CCminer(SP-MOD) Modded GPU kernels.

[pokeytex]

Maybe for a donate?

I already did +5%

My version does 840MHASH. tvpruvots version does 800MHASH on the 750ti.


here are the changesets:

https://github.com/sp-hash/ccminer/commit/bd09f337b177ae495dbbff40655338b0d5663792

https://github.com/sp-hash/ccminer/commit/cd7e2b5a91491e468e58fc9a97b390d1eac7d969

Hrmmm- I am getting over 900mh/s on tpruvots versions 750ti, 2.6 gh/s 970, 1.7 gh/s on 960 - modded last night by xCore.  I believe he added a file to your version.  Was hoping for faster than that!

Than again i am not a programmer so I don't know who's version is who's lol.

[sp_]

There are many 750ti cards. with different clocks.

Download release 78 and write

ccminer -a blakecoin --benchmark

Then you will see.. I just tested the tvpruvot release and it was slower on all cards.

https://github.com/sp-hash/ccminer/releases/

[pokeytex]

There are many 750ti cards. with different clocks.

Download release 78 and write

ccminer -a blakecoin --benchmark

Then you will see.. I just tested the tvpruvot release and it was slower on all cards.

https://github.com/sp-hash/ccminer/releases/

Yeah xCore modified it last night so we had a faster version with the Vanillacoin fork.  I am not sure how he did it.

[sp_]

Yeah xCore modified it last night so we had a faster version with the Vanillacoin fork.  I am not sure how he did it.

You meen this version:

https://talk.vanillacoin.net/uploads/files/tools/ccminer-1.7.1-blake256-standard-uncompressed-sp.zip

It is not faster than my opensource..

[Epsylon3]

1.7.1 in this archive is a typo when he made the folder zip... its yours

i grabbed your kernel changes and cleanup the old ifdef (initial implementation for neos)... didnt had the time/need to focus on blake recently

so my 1.7.2 will be close to you.... cuda 7.5 is better for this algo

[sp_]

1.7.1 in this archive is a typo when he made the folder zip... its yours

i grabbed your kernel changes and cleanup the old ifdef (initial implementation for neos)... didnt had the time/need to focus on blake recently

so my 1.7.2 will be close to you.... cuda 7.5 is better for this algo

I tested again.. same speed. (1.7.2) (ccminer -a blake --benchmark -q)

[sp_]

I can make this faster..

[joblo]

I just improved quark on r74. Do you want it?

1.5.74-jdd+

Code:

[2016-01-27 19:00:14] accepted: 33/33 (100.00%), 34643 kH/s yes!
[2016-01-27 19:00:18] GPU #0: GeForce GTX 970, 17825
[2016-01-27 19:00:18] accepted: 34/34 (100.00%), 34643 kH/s yes!
[2016-01-27 19:00:19] GPU #1: GeForce GTX 970, 17134
[2016-01-27 19:00:19] accepted: 35/35 (100.00%), 34660 kH/s yes!
[2016-01-27 19:00:22] GPU #1: GeForce GTX 970, 17085
[2016-01-27 19:00:22] accepted: 36/36 (100.00%), 34673 kH/s yes!
[2016-01-27 19:00:28] GPU #1: GeForce GTX 970, 17114
[2016-01-27 19:00:28] accepted: 37/37 (100.00%), 34686 kH/s yes!
[2016-01-27 19:00:29] GPU #0: GeForce GTX 970, 17719
[2016-01-27 19:00:29] accepted: 38/38 (100.00%), 34682 kH/s yes!

1.5.74-SP

Code:

[2016-01-27 19:02:07] accepted: 19/19 (100.00%), 33334 kH/s yes!
[2016-01-27 19:02:08] GPU #0: GeForce GTX 970, 17793
[2016-01-27 19:02:08] accepted: 20/20 (100.00%), 33486 kH/s yes!
[2016-01-27 19:02:11] GPU #1: GeForce GTX 970, 17146
[2016-01-27 19:02:11] accepted: 21/21 (100.00%), 33482 kH/s yes!
[2016-01-27 19:02:12] GPU #1: GeForce GTX 970, 17139
[2016-01-27 19:02:12] accepted: 22/22 (100.00%), 33478 kH/s yes!

You should have told the truth.

[sp_]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

[joblo]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

Exactly what I told you. You can have it if you open some of you private stash.

[sp_]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

Exactly what I told you. You can have it if you open some of you private stash.

Release 78 is faster.

[joblo]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

Exactly what I told you. You can have it if you open some of you private stash.

Release 78 is faster.

Want me to do 78 too?

[sp_]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

Exactly what I told you. You can have it if you open some of you private stash.

Release 78 is faster.

Want me to do 78 too?

Yes I want you to do +30%

[joblo]

I just improved quark on r74. Do you want it?
1.5.74-jdd+

What did you do?

Exactly what I told you. You can have it if you open some of you private stash.

Release 78 is faster.

Want me to do 78 too?

Yes I want you to do +30%

when is the last time you delivered 30% in less than an hour?

[sp_]

when is the last time you delivered 30% in less than an hour?

Today. your quark kernel.

Since skein is much faster than groestl we only do skein and throw away 50% of the hashes.

    if (hash[0] & 0x8)
    {
        sph_groestl512_init(&ctx_groestl);
        sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
        sph_groestl512_close(&ctx_groestl, (void*) hash);
    }
    else
    {
        sph_skein512_init(&ctx_skein);
        sph_skein512 (&ctx_skein, (const void*) hash, 64);
        sph_skein512_close(&ctx_skein, (void*) hash);
    }

    if (hash[0] & 0x8)
    {
        don't hash the rest
    }
    else
    {
        sph_skein512_init(&ctx_skein);
        sph_skein512 (&ctx_skein, (const void*) hash, 64);
        sph_skein512_close(&ctx_skein, (void*) hash);
    }

[joblo]

when is the last time you delivered 30% in less than an hour?

Today. your quark kernel.

That doesn't make sense.

Anyway here it is. This one's on me.

https://drive.google.com/file/d/0B0lVSGQYLJIZbllYWENUV0l0VnM/view?usp=sharing

[joblo]

when is the last time you delivered 30% in less than an hour?

Today. your quark kernel.

Since skein is much faster than groestl we only do skein and throw away 50% of the hashes.

    if (hash[0] & 0x8)
    {
        sph_groestl512_init(&ctx_groestl);
        sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
        sph_groestl512_close(&ctx_groestl, (void*) hash);
    }
    else
    {
        sph_skein512_init(&ctx_skein);
        sph_skein512 (&ctx_skein, (const void*) hash, 64);
        sph_skein512_close(&ctx_skein, (void*) hash);
    }

There was an optimization made in cpuminer that  if it was determined that a second
round of groestl was necessary the existing hashes would be thrown away on the belief
it would take longer to complete the second groestl than to start over. It didn't work.

However, I might try ccminer's logic. cpuminer uses a state machine as
the engine. ccminer just uses a simple if.

I'm also going to look at other contexts. selctively reinitializing necessary fields may be
quicker thn the current implementation of copying a saved initialiazed context.
Both are quicker than what ccminer does.

[sp_]

this is in the cpu verification code. The gpu code is different. There we have precalc tables of the states to avoid conditional branches.

[joblo]

this is in the cpu verification code. The gpu code is different. There we have precalc tables of the states to avoid conditional branches.

Whatever it is it's faster.

[sp_]

this is in the cpu verification code. The gpu code is different. There we have precalc tables of the states to avoid conditional branches.

Whatever it is it's faster.

The cpu verification is only done when the gpu find a solution.