[Dev] NeoScrypt GPU Miner - Public Beta Test

vehre

No, I haven’t used that nvidia software yet. Most pressing was to get the kernel running at all and to fix the endianess issues in the protocols.

The kernel does not know anything about driver versions and currently there is nothing that could depend on that yet. So I guess there is some compiler issue.

The good news is, my radeon arrived today. So I can figure what is special about these cards. Unfortunately I do not run windows, or should I say fortunately :-)

DemocraticRepublicOfDave

yyep…
when i tried to create bin file with tmuir12

4.xx drivers:
i got double hashrate(100khs/gpu) but only HW errors.
all other setting where just as using 13.xx drivers.
14.xx created bin used at 13.xx drivers gave same results:double hash rate but only HW errors.

could there be a change to get valid hashes at this double hashrate with 14.xx bin file tweaks or something?
i could take a look how much power usage is when using 14.xx bins… if its 210W… we maybe have something

I can’t get it working at all with 14.4 on ubuntu 14.04.

@tmuir12. Thanks for the catalyst info. I was unaware that cgminer has issues with catalyst 14.x. I will downgrade my rig to 13.12 and see what I get.

Cheers

Dave

tmuir12

I’ve not tried cgminer on Linux yet. I have a VM machine with Ubuntu 14.04 on it and plan to try it this week.

Wellenreiter

…

The good news is, my radeon arrived today. So I can figure what is special about these cards. Unfortunately I do not run windows, or should I say fortunately :-)

It’s the later ;)

daimyo

win7, amd driver 13.something, radeon hd 7990, not good. What can I do?

Wellenreiter

Get an error message, when trying to start the windows version on a win7/64 PC: libgcc_s_dw2-1.dll is missing.

is it possible to link all libraries statically into the windows version?

vehre

As far as I know, are there some libraries that are not available for static linking. Anyway, why would one want to do the bother? Install the lib and be done.

Anyway here is the second google answer to the question:

http://stackoverflow.com/questions/4702732/the-program-cant-start-because-libgcc-s-dw2-1-dll-is-missing

daimyo

thats better :)

7990 default settings. Strangely, -I 13 = 35 mhs, -I 12 = 46 mhs. If the gpu temp is correct then it seems that they are not loaded at all cause thats almost idle temps

einkerl

What do i need to add the nvidia components to the build ?

vehre

What do i need to add the nvidia components to the build ?

https://developer.nvidia.com/gpu-deployment-kit

Google does the trick!

Linux: Install it to default location, or like on

Windows: add --with-nvml-dir= to the autogen.sh call.

is the path containing the include and lib directory of nvml. Make sure the .so or .dll, respectively, is on the LD_LIBRARY_PATH (Linux) or PATH (Windows) or cgminer.exe’s directory (Windows).

einkerl

i’ve installed the deployment kit but i do not use visual studio or have any nvidia hardware installed. Alot of the kit is not being installed.

I’m building in win8.1 in mingw with amd sdk 2.9 and adl sdk 6. I have found a nvml.dl_ in anothers computer divers dir. Is that the one i need ?

vehre

I am on Linux, so I can’t verify what I say:

Most of all you need the includes of the NVML. The configuration process does not check for the nvml.dll anyway, because that is only tried at runtime and the initialisation is simply skipped if the .dll is not found. Furthermore, there is no need to distribute the .dll along with cgminer, because nvml.dll comes with all CUDA toolkits.

That said: When autogen.sh aka configure reports it has found NVML support, then everything is fine for compilation.

Alpha Wolf

Was anyone ever able to get 3.7.5 to work as a solo miner?

I was unable to find a way to make it locate the wallet while trying to solo mine.

einkerl

Ok found it, in windows you only need to install nvidia gpu deployment kit

Install path is C:\Program Files\NVIDIA Corporation\GDK\nvml\

How do i link this to MinGW ?

Thank you

daimyo

So folks, assuming the gpu miner is operational, is this going livenet?

kris_davison

PXC is live with NeoScrypt now. But Feathercoin is reasonably close to announcing a fork date / block in the not too distant future. :)

We are aiming towards late Oct hopefully. Maybe sooner.

Wellenreiter

if u are using pcx wallet it does not work.

or u must compile “updated” wallet to solo mining.

miner is ok and it does solo very well…

i have compiled unofficial ufo wallet to neo fork and it does solo without any problems.

Version 3.7.6a will work with the PXC wallet for solomining, too

bye2003

Hi All,

I think it’s better to look into the current code:

If you compile it using CodeXL 1.5 with optimization disabled, the strachreg number is huge: more than 1500.

If open the optimization, the strachreg unumber is more than 400.

The key issue for neoscrypt is: it uses too much dynamic copy: calculating the buffer position for B and A.

It’s hard to make full use of the uint4 and opencl compiler nees lots of VREGS to calculate the next buffer position.

The second difficulty is: It uses too much local array and the result is indexing local array will hurt the opencl code’s performance.

Because of the above 2, the reginsers will be used very quickly and memory will spill to global memory.

Directly convert the c code to opencl is just the very first step, need to reduce the strachreg number to 0.

I have rewrite the original code and reduce the strachreg number to 372 in non-optimization mode.

I’m not modifying the cgminer3.7.2, but adding code into original cpuminer and simpliy it. (Current, working on Win8.1 only)

Without the optimization, my code on R9290 can run 95K/s and with 5 R9290, I got 440k/s on pxc.theblocksfactory.

The code can run 145k/s if open the optimization and the Strachreg number reduce to around 220.

But unfortunely, it will reproduce wrong nonce with very wired mid value. I am still testing it.

Like to discuss any techknowledge with all of you.

Thanks

Ralph

vehre

Hi,

I am the current maintainer of the neoscrypt cgminer. I am fully aware, that the kernel is sub-optimal. Actually the focus upto now was on getting it running and working with pools and wallets. Optimizing the kernel is done in the not so far feature. I already have some ideas.

Neoscrypt is especially designed to be memory (access) intensive. First of all to harden the use of ASICS. I know that the GPU has a burden with that kernel and I already had a struggle to make the neoscrpyt kernel fit for running there.

I haven’t done any profiling yet and it will be some more days before I have the time to do it.

But a question arises: You write, that you add the kernel to a cpuminer. How that? Did you also add all opencl-management code to the cpuminer? Why that double burden? I know that cgminer from a software design/ architecture perspective is bad, but adding everything anew to cpuminer seems like duplicated work.

Regards,

Andre