Open source OpenCL GPU miner by girino - Binaries for win, linux, osx (1% fee)

[archit]

girino, iruu told me that the radeons before r9x didn't allow continuous buffers to be larger than 512 MB. Did you implement a work around for that?

[girino]

That's a very good suggestion and i might implement it in a later version. Or someone else might do it, since the project is open source. Now, for the moment being, I don't have this option and if you really want to make this work, i need you to send me the exact parameters AND the exact output of the program. So please, please, pretty please, pretty pretty please, be so kind as to send me a screenshot or to copy and paste the contents of your terminal.

[mich431]

its cygwin64 and i used it with different parameters - all led to the same result - tryed the latest version - the amd sdk is installed (otherwise i wouldnt be able to build it) and the output is something like "segmentation fault (core dumped)" or something. Maybe catching all exceptions and priniting the stacktrace to the console could help to debug it, so that you at least know where in the code the exception occured...

[girino]

i get segmentation faults... any idea why?

happens with any setting .. using cygwin under win7 64bit with an ATI card (Vram 512 mb) - from what i understand at least the normal mode should work....

or is there a chance to get windows binaries maybe? :)

can you post the output of the program and the parameters used to launch it? Did you try with the latest version or some previous one? Are you using cygwin 32 or 64 bits? What about the SDK?

[mich431]

Soon, I am implementing this in my miner

your aropencl miner is crashing when the actual work should begun :/

i can encourage you to at least catch all exceptions and print them to the console, so that we can give you at least some kind of feedback and not just the "it crashed" feedback... just an idea

[archit]

Soon, I am implementing this in my miner

[mich431]

i get segmentation faults... any idea why?

happens with any setting .. using cygwin under win7 64bit with an ATI card (Vram 512 mb) - from what i understand at least the normal mode should work....

or is there a chance to get windows binaries maybe? :)

[girino]

What's the memory requirement for v4?

1.3 Gb. Not sure how much of it is allocated in main memory and how much in the graphic card, but my 32 bit Linux with an old GeForce couldn't run it.

[archit]

Hi. thank you!
can you release any windows binary for test?
cygwin installing taking a lot of time.

I can, soon

[renesoft]

Hi. thank you!
can you release any windows binary for test?
cygwin installing taking a lot of time.

[archit]

What's the memory requirement for v4?

[girino]

What's gpuv2, gpuv3, gpuv4 and gpuv5?

hidden options for testing different algorithms.

v4 is faster but uses huge ammounts of memory (3 passes: 1 precalculates all the hashes, 2 fills the hash table, and 3 finds collisions).
v3 is slower but uses way less memory (calculates the hash and populates the DB in the same step).
v2 is even slower (does everything in a single step, just like the original algorithm from ptsminer).
v5 is garbage (i as experimenting with binary search instead of hash table, but sorting is still too slow).

Default is gpuv4, which should work for all high end GPUs.

[archit]

What's gpuv2, gpuv3, gpuv4 and gpuv5?

[girino]

I have posted some time ago in the beeeeer.org thread about my OpenCL version of ptsminer, and now I think it is mature enough to be "released". So here it goes. You can find it at the following address:

http://github.com/girino/ptsminer

Binaries are available here: https://www.dropbox.com/sh/n4ta5olqp2g5i9l/xkr0sCTrUu

This is even faster than the old version. (it is faster than CUDA miners on amazon ec2). Same options are available. Added the new modes: gpuv7 (default), gpuv8 and gpuv9 (low memory version). gpuv7 is generally faster, but choose the one most apropriate for you.

This is an experimental miner, use at your own risk. Fully open source, no hidden "mining for the author" tricks. So please, donate some PTS if you like it or if it is useful to you. Edit: There is a 1% fee now. (30 seconds every hour or so). You can still remove it in the source code and recompile.

It is preconfigured to mine at beeeeer.org pool, but this can be changed with the -o command line switch.

Since this is a modified CPU miner, default is to mine using CPU. To mine using GPU, use the "-a gpu" switch.

Please read installation instructions, and good luck.

Edit: Added the option "-device" where you can choose which devices will be used. Use a comma separated list, with no spaces, like this:

Code Select Expand


./ptsminer -u PkyeQNn1yGV5psGeZ4sDu6nz2vWHTujf4h  -a gpu -device 0,3,4


The old way (assigning threads sequentially to each device) still works, so if you want to use the first devices or all of them you can still use "-t" option instead.