Changing the hashing algorithm

lizhi

Not connect http, why ?

j_scheibel

i had the same problem. no idea how to fix it though. i just used one of the 2 pools listed.

Lukey_p

intel xeon cpu e5-2697 v2 @2.97 ghz

solo mining - kind of interesting how it degrades despite not running out of logical processors.

1 thread - 2655 h/s

2 threads - 5230 h/s

4 threads - 10kh/s

8 threads - 20 kh/s

12 threads - 30 kh/s

24 threads - 45 kh/s

32 threads - 64 kh/s

44 threads - 72 kh/s

pool mining

1 thread - 8.9kh/s

2 thread - 17.2kh/s

4 threads - 33.3 kh/s

8 threads - 67.6 kh/s

12 threads - 100 kh/s *pool miner tends to hang beyond here if I let it run for more than just a few seconds*

Hi, i assume you are using Hyper-V or VMware and assigning vCPUs, At a guess, i would imagine using HT cores could contend very quickly. Are you in a position to disable HT on the CPU and assign cores only? just to see if a single core gives more that 2 threads with HT on?

Just a thought.

Luke

lizhi

Yes. Blank or whatever. You can also use “Pool Mining” for solo mining (server = http://127.0.0.1, port = 19554, username and password = what you have in phoenixcoin.conf). There will be a difference in performance though.

The wallet uses the reference C code for its solo miner. CPUminer (minerd) is powered by the SSE2 assembly code. It is also used for pool mining by the wallet. Try both and report your KH/s. For example, my 3GHz 4-core A8-3870K delivers 10KH/s using the built-in miner and 8KH/s using CPUminer (yes, it’s slower in my case).

Please give me a screen-shot .

FeatherKing

I got 3.4 KH/s with 2 threads on I3-2330 2.2 GHz at pool mining.

lizhi

Intel G2020 CPU 3.1G is6.5Kh.

Lukey_p

So, i have a question… When the move is complete, do we think CPU mining will actually be viable? Is CPU mining going to be exclusive for a while before GPU? If GPU mining neoscrypt is working on release, will anyone actually use CPU?

Just throwing it out there, be nice to have an idea, Cheers

ghostlander

Not connect http, why ?

Maybe add rpcallowip=127.0.0.1, rpcuser=lizhi and rpcpassword=123 to data/testnet/phoenixcoin.conf?

j_scheibel

Hi, i assume you are using Hyper-V or VMware and assigning vCPUs, At a guess, i would imagine using HT cores could contend very quickly. Are you in a position to disable HT on the CPU and assign cores only? just to see if a single core gives more that 2 threads with HT on?

Just a thought.

Luke

I wasn’t using any kind of virtual machine. I could go in to the bios and turn off the hyper threading but I can’t imagine that helping overall since you cut the overall power by half. I probably wont be in front of the computer to play with it for a day or so, regardless, I really expect this was more of a bottleneck imposed by the managing process (ie the wallet). I mean the wallet had locking issues when using pool settings with 12 or more threads . so there might be some bugs in there. however, the underlining code that does the new scrypt seems to be running ok though. (which I thought was the point of the test :) )

mirrax

I don’t know if Lizhi is successfull with modification of cgminer?

But now two scenarios trouble me a little:

1st, botnets targeting Neoscrypt in CPU era

2nd, if first GPU miner will be private

Is this valid lamentation or not much?

lizhi

Hi guys, I’m still not successful . I was caught in the kernel . I refer to the KECCAK, try to write some code. But no progress. :'(

I cannot write similar code for Nerscrypt in driver-opencl.c and ocl.c

#ifdef USE_KECCAK
static cl_int queue_keccak_kernel(_clState *clState, dev_blk_ctx *blk, __maybe_unused cl_uint threads)
{
        cl_kernel *kernel = &clState->kernel;
        unsigned int num = 0;
        cl_int status = 0;

        status = clEnqueueWriteBuffer(clState->commandQueue,
                clState->keccak_CLbuffer, true, 0,
                KECCAK_BUFFER_SIZE, &blk->keccak_data[0],
                0, NULL,NULL);

        CL_SET_ARG(clState->keccak_CLbuffer);
        CL_SET_ARG(clState->outputBuffer);

        return status;
}
#endif

The biggest difficulty is to write neoscrypt140707.cl, This is a binary code .

		case KL_NEOSCRYPT:
			strcpy(filename, NEOSCRYPT_KERNNAME".cl");
			strcpy(binaryfilename, NEOSCRYPT_KERNNAME);
			/* Neoscrypt only supports vector 1 */
			cgpu->vwidth = 1;
			break;

lizhi

This is my project, It comes from cgminer 3.6.6

How to build:

cd /D/FTC/neoscrypt/max-cgminer-master

autoreconf -fvi

export LIBCURL_CFLAGS=-I/usr/local/curl/include

export LIBCURL_LIBS=-L/usr/local/curl/lib

CFLAGS=“-O2 -msse2” ./configure --enable-scrypt --enable-keccak --enable-neoscrypt

make

if have error information "undefined reference to `_imp__curl_easy_init’ "

you edit Makefile file , add " LIBS = -lpdcurses -lpthread -lcurl -lcurldll "

DemocraticRepublicOfDave

Hi there.

Can anyone give me the github links to the “standard” neoscrypt cpuminer and wallet?

Lizhi I have your source and am studying it. I am a newbie to opengl but I have been intending for a long time to get into it.

Lizhi, am I correct in that the neoscrypt.c/.h files in the standard reference implementation?

The main issue for me is to understand the neoscrypt algorithm first.

Cheers

Dave.

slavco

People, I’m following this discussion quite long and I want to ask if audit was performed on the neoscrypt algo?

ghostlander

People, I’m following this discussion quite long and I want to ask if audit was performed on the neoscrypt algo?

What kind of audit?

slavco

Example: What if someone manage to lower the complexity of the algorithm?
Audit in the form of public review of the algorithm by crypto comunity.

mirrax

Example: What if someone manage to lower the complexity of the algorithm?
Audit in the form of public review of the algorithm by crypto comunity.

It will be opensource…

ghostlander

Example: What if someone manage to lower the complexity of the algorithm?
Audit in the form of public review of the algorithm by crypto comunity.

I doubt they can get any sensible advantage, but it’s going public with the source code in a couple of days anyway. Anyone can test it until our livenets switch in 2 weeks or so.

slavco

Great, I’ll check it and share it! Thumbs up!

bacobob

So if i get it right the new algo in 2 weeks ;)