Jules81
Newbie
Offline
Activity: 32
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 08:38:41 AM |
|
do you own such multiplier already? will this really work to load 4 gpu without any performance/bandwidth disadvantages?
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 10:43:55 AM |
|
I can't seem to compile the latest commit for blake and I'm not familiar with how to handle nvcc errors like the following: nvcc -g -O2 -Xptxas "-abi=no -v" -arch=compute_10 --maxrregcount=64 --ptxas-options=-v -o blake.o -c blake.cu
### Assertion failure at line 1917 of ../../be/cg/cgemit.cxx: ### Compiler Error in file /tmp/tmpxft_000010b1_00000000-9_blake.cpp3.i during Assembly phase: ### operand 0 is not a register nvopencc INTERNAL ERROR: /usr/lib/nvidia-cuda-toolkit/lib//be returned non-zero status 1 you might want to get the latest tagged release instead. by the way, compilation on windows seemed to work fine for blake.cu EDIT: I can repro the problem now under Linux. I think I am going to replace the entire blake implementation anyway, as yesterday's approach with some simple and clean blake256 implementation didn't produce compatible hashes. So it's likely the problem will go away as a result. Christian Wouldn't be easier to start from the opencl (sgminer) code ?
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
coercion
Newbie
Offline
Activity: 34
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 10:58:25 AM |
|
you might want to get the latest tagged release instead.
by the way, compilation on windows seemed to work fine for blake.cu
EDIT: I can repro the problem now under Linux. I think I am going to replace the entire blake implementation anyway, as yesterday's approach with some simple and clean blake256 implementation didn't produce compatible hashes. So it's likely the problem will go away as a result.
Cool beans. I'll look forward to it. I have been using a commit from just a few days ago, but I want to play with blake. Not worth my time to set Windows up. p.s. check your btc wallet.
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 11:58:09 AM Last edit: March 07, 2014, 12:08:27 PM by cbuchner1 |
|
Wouldn't be easier to start from the opencl (sgminer) code ?
nope. Trying to not poison my mind with other people's optimizations for other architectures. ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) I might consider this when I hit a dead-end in my optimization efforts.
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 11:59:06 AM Last edit: March 07, 2014, 01:09:01 PM by cbuchner1 |
|
Cool beans. I'll look forward to it. I have been using a commit from just a few days ago, but I want to play with blake. Not worth my time to set Windows up.
a slow, but working blake implementation is now committed, and this time it's meant for Linux. Today I used the sph_blake code as a basis for starting out, just like the original blakecoin cpuminer does. Note that this will become much faster, once I manage to move all the data from local memory to registers. yesterday's attempt was based on blake256_light.c, with code lifted from here (except for HMAC parts) https://github.com/davidlazar/BLAKE/blob/master/blake256.cChristian
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 02:14:49 PM Last edit: March 07, 2014, 03:21:53 PM by cbuchner1 |
|
guys, there is a Chinese cryptocurrency called huitongbi and someone asked if I can make a cuda miner for it. I've looked at the wallet's source code and I find this. void scrypt_hash(const void* input, size_t inputlen, uint32_t *res, unsigned char Nfactor) { Nfactor = 9; return scrypt((const unsigned char*)input, inputlen, (const unsigned char*)input, inputlen, Nfactor, 5, 0, (unsigned char*)res, 32); }
a bit later there is this scrypt((const unsigned char*)&data, 80, (const unsigned char*)&data, 80, Nfactor, 8, 0, (unsigned char*)hash, 32); hash_count += 1;
and yet the scrypt function only takes 4 arguments static void scrypt(const void* input, size_t inputlen, uint32_t *res, void *scratchpad)
much confused... code is here: http://pan.baidu.com/wap/share/home?uk=2923272396&third=0Chinese web page is http://huitongbi.comand through google translate: http://translate.google.com/translate?sl=zh-CN&tl=en&js=n&prev=_t&hl=de&ie=UTF-8&u=http%3A%2F%2Fhuitongbi.com%2F
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 02:22:05 PM |
|
guys, there is a Chinese cryptocurrency called huitongbi and someone asked if I can make a cuda miner for it. I've looked at the wallet's source code and I find this. void scrypt_hash(const void* input, size_t inputlen, uint32_t *res, unsigned char Nfactor) { Nfactor = 9; return scrypt((const unsigned char*)input, inputlen, (const unsigned char*)input, inputlen, Nfactor, 5, 0, (unsigned char*)res, 32); }
a bit later there is this scrypt((const unsigned char*)&data, 80, (const unsigned char*)&data, 80, Nfactor, 8, 0, (unsigned char*)hash, 32); hash_count += 1;
and yet the scrypt function only takes 4 arguments static void scrypt(const void* input, size_t inputlen, uint32_t *res, void *scratchpad)
much confused... code is here: http://pan.baidu.com/wap/share/home?uk=2923272396&third=0Chinese web page is http://huitongbi.comdoes it compile ?
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
bigjme
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 02:22:34 PM |
|
Hmm so it's passing 9 values through to a 4 value array?
That seems a little strange, I will go through it and check through the code to see where it might be parsed out separately if you want? Haven't gone through the cudaminer code before. Or any of these code but is seems stupidly similar to language used for arduino
|
Owner of: cudamining.co.uk
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 03:20:38 PM |
|
does it compile ?
will try when I am at home. This could be one of the coins where having a CUDA miner first would be a nice advantage (provided that it isn't just standard scrypt mining) Christian
|
|
|
|
lagur
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 03:46:36 PM Last edit: March 07, 2014, 04:57:15 PM by lagur |
|
Is there a plan for DRK (Darkcoin) support? ![Roll Eyes](https://bitcointalk.org/Smileys/default/rolleyes.gif)
|
|
|
|
Omega0255
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 03:51:49 PM |
|
guys, there is a Chinese cryptocurrency called huitongbi and someone asked if I can make a cuda miner for it. I've looked at the wallet's source code and I find this. void scrypt_hash(const void* input, size_t inputlen, uint32_t *res, unsigned char Nfactor) { Nfactor = 9; return scrypt((const unsigned char*)input, inputlen, (const unsigned char*)input, inputlen, Nfactor, 5, 0, (unsigned char*)res, 32); }
a bit later there is this scrypt((const unsigned char*)&data, 80, (const unsigned char*)&data, 80, Nfactor, 8, 0, (unsigned char*)hash, 32); hash_count += 1;
and yet the scrypt function only takes 4 arguments static void scrypt(const void* input, size_t inputlen, uint32_t *res, void *scratchpad)
much confused... code is here: http://pan.baidu.com/wap/share/home?uk=2923272396&third=0Chinese web page is http://huitongbi.comand through google translate: http://translate.google.com/translate?sl=zh-CN&tl=en&js=n&prev=_t&hl=de&ie=UTF-8&u=http%3A%2F%2Fhuitongbi.com%2FCute name "Gray Copper Coin"
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 03:56:10 PM Last edit: March 07, 2014, 05:07:09 PM by cbuchner1 |
|
does it compile ?
buchner@Rennsemmel:~/huitongbi_src_r2$ qmake huitongbi-qt.pro Project MESSAGE: Building with UPNP support Removed plural forms as the target language has less forms. If this sounds wrong, possibly the target language is not set or recognized. ... buchner@Rennsemmel:~/huitongbi_src_r2$ make cd /home/buchner/huitongbi_src_r2/src/leveldb && CC=gcc CXX=g++ make OPT="-m64 -pipe -fstack-protector-all -D_FORTIFY_SOURCE=2 -O2" libleveldb.a libmemenv.a /bin/sh: 1: ./build_detect_platform: Permission denied make[1]: Betrete Verzeichnis '/home/buchner/huitongbi_src_r2/src/leveldb' Makefile:18: build_config.mk: Datei oder Verzeichnis nicht gefunden make[1]: *** Keine Regel, um »build_config.mk« zu erstellen. Schluss. make[1]: Verlasse Verzeichnis '/home/buchner/huitongbi_src_r2/src/leveldb' make: *** [/home/buchner/huitongbi_src_r2/src/leveldb/libleveldb.a] Fehler 2
the same computer compiled the Blakecoin wallet just fine minutes ago. but hey, copying in build_config.mk from the same folder in the Blakecoin wallet worked. Now I get this: g++ -c -m64 -pipe -fstack-protector-all -D_FORTIFY_SOURCE=2 -O2 -D_REENTRANT -fdiagnostics-show-option -Wall -Wextra -Wformat -Wformat-security -Wno-unused-parameter -Wstack-protector -DQT_WEBKIT -DBOOST_THREAD_USE_LIB -DBOOST_SPIRIT_THREADSAFE -DUSE_UPNP=1 -DSTATICLIB -DUSE_IPV6=1 -DHAVE_BUILD_INFO -DLINUX -D_FILE_OFFSET_BITS=64 -DQT_NO_DEBUG -DQT_GUI_LIB -DQT_NETWORK_LIB -DQT_CORE_LIB -DQT_SHARED -I/usr/share/qt4/mkspecs/linux-g++-64 -I/usr/include/qt4/QtCore -I/usr/include/qt4/QtNetwork -I/usr/include/qt4/QtGui -I/usr/include/qt4 -Isrc -Isrc/json -Isrc/qt -Isrc/leveldb/include -Isrc/leveldb/helpers -Ibuild -Ibuild -o build/bitcoin.o src/qt/bitcoin.cpp In file included from src/bignum.h:12:0, from src/chainparams.h:9, from src/base58.h:21, from src/walletdb.h:9, from src/wallet.h:8, from src/qt/walletmodel.h:7, from src/qt/bitcoin.cpp:7: src/util.h:125:2: Fehler: #error missing boost sleep implementation make: *** [build/bitcoin.o] Fehler 1
It's a Chinese obstacle course! the code snippet it barfs on is this: #if defined(HAVE_WORKING_BOOST_SLEEP_FOR) boost::this_thread::sleep_for(boost::chrono::milliseconds(n)); #elif defined(HAVE_WORKING_BOOST_SLEEP) boost::this_thread::sleep(boost::posix_time::milliseconds(n)); #else //should never get here #error missing boost sleep implementation #endif
I will simply keep patching bits and pieces with code I find in the Blakecoin wallet ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) NOW we're getting somewhere. The wallet contains some scrypt-jane code, and apparently we're lacking a global #define that sets which hash function we need. This is kind of bad, because it keeps me guessing. In file included from src/scrypt-jane/scrypt-jane.c:11:0: src/scrypt-jane/code/scrypt-jane-hash.h: In Funktion »scrypt_hash_init«: src/scrypt-jane/code/scrypt-jane-hash.h:19:50: Warnung: unbenutzter Parameter »S« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h: In Funktion »scrypt_hash_update«: src/scrypt-jane/code/scrypt-jane-hash.h:20:52: Warnung: unbenutzter Parameter »S« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h:20:70: Warnung: unbenutzter Parameter »in« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h:20:81: Warnung: unbenutzter Parameter »inlen« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h: In Funktion »scrypt_hash_finish«: src/scrypt-jane/code/scrypt-jane-hash.h:21:52: Warnung: unbenutzter Parameter »S« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h:21:64: Warnung: unbenutzter Parameter »hash« [-Wunused-parameter] src/scrypt-jane/code/scrypt-jane-hash.h: Auf höchster Ebene: src/scrypt-jane/code/scrypt-jane-hash.h:23:3: Fehler: #error must define a hash function!
In file included from src/scrypt-jane/scrypt-jane.c:12:0: src/scrypt-jane/code/scrypt-jane-romix.h: Auf höchster Ebene: src/scrypt-jane/code/scrypt-jane-romix.h:16:3: Fehler: unbekannter Typname: »scrypt_ROMixfn« src/scrypt-jane/code/scrypt-jane-romix.h: In Funktion »scrypt_getROMix«: src/scrypt-jane/code/scrypt-jane-romix.h:16:45: Warnung: return erzeugt Ganzzahl von Zeiger ohne Typkonvertierung [standardmäßig aktiviert] src/scrypt-jane/code/scrypt-jane-romix.h: Auf höchster Ebene: src/scrypt-jane/code/scrypt-jane-romix.h:21:3: Fehler: #error must define a mix function!
from what I gather looking at the strings in the windows wallet's EXE file, it seems to use keccak hash and chacha mix function. So maybe it's a scrypt-jane coin... what isn't clear to me yet is whether it uses keccak256 or keccak512.... From the web page (through Google Translate) "Remittance currency using the original Scrypt-HTC workload proved algorithm, its enormous amount of computation". They call it Scrypt-HTC and original. hmm.... I am joining the opinion here: https://bitcointalk.org/index.php?topic=308713.msg3329194#msg3329194scrypt-htc is reminscent of scrypt-pgc (Pangu) and scrypt-new (Kingcoin) - just another scrypt-jane that is disguised as a CPU-mining algo while the creators mine on their GPUs
Another interesting quote https://bitcointalk.org/index.php?topic=308713.msg3347456#msg3347456From sources - there is scrypt-jane, forked from yacoin, so it must be scrypt-chacha + pbkdf with keccak 512. Nfactor is 9, rfactor is 5, so for one thread must be used: for V array 2^(9+1) * 2^5 * 128 / LOOKUP_GAP = 4194304 bytes (for lookup_gap = 1) + for X array = 4096 bytes + for Y array = 4096 bytes
The ybc-miner will not work with it, as it hardcoded for rfactor = 0.
Theriotically hashing can be run on GPU, but there no gpu-miner available on the internet, maybe someone already has it.
Looking at the pool stats of the ONLY mining pool (which has 454 kHash/s) I would say the top miners are most likely GPU farms. The kHash rating per GPU would be in a similar area as for Yacoin, given that the total scratchpad size is about 4 MB. http://pool.huitongbi.com:8088/index.php?page=statistics&action=poolChristian
|
|
|
|
debido666
Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 68
Merit: 10
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 04:32:57 PM |
|
|
|
|
|
eLeMe
Newbie
Offline
Activity: 37
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:31:14 PM |
|
If (when) you get this working someone should put up a pool for us nvidia miners to bring some healthy hash to the network. ![Wink](https://bitcointalk.org/Smileys/default/wink.gif)
|
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:36:00 PM |
|
If (when) you get this working someone should put up a pool for us nvidia miners to bring some healthy hash to the network. ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) Yeah, we Westerners should just hijack that coin ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) The challenge would be to get a working stratum module for this coin (maybe forking some existing stratum repo for some scrypt-jane coin would work). but what good is that coin if I cannot find any exchanges... BEWARE: there is another coin under the HTC moniker (Hotcoin). Maybe some of you could scour the Chinese exchanges for any markets trading HTC. I've checked on bter.com and there is nada....zilch....niente I am looking into how much effort it would be to make this coin cuda-mineable....
|
|
|
|
bigjme
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:36:16 PM |
|
I've been thinking that we need our own pool for a while now
|
Owner of: cudamining.co.uk
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:47:03 PM |
|
p.s. check your btc wallet.
I just did and my eyes popped. OMG WHY? (no, I wasn't robbed, quite the opposite) Christian
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:49:05 PM |
|
I've been thinking that we need our own pool for a while now
we need our coin... something which would work with cuda but would create a huge bottleneck in opencl... (then complain about about fairness, once it is resolved like all the whiners which populate that forum)
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
bigjme
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:51:59 PM |
|
I've been thinking that we need our own pool for a while now
we need our coin... something which would work with cuda but would create a huge bottleneck in opencl... (then complain about about fairness, once it is resolved like all the whiners which populate that forum) Now this I would like to see. Christian you up for writing a special algo that works well on cuda and crappy for opencl
|
Owner of: cudamining.co.uk
|
|
|
cbuchner1 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
March 07, 2014, 06:58:06 PM Last edit: March 07, 2014, 07:18:53 PM by cbuchner1 |
|
I've been thinking that we need our own pool for a while now
we need our coin... something which would work with cuda but would create a huge bottleneck in opencl... (then complain about about fairness, once it is resolved like all the whiners which populate that forum) Now this I would like to see. Christian you up for writing a special algo that works well on cuda and crappy for opencl Not sure if this Chinese coin is "it". I mean, you've got to trade this stuff and cash out, otherwise it's worthless. But creating a particularly specialized scrypt algo like this... why not. One could fork that coin also into something much less China-centric. I lol'd at the google translation for this page: http://huitongbi.com/?cat=2"no pre-drilling team" - hear hear, the coin hasnt pre drilled (or motorboated, or whatever...) and "mineral pools"... yeah I love my daily mineral bath. Christian
|
|
|
|
|