Changing the hashing algorithm

voingiappone

So, ghostlander… now we need to modify the scrypt module in cg/sg-miner.

I’ve been watching/messing out with the sph-sgminer for Darkcoin as I wanted to modify the algo (I’ll do wathever if it can distract me from my thesis! ;) ). For what I see, it’s enough to exactly “replicate” the algo from the hashing.cpp (in the wallet) to the algo.c file in the root directory. But I may be off-track… Do you think that’s the right path?

I mean, some optimization should be done (I don’t know where to start) but having a working GPU miner (albeit a non-opt one) from day one looks like a very good start, don’t you think?

So, the big question is: who is doing this? Is the algo accessible or is it still in a top secret state?

lizhi

Suffered a setback . :(

I’m compiling cgminer on win32(x86) . My IDE is Visual Studio 2013 . After I add neoscrypt.cpp and neoscrypt.h to the cgminer.vcxproj, then compily it .

But I get a lot of errors:

ROTL32, Identifier not found

SCRYPT_BLOCK_SIZE ,Undeclared identifier

stack, Unknown size

ghostlander

This is because NeoScrypt is supposed to be compiled by a C compiler, not a C++ one. The work-around is in a PM.

wrapper

Great work on the algo change. Looks like it might be worth haveing a p2pool test site, with all miner updates needed.

vizay

I believe it’s of very great importance that we don’t forget the nvidia cards out there, we want to be open and welcoming to all miners as fast as possible after release :)

wrapper

most of my GPUs have blown up in the last 2 months.

wrapper

One thing I have noticed running feathercoind and other coin daemons on a p2pool with tight memory. The 0.6 series daemon takes 190 MB compared to the 0.8 which are more like 230 MB. (can vary higher)

Both of which seems higher than they need…

That won’t be directly related to the algorythm, but the pool daemon pass round the work, so its something that might be “optimisation” In the builds.

Whilst we’re considering testing / improving performance.

Wellenreiter

I had a mail chat with one of the authors of cudaminer.
He is interested to add neoscrypt to cudaminer, but he can’t make it happen soon.

ghostlander

One thing I have noticed running feathercoind and other coin daemons on a p2pool with tight memory. The 0.6 series daemon takes 190 MB compared to the 0.8 which are more like 230 MB. (can vary higher)

Block index is held in memory. More blocks in main chain = more memory consumed.

vizay

I had a mail chat with one of the authors of cudaminer.
He is interested to add neoscrypt to cudaminer, but he can’t make it happen soon.

Is there someone official behind the ccminer? That’s also a viable option for nvidia and perhaps he/she/they have more time to add NS? :)

lizhi

I had compily cgminer 3.6.6 with MinGW . It can work with scrypt and keccak. But that work add neoscrypt is very complex and arduous . May God bless us !

------------------------------------------------------------------------
cgminer 3.6.6
------------------------------------------------------------------------


Configuration Options Summary:

  libcurl(GBT+getwork).: Enabled: -L/usr/local/curl/lib
  curses.TUI...........: FOUND: -lpdcurses
  OpenCL...............: FOUND. GPU mining support enabled
  scrypt...............: Enabled
  keccak...............: Enabled
  ADL..................: SDK found, GPU monitoring support enabled

  Avalon.ASICs.........: Disabled
  BFL.ASICs............: Disabled
  BitForce.FPGAs.......: Disabled
  BitFury.ASICs........: Disabled
  Icarus.FPGAs.........: Disabled
  Klondike.ASICs.......: Disabled
  ModMiner.FPGAs.......: Disabled

Compilation............: make (or gmake)
  CPPFLAGS.............:
  CFLAGS...............: -O2 -msse2
  LDFLAGS..............:
  LDADD................:  -L/usr/local/curl/lib compat/jansson-2.5/src/.libs/lib
jansson.a -lpthread -lOpenCL   -lws2_32 -lm

Installation...........: make install (as root if needed, with 'su' or 'sudo')
  prefix...............: /usr/local

voingiappone

Ok… so I definitely missed something: where do I find the files for the neoscrypt?

lizhi

Step by Step ,

I had compily cgminer 3.6.6 with neoscrypt on MinGW. I set --enable-neoscrypt , Compile support for neoscrypt feathercoin mining.

I get cgminer-neoscrypt.o and cgminer.exe .

Next, there is still a very large number of hard work. Wish me good luck.

kris_davison

wow thats amazing progress well done. I can feel a tip coming on!

lizhi

wow thats amazing progress well done. I can feel a tip coming on!

We have not an OpenCL kernel, there is still a very hard work.

ChekaZ

Great job lizihi! :)

Wellenreiter

Is there someone official behind the ccminer? That’s also a viable option for nvidia and perhaps he/she/they have more time to add NS? :)

yes ccminer is the second miner for NVIDIA cards.

I have no contact there, but will check on Github, if there is a contact email. Otherwise I’l lcreate a feature request.

Question: can we provide neoscrypt code to the programmer of ccminer and cudaminer, or should it remain private for the time beeing?

-------Edit--------

The owner of the ccminer repository on Github is the same than for Cudaminer.

I think the answer ill be the same: yes, but not very soon…

ghostlander

Question: can we provide neoscrypt code to the programmer of ccminer and cudaminer, or should it remain private for the time beeing?

Yes if they agree to keep it confidential until the official public release. On the other hand, if they say “yes, but not very soon”, they get the code after the public release anyway.

Flobdeth

Sorry for the post, but are there any links to what Neo is about or are you guys trying to develop it from a bastardisation of blake or something?

What are the main benefits we can look forward too for GPU mining or not 100% on stats yet? (other than maybe lower power and asic _resistance _short term)

ghostlander

It is what Scrypt has been used to be: memory intensive, ASIC resistant, more secure. It isn’t a bastardisation of anything. The complete specification will be in the press release.