Vanitygen: Vanity bitcoin address generator/miner [v0.22]

[OgNasty]

I might take a look at it later, but since my experience with OpenCL is very little, no promises.

It would be awesome if somebody got this working properly with current drivers or gave a combination that will work.  So far I've tried Catalyst 11.11, 12.7, 12.8, 13.1, & the latest.  I tried them all after clean driver wipes and several different versions of the SDK to no avail.  I had been using the CPU version without issues for the longest time.  I think I opened a can of worms trying to get the GPU version to work, as I can't get past the "cpu hash gpu hash" issue that seems to have been plaguing people with this software for a couple years now.

[TheRealSteve]

If you are going to try to fix bits, at least try to use a more recent code base (with speed-ups, better altcoin support, etc.) included

[RainMan28]

Trying to generate an address on a Win7 machine with 3 R9 280x's on it. Running AMD 14.4

Whether I run 3 separate instances (one for each GPU) or just one instance (with all 3 GPUs), the machine crashes with "AMD Display Driver has stopped responding" error after running for some time.

The same happens even if I just run on GPU, after a couple of hours, or sometimes only a few minutes, it will crash.

Is there a way to "reduce" each GPU from how much it is hashing? Or run it at less than full power?

Thanks.

[MICRO]

Trying to generate an address on a Win7 machine with 3 R9 280x's on it. Running AMD 14.4

Whether I run 3 separate instances (one for each GPU) or just one instance (with all 3 GPUs), the machine crashes with "AMD Display Driver has stopped responding" error after running for some time.

The same happens even if I just run on GPU, after a couple of hours, or sometimes only a few minutes, it will crash.

Is there a way to "reduce" each GPU from how much it is hashing? Or run it at less than full power?

Thanks.

Is ur driver like newest version. I had problems with my Radeon, i needed to downgrade my driver to some older version .

[RainMan28]

I am using the 12.10 SDK, and I don't know if its the standard driver issue since I am able to get addresses to generate, but what happens is that for addresses that take longer, after a few minutes, the system reboots or the display driver crashes.

I've tried to use the -g 1024x1024 variable also but it seems to make no difference, the grid size is always 4096x2048.

Here's my output:

Code:

C:\Users\Desktop\oclvanitygen>oclvanitygen64 -D 0:1 -k -v -o output.txt
-F compressed 1RainMan
Prefix difficulty:       51529903411245 1RainMan
Difficulty: 51529903411245
Device: Tahiti
Vendor: Advanced Micro Devices, Inc. (1002)
Driver: 1016.4 (VM)
Profile: FULL_PROFILE
Version: OpenCL 1.2 AMD-APP (1016.4)
Max compute units: 32
Max workgroup size: 256
Global memory: -1073741824
Max allocation: -1343225856
OpenCL compiler flags: -DDEEP_PREPROC_UNROLL -DVERY_EXPENSIVE_BRANCHES -DDEEP_VL
IW -DCOMPRESSED_ADDRESS
Loading kernel binary 5599b51694bf79accd1be695478549f0.oclbin
Grid size: 4096x2048
Modular inverse: 8192 threads, 1024 ops each
Using OpenCL prefix matcher
GPU idle: 10.20%
[11.35 Mkey/s][total 218103808][Prob 0.0%][50% in 36.4d]

After 5-10 mins, it usually crashes. This is on a R9 280x (I have 3 total cards in the system but even just trying on one, makes the system crash).

[LtPaxIV]

I might take a look at it later, but since my experience with OpenCL is very little, no promises.

It would be awesome if somebody got this working properly with current drivers or gave a combination that will work.  So far I've tried Catalyst 11.11, 12.7, 12.8, 13.1, & the latest.  I tried them all after clean driver wipes and several different versions of the SDK to no avail.  I had been using the CPU version without issues for the longest time.  I think I opened a can of worms trying to get the GPU version to work, as I can't get past the "cpu hash gpu hash" issue that seems to have been plaguing people with this software for a couple years now.

catalyst 13.1 and 2.6 SDK should be ok,as it was the last combination i had, or are u in need to use the latest SDK ?

[Remember remember the 5th of November]

If you are going to try to fix bits, at least try to use a more recent code base (with speed-ups, better altcoin support, etc.) included

I am not aware of any. Also, nobody should be forced to run severely old drivers to run this program, it's just not right.

EDIT:This problem was known way back with Catalyst 13 and prior, user wyzhe provided a fix, but somehow it got lost I guess, here is the repo with said fixes https://github.com/wyuzhe/vanitygen

When you download that repository's source code, you will need to apply just a small patch to file oclengine.c for this to work on latest drivers. Open file oclengine.c, find this

Code:

if (drv && (strstr(drv, "1112") || strstr(drv, "1113")))

, for simplicity's sake, just remove that line of code, recompile and test. Don't forget to remove old .oclbin files

Here is the binary for those that cannot compile. https://www.dropbox.com/s/o5genfs5ul4ryuw/OCLVanitygen.zip. For those that express concern over binaries from random people, I've provided binaries for miners in the past, but it's ultimately your choice, you've got the source, you know what to do.

Tips are appreciated(if it ends up working for you). Personally, for my 5850 and a 14.x catalyst, seems to work.

[RoadStress]

Can this be modified to run on SHA256 ASICS? If it's not fully compatible what about splitting the job half ASIC and the rest the video card?

[Bizmark13]

Can this be modified to run on SHA256 ASICS? If it's not fully compatible what about splitting the job half ASIC and the rest the video card?

I'd be interested to know this as well. Generating vanity addresses would be a great way to make use of a Block Erupter or any other obsolete SHA-256 miner.

[MICRO]

Can this be modified to run on SHA256 ASICS? If it's not fully compatible what about splitting the job half ASIC and the rest the video card?

I'd be interested to know this as well. Generating vanity addresses would be a great way to make use of a Block Erupter or any other obsolete SHA-256 miner.

I think that could be posible. Unfortunately i dont have any sha256 asic to test it out.

[Kangaderoo]

Can this be modified to run on SHA256 ASICS? If it's not fully compatible what about splitting the job half ASIC and the rest the video card?

I'd be interested to know this as well. Generating vanity addresses would be a great way to make use of a Block Erupter or any other obsolete SHA-256 miner.

I think that could be posible. Unfortunately i dont have any sha256 asic to test it out.

I doubt that will work. SHA256 is only one step of generating vanity addresses.
I played with the cpu software, porting both the SHA256 and RMD160 (probably not part the asic command set) routines to use the SIDM part of the CPU.
In theory I would now hash 4* as fast, but the speed increase was only 30%.
Generating a batch of valid ECSDA keys (also not part of the asic command set) requires the most time.

So GPU would be the best solution for speed, with the openCL code to do ECSDA generation, sha256 and RMD160 hashing as parallel as possible......

[nonnakip]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

[OgNasty]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

Thank you very much for this!

[Remember remember the 5th of November]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

A similar patch for Windows exist, I've detailed it in my post above.

[RainMan28]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

A similar patch for Windows exist, I've detailed it in my post above.

I've downloaded your patch and overwrote the files that were included in my oclvanitygen folder with those that were in your binary dropbox zip. I'm running a Windows 7 rig with 3x R9 280x with Catalyst 14.4 and it is hashing much slower now.

Previously I was using this command:

Code:

oclvanitygen64 -D 0:0 -k -v -o output.txt -F compressed 1RainMan

Now I'm running it with:

Code:

oclvanitygen -D 0:0 -k -v -o output.txt 1Rainman

and am getting around 5.6 Mkeys/s on a r9 280x ... that's significantly slower than what I was getting before (around 15 Mkeys/s+).

Any idea what I'm doing wrong?

Thanks!

[Remember remember the 5th of November]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

A similar patch for Windows exist, I've detailed it in my post above.

I've downloaded your patch and overwrote the files that were included in my oclvanitygen folder with those that were in your binary dropbox zip. I'm running a Windows 7 rig with 3x R9 280x with Catalyst 14.4 and it is hashing much slower now.

Previously I was using this command:

Code:

oclvanitygen64 -D 0:0 -k -v -o output.txt -F compressed 1RainMan

Now I'm running it with:

Code:

oclvanitygen -D 0:0 -k -v -o output.txt 1Rainman

and am getting around 5.6 Mkeys/s on a r9 280x ... that's significantly slower than what I was getting before (around 15 Mkeys/s+).

Any idea what I'm doing wrong?

Thanks!

Hmm, now that I look at it, I don't think it's producing valid addresses. Try testing with an address say 1abc and see if the output starts with 1abc. Looks like Mr. Yu wasn't correct. But I wasn't paying attention myself.

[RainMan28]

I spend several hours getting oclvanitygen working with AMD Radeon 7xxx on 64-bit Linux - Ubuntu 12.04 and Ubuntu 14.04. In the end only the following is needed:

• Ubuntu drivers
• vanitygen patch
• 1 userspace library file from AMD APP SDK v2.7

I created a minimal instruction text file. Maybe this will save someone some hours of searching the internet.

https://nastyfans.org/download/oclvanitygen.txt

A similar patch for Windows exist, I've detailed it in my post above.

I've downloaded your patch and overwrote the files that were included in my oclvanitygen folder with those that were in your binary dropbox zip. I'm running a Windows 7 rig with 3x R9 280x with Catalyst 14.4 and it is hashing much slower now.

Previously I was using this command:

Code:

oclvanitygen64 -D 0:0 -k -v -o output.txt -F compressed 1RainMan

Now I'm running it with:

Code:

oclvanitygen -D 0:0 -k -v -o output.txt 1Rainman

and am getting around 5.6 Mkeys/s on a r9 280x ... that's significantly slower than what I was getting before (around 15 Mkeys/s+).

Any idea what I'm doing wrong?

Thanks!

Hmm, now that I look at it, I don't think it's producing valid addresses. Try testing with an address say 1abc and see if the output starts with 1abc. Looks like Mr. Yu wasn't correct. But I wasn't paying attention myself.

Yep, you're right, trying to generate 1ABC spits out stuff like:

Address: 16qR3a1...

[Remember remember the 5th of November]

Alright, looks like user nonnakip was on the right path, it's the BFI INT patch that causes this problem. Try these binaries https://www.dropbox.com/s/o5genfs5ul4ryuw/OCLVanitygen.zip.

For an actual source code patch, find this

Code:

			//if (dvn && strstr(dvn, "cl_amd_media_ops"))
				//quirks |= VG_OCL_AMD_BFI_INT;

and

Code:

		//if (drv && (strstr(drv, "1112") || strstr(drv, "1113")))
			//quirks |= VG_OCL_CATALYST_WORKAROUND;

remove them and recompile.

Don't forget to remove old .oclbin files.

[RainMan28]

That seems to have gotten the speed back up to what it was.

I still have the system crash if I run two instances of it with different video cards...any idea what is causing that?

I map one instance to -D 0:0 and one to -D 0:1 and still have a third un-used GPU in the system but the system crashes.

[Remember remember the 5th of November]

That seems to have gotten the speed back up to what it was.

I still have the system crash if I run two instances of it with different video cards...any idea what is causing that?

I map one instance to -D 0:0 and one to -D 0:1 and still have a third un-used GPU in the system but the system crashes.

This sounds like a driver issue to me.