Topic: [ANN]: cpuminer-opt v3.8.8.1, open source optimized multi-algo CPU miner - page 110. (Read 444067 times)

you can also toggle warnings per kind with CFLAGS :

-Wimplicit-function-declaration vs -Wno-implicit-function-declaration

but they are there for a good reason in general

Since no one has reported it It's probably not a big issue but I have just discovered x11evo is broken
in the latest release. If you have a working version don't upgrade until this is fixed.

It works in 3.5.1, broken in 3.5.2.

Edit: I've got a fix plus a surprise. I still need to finish the other stuff I was working on before the next release.

Tired of seeing all those compiler warnings from algo-gate-api.c? Well the're going away. I found the
way to supress them for selected code segments.

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wimplicit-function-declaration"
// ignore these warnings
[...]
// until here
#pragma GCC diagnostic pop

These warnings get displayed because algos don't, are for convenience are not required, to declare
their registration functions in base code.

On another note I still hate c pointer arithmetic. Pointer arithmetic should always use byte offsets,
arrays indexes scale to the target size. p+n should always have the same result regardless of
the size of p's target. I wasted the better part of a day fighting with a problem that turned out to
be bad pointer arithmetic.

thx a lot , works now , step by step increasing against win version +7k/h   woow old crap digging
Xubuntu 16.04 LTS @ Core2 Duo E6600 @ 83k/h
                                                     E6700 @ 89k/h
Eeeeeennndlich      last machine compiled   uffff  !!!!

CPU:         Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
CPU features: SSE2 AES AVX
SW built on Jan 30 2017 with GCC 5.4.0
SW features: SSE2 AES AVX
Algo features: SSE2 AES AVX AVX2
Start mining with SSE2 AES AVX

@  ~520 k/h        BIG thx for the great work to Joblo & Epsylon3 , my race is over for now

miner3@miner3-HP-500B:~$ cd cpuminer-opt-3.5.3
miner3@miner3-HP-500B:~/cpuminer-opt-3.5.3$ ./cpuminer

         **********  cpuminer-opt 3.5.3  ***********
     A CPU miner with multi algo support and optimized for CPUs
     with AES_NI and AVX extensions.
     BTC donation address: 12tdvfF7KmAsihBXQXynT6E6th2c2pByTT
     Forked from TPruvot's cpuminer-multi with credits
     to Lucas Jones, elmad, palmd, djm34, pooler, ig0tik3d,
     Wolf0, Jeff Garzik and Optiminer.

[2017-01-30 10:45:40] FAIL: algo_gate registration failed, unknown algo (null).

miner3@miner3-HP-500B:~/cpuminer-opt-3.5.3$

conf:

{
   "_comment1" : "Any long-format command line argument ",
   "_comment2" : "may be used in this JSON configuration file",

   "api-bind" : "127.0.0.1:4048",

   "url" : "stratum+tcp://fiber.ccminer.org:3555",
   "user" : "MTRLHnKQpQby52DHtCYHsXFeJbwCsdf7Wx",
   "pass" : "x,c=MAC",

   "algo" : "timetravel",
   "threads" : 2,
   "cpu-priority" : 0,
   "cpu-affinity" : -1,

   "benchmark" : false,
   "debug" : false,
   "protocol": false,
   "quiet" : false
}

No.

Instability on the amd fx-6300, 50-52 accept on and off ((
I can see 3.5.4 is tested, can it be stable

hi all ...

ive finally got my old profile back ...

tanx to cyrus ( bct mod ) and a few others - i now have full access to my old profile ... so will be using this one as much as possible from now on ...

the 'crysx' profile will still be in use - but will be only when necessary - not as my standard ...

a lot to sort out now ...

update ...
https://bitcointalksearch.org/topic/m.17653092 ...

#crysx

Thanks.

Nice Work bro!

It works for me, need more info.

Timetravel benchmark is broken

Great Job Joblo  

big thank´s ,  increase on i7 2600  +70k/h  means atm up to 475K/h  for this old thing. at Timetravel  , others not tested yet

btw  may i ask you why i cant use the avx ?  is that a limitation of Windows , cpu works only with the sse41 , or corei7 exe´s

Processors Information
-------------------------------------------------------------------------

Processor 1         ID = 0
   Number of cores      4 (max
   Number of threads   8 (max 16)
   Name         Intel Core i7 2600
   Codename      Sandy Bridge
   Specification      Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
   Package (platform ID)   Socket 1155 LGA (0x1)
   CPUID         6.A.7
   Extended CPUID      6.2A
   Core Stepping      D2
   Technology      32 nm
   TDP Limit      95.0 Watts
   Tjmax         98.0 °C
   Core Speed      3492.9 MHz
   Multiplier x Bus Speed   35.0 x 99.8 MHz
   Stock frequency      3400 MHz
   Instructions sets   MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, EM64T, VT-x, AES, AVX
   L1 Data cache      4 x 32 KBytes, 8-way set associative, 64-byte line size
   L1 Instruction cache   4 x 32 KBytes, 8-way set associative, 64-byte line size
   L2 cache      4 x 256 KBytes, 8-way set associative, 64-byte line size
   L3 cache      8 MBytes, 16-way set associative, 64-byte line size
   FID/VID Control      yes

cpuminer-opt v3.5.3 is released.

Timetravel +16%
Xevan +3%
Qubit +12%

https://bitcointalksearch.org/topic/ann-cpuminer-opt-v3881-open-source-optimized-multi-algo-cpu-miner-1326803

The optimized Luffa is now working for various data lengths where previously it only worked with 64 byte length.
It also now works with 512 or 1024 bit hash length.

I now have timetravel, xevan and qubit working with the optimized Luffa. This marks a significant milestone.
Except for the hard coded macros, all of the original optimized x11 functions have now been implemented in
all algos that use them.

There is still some cleanup and testing to do before release.

It depend on the algo. Memory hard algos are most affected.

is memory mode has such influance on hashing speed?have one channel conf.

I think I've found the problem with the SSE2 version of Luffa. It simply doesn't support 80 byte
input. The datalen arg is ignored and it always hashes just 64 bytes. This explains why it worked for
a while in Timetravel. It also explains why it never worked in Qubit even though I passed the correct
datalen.

It looks fairly straighforward to change based on how other functions do it.

I've also been looking at the AES Groestl function. It has a lot of code that isn't used, much of it to support
odd hash lengths. Since only two hash lengths are ever used, 64 and 80 bytes, Some cleanup might give
a little boost in speed.

I cant blame you for not being a fan of the coin as it is plagued by multiple issues at the moment.

I was just curious about the mining speeds with a quad channel setup more than anything.

I am guessing you are doing about 25 h/s with your setup