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

[joblo]

if someone have troubles to compile the right Arch

different CPU architecture:

nehalem = 1st Generation Intel® Core™ (also westmere)
sandybridge = 2nd Generation Intel® Core™ (also ivybridge)

I little correction, the family name doesn't always work. On Intel native should always work. One exception
is Westmeme where the native compile may fail to generate AES code. Any Westmere users who compile
there own could confirm. AMD is more difficult, the brand names sometimes work but don't seem to
generate optimum code. My best advoce for AMD users is to choose the arch that best matches the CPUs features.
AMD users who compile may have better advice.

The Windows binaries are all built as follows, most users can find the right fit.

Core2 is "-march=core2".
Nehalem does not have AES or AVX, the arch name is -march=corei7
There aren't any feature differences between core2 and Nehalem that affect cpouminer.
Westemere does have AES but does not have AVX, there is no compile arch name, use "-maes, -msse4.2"
Sandybridge has AES and AVX, the archname is "-march=corei7-avx"
Ivybridge has AES and AVX but it's own arch name "-march=core-avx-i", not built
There are no feature differences between Sandybridge and Ivybridge that affect cpuminer, I build corei7-avx.
Haswell and above use "-march=core-avx2"

[Epsylon3]

pentium G can have problems also, native seems to enable AVX which is not on these... but was maybe fixed in the recent gcc.. was the case with 4.8/4.9

[joblo]

pentium G can have problems also, native seems to enable AVX which is not on these... but was maybe fixed in the recent gcc.. was the case with 4.8/4.9

Thanks. It looks like Pentium G may be like Westmere, AES without AVX.
The fallback in any case where native compile doesn't match with the CPUs features is to specify
them manually with -m.

[integrale]

have also expirenced with my i7 2nd gen (sandybridge)   but wont work when i chose Sandybridge executables ?
on Windows where given binary aviable are and discribed for  this core family
when im using cpum....i7-sse41.exe it works on some miner´s

[joblo]

have also expirenced with my i7 2nd gen (sandybridge)   but wont work when i chose Sandybridge executables ?
on Windows where given binary aviable are and discribed for  this core family
when im using cpum....i7-sse41.exe it works on some miner´s

This doesn't seem right. All Sandybridge core-i should have AVX. What happens when you compile native?
What happens when you run the AVX binary? Is this the i7-2600 you mentioned earlier? I have an i5-2400
and it perfoms with AVX as expected.

[joblo]

I found this regarding the difference between Sandybridge and Ivybridge. Not much. It doesn't
affect targetted code and I have no idea what opimizations the compiler may do to portable
code. It does only apply to the core series, not Pentium G or Celeron.

CORE-AVX-I    May generate Intel® AVX, SSE4.2, SSE4.1, SSSE3, SSE3, SSE2 and SSE instructions for
Intel® processors, including instructions for 3rd generation Intel® Core™ processors. Optimizes for
3rd generation Intel® Core™ processors and the Intel® Xeon® Processor E3 v2, E5 v2 and E7 v2 families.

AVX    May generate Intel® AVX, SSE4.2, SSE4.1, SSSE3, SSE3, SSE2 and SSE instructions for Intel® processors.
Optimizes for 2nd generation Intel® Core™ i7, i5 and i3 processor families and the Intel® Xeon® Processor
E5 and E3 families.

[integrale]

if i compile from source on Linux by myself its ok . it happens only sometimes with precompiled binary on windows

need to find some cheap Dual cpu board ,socket 771 , have here two E5462 2.8 ghz 12Mb  laying around   lazy things   so i cant test something with them
just now found this one :Serverboard Supermicro X7DBP-I Dual Socket 771   inkl heatsinks ram and  2x E5405  affordable price    holy shi...  75€

[Epsylon3]

beware with the memory types of these :p a mistake could double the price

[integrale]

dont know what you mean   , pls advice me   i know  they are only sse3 able  but  what shall i do   the package is cheap

[joblo]

After a year of trying I've finally made progress improving cryptonight. I think I can get 5% with
a combination of aligning data on a cache line boundary and aggressive prefetching.

Also a reminder the next release will use a new naming scheme for the Windows binaries. The names
indicates the CPU features required to use them. Intel or AMD, choose the one that best matches your CPU's
features. Or you can use trial and error to find the best one.

Code:

cpuminer-sse2.exe         -march=core2,      Core2  
cpuminer-sse42.exe        -march=corei7,     Nehalem
cpuminer-aes-sse42.exe    -maes -msse4.2     Westmere
cpuminer-aes-avx.exe      -march=corei7-avx, Sandybridge, Ivybridge
cpuminer-aes-avx2.exe     -march=core-avx2,  Haswell, Broadwell, Skylake, Kabylake

Prepare to update your scripts, that's bat for Windows users.

[m1n1ngP4d4w4n]

nice work joblo ! Maybe we'll get up par to that private miner from minergate I'll be more than willing to test that out !

[joblo]

cpuminer-opt-3.5.7 is released. See OP, link in sig.

After a year of trying I've finally made progress improving cryptonight. I think I can get 5% with
a combination of aligning data on a cache line boundary and aggressive prefetching.

Also a reminder the next release will use a new naming scheme for the Windows binaries. The names
indicates the CPU features required to use them. Intel or AMD, choose the one that best matches your CPU's
features. Or you can use trial and error to find the best one.

Code:

cpuminer-sse2.exe         -march=core2,      Core2  
cpuminer-sse42.exe        -march=corei7,     Nehalem
cpuminer-aes-sse42.exe    -maes -msse4.2     Westmere
cpuminer-aes-avx.exe      -march=corei7-avx, Sandybridge, Ivybridge
cpuminer-aes-avx2.exe     -march=core-avx2,  Haswell, Broadwell, Skylake, Kabylake

Prepare to update your scripts, that's bat for Windows users.

[m1n1ngP4d4w4n]

Excellent work man, it's finally matching minergate private cpu-miner !

Mad props to you  

cpuminer-opt v3.5.7

Code:

[2017-02-14 21:58:47] CPU #2: 508 H, 8.45 H/s
[2017-02-14 21:58:47] CPU #7: 513 H, 8.54 H/s
[2017-02-14 21:58:47] CPU #3: 503 H, 8.38 H/s
[2017-02-14 21:58:47] CPU #0: 515 H, 8.56 H/s
[2017-02-14 21:58:47] CPU #1: 511 H, 8.50 H/s
[2017-02-14 21:58:47] CPU #5: 501 H, 8.34 H/s
[2017-02-14 21:58:47] CPU #4: 503 H, 8.36 H/s
[2017-02-14 21:58:47] CPU #6: 502 H, 8.35 H/s
[2017-02-14 21:59:11] CPU #2: 201 H, 8.45 H/s
[2017-02-14 21:59:11] Accepted 3/3 (100%), 3749 H, 67.47 H/s, 37C

Minergate cli 4.04

Code:

[2017-02-14 22:01:42.485] [ info]  XMR hashrate: 67.3 H/s
[2017-02-14 22:01:52.486] [ info]  XMR hashrate: 67.1933 H/s
[2017-02-14 22:02:00.094] [ info]  New difficulty: 1063
[2017-02-14 22:02:02.485] [ info]  XMR hashrate: 67.435 H/s

I can finally get rid of minergate, you made me a happy man, donation coming your way 

[joblo]

Excellent work man, it's finally matching minergate private cpu-miner !

Mad props to you  

cpuminer-opt v3.5.7

Code:

[2017-02-14 21:58:47] CPU #2: 508 H, 8.45 H/s
[2017-02-14 21:58:47] CPU #7: 513 H, 8.54 H/s
[2017-02-14 21:58:47] CPU #3: 503 H, 8.38 H/s
[2017-02-14 21:58:47] CPU #0: 515 H, 8.56 H/s
[2017-02-14 21:58:47] CPU #1: 511 H, 8.50 H/s
[2017-02-14 21:58:47] CPU #5: 501 H, 8.34 H/s
[2017-02-14 21:58:47] CPU #4: 503 H, 8.36 H/s
[2017-02-14 21:58:47] CPU #6: 502 H, 8.35 H/s
[2017-02-14 21:59:11] CPU #2: 201 H, 8.45 H/s
[2017-02-14 21:59:11] Accepted 3/3 (100%), 3749 H, 67.47 H/s, 37C

Minergate cli 4.04

Code:

[2017-02-14 22:01:42.485] [ info]  XMR hashrate: 67.3 H/s
[2017-02-14 22:01:52.486] [ info]  XMR hashrate: 67.1933 H/s
[2017-02-14 22:02:00.094] [ info]  New difficulty: 1063
[2017-02-14 22:02:02.485] [ info]  XMR hashrate: 67.435 H/s

I can finally get rid of minergate, you made me a happy man, donation coming your way 

Thanks.

Wow the're pretty close. You might do better with 4 threads depending on your cache size.
threads = cache size MB / 2 MB. Four threads makes a big difference on my i7-6700K.

[m1n1ngP4d4w4n]

Excellent work man, it's finally matching minergate private cpu-miner !

Mad props to you  

cpuminer-opt v3.5.7

Code:

[2017-02-14 21:58:47] CPU #2: 508 H, 8.45 H/s
[2017-02-14 21:58:47] CPU #7: 513 H, 8.54 H/s
[2017-02-14 21:58:47] CPU #3: 503 H, 8.38 H/s
[2017-02-14 21:58:47] CPU #0: 515 H, 8.56 H/s
[2017-02-14 21:58:47] CPU #1: 511 H, 8.50 H/s
[2017-02-14 21:58:47] CPU #5: 501 H, 8.34 H/s
[2017-02-14 21:58:47] CPU #4: 503 H, 8.36 H/s
[2017-02-14 21:58:47] CPU #6: 502 H, 8.35 H/s
[2017-02-14 21:59:11] CPU #2: 201 H, 8.45 H/s
[2017-02-14 21:59:11] Accepted 3/3 (100%), 3749 H, 67.47 H/s, 37C

Minergate cli 4.04

Code:

[2017-02-14 22:01:42.485] [ info]  XMR hashrate: 67.3 H/s
[2017-02-14 22:01:52.486] [ info]  XMR hashrate: 67.1933 H/s
[2017-02-14 22:02:00.094] [ info]  New difficulty: 1063
[2017-02-14 22:02:02.485] [ info]  XMR hashrate: 67.435 H/s

I can finally get rid of minergate, you made me a happy man, donation coming your way  

Thanks.

Wow the're pretty close. You might do better with 4 threads depending on your cache size.
threads = cache size MB / 2 MB. Four threads makes a big difference on my i7-6700K.

I was compiling it to other cpu's to test if the speed increase still the same, will try optimizing after this, i'll get back to you with more numbers in a few

Edit :

so i have 2 cpus

Intel(R) Xeon(R) CPU L3426  @ 1.87GHz - SSE2 - 8MB cache L2 - Non fixed thread count (use as much as logical core) - Thread 8

Minergate CLI : ~60 H/s
Cpuminer 3.5.7 :  ~55-56 H/s

Intel(R) Atom(TM) CPU  C2750  @ 2.40GHz - SSE2 / AES - 4MB cache L2 - Non fixed thread count (use as much as logical core) - Thread 8

Minergate CLI : 67-67.5H/s
Cpuminer 3.5.7 : ~68 H/s

So it seems that on AES/SS2 cpu it match and even go a tad above in speed. But for SS2 CPU only, it's better but not on par yet. One thing to note, when i use a lesser thread count than my logical core, it only hash on as much core as the thread allow so it give me less overall hash.

[oldDIN]

joblo, not critical, but....forgotten lyra2re
http://prntscr.com/e8sqcd  ore  http://prntscr.com/e8sr0b
same on intel cpuminer-sse42.exe

[joblo]

joblo, not critical, but....forgotten lyra2re
http://prntscr.com/e8sqcd  ore  http://prntscr.com/e8sr0b
same on intel cpuminer-sse42.exe

Oh it crashes. You could have said so.

Seems to be a Windows problem with lyra2RE. I hate it when it works on Linux but not on Windows.
I'm looking into it.

[joblo]

So it seems that on AES/SS2 cpu it match and even go a tad above in speed. But for SS2 CPU only, it's better but not on par yet. One thing to note, when i use a lesser thread count than my logical core, it only hash on as much core as the thread allow so it give me less overall hash.

Correct I only optimized the AES version. I can port it to the non-aes version, it's a simple copy paste.

[m1n1ngP4d4w4n]

That would be great, i have a few old servers that would benefit from that, if it's not too much of a hassle for you  

Sent you a beer

Address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
TxId: 1c55626d7c10feacec581988396b367171b18810dadf90ea110038701d274f7c

Keep up the good work 

[integrale]

V.3.5.7 Linux  i7 2600  4 threads

confirming increase + 20 h/s  from previous release

[2017-02-14 23:03:16] CPU #1: 1918 H, 63.90 H/s
[2017-02-14 23:03:20] CPU #1: 266 H, 64.07 H/s
[2017-02-14 23:03:20] Accepted 11/11 (100%), 5431 H, 254.85 H/s, 59C
[2017-02-14 23:03:21] CPU #2: 356 H, 64.00 H/s
[2017-02-14 23:03:21] Accepted 12/12 (100%), 4423 H, 254.74 H/s, 60C
[2017-02-14 23:03:22] CPU #3: 421 H, 63.04 H/s
[2017-02-14 23:03:22] Accepted 13/13 (100%), 2947 H, 254.56 H/s, 60C
[2017-02-14 23:03:25] CPU #0: 568 H, 63.05 H/s
[2017-02-14 23:03:25] Accepted 14/14 (100%), 1611 H, 254.15 H/s, 61C


Great job , thx Joblo