[LOCKED] cpuminer-opt v3.12.3, open source optimized multi-algo CPU miner

[joblo]

cpuminer-opt v3.1 is out. It icludes th enew algo gate and a 5% increase in
performance in all algos.  There is an addtional 10% increase in SSE2 performance.

Quick turnaround, V3.1.1 is out. Nothing wrong with 3.1, just fixed aliases and hash display
and some optimisations.

https://drive.google.com/file/d/0B0lVSGQYLJIZT29nc1dEUlBXNzQ/view?usp=sharing

[kangjooe]

cpuminer-opt v3.1 is out. It icludes th enew algo gate and a 5% increase in
performance in all algos.  There is an addtional 10% increase in SSE2 performance.

Quick turnaround, V3.1.1 is out. Nothing wrong with 3.1, just fixed aliases and hash display
and some optimisations.

https://drive.google.com/file/d/0B0lVSGQYLJIZT29nc1dEUlBXNzQ/view?usp=sharing

sir. is this cpuminer support vanilla coin mining?

[joblo]

cpuminer-opt v3.1 is out. It icludes th enew algo gate and a 5% increase in
performance in all algos.  There is an addtional 10% increase in SSE2 performance.

Quick turnaround, V3.1.1 is out. Nothing wrong with 3.1, just fixed aliases and hash display
and some optimisations.

https://drive.google.com/file/d/0B0lVSGQYLJIZT29nc1dEUlBXNzQ/view?usp=sharing

sir. is this cpuminer support vanilla coin mining?

It should be the same as blakecoin.

Edit: I just found a bug in blakecoin, download 3.1.2 here. I couldn't test it so please report
back results using blakecoin on vanilla.

https://drive.google.com/file/d/0B0lVSGQYLJIZUlJaTlNqNFpMcWM/view?usp=sharing

[bobben2]

Hi joblo,
Thanks for your continuing efforts with the miner.

It seems like neosrypt mining is broken.  The 3.1 version core dumps
with -a neoscrypt.

Code:

 4020 Segmentation fault      (core dumped) cpuminer -t 1 -a neoscrypt  --api-bind 127.0.0.1:4044 $SITE1

I tried to recompile the code with -g switch so I could have gdb locate where it crashes.
But with -g the compile fails as well (This is on an i5 Haswell with Ubuntu 14.04.3):

Code:

gcc -std=gnu99 -DHAVE_CONFIG_H -I.  -Iyes/include -Iyes/include -fno-strict-aliasing  -I. -Iyes/include -Iyes/include -Wno-pointer-sign -Wno-pointer-to-int-cast   -march=native -g  -Iyes/include -Iyes/include -MT algo/groestl/sse2/cpuminer-grso-asm.o -MD -MP -MF algo/groestl/sse2/.deps/cpuminer-grso-asm.Tpo -c -o algo/groestl/sse2/cpuminer-grso-asm.o `test -f 'algo/groestl/sse2/grso-asm.c' || echo './'`algo/groestl/sse2/grso-asm.c
algo/groestl/sse2/grso-asm.c: In function ‘grsoP1024ASM’:
algo/groestl/sse2/grso-asm.c:6:3: error: ‘asm’ operand has impossible constraints
   asm (
   ^
make[2]: *** [algo/groestl/sse2/cpuminer-grso-asm.o] Error 1

EDIT: I managed to eke out a location where it apparently crashes without -g'ing an executable:
 0x00000000004be530 in absorbBlockBlake2Safe ()

[joblo]

Hi joblo,
Thanks for your continuing efforts with the miner.

It seems like neosrypt mining is broken.  The 3.1 version core dumps
with -a neoscrypt.

Code:

 4020 Segmentation fault      (core dumped) cpuminer -t 1 -a neoscrypt  --api-bind 127.0.0.1:4044 $SITE1

I tried to recompile the code with -g switch so I could have gdb locate where it crashes.
But with -g the compile fails as well (This is on an i5 Haswell with Ubuntu 14.04.3):

Code:

gcc -std=gnu99 -DHAVE_CONFIG_H -I.  -Iyes/include -Iyes/include -fno-strict-aliasing  -I. -Iyes/include -Iyes/include -Wno-pointer-sign -Wno-pointer-to-int-cast   -march=native -g  -Iyes/include -Iyes/include -MT algo/groestl/sse2/cpuminer-grso-asm.o -MD -MP -MF algo/groestl/sse2/.deps/cpuminer-grso-asm.Tpo -c -o algo/groestl/sse2/cpuminer-grso-asm.o `test -f 'algo/groestl/sse2/grso-asm.c' || echo './'`algo/groestl/sse2/grso-asm.c
algo/groestl/sse2/grso-asm.c: In function ‘grsoP1024ASM’:
algo/groestl/sse2/grso-asm.c:6:3: error: ‘asm’ operand has impossible constraints
   asm (
   ^
make[2]: *** [algo/groestl/sse2/cpuminer-grso-asm.o] Error 1

EDIT: I managed to eke out a location where it apparently crashes without -g'ing an executable:
 0x00000000004be530 in absorbBlockBlake2Safe ()

It looks like you are compileing for SSE2 but a Haswell has AES_NI. This should still work though,
but slower.

Either way I retested neoscrypt with both AES_NI and SSE2 in v3.1.

Edit: I tested on a pool, it looks like you're mining solo.

Edit: Upon further thought the compikle may have been coreect but it looks like GDB couldn't handle
some of the SSE2 code

If the problems only occur solo it looks like a wallet/solo mining issue.

[bobben2]

This is my build command:

Code:

./autogen.sh
make clean
./configure CFLAGS="-march=native -Ofast" CXXFLAGS=$CFLAGS --with-crypto --with-curl
make

and
cat /proc/cpuinfo | grep model
model name      : Intel(R) Core(TM) i5-4570S CPU @ 2.90GHz

I tried both of these pools:
neoscrypt.eu.nicehash.com:3341
hashpower.co:4233

[joblo]

This is my build command:

Code:

./autogen.sh
make clean
./configure CFLAGS="-march=native -Ofast" CXXFLAGS=$CFLAGS --with-crypto --with-curl
make

and
cat /proc/cpuinfo | grep model
model name      : Intel(R) Core(TM) i5-4570S CPU @ 2.90GHz

I tried both of these pools:
neoscrypt.eu.nicehash.com:3341
hashpower.co:4233

It works for me. what can I do?

[pallas]

Vanilla coin needs a specific algo gate.

[bobben2]

This is my build command:

Code:

./autogen.sh
make clean
./configure CFLAGS="-march=native -Ofast" CXXFLAGS=$CFLAGS --with-crypto --with-curl
make

and
cat /proc/cpuinfo | grep model
model name      : Intel(R) Core(TM) i5-4570S CPU @ 2.90GHz

I tried both of these pools:
neoscrypt.eu.nicehash.com:3341
hashpower.co:4233

It works for me. what can I do?

Whoops...
I figured it out.  I had an old version of cpuminer in folder in my $PATH.   So once I removed that, my script finally invoked the
correct cpuminer version.  And it works as I hoped!
So the next beer is on me!

[joblo]

Vanilla coin needs a specific algo gate.

Ahah, I see a sublte difference. Just a minor tweak to the blakecoin gate function should do it.

[JuanHungLo]

Vanilla coin needs a specific algo gate.

Ahah, I see a sublte difference. Just a minor tweak to the blakecoin gate function should do it.

Any links to a windows binary for this?  Just got a new i7 4790K and I wanna burn it in real good.

[joblo]

Vanilla coin needs a specific algo gate.

Ahah, I see a sublte difference. Just a minor tweak to the blakecoin gate function should do it.

Any links to a windows binary for this?  Just got a new i7 4790K and I wanna burn it in real good.

Sorry no windows yet.

[joblo]

cpuminer-opt v3.1.3 contains fixes for vanilla. It compiles and
works in benchmark mode.  Not tested on a live pool. YMMV.

https://drive.google.com/file/d/0B0lVSGQYLJIZNTFic1ZwVHVVUE0/view?usp=sharing

[joblo]

I've noticed some of the lesser known algos are having some problems since v3.1.

If any users find problems with algos that are listed as supported please report and
retest with v3.0.7.

[joblo]

cpuminer-opt v3.1.3 contains fixes for vanilla. It compiles and
works in benchmark mode.  Not tested on a live pool. YMMV.

https://drive.google.com/file/d/0B0lVSGQYLJIZNTFic1ZwVHVVUE0/view?usp=sharing

Tested working at 33.8 MH/s on 4790K.

There is a swerr message displayed that can be ignored. It will be fixed in the next release.

I wrote too soon. After two accepts nothing but rejects.

[PeaMine]

Very interesting thread and progress.
I have a Haswell and Ivy Linux and Windows available if ya need me to test anything.

[joblo]

Very interesting thread and progress.
I have a Haswell and Ivy Linux and Windows available if ya need me to test anything.

Can't do windows yet, buty if you can do linux...

[pallas]

I can confirm [the algo gate version] works on SSE2 and is slightly faster.
Good job! :-)

[zTheWolfz]

I've noticed some of the lesser known algos are having some problems since v3.1.

If any users find problems with algos that are listed as supported please report and
retest with v3.0.7.

Checking in: I tested v3.1 & 3.1.1 on 4 different rigs, some still using 3.0.2 others was using 3.07.
I'm seeing a great deal more rejects with v3.1/.1 on x11 and is the only algo I've tested.
I'm back on the older versions for now, will test newer release as soon as I can find some time.
3.02/3.0.7 only maybe 10 rejects out of 100,000 3.1/.1 somewhere between 8 to 20% rejects.

Oh & hash was just a tad lower than v3.02/.07 on x11.

[joblo]

I've noticed some of the lesser known algos are having some problems since v3.1.

If any users find problems with algos that are listed as supported please report and
retest with v3.0.7.

Checking in: I tested v3.1 & 3.1.1 on 4 different rigs, some still using 3.0.2 others was using 3.07.
I'm seeing a great deal more rejects with v3.1/.1 on x11 and is the only algo I've tested.
I'm back on the older versions for now, will test newer release as soon as I can find some time.
3.02/3.0.7 only maybe 10 rejects out of 100,000 3.1/.1 somewhere between 8 to 20% rejects.

Oh & hash was just a tad lower than v3.02/.07 on x11.

Thanks for testing.

This is SSE2 right?

I have noticed some rejects at startup somretimes but then things settle down. Do the rejects affects the
hash rate at the pool?

Things are settling down a bit so I will do a thorough performance test to find the best release.
There are also some regressions that need to be fixed.

My availability will be limited as I just bought a new PC and having problems with it.