CCminer(SP-MOD) Modded GPU kernels.

[Travis9x]

Code:

@ECHO off
setx GPU_MAX_ALLOC_PERCENT 100
ccminer.exe -a x11 -o stratum+tcp://us1.coinking.io:6666 -u Travis9x.ASRockX11 -p x -D
PAUSE

Code:

SUCCESS: Specified value was saved.
*** ccMiner for nVidia GPUs by Christian Buchner and Christian H. ***
         This is the forked version 1.4.7.SP (sp-hash@github)
          Built with VC++ 2013 and nVidia CUDA SDK 6.5

          based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler
            and HVC extension from http://hvc.1gh.com/

        Cuda additions Copyright 2014 Christian Buchner, Christian H.

        Include some of djm34 additions, cleaned by Tanguy Pruvot
                  Optimized Kernals By SP^Cryptoburnes

[2014-11-08 19:18:16] 2 miner threads started, using 'x11' algorithm.
[2014-11-08 19:18:16] Starting Stratum on stratum+tcp://us1.coinking.io:6666
[2014-11-08 19:18:16] Binding thread 0 to cpu 0
[2014-11-08 19:18:16] Binding thread 1 to cpu 1
[2014-11-08 19:18:17] Failed to get Stratum session id
[2014-11-08 19:18:17] Stratum difficulty set to 0.004
[2014-11-08 19:18:17] DEBUG: job_id=45ecd6d dc92 xnonce2=00000000 time=19:11:57
[2014-11-08 19:18:17] us1.coinking.io:6666 sent x11 block 2849
[2014-11-08 19:18:17] sleeptime: 500 ms
[2014-11-08 19:18:17] job 45ecd6d dc92 target change: f9ff060000 (1.0)
[2014-11-08 19:18:17] sleeptime: 500 ms
[2014-11-08 19:18:17] job 45ecd6d dc92 00000000
[2014-11-08 19:18:17] job 45ecd6d dc92 target change: f9ff060000 (1.0)
[2014-11-08 19:18:17] job 45ecd6d dc92 7fffffff
Cuda error in file 'C:/code/ccminer-sp/x11/cuda_x11_simd512.cu' in line 647 : in
valid texture reference.
Press any key to continue . . .

Any thoughts about this texture reference error?   

[sp_]

You get this error if you try to run the compute52.exe on a 750ti. The 5.2 is for the 970 and 980 cards only.

[sp_]

perf is reduced on the 750Ti (linux), else i found the way to enhance a bit your groestl change (commited)
EDIT: hmm in fact not exactly, but... its hard to compare

This version was tweaked to run fast on the 980. Launchbounds/threads/code. I think I will need to do a test for compute 50, and    do seperate code for the 750ti some places.

[sp_]

Of all the betas I have sendt out, version 6 seems to be the most stable. Been running stable all night with +150 on the gpu clock and no booos.

[sp_]

I found the way to enhance a bit your groestl change (commited)
EDIT: hmm in fact not exactly, but... its hard to compare

I didn't get any speedups when testing your changes, but I improved my groestl  now should be 2-3% faster.

this:
int andmask1 = ((~((threadIdx.x & 0x03) - 3)) & 0xffff0000);

rewritten to this:

uint32_t andmask1 =-((threadIdx.x & 0x03) == 3) & 0xffff0000;

and some small other changes.

The % operator (modulo) is more expensive than an and.

[Epsylon3]

weird results with your last changes, so i will do nothing more on groestl

I added the -i param to prevent these useless throughput commits
https://github.com/tpruvot/ccminer/commit/9f62014690a479976c6ba6af1ed5f51c07f7e86a

[sp_]

weird results with your last changes, so i will do nothing more on groestl
I added the -i param to prevent these useless throughput commits
https://github.com/tpruvot/ccminer/commit/9f62014690a479976c6ba6af1ed5f51c07f7e86a

This is nice. Note that also the throughput should be editable in the commandline as well. Quark is running 10% faster after I multiplied the througput by 10. (750ti/970/980) Same for NIST-5. The only problem is when mining with low diff.

[sp_]

Been working on Neoscrypt, so still at 9.6MH/s on my 290X. 980 hasn't caught up yet, though.

My checkins today improves the X11 hash another 50-100 KHASH on the stock clocked 980.(7850-7900 KHASH). With watercooling perhaps we can reach 9200 KHASH.

[Epsylon3]

weird results with your last changes, so i will do nothing more on groestl
I added the -i param to prevent these useless throughput commits
https://github.com/tpruvot/ccminer/commit/9f62014690a479976c6ba6af1ed5f51c07f7e86a

This is nice. Note that also the throughput should be editable in the commandline as well. Quark is running 10% faster after I multiplied the througput by 10. (750ti/970/980) Same for NIST-5. The only problem is when mining with low diff.

-i param = command line :/

[th00ber]

what i parameter is the best for 750 ti ?

[djm34]

what i parameter is the best for 750 ti ?

the smallest value which gives you 99-100% gpu usage (I am speaking about the throughput not exactly the i parameter) also must be a multiple of 256.

[Epsylon3]

-i 19 = 1 << 19 (left shift) = 524288

Will be in next 1.4.7 (-i 0 is the default, fixed for each algo, often based on the 750 Ti best perf)

Here is a shared spreadsheet to complete for 970/980 users. To be able to tune the default values for these cards :

https://docs.google.com/spreadsheets/d/1dI1Cc3JhhsA-UdIRlndvQ9huX8FmoJT2TXFDB0wdp54/edit?usp=sharing

i will link the 1.4.7 test version here soon

Here it is : https://dl.dropboxusercontent.com/u/31689596/ccminer/ccminer-rel1.4.7-test-x86.7z

[italeffect]

-i 19 = 1 << 19 (left shift) = 524288

Will be in next 1.4.7 (-i 0 is the default, fixed for each algo, often based on the 750 Ti best perf)

Here is a shared spreadsheet to complete for 970/980 users. To be able to tune the default values for these cards :

https://docs.google.com/spreadsheets/d/1dI1Cc3JhhsA-UdIRlndvQ9huX8FmoJT2TXFDB0wdp54/edit?usp=sharing

i will link the 1.4.7 test version here soon

Here it is : https://dl.dropboxusercontent.com/u/31689596/ccminer/ccminer-rel1.4.7-test-x86.7z

Thanks for this. Is there any difference between ccminer-50 and ccminer-50-52 when running on a 750ti?

[Epsylon3]

Thanks for this. Is there any difference between ccminer-50 and ccminer-50-52 when running on a 750ti?

Good question, test and tell me Seems to be variable against driver version, the 50 52 is for people with a 750Ti and a 9xx on the same Machine

[italeffect]

Thanks for this. Is there any difference between ccminer-50 and ccminer-50-52 when running on a 750ti?

Good question, test and tell me Seems to be variable against driver version, the 50 52 is for people with a 750Ti and a 9xx on the same Machine

Hash rate jumps around so much it's hard to tell. After a few hours on each I can't see any meaningful difference.

[AliMan]

Thanks for this. Is there any difference between ccminer-50 and ccminer-50-52 when running on a 750ti?

Good question, test and tell me Seems to be variable against driver version, the 50 52 is for people with a 750Ti and a 9xx on the same Machine

Yep, runs great with both, 750 ti and 970's together!

[sp_]

Looks like the wolf miner(?)  (AMD) has been published too some ppl. 10MHASH on 290x. I will checkin some more kernals soon, hopefully the 980 will pass 10MHASH as well.

https://bitcointalk.org/index.php?topic=854257.0

[Epsylon3]

final 1.4.7 released : https://github.com/tpruvot/ccminer/releases/tag/1.4.7-tpruvot

[jpouza]

final 1.4.7 released : https://github.com/tpruvot/ccminer/releases/tag/1.4.7-tpruvot

There is something wrong, it's not adding the 3 cards hash, it counts as one only.

3x980@9MH/s = 27MH/s
Counts only 9MH/s with the 03 cards...

[Epsylon3]

yep, just fixed that :/ for this version just multiply by 3

will be in 1.4.8 (with a -N param to specify the number of samples for the average, if wanted)

im starting the work on the network json interface (like cgminer, but read-only to report some important stats)