VanitySearch (Yet another address prefix finder)

[citb0in]

$ ./VanitySearch -gpu -stop 1cit

Code:

VanitySearch v1.19
Difficulty: 4553521
Search: 1cit [Compressed]
Start Sun Aug 28 12:02:43 2022
Base Key: FF4374C8A994CB92A73E269619F6019E79F940A6617B9B0AF8E79179A38BE6B9
Number of CPU thread: 1
GPU: GPU #0 NVIDIA GeForce RTX 3070 Laptop GPU (40x128 cores) Grid(320x128)

PubAddress: 1citMwSBYgWGTNtgmRsrb89nKbt3jDzPK
Priv (WIF): p2pkh:KwF9Dkr1yn25tZS9RoD5pDZiZMKcZNqCUz6qAUbfv414DguNxe6z
Priv (HEX): 0xBC8B37566B346D58C1D969E609FE6040B59C404DCD0530C6EACD132CA2C723

==> Only Priv (HEX) is correct, the rest is wrong.

$ ./VanitySearch -stop 1cit

Code:

VanitySearch v1.19
Difficulty: 4553521
Search: 1cit [Compressed]
Start Sun Aug 28 12:02:23 2022
Base Key: 9EA70CAD422B49B5BC29E1112C41FE14EEAAB16E971E1DCE172B8322984C3775
Number of CPU thread: 1

PubAddress: 1citmgfDZuA4GYMjRTcRVStBL4nirQpuN
Priv (WIF): p2pkh:L2Y7NazcDkpU9HFSCXGv34CK7Bjc2Wj8XVWFhJMmhDwr61s7R8rr
Priv (HEX): 0x9EA70CAD422B49B5BC29E1112C41FE14EEAAB16E971E1DCE172B8322984F580F

==> Only Priv (HEX) is correct, the rest is wrong.

According to the output on the second search while omitting the GPU flag it shows only "Number of CPU thread used" = 1 and no hint about GPU. But the error is still there, the calculation is wrong.

[1715136374]

1715136374

[nc50lc]

According to the output on the second search while omitting the GPU flag it shows only "Number of CPU thread used" = 1 and no hint about GPU. But the error is still there, the calculation is wrong.

Hmm, that's weird since I'm not experiencing that using Vanitysearch v1.19 (release).

Here's the result of ./vanitysearch -check command:

Code:

GetBase10() Results OK
Add() Results OK : 666.667 MegaAdd/sec
Mult() Results OK : 44.209 MegaMult/sec
Div() Results OK : 5.368 MegaDiv/sec
ModInv()/ModExp() Results OK
ModInv() Results OK : 531.457 KiloInv/sec
IntGroup.ModInv() Results OK : 9.421 MegaInv/sec
ModMulK1() Results OK : 39.186 MegaMult/sec
ModSquareK1() Results OK : 47.253 MegaSqr/sec
ModInv() Cost : 88.9 S
ModMulK1order() Results OK : 6.737 MegaMult/sec
ModSqrt() Results OK !
Check Generator :OK
Check Double :OK
Check Add :OK
Check GenKey :OK
Adress : 15t3Nt1zyMETkHbjJTTshxLnqPzQvAtdCe OK!
Adress : 1BoatSLRHtKNngkdXEeobR76b53LETtpyT OK!
Adress : 1Test6BNjSJC5qwYXsjwKVLvz7DpfLehy OK!
Adress : 16S5PAsGZ8VFM1CRGGLqm37XHrp46f6CTn OK!
Adress : 1Tst2RwMxZn9cYY5mQhCdJic3JJrK7Fq7 OK!
Adress : 3CyQYcByvcWK8BkYJabBS82yDLNWt6rWSx OK!
Adress : 31to1KQe67YjoDfYnwFJThsGeQcFhVDM5Q OK!
Adress : bc1q6tqytpg06uhmtnhn9s4f35gkt8yya5a24dptmn OK!
Check Calc PubKey (full) 1ViViGLEawN27xRzGrEhhYPQrZiTKvKLo :OK
Check Calc PubKey (even) 385cR5DM96n1HvBDMzLHPYcw89fZAXULJP:OK
Check Calc PubKey (odd) 18aPiLmTow7Xgu96msrDYvSSWweCvB9oBA:OK
GPUEngine: CudaGetDeviceCount CUDA driver version is insufficient for CUDA runtime version 35

A quick test using your command: ./vanitySearch -stop 1cit:

Code:

PubAddress: 1citZrQ1cZ5Hhx62v4uzv5rVHvv84hCUc
Priv (WIF): p2pkh:L2yxjxkwLKAoanYJ7f5SE2z82UVoPmGQdksjyWdsBGDn3dwLHasP
Priv (HEX): 0xABF38C98E8F446012CA1C5F40D8E96761DE1E2F10A1B06A8A642316F20EB04B6

WIF private key and address are valid and correct.

Have you compiled directly from the master branch?
If so, try to build from the 'releases tags' instead, specifically v1.19 (link).
The issue may have been introduced by the last few commits after Aug 10, 2020.

[citb0in]

Just tried this. No luck! I also tried using all the available codes for my GPU which I got with

Code:

/usr/local/cuda-11.7/bin/nvcc --list-gpu-code[/quote]

I compiled and tried all these  available GPU codes:
[code]
$ /usr/local/cuda-11.7/bin/nvcc --list-gpu-code
sm_35
sm_37
sm_50
sm_52
sm_53
sm_60
sm_61
sm_62
sm_70
sm_72
sm_75
sm_80
sm_86
sm_87

VanitySearch -check always returning errors.

Which Nvidia drivers and CUDA version do you use, can you post some more details ?[/code]

[NotATether]

Just like the result in your vanitySearch -check command, the checksum here is also invalid.

The first four Bytes of SHA256[SHA256(80F30DBBB1FFF76EBF9741202CCE4676B6258ACCAF648C040844D9796F239AD99801)] should be: 0xA3B0A194
while the invalid WIF private key has: 0x6B4E251C, so it didn't pass the checksum.

Same with the address, 1citboYAWtgbJCySZe8C7U7y36cxGfZUZ has an invalid checksum of 0xB071873E
The first four Bytes of SHA256[SHA256(0006C186C1421FF64024B3FC356AF312061DE52688)] should be: 0x054A4884
Someone has reported a similar issue (also using a GPU): https://github.com/JeanLucPons/VanitySearch/issues/124
BTW, I'm not experiencing this when only using CPU.

That means the CUDA code for generating address checksums is wrong, which is not surprising to me since VanityGen was retrofitted using Kangaroo source code (which does not have this subsystem), so probably wasn't tested thoroughly enough.

[nc50lc]

Just tried this. No luck! I also tried using all the available codes for my GPU which I got with
-snip-

You also tried it without a -gpu parameter and it still produced an invalid WIF PrvKey and address...
Anyways, have you tried to compile the source code from vanitysearch's releases instead of the master branch?

I think the best course of action is to support the existing issue in GitHub by adding more info.
The developer may need it to fix the issue.

[citb0in]

No luck here. I did

Code:

make clean && make all

without providing the argument to compile with GPU capabilities. But when I run

Code:

./VanitySearch -check

it still throws errors. Tried the release and master branch. IMHO the bug is not in the GPU/Cuda part but somewhere else ?

[nc50lc]

without providing the argument to compile with GPU capabilities. -snip-
it still throws errors. Tried the release and master branch. IMHO the bug is not in the GPU/Cuda part but somewhere else ?

I think you should start with the compiler or other requirement, see if you satisfied the required versions, specially gcc since
you already tried the release which should work without an issue, also compiled without CUDA.

Other that that, I have no other comment since I can't reproduce it.

[citb0in]

Well, I did some more checks and troubleshooting and have found maybe some valuable information that may leed finding the culprit.

What I simply did: downloading a specific release version, unzip and then just run "make" so it won't get compiled with gpu support. Then I execute "./VanitySearch -check" to see if it works or fails.

I started from latest release 1.19 and went down step by step. Release version 1.13 is the one that works without errors. Everything later than 1.13 will throw an error on my computer.

I continued playing around with 1.13, mean I first reverted everything back

Code:

make clean

then modified the Makefile and replaced lines 33-36 to fit to my system

CXX        = g++
CUDA       = /usr/local/cuda-11.7
CXXCUDA    = /usr/bin/g++-11
NVCC       = $(CUDA)/bin/nvcc

and the compiled again with GPU support. Please note that you need to specify the ccap variable name in lower-case, else it won't work. This was corrected in later version in future.

Code:

make gpu=1 ccap=86 all

Code:

./VanitySearch -check

GetBase10() Results OK
Add() Results OK : 1.250 GigaAdd/sec
Mult() Results OK : 60.606 MegaMult/sec
Div() Results OK : 8.000 MegaDiv/sec
ModInv()/ModExp() Results OK
ModInv() Results OK : 430.389 KiloInv/sec
IntGroup.ModInv() Results OK : 14.674 MegaInv/sec
ModMulK1() Results OK : 21.432 MegaMult/sec
ModSquareK1() Results OK : 20.982 MegaMult/sec
ModMulK1order() Results OK : 9.701 MegaMult/sec
ModSqrt() Results OK !
Check Generator :OK
Check Double :OK
Check Add :OK
Check GenKey :OK
Adress : 15t3Nt1zyMETkHbjJTTshxLnqPzQvAtdCe OK!
Adress : 1BoatSLRHtKNngkdXEeobR76b53LETtpyT OK!
Adress : 1Test6BNjSJC5qwYXsjwKVLvz7DpfLehy OK!
Adress : 16S5PAsGZ8VFM1CRGGLqm37XHrp46f6CTn OK!
Adress : 1Tst2RwMxZn9cYY5mQhCdJic3JJrK7Fq7 OK!
Adress : 3CyQYcByvcWK8BkYJabBS82yDLNWt6rWSx OK!
Adress : 31to1KQe67YjoDfYnwFJThsGeQcFhVDM5Q OK!
Adress : bc1q6tqytpg06uhmtnhn9s4f35gkt8yya5a24dptmn OK!
Check Calc PubKey (full) 1ViViGLEawN27xRzGrEhhYPQrZiTKvKLo :OK
Check Calc PubKey (even) 1Gp7rQ4GdooysEAEJAS2o4Ktjvf1tZCihp:OK
Check Calc PubKey (odd) 18aPiLmTow7Xgu96msrDYvSSWweCvB9oBA:OK
GPU: GPU #0 NVIDIA GeForce RTX 3070 Laptop GPU (40x0 cores) Grid(320x128)
Seed: 63852
1.488 GigaKey/sec
ComputeKeys() found 7712 items , CPU check...
^C

Then I ran a real test and the results were all correct. Works!

Summary so far: Whatever change were made after 1.13 obviously leads to the mentioned error.. Anyone is welcome to test and try to reproduce. Maybe Jean_Luc can test and proof and fix this ?

EDIT: On a different computer I download the latest master 1.19 version, unzipped, executed "make && ./VanitySearch -check" and it worked like a charm. That means on another computer there's no issue. I am not sure what the error is related to. Any hints? Jean_Luc, do you happen to be there and give us some hints?

[citb0in]

No, the second computer I used is something completely different. But I want to emphasize that I compiled with a simple

Code:

make

and thus no GPU support. The test I run after the successful compile is just

Code:

./VanitySearch -check

again, no GPU is in use here.

[citb0in]

Workaround:

Instead of using g++-11 compiler you should install and use g++-9 and re-compile from scratch. There's obviously going something wrong when using version 11. But Version9 works!!!

Code:

$ ./VanitySearch -check

$ ./VanitySearch -check
GetBase10() Results OK
Add() Results OK : 400.000 MegaAdd/sec
Mult() Results OK : 48.077 MegaMult/sec
Div() Results OK : 15.152 MegaDiv/sec
R1=1000003D1
R2=1000007A2000E90A1
Field characteristic size: 256bits
ModInv()/ModExp() Results OK
ModInv() Edge cases Results OK
.Avg = 6.13
ModInv() Results OK : 929.515 KiloInv/sec
ModInv() cycles : 2685.26
ModSqrt() Results OK !
IntGroup.ModInv() Results OK : 12.757 MegaInv/sec
ModMulK1() Results OK : 47.549 MegaMult/sec
ModSquareK1() Results OK : 53.285 MegaSqr/sec
ModInv() Cost : 114.7 S
ModMulK1order() Results OK : 8.427 MegaMult/sec
Check Generator :OK
Check Double :OK
Check Add :OK
Check GenKey :OK
Adress : 15t3Nt1zyMETkHbjJTTshxLnqPzQvAtdCe OK!
Adress : 1BoatSLRHtKNngkdXEeobR76b53LETtpyT OK!
Adress : 1Test6BNjSJC5qwYXsjwKVLvz7DpfLehy OK!
Adress : 16S5PAsGZ8VFM1CRGGLqm37XHrp46f6CTn OK!
Adress : 1Tst2RwMxZn9cYY5mQhCdJic3JJrK7Fq7 OK!
Adress : 3CyQYcByvcWK8BkYJabBS82yDLNWt6rWSx OK!
Adress : 31to1KQe67YjoDfYnwFJThsGeQcFhVDM5Q OK!
Adress : bc1q6tqytpg06uhmtnhn9s4f35gkt8yya5a24dptmn OK!
Check Calc PubKey (full) 1ViViGLEawN27xRzGrEhhYPQrZiTKvKLo :OK
Check Calc PubKey (even) 385cR5DM96n1HvBDMzLHPYcw89fZAXULJP:OK
Check Calc PubKey (odd) 18aPiLmTow7Xgu96msrDYvSSWweCvB9oBA:OK
GPU: GPU #0 NVIDIA GeForce RTX 3070 Laptop GPU (40x128 cores) Grid(320x128)
Seed: 1663082517
1.553 GigaKey/sec
ComputeKeys() found 7711 items , CPU check...
^C

and the calculated pub addresses are correct now. I'm happy so far  

[NotATether]

Workaround:

Instead of using g++-11 compiler you should install and use g++-9 and re-compile from scratch. There's obviously going something wrong when using version 11. But Version9 works!!!

I'm pretty sure version 11 or perhaps 10 introduced some default optimization flag which breaks the checksum computation algorithm for the GPU code. Might want to look into compiling with -g or -O0.

[Dr88]

Workaround:

Instead of using g++-11 compiler you should install and use g++-9 and re-compile from scratch. There's obviously going something wrong when using version 11. But Version9 works!!!

I'm pretty sure version 11 or perhaps 10 introduced some default optimization flag which breaks the checksum computation algorithm for the GPU code. Might want to look into compiling with -g or -O0.

have you python version of group addition ? i tryed to understund how VanitySearch group addition works but c+ is hard to me

[citb0in]

Sorry if this seems like a dumb question. But wouldn't it be more effective and much more performant to use an ASIC miner for the computational tasks of VanitySearch instead of a modern GPU? My GPU does 1,750 MKeys/sec and in comparison to that a GekkoScience Compac F USBminer does about 300 GHashes per second. Is there possibly a VanitySearch version that uses such a Compac F for the calculation and thus works much faster, any clues ?

[citb0in]

You're right. I thought the most difficult work is the sha256 part. I imagined running the sha256 calcs on a ASIC miner and perform the rest through GPU to save time and thus increase speed.

[citb0in]

VanitySearch may not compute a good gridsize for your GPU, so make several tries using -g options in order to find best performances.

What is the best and most efficient way to determine and use the optimum gridsize for a GPU?

[ABCbits]

VanitySearch may not compute a good gridsize for your GPU, so make several tries using -g options in order to find best performances.

What is the best and most efficient way to determine and use the optimum gridsize for a GPU?

The author was asked similar question in past and he only recommend to try different number[1]. Personally i'd recommend you to see benchmark list created by DaveF[2] as reference. Of course there are general guide to choose total block/grid on CUDA[3-4], but i've no idea whether it can be applied to VanitySearch's gridsize.

[1] https://bitcointalk.org/index.php?topic=5112311.msg54144836#msg54144836
[2] https://bitcointalk.org/index.php?topic=5112311.msg50823897#msg50823897
[3] https://stackoverflow.com/a/9986748
[4] http://selkie.macalester.edu/csinparallel/modules/CUDAArchitecture/build/html/2-Findings/Findings.html

[citb0in]

VanitySearch seems to not support the latest Nvidia RTX 4090 GPU. This list says it has CCAP 8.9 and I did edit Makefile in Vanitysearch accordingly. Unfortunately the make process fails with

8.9 unsupported compute capability

is there any known fix to this how we can have run VanitySearch with the 8.9 CCAP ?

[PrivatePerson]

Which card is faster in computing RTX 2060 Super 8GB or RTX 3060 12GB?

[citb0in]

Which card is faster in computing RTX 2060 Super 8GB or RTX 3060 12GB?

RTX 3060 is slightly better

VanitySearch seems to not support the latest Nvidia RTX 4090 GPU. This list says it has CCAP 8.9 and I did edit Makefile in Vanitysearch accordingly. Unfortunately the make process fails with

8.9 unsupported compute capability

is there any known fix to this how we can have run VanitySearch with the 8.9 CCAP ?

what does it need to make the RTX 4090 compatible with tools like VanitySearch, BitCrack, etc... ? Is this something we can fix ourselves by just adding a particular ccap file into the sources of the program so it can compile? Or does it need something on CUDA side? Any clues ?

[_Counselor]

what does it need to make the RTX 4090 compatible with tools like VanitySearch, BitCrack, etc... ? Is this something we can fix ourselves by just adding a particular ccap file into the sources of the program so it can compile? Or does it need something on CUDA side? Any clues ?

It's the compiler that doesn't support the 8.9 compute capability, not the program. Just upgrade CUDA toolkit to version 11.8 or newer.