I'm presuming that performance on Intel CPUs is staying the same?
Yes, some improvement for CPU can be made only for: 1. x64 because it has more registers. 2. CPUs with AVX |
|
|
|
Could you please update release dates and version numbers in the first post of the thread to make it easier to keep track of whether or not people have the latest version? I'm currently running version 0.4.
Version 0.6 released. Added: Improved performance on AMD Evergreen+ GPUs. The first Post will show version and release month. |
|
|
|
So far, I get the windows installer from the site listed, and it opens my cmd, very quickly, then exits. Quick flash.
bincoin-miner.exe is not installer, it is program itself. Just run cmd.exe, change dir to where you have downloaded the program and run: bincoin-miner.exe -o url -u user -p password |
|
|
|
bitcoin-miner.exe -o http://bitpenny.dyndns.biz:8332/ -u 1B8M8C48PBGdKtsGSVnCAPMgnxFTVxGXUB -p xyzzy i have played around with it a bit but invariably it just hangs, displaying 0 MHash/s. i can use deepbit without an issue though. WinInet firstly makes request without Login/Password. But this pool incorrectly returns "OK" HTTP code in Response Headers. |
|
|
|
Ufa: is there a place to go (other than here in the forums) to keep up on your miner updates?
I think this forum is best place for all btc-related info. And updates will be rare. |
|
|
|
There appears to be a small memory leak somewhere in this program. I haven't looked at the code yet to try and find it.
Version 0.4 uploaded: - The Memory leak fixed.
- ATI/AMD GPU adapters now supported, but performance is not better than on other OpenCL miners
|
|
|
|
I'm trying to use your miner with my pool http://deepbit.net but looks like your implementation of HTTP protocol differs from other miners. Why aren't you sending "Authorization" field in http headers ? This is WinInet HTTP implementation. It sends "Authorization" field after failed "without authorization" request. |
|
|
|
-Make the delay between work requests settable by the user through the command line instead of it being set statically at 30 seconds as it is now. This would allow the period to be decreased and prevent the unaccepted work problems that I've been having in slush's pool while using this miner.
-Make different levels of output verbosity so that the program doesn't have to throw out a huge amount of hex into the terminal. This is obviously mostly cosmetic, but it would be nice.
Implemented both these options: bitcoin-miner -a seconds -v ... |
|
|
|
Thanks for your help, but unfortunately it looks like I'll try to use your code as a guide to re-implement it as part of jgarzik's miner instead of trying to build it myself.
That is the best way. BTW I have reimplemented entire SHA-256 SSE algorithm on Assembly Language. It saves at least 10% cycles. So I think you need only this .asm module. |
|
|
|
By the way, which libraries are you using?
Our own libs. All necessary Sources of libraries you can find at http://ufasoft.com/files/ufasoft_lisp-4.30.tar.lzmaIt is other project, but uses the same framework. Additionally: bitcoin-miner uses BigInteger to compare with target. Really it is not necessary, memcmp() function is enough |
|
|
|
FYI, I downloaded the newest version (Windows .exe) from the link at the head of this topic. No timestamp appears. This is the result I get:
It should be version 0.3. Probably your browser cache gives you old file. |
|
|
|
Feature Request: a timestamp for an accepted string.
Implemented |
|
|
|
Is the MHash/s rate that displays on an Accepted the average for that acceptance, or the hash rate at the instant it was solved?
At the instant it was solved. "Rate" here is just last printed line, because after sending solution to server the NewLine character printed. |
|
|
|
I'd love to bro. I'll take a look at his control code and see where it differs from yours, and then see what I can do with integrating the two systems for maximum positive effect.
most important differences are in file bitcoin-sha256.h: 1. I precalculate first 3 SHA-256 rounds, because their results don't change on every iteration. and I skip last 3 rounds, because we need not them to calc last 32-bit word of hash [ h word]. 2. I use _mm_andnot_si128() to calc expresssions (~a & b) 3. I don't unfold SHA-256 loop. I think CPU can predict loop jumps very good. |
|
|
|
If you could post the rest of the code,
More sources was added to archive: http://ufasoft.com/files/open/bitcoin-miner.zip It still don't include some framework classes, like String, DateTime.... they have obvious implementation, but uploading them will require me to check many dependencies between my sources. |
|
|
|
So, how about the rest of the source code? I'd like to build this for linux to see how it runs there.
I know that jgarzik's will only ocmpute a certain number of hashes before returning the "I couldn't find anything" result to be sure that the result is returned in a reasonable amount of time. Does this employ the same tactic?
Source code is not secret, but it uses WinInet API, and cannot be directly recompiled. Tactic is following: Now "getwork period" == 30 sec is hardcoded, and each thread takes portions by "4*1024*1024 nonces". After trying aver 3 millions nonces thread chexk timeout (30 sec). and takes another portion of nonces or request new "GETWORK". Probably 30 sec is very long period and is the reason of "Not accepted" results. I have tested it with Pool http://mining.bitcoin.cz/. |
|
|
|
|
Show Posts
