Running Search on M1 gpu
Web3Paradise opened this issue · comments
- Build the cache
- Consider running with 50k - 500k steps
- The number of steps for a key is fixed and unknown, so the GPU choice does not affect it in any way
what are you even talking about? chatgpt?
youre not trying to brute force the seeds initial public key but a derivative and steps tells you how deep it is
read the blog post again
skip-x skip-y and enhanced all work as intended
rebryk's code is a work of art
no more updates are needed
if you cant get it to work after following the directions your gpu opencl drivers are broken or you dont have enough vram
the example public key should have a collision with seed 2108662006 and 2804 steps which is in batch 5
@Web3Paradise : how did you solve your problem of not finding a solution ? I’m in the same case : the command line of readMe never works, even with 4millions steps…
@b006 : state of art for osx and without ‑fsanitize=address for all the buffer overflows, but not every person can afford an Apple. Please respect all those across issues who can’t even build it on Linux.
@Web3Paradise : how did you solve your problem of not finding a solution ? I’m in the same case : the command line of readMe never works, even with 4millions steps…
Are you on Mac? Enable Roxetta 64 by default for your terminal, download x-tool for Mac, clone the repo, install opencl headers, install the dependencies, follow instructions in the Readme to install the tool.
Run test with the example in Readme with 3000 steps, it should work
@Web3Paradise : I don’t own an Apple product. Beside changing to CL_DEVICE_TYPE_ALL and the size of the number of devices for my Xeon Phis, I followed instructions and the example of the Readme. Peoples that didn’t to that have that problem too here.
the simple case is to try in a Linux VM to see how things are quite different.
@Web3Paradise : I don’t own an Apple product. Beside changing to
CL_DEVICE_TYPE_ALLand the size of the number of devices for my Xeon Phis, I followed instructions and the example of the Readme. Peoples that didn’t to that have that problem too here.the simple case is to try in a Linux VM to see how things are quite different.
A lot of issues with windows, try dual boot with Linux native OS it work better. Most Linux vm doesn't support graphic acceleration, so you'll only get issues running vms or windows.
@Web3Paradise : Xeons phis are ɢᴘᴜs treated as ᴄᴘᴜ, so no driver problem in addition I now have to use Linux natively. Can your confirm switching in the code CL_DEVICE_TYPE_CPU to CL_DEVICE_TYPE_GPU works on Apple along decreasing the number of devices by 1 ? If yes, you can try in a Linux ᴠᴍ using only ᴄᴘᴜꜱ though I recognize the Intel architecture is better for that type of tasks than the aarch64.