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

[mikeirvine]

 **********  cpuminer-opt 3.8.0  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI and AVX2 and SHA extensions.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT

CPU: Intel(R) Core(TM) i3 CPU       M 380  @ 2.53GHz.
SW built on Jan 24 2018 with GCC 7.2.0.
CPU features: SSE2.
SW features: SSE2.
Algo features: AVX AVX2.
Start mining with no optimizations.




Why Start mining with ""no optimizations.""?

Tried -march=haswell and -march=native

But still the same.

[1714976606]

1714976606

[1714976606]

1714976606

[1714976606]

1714976606

[joblo]

**********  cpuminer-opt 3.8.0  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI and AVX2 and SHA extensions.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT

CPU: Intel(R) Core(TM) i3 CPU       M 380  @ 2.53GHz.
SW built on Jan 24 2018 with GCC 7.2.0.
CPU features: SSE2.
SW features: SSE2.
Algo features: AVX AVX2.
Start mining with no optimizations.




Why Start mining with ""no optimizations.""?

Tried -march=haswell and -march=native

But still the same.

You didn't specify the algo (it's important) but the issue is very clear. Your CPU only supports SSE2 but the algo
does have any SSE2. BTW you don't have a haswell.

[mikeirvine]

**********  cpuminer-opt 3.8.0  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI and AVX2 and SHA extensions.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT

CPU: Intel(R) Core(TM) i3 CPU       M 380  @ 2.53GHz.
SW built on Jan 24 2018 with GCC 7.2.0.
CPU features: SSE2.
SW features: SSE2.
Algo features: AVX AVX2.
Start mining with no optimizations.




Why Start mining with ""no optimizations.""?

Tried -march=haswell and -march=native

But still the same.

You didn't specify the algo (it's important) but the issue is very clear. Your CPU only supports SSE2 but the algo
does have any SSE2. BTW you don't have a haswell.

I am mining on lyra2zoin or lyra2z330

[joblo]

**********  cpuminer-opt 3.8.0  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI and AVX2 and SHA extensions.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT

CPU: Intel(R) Core(TM) i3 CPU       M 380  @ 2.53GHz.
SW built on Jan 24 2018 with GCC 7.2.0.
CPU features: SSE2.
SW features: SSE2.
Algo features: AVX AVX2.
Start mining with no optimizations.




Why Start mining with ""no optimizations.""?

Tried -march=haswell and -march=native

But still the same.

You didn't specify the algo (it's important) but the issue is very clear. Your CPU only supports SSE2 but the algo
does have any SSE2. BTW you don't have a haswell.

I am mining on lyra2zoin or lyra2z330

Typo in my first reply: algo doesn't have SSE2.
You'll need a newer CPU to get optimizations on lyra2z330 but they don't help much because that algo is I/O bound.

[zazaxxx]

v3.8.0 AVX2 build crashes on neoscrypt a few seconds after starting.

[zazaxxx]

Any plans to add 4way to yescrypt?

I deliberately did not mention the algos that I'm looking at because I have no idea if I can do anything
with them yet.

It's already good to know the idea hasn't been discarded. Thanks.

[4ward]

Strange observations about HODL algo:
1. It almost never use 100% cpu. Most of the time it floats somewhere in 70-90
2. It doesn't like some thread counts (3, 6, 7) in avx/avx2 versions

any ideas? ))

[joblo]

Strange observations about HODL algo:
1. It almost never use 100% cpu. Most of the time it floats somewhere in 70-90
2. It doesn't like some thread counts (3, 6, 7) in avx/avx2 versions

any ideas? ))

Hodl is memory hard so it is likely the memory bandwifth is exhausted causing the cpu
to stall waiting for data. It also two unique features not present in any other algo.
It has thread synchronization so threads that find a solution quicker have to wait for
the others to catch up. Hodl threads also work on slices of a shared buffer, it must be divided
by 2, 4, 8 etc, to work properly.

[lncm]

Could optimizations for Scrypt (Verium) be expected?

Official wallet miner says it supports AVX/AVX2.

Thanks.

[Slkp]

what about avx-512 support?

[joblo]

Good afternoon.
I got 2x Opteron 6282 Se
http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron%206282%20SE.html
how do I correctly set --cpu-affinity (CPU cores 32)?
http://store.fsxtimes.com/cal-cam.php
 yescrypt algorithm
FX 8350 - cpu-affinity 255 (CPU cores - 3900 h / s
xeon E5 2670 - cpu-affinity 65535 (CPU cores 16) - 4700 h / s
2x Opteron 6282 Se - cpu-affinity 0x5555555555555555 (CPU cores 32) -10280 h / s (Windows Server 2012 L3 24 mb instead of 32, there is such problemma!BIOS sees Level 3 cache size 32 mb, ie for each processor of 16 MB, problem in Windows!
https://support.microsoft.com/en-us/help/3019274/l3-cache-size-is-reported-incorrectly-from-windows-management-instrume
You should try Windows server 2008
Joblo on version 3.8.0 - 2x Opteron 6282 Se --cpu-affinity 0x5555555555555555 - 3500 h / s :-( something you broke (why is it so?
Thank you

I can't sort through this. If there's a problem in there please describe it and leave everything else out.

[akuci]

Any plan for scrypt² (Verium) optimization?

[robminer80]

Could you update Windows compilation instructions?

[joblo]

Could you update Windows compilation instructions?

Can you be more specific? The documented procedure is the one that works to compile on Wndows
for Windows. It's old but is the only one I've found that works.

A better way was found but it is compiled on Linux for Windows. This isn't practical for Windows users
as they would need Linux as well.

Unfortunately there is no good solution at this time for Windows users to compile from source..

[joblo]

Sorry, but I do not show the pool and coins (

cpuminer-sse2 -a yescrypt -t 32 -o stratum+tcp://pool -u adres -p x --cpu-affinity --cpu-affinity 0x5555555555555555

**********  cpuminer-opt 3.7.10  ***********

Start mining with SSE2.

[2018-01-27 10:36:33] Starting Stratum on stratum+tcp://pool
[2018-01-27 10:36:33] 32 miner threads started, using 'yescrypt' algorithm.
[2018-01-27 10:36:34] Stratum difficulty set to 0.5
[2018-01-27 10:36:39] CPU #25: 512 H, 318.14 H/s


cpuminer-sse2 -a yescrypt -t 32 -o stratum+tcp://pool -u adres -p x --cpu-affinity --cpu-affinity 0x5555555555555555

 **********  cpuminer-opt 3.8.0  ***********

Start mining with SSE2.

[2018-01-27 10:37:53] Binding process to cpu mask 23fcb0
[2018-01-27 10:37:53] Starting Stratum on stratum+tcp://pool
[2018-01-27 10:37:53] 32 miner threads started, using 'yescrypt' algorithm.
[2018-01-27 10:37:55] Stratum difficulty set to 0.5
[2018-01-27 10:37:58] yescrypt block 64773, diff 0.104
[2018-01-27 10:38:01] Share submitted.
[2018-01-27 10:38:01] CPU #7: 412 H, 225.37 H/s
[2018-01-27 10:38:01] Accepted 1/1 (100%), diff 1.6e-005, 225.37 H/s


I noticed in version 3.7.10 does not understand the cpu mask
I use the same batch file)
Thank you

Yes it appears to be a problem with cpu-affinity. I made a change in 3.8.0 but it appears it doesn't work.
I'll remove it in the next release and go back to the old way.

When you're testing you can get more info by using --debug, it will give more details about cpu-affinity.

Edit: Why are you using cpu-affinity with 32 threads? Two 6282 SE is 32 cores. With the affinity mask you
used you are only allowing 16 cores to be used for 32 threads.

Don't ever use --cpu-affinity with the same number of threads as cpu cores available, it will only make
things worse.

Please retest v3.7.10 and v3.8.0 with default threads and affinity and with --debug

[robminer80]

Could you update Windows compilation instructions?

Can you be more specific? The documented procedure is the one that works to compile on Wndows
for Windows. It's old but is the only one I've found that works.

A better way was found but it is compiled on Linux for Windows. This isn't practical for Windows users
as they would need Linux as well.

Unfortunately there is no good solution at this time for Windows users to compile from source..

That procedure results in use old version of GCC (I tried to update it with no luck, build not works). I need to compile march native for Kaveri architecture. In XMRig (for example) I have 10% boost on precompiled version and 20% boost compared to cpuminer-opt using cryptonight and cryptonight-lite algo.

[cryptonitro]

Hey friends , can we have a line to use a percent of cpu ?
like
cpu max alloc percent ? or something like that in batch file ?

[4ward]

Hey friends , can we have a line to use a percent of cpu ?
like
cpu max alloc percent ? or something like that in batch file ?

use "-t" param to use specific amount of cores (i.e. "-t 3" on 4 core cpu)

[4ward]

Could you update Windows compilation instructions?

A better way was found but it is compiled on Linux for Windows. This isn't practical for Windows users
as they would need Linux as well.

Unfortunately there is no good solution at this time for Windows users to compile from source..

I actually use the cross compile procedure on Ubuntu shell in Windows 10.
But it still does require some basic linux knowledge

[joblo]

Could you update Windows compilation instructions?

A better way was found but it is compiled on Linux for Windows. This isn't practical for Windows users
as they would need Linux as well.

Unfortunately there is no good solution at this time for Windows users to compile from source..

I actually use the cross compile procedure on Ubuntu shell in Windows 10.
But it still does require some basic linux knowledge

This is something I could add to the procedure. I don't have Win10 so I can't try it myself.