[ANN] [SKC] Skeincoin 0.9.3.1 | Skein-SHA2

[reorder]

reorder, what are driver version and openCL SDK version your kernel compile/work with?

Should work with any, but to get BFI_INT for bitselect without binary patching you are going to need 11.9+ and 2.6+ respectively.

AMD Kernel Analizer 2 cannot compile it on 13.10 and 2.8 - unhandled exception =( maybe you can tell what versions (catalist and SDK) do you use? Thanks

Catalyst 13.8 and SDK 2.8.1. I have not tried loading it in analyser though, but have seen it crashing on different kernels way too often, so do not expect much from it. You can try preprocessing with gcc -E and loading the result, this has worked for me sometimes.

[1714843921]

1714843921

[1714843921]

1714843921

[1714843921]

1714843921

[glitchboy]

You will also probably need this, this is how the midstate for kernel is computed:

Thanks for sharing! I'm trying to use your kernel in cgminer. Could you please take a look at this:

Code:

int skeinhashmid(unsigned char *out, const unsigned char *in)
{
	Skein_512_Ctxt_t ctx;
	Skein_512_Init  (&ctx,8*64);
	Skein_512_Update(&ctx,in,(size_t) 80);
	memcpy(out, ctx.X, 64);

	return 0;
}

#define CL_SET_BLKARG(blkvar) status |= clSetKernelArg(*kernel, num++, sizeof(uint), (void *)&blk->blkvar)
#define CL_SET_ARG(var) status |= clSetKernelArg(*kernel, num++, sizeof(var), (void *)&var)

cl_int queue_skein_kernel(_clState *clState, dev_blk_ctx *blk, __maybe_unused cl_uint threads)
{
	cl_kernel *kernel = &clState->kernel;
	unsigned int i, num = 0;
	cl_int status = 0;

	unsigned char swap[80];
	uint32_t *swap32 = swap;
	flip80(swap32, (uint32_t *)blk->work->data); // convert getwork data to big-endian

	unsigned char mid[64];
	skeinhashmid(mid, swap); // calc skein midstate?

	uint64_t *state64 = (uint64_t *)mid;
	uint32_t *data32 = (uint32_t *)swap;

	CL_SET_ARG(state64[0]); // kernel state0
	CL_SET_ARG(state64[1]); // kernel state1
	CL_SET_ARG(state64[2]); // kernel state2
	CL_SET_ARG(state64[3]); // kernel state3
	CL_SET_ARG(state64[4]); // kernel state4
	CL_SET_ARG(state64[5]); // kernel state5
	CL_SET_ARG(state64[6]); // kernel state6
	CL_SET_ARG(state64[7]); // kernel state7

	CL_SET_ARG(data32[16]); // kernel data16
	CL_SET_ARG(data32[17]); // kernel data17
	CL_SET_ARG(data32[18]); // kernel data18

	CL_SET_BLKARG(nonce); // kernel base, i.e. base nonce

	CL_SET_ARG(clState->outputBuffer);

	return status;
}

Am i doing this right?
Thanks!

[reorder]

You will also probably need this, this is how the midstate for kernel is computed:

Thanks for sharing! I'm trying to use your kernel in cgminer. Could you please take a look at this:

Code:

int skeinhashmid(unsigned char *out, const unsigned char *in)
{
	Skein_512_Ctxt_t ctx;
	Skein_512_Init  (&ctx,8*64);
	Skein_512_Update(&ctx,in,(size_t) 80);
	memcpy(out, ctx.X, 64);

	return 0;
}

#define CL_SET_BLKARG(blkvar) status |= clSetKernelArg(*kernel, num++, sizeof(uint), (void *)&blk->blkvar)
#define CL_SET_ARG(var) status |= clSetKernelArg(*kernel, num++, sizeof(var), (void *)&var)

cl_int queue_skein_kernel(_clState *clState, dev_blk_ctx *blk, __maybe_unused cl_uint threads)
{
	cl_kernel *kernel = &clState->kernel;
	unsigned int i, num = 0;
	cl_int status = 0;

	unsigned char swap[80];
	uint32_t *swap32 = swap;
	flip80(swap32, (uint32_t *)blk->work->data); // convert getwork data to big-endian

	unsigned char mid[64];
	skeinhashmid(mid, swap); // calc skein midstate?

	uint64_t *state64 = (uint64_t *)mid;
	uint32_t *data32 = (uint32_t *)swap;

	CL_SET_ARG(state64[0]); // kernel state0
	CL_SET_ARG(state64[1]); // kernel state1
	CL_SET_ARG(state64[2]); // kernel state2
	CL_SET_ARG(state64[3]); // kernel state3
	CL_SET_ARG(state64[4]); // kernel state4
	CL_SET_ARG(state64[5]); // kernel state5
	CL_SET_ARG(state64[6]); // kernel state6
	CL_SET_ARG(state64[7]); // kernel state7

	CL_SET_ARG(data32[16]); // kernel data16
	CL_SET_ARG(data32[17]); // kernel data17
	CL_SET_ARG(data32[18]); // kernel data18

	CL_SET_BLKARG(nonce); // kernel base, i.e. base nonce

	CL_SET_ARG(clState->outputBuffer);

	return status;
}

Am i doing this right?
Thanks!

cgminer must be parsing the output buffer differently from poclbm (see SETFOUND macro in its kernels), otherwise looks good, at a glance. Please also note the kernel only returns diff 1 shares, for pool you are going to need to either hardcode the target or pass it to kernel.

[glitchboy]

cgminer must be parsing the output buffer differently from poclbm (see SETFOUND macro in its kernels), otherwise looks good, at a glance. Please also note the kernel only returns diff 1 shares, for pool you are going to need to either hardcode the target or pass it to kernel.

Yes, cgminer parses it differently. I added this

Code:

// output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
    #define FOUND (0x0F)
    #define SETFOUND(Xnonce) output[output[FOUND]++] = Xnonce

    SETFOUND(nonce);

  to your kernel and it should work.
Unfortunately it never reaches this code (on Geforce GT 650) and I don't know why...

[reorder]

cgminer must be parsing the output buffer differently from poclbm (see SETFOUND macro in its kernels), otherwise looks good, at a glance. Please also note the kernel only returns diff 1 shares, for pool you are going to need to either hardcode the target or pass it to kernel.

Yes, cgminer parses it differently. I added this

Code:

// output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
    #define FOUND (0x0F)
    #define SETFOUND(Xnonce) output[output[FOUND]++] = Xnonce

    SETFOUND(nonce);

  to your kernel and it should work.
Unfortunately it never reaches this code (on Geforce GT 650) and I don't know why...

You can try a lower diff (uncomment the '& 0xc0ffffff' for example) and see what happens. You can also use printf() in the kernel for debugging if your SDK is at least OpenCL 1.1 compliant.

[broketech]

I  tried patching this work into cgminer 3.7.2 as it's own kernel switch.  The one line in the kernel queue causes problems:

flip80(swap32, (uint32_t *)blk->work->data); // convert getwork data to big-endian

If I don't set scrypt in configure it will hang on work not being available in dev_blk_ctx without enabling scrypt.  All other issues handled, like I said I enable scrypt as a hack, it compiles fine, but segfaults at (you can almost see it when you have debug/verbose on) somewhere in the neighborhood of the swap line, and flipping the work data.  

Again, as a hack to see if it would get off the ground, for giggles I enabled scrypt on the command line. I didn't add it in the 'proper' place, at the front of the command line, but added it at the end.  It fired up, and hashed, but never claimed any diff at all. Best stayed at 0, even after monkying with the result diff in the cl.  Thats as far as I got.  Soon as that machine is available to me again, I was going to see about that work line..

edit: 200-300mhash sec, on a 5870, with sha settings, 966/300.  It felt like it was doing what it was supposed to, just no results, probably due to me setting scrypt and the result getting fumbled.

[snoopdoug]

I  tried patching this work into cgminer 3.7.2 as it's own kernel switch.  The one line in the kernel queue causes problems:

flip80(swap32, (uint32_t *)blk->work->data); // convert getwork data to big-endian

I believe that "work" only exists in the "blk" struct when the scrypt kernel is enabled.

I added:

Code:

work->blk.work = work;

just before the call to thrdata->queue_kernel_parameters in driver-opencl.c, and cgminer seemed to start working for me (but no solved blocks yet).

Warning: if you add that line like I did, cgminer may have trouble mining other coins.  The function opencl_prepare_work() would be a better place to handle that.  I'm just trying to get something working.

Hope that helps,

Doug

[Gunther]

Strange. This coin still isn't on any exchange ... i don't think this will work. Too bad.

[D4T]

Theres still time 

[barwizi]

considering all the work being done on this coin and it's algo, it's a shame it has been raped.

[manawenuz]

considering all the work being done on this coin and it's algo, it's a shame it has been raped.

i believe it's mostly the fault of the developer, as he/she doesn't show any signs of activities !!
seems like that it was a coin which got dropped from day one , maybe algo will be used someday for a successful coin but not here ! not now !

[reecelander]

anybody interested in buying my 416 skeincoins?

[asor]

I'm still selling 2500 SKC for 0.3 LTC...

[Gunther]

Come on dev, respond. Aargh.
LTC it is then. Mining .... now !

[Red Kendra]

Strange. This coin still isn't on any exchange ... i don't think this will work. Too bad.

Patience The coin is barely three weeks old. I have contacted the major exchanges and I'm sure one will pick up Skeincoin in due time.

[barwizi]

i now buys @ 0.5 LTC / 10k what with money supply jumping due to that GPU rush

[broketech]

GPU rush

Last block diff 3.0225129. Some rush. I've never seen it that low.

[Gunther]

Strange. This coin still isn't on any exchange ... i don't think this will work. Too bad.

Patience The coin is barely three weeks old. I have contacted the major exchanges and I'm sure one will pick up Skeincoin in due time.

Hope so. I'm currently mining LTC again, but when this coin will appear on some exchange, i'll be mining this again.
Have >7k SKC right now

[barwizi]

come on guys, sell me some skein!!!

[surfer43]

come on guys, sell me some skein!!!

I sell 500 SKC for .1 BTC