Topic: [ANN] [SKC] Skeincoin 0.9.3.1 | Skein-SHA2 - page 68. (Read 161541 times)
anybody interested in buying my 416 skeincoins?
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 !
considering all the work being done on this coin and it's algo, it's a shame it has been raped.
Theres still time 
Strange. This coin still isn't on any exchange ... i don't think this will work. Too bad.
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:
Quote
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
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:
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.
Quote
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.
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 thisCode:
// output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
#define FOUND (0x0F)
#define SETFOUND(Xnonce) output[output[FOUND]++] = Xnonce
SETFOUND(nonce);
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.
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 thisCode:
// output[OUTPUT_SIZE] = output[nonce & OUTPUT_MASK] = nonce;
#define FOUND (0x0F)
#define SETFOUND(Xnonce) output[output[FOUND]++] = Xnonce
SETFOUND(nonce);
Unfortunately it never reaches this code (on Geforce GT 650) and I don't know why...
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;
}
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.
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;
}
Thanks!
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.
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.
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.
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.
Guys - just let me know when you want to test your GPU miners on the pool as I would need to change share difficulty - otherwise you will flood my pool with low diff shares
probably because of low diff for shares - I will change it now so you can try again.
It is probably due to poclbm not able to restart work immediately, but still it is like this now:
Code:
skc.coinmine.pl:6400 0:1:Pitcairn 72.0[277.086 MH/s (~181 MH/s)] [Rej: 96/275 (34.91%)]
I am sure cgminer can do better.
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.
Guys - just let me know when you want to test your GPU miners on the pool as I would need to change share difficulty - otherwise you will flood my pool with low diff shares
you are sending quite a lot of duplicates...
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.
Guys - just let me know when you want to test your GPU miners on the pool as I would need to change share difficulty - otherwise you will flood my pool with low diff shares
probably because of low diff for shares - I will change it now so you can try again.
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.
Guys - just let me know when you want to test your GPU miners on the pool as I would need to change share difficulty - otherwise you will flood my pool with low diff shares
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.
Guys - just let me know when you want to test your GPU miners on the pool as I would need to change share difficulty - otherwise you will flood my pool with low diff shares
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.
reorder, what are driver version and openCL SDK version your kernel compile/work with?
Jump to: