taken when VanitySearch detects a full address that check the ripemd160 hash for a match instead of a lengthy strncmp.

I think that is part of the problem, at least with the 30xx series. I can run 20Mil addresses to 50000000 billion times 3.14 on the 20xx cards with no issues but you have to choke addresses down to a strncmp in order for 30xx cards to run it. And then you can only run one partial address. If you have more than one, the 30xx cards throw the misaligned error. Also, I am running v17 but if I try running the newest version, the 20xx cards throw that error; on cuda 10.2 and cuda 11.  So something changed in the code from 17 til now plus maybe something in cuda, has caused issues with 20xx cards.

Do you know off the top of your head if bitcrack uses RIPE vs strncmp? I know it initially checks to make sure address is valid but I don't know how it searches for them; haven't got deep into that part of code with bitcrack.

[NotATether]

Do you know off the top of your head if bitcrack uses RIPE vs strncmp? I know it initially checks to make sure address is valid but I don't know how it searches for them; haven't got deep into that part of code with bitcrack.

It's using RIPE after it hashes down the pubkeys into SHA256 first. Thankfully there is no strncmp anywhere in Bitcrack, else with the number of addresses it supports searching for, runtime would be madness 

[WanderingPhilospher]

Do you know off the top of your head if bitcrack uses RIPE vs strncmp? I know it initially checks to make sure address is valid but I don't know how it searches for them; haven't got deep into that part of code with bitcrack.

It's using RIPE after it hashes down the pubkeys into SHA256 first. Thankfully there is no strncmp anywhere in Bitcrack, else with the number of addresses it supports searching for, runtime would be madness 

I did a test a few days ago, VS loaded and ran a 20 Mil + address list, searching for both uncomp and comp, BitCrack could not. It would sit and sit and eventually "time out" with no code or explanation.  I was getting 1200Mkey/s with 20 Mil + addresses while searching for u and c...just can't run full addresses with 30xx cards.

[WhyFhy]

Also some benchmark results for my K520 in lieu of DaveF's list:

Number of CPU thread: 8 <--- Xeon 2650 8 threads
GPU: GPU #0 GRID K520 (8x192 cores) Grid(64x128)  
128.80 Mkey/s (GPU 117.41 Mkey/s)

So Volta is already 10x faster than equivalent Kepler datacenter GPUs  

I got a good deal for a Tesla T4 that's better than a V100 so I'm eagerly awaiting for that to come and I'll post benchmarks of that too.

how much did you pay for your t4 cause I'm selling them for half here what I am on ebay. I also have rtx6000 and pci v100s
Also t4 isnt better then v100 in my use case scenarios

[NotATether]

how much did you pay for your t4 cause I'm selling them for half here what I am on ebay. I also have rtx6000 and pci v100s
Also t4 isnt better then v100 in my use case scenarios

Leaseweb wants to sell one to me for 300€ a month, this includes the cost of the SSD, CPU and memory sticks so I'm inclined to think that the T4 by itself is worth ~180€/month. So assuming it matches the MSRP in 6 months the T4 could be 1,080€.

Is rtx6000 a typo?