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

[joblo]

Having some error when compiling:

These files do not exist in GIT.
What are the solutions?

Obviously those files do exist, you can look for yourself. It would help it you said what you did.

Here - https://github.com/JayDDee/cpuminer-opt I have not found these files.
I have tried to assemble these sources in VC + +2015.
Previously download them via GIT.
As well as simply by downloading the archive.
The content was identical.

Read the instructions before posting stupid questions.

[integrale]

Report as been asked for

Xubuntu 16.04LTS GCC 5.4.1  AMD Athlon II x2 240

Cpuminer-opt 3.5.10

Built with -march=btver1    compiling with -march=core2 makes no difference , same result   both arches choosen working on this CPU

algos they work or not    

timetravel  OK
neoscrypt   OK
decred        OK
blake2s      OK
x17             Fail
m7m          OK
c11             Fail
blakecoin   OK
sib              Fail
x11evo       Fail
veltor         OK
yescrypt     Fail
nist5           OK
groestl        OK
Qubit          Fail
x13             Fail
x14             Fail
x15             Fail
Quark         OK
myr-gr        OK
skein          OK
keccak       OK
xevan        Fail
hmq1725   Fail
lyra2rev2   OK
lyra2z        Fail

hope this help someone

[joblo]

Report as been asked for

Xubuntu 16.04LTS GCC 5.4.1  AMD Athlon II x2 240

Cpuminer-opt 3.5.10

Built with -march=btver1    compiling with -march=core2 makes no difference , same result   both arches choosen working on this CPU

algos they work or not    

timetravel  OK
neoscrypt   OK
decred        OK
blake2s      OK
x17             Fail
m7m          OK
c11             Fail
blakecoin   OK
sib              Fail
x11evo       Fail
veltor         OK
yescrypt     Fail
nist5           OK
groestl        OK
Qubit          Fail
x13             Fail
x14             Fail
x15             Fail
Quark         OK
myr-gr        OK
skein          OK
keccak       OK
xevan        Fail
hmq1725   Fail
lyra2rev2   OK
lyra2z        Fail

hope this help someone

Thanks. That was a lot of work, very much appreciated. It will help with my testing.
I don't often test for these architectures.

Edit:

I've tested the algos that failed for you on a i7-6700K compiled with gcc 4.8 using -march=core2 and they work.
You can retest hmq1275 in the next release. Some of the other algos have the same bug that caused x11 to crash
on Windows (The x11 bug affected only AVX2, not an issue on your x2). How did those algos fail?
Did the miner crash? Produce rejects? Do something else?

[joblo]

cpuminer-opt 3.5.11 is released. No optimizations, just bug fixes. Details in OP.

[denravonska]

Thanks. That was a lot of work, very much appreciated. It will help with my testing.
I don't often test for these architectures.

Edit:

I've tested the algos that failed for you on a i7-6700K compiled with gcc 4.8 using -march=core2 and they work.
You can retest hmq1275 in the next release. Some of the other algos have the same bug that caused x11 to crash
on Windows (The x11 bug affected only AVX2, not an issue on your x2). How did those algos fail?
Did the miner crash? Produce rejects? Do something else?

I'm trying on an Athlon64 X2 4000+ with gcc-5.2.1 using CFLAGS "-fomit-frame-pointer -Ofast -march=core2 -Wall". I made a super quick and dirty Python script which assumes that the algo works if it can run in the benchmark for 10 seconds:

- axiom: Fail
- bastion: Fail
- blake: OK
- blakecoin: Fail
- blake2s: OK
- bmw: Fail
- c11: Fail
- cryptolight: OK
- cryptonight: OK
- decred: OK
- deep: OK
- drop: OK
- fresh: Fail
- groestl: OK
- heavy: Fail
- hmq1725: Fail
- hodl: OK
- keccak: Fail
- lbry: OK
- luffa: Fail
- lyra2re: OK
- lyra2rev2: OK
- lyra2z: Fail
- lyra2zoin: Fail
- m7m: OK
- myr-gr: OK
- neoscrypt: OK
- nist5: OK
- pluck: Fail
- pentablake: OK
- quark: OK
- qubit: Fail
- scrypt: OK
- scryptjane:nf: Fail  <- Not sure how to test this.
- sha256d: OK
- shavite3: Fail
- skein: OK
- skein2: OK
- timetravel: Fail
- vanilla: Fail
- veltor: Fail
- whirlpool: Fail
- whirlpoolx: Fail
- x11: Fail
- x11evo: Fail
- x11gost: Fail
- x13: Fail
- x14: Fail
- x15: Fail
- x17: Fail
- xevan: Fail
- yescrypt: Fail
- zr5: OK

When failing it's always "Illegal instruction (core dumped)".

CPU:

CPU features: SSE2
SW built on Mar  1 2017 with GCC 5.2.1
SW features: SSE2
Algo features: SSE2
Start mining with SSE2

[ocminer]

I'm getting this error when trying to compile:

cpuminer-cpu-miner.o: In function `check_cpu_capability':
cpu-miner.c:(.text+0x8696): undefined reference to `set_incl'
cpu-miner.c:(.text+0x86a8): undefined reference to `set_incl'
cpu-miner.c:(.text+0x86ba): undefined reference to `set_incl'
algo/argon2/cpuminer-argon2a.o: In function `argon2hash':
argon2a.c:(.text+0x256): undefined reference to `argon_call'
algo/luffa/sse2/cpuminer-luffa_for_sse2.o: In function `update_luffa':
luffa_for_sse2.c:(.text+0x1dc): undefined reference to `mm_byteswap_epi32'
luffa_for_sse2.c:(.text+0x1f2): undefined reference to `mm_byteswap_epi32'
luffa_for_sse2.c:(.text+0x237): undefined reference to `mm_byteswap_epi32'
algo/luffa/sse2/cpuminer-luffa_for_sse2.o: In function `update_and_final_luffa':
luffa_for_sse2.c:(.text+0x595): undefined reference to `mm_byteswap_epi32'
luffa_for_sse2.c:(.text+0x5ab): undefined reference to `mm_byteswap_epi32'
algo/luffa/sse2/cpuminer-luffa_for_sse2.o:luffa_for_sse2.c:(.text+0x5ec): more undefined references to `mm_byteswap_epi32' follow
algo/luffa/sse2/cpuminer-luffa_for_sse2.o: In function `finalization512':
luffa_for_sse2.c:(.text+0x17853): undefined reference to `mm256_byteswap_epi32'
luffa_for_sse2.c:(.text+0x179ef): undefined reference to `mm256_byteswap_epi32'
algo/lyra2/cpuminer-lyra2.o: In function `LYRA2RE':
lyra2.c:(.text+0xd9e): undefined reference to `memset_zero_m256i'
algo/lyra2/cpuminer-sponge.o: In function `squeeze':
sponge.c:(.text+0x1d6): undefined reference to `memcpy_m256i'
sponge.c:(.text+0xd9d): undefined reference to `memcpy_m256i'
algo/lyra2/cpuminer-lyra2rev2.o: In function `lyra2rev2_thread_init':
lyra2rev2.c:(.text+0xb43): undefined reference to `memset_zero_m256i'
algo/lyra2/cpuminer-zcoin.o: In function `zcoin_thread_init':
zcoin.c:(.text+0x752): undefined reference to `memset_zero_m256i'
algo/lyra2/cpuminer-zoin.o: In function `zoin_thread_init':
zoin.c:(.text+0x564): undefined reference to `memset_zero_m256i'
algo/cpuminer-m7m.o: In function `swit2_':
m7m.c:(.text+0x1b3): undefined reference to `exp_n'
m7m.c:(.text+0x1e4): undefined reference to `exp_n'
algo/cpuminer-timetravel.o: In function `next_permutation':
timetravel.c:(.text+0x189): undefined reference to `tt_swap'
timetravel.c:(.text+0x19c): undefined reference to `reverse'
timetravel.c:(.text+0x1bb): undefined reference to `reverse'
collect2: error: ld returned 1 exit status
Makefile:1347: recipe for target 'cpuminer' failed
make[2]: *** [cpuminer] Error 1



Really weird.. What am I missing there ?

[ocminer]

I was able to compile the miner using the "-O3" flag in the compiler.. Whatever that means

I've created a pull request to add support for onecoin (OC):

https://github.com/JayDDee/cpuminer-opt/pull/5

Coin:
https://bitcointalk.org/index.php?topic=1801129

[joblo]

I was able to compile the miner using the "-O3" flag in the compiler.. Whatever that means

I've created a pull request to add support for onecoin (OC):

https://github.com/JayDDee/cpuminer-opt/pull/5

Coin:
https://bitcointalk.org/index.php?topic=1801129

I'm not sure what went wrong with your compile, but it was really messed up. -O3  is optimizaration
level 3, just a short cut to enable a bunch of individual options.

I'll take a look at your pull request.

[integrale]

the algos which gonna fail

miner stops after checking first block , crash is then reported by miner as illegal instruction

3.5.11 tested , same results as in 3.5.10 nothing changes for Amd cpu

[oldDIN]

cpuminer-opt 3.5.11 is released. No optimizations, just bug fixes. Details in OP.

He's lying, lying  
I'm on my 3.5.11 compilation gcc 6.3 speed became higher by 3-5% than at the same 3.5.10

[joblo]

cpuminer-opt 3.5.11 is released. No optimizations, just bug fixes. Details in OP.

He's lying, lying  
I'm on my 3.5.11 compilation gcc 6.3 speed became higher by 3-5% than at the same 3.5.10

LOL. Which algo?

[integrale]

fine , then you can count yourself to the lucky one´s

we´ve talked about this config

Xubuntu 16.04LTS GCC 5.4.1  AMD Athlon II x2 240

Cpuminer-opt 3.5.10

hmq1725 and some other x-variants , also the hashrate is lower at about 10kh/s compared to 3.5.9 on timetravel makes around 8% in my case

[joblo]

the algos which gonna fail

miner stops after checking first block , crash is then reported by miner as illegal instruction

3.5.11 tested , same results as in 3.5.10 nothing changes for Amd cpu

Thanks for testing. I think I've done all I can to get thosre old AMDs to work.

An illegal instruction usually means the CPU hasn't implemented it, often when trying to
ue AES instructions on a CPU that doesn't support AES. In this case it appears these AMD
CPUs don't have the full SSE2 implementation.

Even if there was a way for the compiler to detect this condition the result would be a build
without any optimizations, ie it would be the same as cpuminer-multi.

My recommendation is to use cpuminer-multi on those algos.

[integrale]

yes  , good advice    only that multiminer doesnt support hmq1725 yet  , maybe in future releases  hopefully

thank you

[oldDIN]

LOL. Which algo?

Yet they do not believe  
AMD FX-6300 t-5 example
hmq1725   90 -  95
zoin        201 - 210
TimeTrav 360 - 385
Although the transition has been with GCC 6.2  GCC 6.3
Is he able to influence?

[integrale]

FX 6300

Features   

    MMX instructions
    Extensions to MMX
    SSE / Streaming SIMD Extensions
    SSE2 / Streaming SIMD Extensions 2
    SSE3 / Streaming SIMD Extensions 3
    SSSE3 / Supplemental Streaming SIMD Extensions 3
    SSE4 / SSE4.1 + SSE4.2 / Streaming SIMD Extensions 4  ?
    SSE4a  ?
    AES / Advanced Encryption Standard instructions
    AVX / Advanced Vector Extensions
    BMI1 / Bit Manipulation instructions 1
    FMA3 / 3-operand Fused Multiply-Add instructions
    FMA4 / 4-operand Fused Multiply-Add instructions
    F16C / 16-bit Floating-Point conversion instructions
    TBM / Trailing Bit Manipulation instructions
    XOP / eXtended Operations instructions
    AMD64 / AMD 64-bit technology  ?
    AMD-V / AMD Virtualization technology
    EVP / Enhanced Virus Protection  ?
    Turbo Core 3.0 technology

[oldDIN]

integrale
Для чeгo ты этo нaпиcaл? Я oтвeчaл нe тeбe. Coмнeний пo твoим пpoблeмaм нe вoзникaлo.

[joblo]

fine , then you can count yourself to the lucky one´s

we´ve talked about this config

Xubuntu 16.04LTS GCC 5.4.1  AMD Athlon II x2 240

Cpuminer-opt 3.5.10

hmq1725 and some other x-variants , also the hashrate is lower at about 10kh/s compared to 3.5.9 on timetravel makes around 8% in my case

Can you clarify the performance loss? Timetravel is dificult to measure because the hash rate is always changing.
You need to compare both immediately after each other and sometimes repeatedly to ensure the same permutation
is tested.

Most of the optimizations I implemented don't help on CPUs without AES and AVX, but shouldn't hurt either.
If Timetravel is slower in 3.5.10 many other algos should also be slower. It's possible some changes may have
actually hurt performance but but wasn't obvious because it was offset by gains from other changes. It may
only become apparent on CPUs that don't get the offsetting gains. The result is I may be bale to restore the
lost performance on your x2 while also improving the performance on more advanced acrhitectures.

TLDR alert.

I am discovering this more and more. My attempts to implemet AVX2 for Luffa have been slower. Switching
back and forth from 128 bit to 256 bit vectors adds overhead. Conversely converting some operations to 256 bit
adds complexity which also adds overhead.

This kind of optimization reduces the instruction count but only if the operations translate directly from 128 to 256 bits.
Anything more complex than simple arithmetic will likley add overhead by requiring more complex (and slower)
instructions or more instructions to perform the same operation. This negates the instruction savings of doubling up
the data size. Also AVX2 doesn't reduce memory accesses, it can just do them with fewer instructions.

As optimizing gets more aggressive there is more potential for unintended side effects. I would like to avoid them
if possible.

[joblo]

LOL. Which algo?

Yet they do not believe  
AMD FX-6300 t-5 example
hmq1725   90 -  95
zoin        201 - 210
TimeTrav 360 - 385
Although the transition has been with GCC 6.2  GCC 6.3
Is he able to influence?

Not even AVX2, pretty good. The thing is there is nothing that changed in those algos in 3.5.11
except fixing the bug in hmq1275, but that had no effect on performance.

[joblo]

yes  , good advice    only that multiminer doesnt support hmq1725 yet  , maybe in future releases  hopefully

thank you

I stole the code from ocminer at suprnova.

https://github.com/ocminer/cpuminer-hmq1725