CCminer(SP-MOD) Modded GPU kernels.

[sp_]

already fixed it..

Recompiled it and when I try neoscript I get
6X750 ti Evga SC Windows 8.1 Pro VS 2013 Driver 353.30
[2015-08-10 14:58:43] neoscrypt.usa.nicehash.com:3341 neoscrypt block 829449
Cuda error in func 'scanhash_neoscrypt' at line 48 : out of memory.
*** ccminer 1.5.59-git(SP-MOD) for nVidia GPUs by sp-hash@github ***

The default intensity is too high on the 960 cards and windows 8. adjust with -i. -i 14.5 or similar.

[joblo]

already fixed it..

The CPU problem is fixed but there's still a problem. I have a linux rig with a 970 and a 780ti, compiled
ccminer for compute versions 3.5, 5.0 & 5.2.
When I start mining neoscrypt on both GPUs I get nothing but HW errors on the 780ti (GPU1). But I can mine
each GPU in seperate instances of ccminer.

Code:

$ ./ccminer -a neoscrypt -o stratum+tcp://neoscrypt.usa.nicehash.com:3341  
*** ccminer 1.5.59-git(SP-MOD) for nVidia GPUs by sp-hash@github ***
	Built with the nVidia CUDA SDK 6.5

  Based on pooler cpuminer 2.3.2 and the tpruvot@github fork
   CUDA support by Christian Buchner, Christian H. and DJM34
  Includes optimizations implemented by sp , klaust, tpruvot and tsiv. 

[2015-08-10 15:56:14] 2 miner threads started, using 'neoscrypt' algorithm.
[2015-08-10 15:56:14] Binding thread 1 to cpu 1 (mask 2)
1 
[2015-08-10 15:56:14] Binding thread 0 to cpu 0 (mask 1)
0 
[2015-08-10 15:56:14] Starting Stratum on stratum+tcp://neoscrypt.usa.nicehash.com:3341
[2015-08-10 15:56:15] Stratum difficulty set to 64
[2015-08-10 15:56:15] neoscrypt.usa.nicehash.com:3341 neoscrypt block 829446
[2015-08-10 15:56:15] GPU #1: result for nonce $00007FAF does not validate on CPU!
[2015-08-10 15:56:15] Hardware errors: 1
[2015-08-10 15:56:15] GPU #1: result for nonce $00007FAF does not validate on CPU!
[2015-08-10 15:56:15] GPU #1: GeForce GTX 780 Ti, 59858503
[2015-08-10 15:56:15] Hardware errors: 2
[2015-08-10 15:56:15] GPU #1: result for nonce $00007FAF does not validate on CPU!

[sp_]

The CPU problem is fixed but there's still a problem. I have a linux rig with a 970 and a 780ti, compiled
ccminer for compute versions 3.5, 5.0 & 5.2.
When I start mining neoscrypt on both GPUs I get nothing but HW errors on the 780ti (GPU1). But I can mine
each GPU in seperate instances of ccminer.

This fork doesn't support compute 3.5. Maxwell only. (5.0 and up). but it might work for some algos. Don't have a 3.5 card anymore.

[joblo]

The CPU problem is fixed but there's still a problem. I have a linux rig with a 970 and a 780ti, compiled
ccminer for compute versions 3.5, 5.0 & 5.2.
When I start mining neoscrypt on both GPUs I get nothing but HW errors on the 780ti (GPU1). But I can mine
each GPU in seperate instances of ccminer.

This fork doesn't support compute 3.5. Maxwell only. (5.0 and up). but it might work for some algos. Don't have a 3.5 card anymore.

This is the first problem I've encountered with 3.5 and it does work when in its
own instance so the 3.5 kernel is ok. I'm guessing it's a problem on startup choosing
the correct kernel when there are multiple GPUs with different compute versions.

[sp_]

Added lyra2rev2 now. copied from the djm34 branch


Hashrates

750ti: 4MHASH
970: 9.5MHASH


use the parameter -a lyra2v2 to mine

[frazier34567]

already fixed it..

Recompiled it and when I try neoscript I get
6X750 ti Evga SC Windows 8.1 Pro VS 2013 Driver 353.30
[2015-08-10 14:58:43] neoscrypt.usa.nicehash.com:3341 neoscrypt block 829449
Cuda error in func 'scanhash_neoscrypt' at line 48 : out of memory.
*** ccminer 1.5.59-git(SP-MOD) for nVidia GPUs by sp-hash@github ***

The default intensity is too high on the 960 cards and windows 8. adjust with -i. -i 14.5 or similar.

Using -i 15 for my 750 Ti's getting 996 Kh/s on my 6 card rig.  (Always runs slower than my single card)

[sp_]

I am building release 60 now with the new lyra2v2 and the fixed neoscrypt(cpu and powerusage)

a 280x is doing only doing 3.5 MHASH. so this miner from djm34 is a goldmine for NVIDIA owners

[Grout]

Thanks SP, beer coming your way :-)
4b8a3d50da011f76a92c80ca19bf146a9001c26ec3f18b19d6d24a18b9c16295

[sp_]

-fixed 100% cpu in neoscrypt and removed powerusage.
-Added lyra2rev2 from the DJM34 fork

1.5.60(sp-MOD) is available here: (10-08-2015)

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

The sourcecode is available here:

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

[sp_]

Thanks SP, beer coming your way :-)
4b8a3d50da011f76a92c80ca19bf146a9001c26ec3f18b19d6d24a18b9c16295

Thanks again

[dga]

-fixed 100% cpu in neoscrypt and removed powerusage.
-Added lyra2rev2 from the DJM34 fork

1.5.60(sp-MOD) is available here: (10-08-2015)

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

The sourcecode is available here:

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

Build is broken under cuda 6.5 on Linux/ubuntu:

Code:

lyra2/lyra2REv2.o: In function `lyra2v2_hash':
tmpxft_00002140_00000000-6_lyra2REv2.compute_52.cudafe1.cpp:(.text+0x65): multiple definition of `lyra2v2_hash'
lyra2/lyra2REv2.o:tmpxft_00002140_00000000-6_lyra2REv2.compute_52.cudafe1.cpp:(.text+0x65): first defined here

Will dig, but you can probably fix it faster than I can find the bug.

update:  Error 1 is that the lyra2REv2.cu line was duplicated in Makefile.am.  But removing it leaves it still not compiling - missing link of bmw.

update 2:  Missing Algo256/cuda_bmw256.cu  Algo256/cuda_cubehash256.cu  in the Makefile.am

The attached patch might break other things, so be careful, but it gets the compile working:

Code:

diff --git a/Makefile.am b/Makefile.am
index eafd385..cb0174e 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -32,11 +32,7 @@ ccminer_SOURCES      = elist.h miner.h compat.h \
                          fuguecoin.cpp Algo256/cuda_fugue256.cu sph/fugue.c uint256.h \
                          groestlcoin.cpp cuda_groestlcoin.cu cuda_groestlcoin.h \
                          myriadgroestl.cpp cuda_myriadgroestl.cu \
-                         lyra2/Lyra2.c lyra2/Sponge.c \
-                         lyra2/lyra2RE.cu lyra2/cuda_lyra2.cu \
-                         lyra2/lyra2REv2.cu lyra2/cuda_lyra2v2.cu \
-                         Algo256/cuda_blake256.cu Algo256/cuda_groestl256.cu Algo256/cuda_keccak256.cu Algo256/cuda_skein256.cu \
-                         lyra2/lyra2REv2.cu lyra2/cuda_lyra2v2.cu \
+                         Algo256/cuda_blake256.cu Algo256/cuda_groestl256.cu Algo256/cuda_keccak256.cu Algo256/cuda_skein256.cu Al
                          Algo256/blake256.cu Algo256/keccak256.cu \
                          JHA/jackpotcoin.cu JHA/cuda_jha_keccak512.cu \
                          JHA/cuda_jha_compactionTest.cu cuda_checkhash.cu \
@@ -45,6 +41,7 @@ ccminer_SOURCES       = elist.h miner.h compat.h \
                          quark/quarkcoin.cu quark/animecoin.cu \
                          quark/cuda_quark_compactionTest.cu  \
                          cuda_nist5.cu pentablake.cu skein.cu\
+                         lyra2/Lyra2.c lyra2/Sponge.c \
                          sph/bmw.c sph/blake.c sph/groestl.c sph/jh.c sph/keccak.c sph/skein.c \
                          sph/cubehash.c sph/echo.c sph/luffa.c sph/sha2.c sph/sph_sha2.c sph/shavite.c sph/simd.c \
                          sph/hamsi.c sph/hamsi_helper.c sph/sph_hamsi.h \
@@ -60,6 +57,8 @@ ccminer_SOURCES       = elist.h miner.h compat.h \
                          x11/c11.cu x11/s3.cu \
                          bitcoin.cu cuda_bitcoin.cu\
                          x15/cuda_whirlpoolx.cu x15/whirlpoolx.cu \
+                         lyra2/lyra2RE.cu lyra2/cuda_lyra2.cu \
+                         lyra2/lyra2REv2.cu lyra2/cuda_lyra2v2.cu \
                          neoscrypt/neoscrypt.cu neoscrypt/cuda_neoscrypt.cu \
               bitcredit/cuda_bitcredit.cu bitcredit/bitcredit.cu
 # scrypt

[scryptr]

-fixed 100% cpu in neoscrypt and removed powerusage.
-Added lyra2rev2 from the DJM34 fork

1.5.60(sp-MOD) is available here: (10-08-2015)

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

The sourcecode is available here:

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

Build is broken under cuda 6.5 on Linux/ubuntu:

Code:

lyra2/lyra2REv2.o: In function `lyra2v2_hash':
tmpxft_00002140_00000000-6_lyra2REv2.compute_52.cudafe1.cpp:(.text+0x65): multiple definition of `lyra2v2_hash'
lyra2/lyra2REv2.o:tmpxft_00002140_00000000-6_lyra2REv2.compute_52.cudafe1.cpp:(.text+0x65): first defined here

Will dig, but you can probably fix it faster than I can find the bug.

BUILD ERRORS OUT--

I am getting a similar error, something about an "unsigned int".       --scryptr

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>PASTE<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
\lyra2/lyra2REv2.o: In function `lyra2v2_hash':
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x20): multiple definition of `lyra2v2_hash'
lyra2/lyra2REv2.o:tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x20): first defined here
lyra2/lyra2REv2.o: In function `scanhash_lyra2v2':
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x200): multiple definition of `scanhash_lyra2v2'
lyra2/lyra2REv2.o:tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x200): first defined here
lyra2/cuda_lyra2v2.o: In function `lyra2v2_gpu_hash_32_v3(unsigned int, unsigned int, uint2*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x20): multiple definition of `lyra2v2_gpu_hash_32_v3(unsigned int, unsigned int, uint2*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x20): first defined here
lyra2/cuda_lyra2v2.o: In function `lyra2v2_gpu_hash_32(unsigned int, unsigned int, uint2*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0xa0): multiple definition of `lyra2v2_gpu_hash_32(unsigned int, unsigned int, uint2*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0xa0): first defined here
lyra2/cuda_lyra2v2.o: In function `lyra2v2_cpu_init(int, unsigned int, unsigned long*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x120): multiple definition of `lyra2v2_cpu_init(int, unsigned int, unsigned long*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x120): first defined here
lyra2/cuda_lyra2v2.o: In function `lyra2v2_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x150): multiple definition of `lyra2v2_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x150): first defined here
lyra2/cuda_lyra2v2.o: In function `__device_stub__Z22lyra2v2_gpu_hash_32_v3jjP5uint2(unsigned int, unsigned int, uint2*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x320): multiple definition of `__device_stub__Z22lyra2v2_gpu_hash_32_v3jjP5uint2(unsigned int, unsigned int, uint2*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x320): first defined here
lyra2/cuda_lyra2v2.o: In function `__device_stub__Z19lyra2v2_gpu_hash_32jjP5uint2(unsigned int, unsigned int, uint2*)':
tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x3a0): multiple definition of `__device_stub__Z19lyra2v2_gpu_hash_32jjP5uint2(unsigned int, unsigned int, uint2*)'
lyra2/cuda_lyra2v2.o:tmpxft_00003577_00000000-3_cuda_lyra2v2.cudafe1.cpp:(.text+0x3a0): first defined here
lyra2/lyra2REv2.o: In function `scanhash_lyra2v2':
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x5a5): undefined reference to `bmw256_setTarget(void const*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x608): undefined reference to `cubehash256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x647): undefined reference to `cubehash256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x661): undefined reference to `bmw256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*, unsigned int*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x7c0): undefined reference to `bmw256_cpu_init(int, unsigned int)'
lyra2/lyra2REv2.o: In function `scanhash_lyra2v2':
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x5a5): undefined reference to `bmw256_setTarget(void const*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x608): undefined reference to `cubehash256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x647): undefined reference to `cubehash256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x661): undefined reference to `bmw256_cpu_hash_32(int, unsigned int, unsigned int, unsigned long*, unsigned int*)'
tmpxft_0000355f_00000000-3_lyra2REv2.cudafe1.cpp:(.text+0x7c0): undefined reference to `bmw256_cpu_init(int, unsigned int)'
collect2: error: ld returned 1 exit status
make[2]: *** [ccminer] Error 1
>>>>>>>>>>>>>>>>>>>>>>>>>>END PASTE<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

[bensam1231]

-fixed 100% cpu in neoscrypt and removed powerusage.
-Added lyra2rev2 from the DJM34 fork

1.5.60(sp-MOD) is available here: (10-08-2015)

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

The sourcecode is available here:

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

Does this fix the 30w increase in power usage?

[pallas]

-fixed 100% cpu in neoscrypt and removed powerusage.
-Added lyra2rev2 from the DJM34 fork

1.5.60(sp-MOD) is available here: (10-08-2015)

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

The sourcecode is available here:

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

Does this fix the 30w increase in power usage?

it should... depending on the wattage difference from idle to 100% of your cpu :-)

[facanferff]

I don't have much knowledge about this new algo lyra2v2 but would want to know if it's normal a "big" increase in speed in comparison to lyra2.

I ran both in benchmark mode (one at a time) and got:
lyra2:

[2015-08-10 22:32:43] GPU #0 Found nounce 00af3bbf
[2015-08-10 22:32:43] GPU #0: GeForce GTX 850M, 683.23
[2015-08-10 22:32:43] CPU #0: 683.23 kH/s

lyra2v2:

[2015-08-10 22:31:43] GPU #0 Found nounce 0dd624a1
[2015-08-10 22:31:43] GPU #0: GeForce GTX 850M, 3755
[2015-08-10 22:31:43] CPU #0: 3755 kH/s

Is it normal this increase?

[scryptr]

I don't have much knowledge about this new algo lyra2v2 but would want to know if it's normal a "big" increase in speed in comparison to lyra2.

I ran both in benchmark mode (one at a time) and got:
lyra2:

[2015-08-10 22:32:43] GPU #0 Found nounce 00af3bbf
[2015-08-10 22:32:43] GPU #0: GeForce GTX 850M, 683.23
[2015-08-10 22:32:43] CPU #0: 683.23 kH/s

lyra2v2:

[2015-08-10 22:31:43] GPU #0 Found nounce 0dd624a1
[2015-08-10 22:31:43] GPU #0: GeForce GTX 850M, 3755
[2015-08-10 22:31:43] CPU #0: 3755 kH/s

Is it normal this increase?

THE INCREASE IS REAL--

DJM34's Lyra2v2 ccminer truly rocks!  Answering your question, however:  The increase is NOT normal!    Increases usually take 5-10-50kh/s steps.  But, DJM34 completely re-did the code, and he is like the main developer for the miner, with the most know-how and experience on vertcoin.

THANKS! DJM34!!!       --scryptr

[facanferff]

I don't have much knowledge about this new algo lyra2v2 but would want to know if it's normal a "big" increase in speed in comparison to lyra2.

I ran both in benchmark mode (one at a time) and got:
lyra2:

[2015-08-10 22:32:43] GPU #0 Found nounce 00af3bbf
[2015-08-10 22:32:43] GPU #0: GeForce GTX 850M, 683.23
[2015-08-10 22:32:43] CPU #0: 683.23 kH/s

lyra2v2:

[2015-08-10 22:31:43] GPU #0 Found nounce 0dd624a1
[2015-08-10 22:31:43] GPU #0: GeForce GTX 850M, 3755
[2015-08-10 22:31:43] CPU #0: 3755 kH/s

Is it normal this increase?

THE INCREASE IS REAL--

DJM34's Lyra2v2 ccminer truly rocks!  Answering your question, however:  The increase is NOT normal!    Increases usually take 5-10-50kh/s steps.  But, DJM34 completely re-did the code, and he is like the main developer for the miner, with the most know-how and experience on vertcoin.

THANKS! DJM34!!!       --scryptr

I see, then it is an excellent gift .

[djm34]

I am building release 60 now with the new lyra2v2 and the fixed neoscrypt(cpu and powerusage)

a 280x is doing only doing 3.5 MHASH. so this miner from djm34 is a goldmine for NVIDIA owners

it should do more, my 290x does 7.5MH/s (it needs small worksize and plenty of thread and to adjust the intensity so it fills up the memory)

[djm34]

I don't have much knowledge about this new algo lyra2v2 but would want to know if it's normal a "big" increase in speed in comparison to lyra2.

I ran both in benchmark mode (one at a time) and got:
lyra2:

[2015-08-10 22:32:43] GPU #0 Found nounce 00af3bbf
[2015-08-10 22:32:43] GPU #0: GeForce GTX 850M, 683.23
[2015-08-10 22:32:43] CPU #0: 683.23 kH/s

lyra2v2:

[2015-08-10 22:31:43] GPU #0 Found nounce 0dd624a1
[2015-08-10 22:31:43] GPU #0: GeForce GTX 850M, 3755
[2015-08-10 22:31:43] CPU #0: 3755 kH/s

Is it normal this increase?

THE INCREASE IS REAL--

DJM34's Lyra2v2 ccminer truly rocks!  Answering your question, however:  The increase is NOT normal!    Increases usually take 5-10-50kh/s steps.  But, DJM34 completely re-did the code, and he is like the main developer for the miner, with the most know-how and experience on vertcoin.

THANKS! DJM34!!!       --scryptr

it isn't an optimization per se but a new set of algos and more gpu friendly settings for lyra2

  however the point was to increase gpu/cpu  ratio from 1~2 to around 10 not nvidia/amd ratio (which actually should be about the same as it was with the optimized lyra2 with the 750ti running at 1500kh/s), also getting rid of groestl is probably more beneficial to the amd (even though it wasn't Christian "killer Groestl").

[fenomenhaa]

With -a lyra2v2 i just get share above target !!! what's wrong?