[ANN] CPU Miner Advanced v5.0.2.0 Beta By Spoetnik

[Spoetnik]

I wanted to share my version of the JackpotCoin CPU miner i have been working on with you all

So for now i will post the change-log here and on the next comment a bunch of preview pic's.
I will also add the Windows 64bit build version i am using optimized for my Intel 4770k CPU.
If your cpu is older than mine the build will most likely crash for you sorry.
I hope to work on more build versions and would appreciate any donations
I recently have had issues with my MingW install after porting it from one machine to another.

edit:
Windows x64 Haswell version - made for the CPU Intel 4770k
compiled using MingW x64 using CFLAGS="-march=native -O3"
So it requires a cpu that can do AVX etc

Other optimized versions coming soon (like 32bit)

NEWEST VERSION:
cma-v5.0.2.0-vs2013-win-x64-spoetnik.rar (644.92 KB) (Windows 64bit SSE2 / SSE4 / AVX2)
http://www.multiupload.nl/QYXHV8Y9JH

cma-v5.0.2.0-mingw-win-x64-spoetnik.rar (696 KB) (Windows 64bit SSE2 / SSE4 / AVX2)
http://www.multiupload.nl/XCEDL7C42H

cma-v5.0.2.0-src-spoetnik.rar (271.28 KB) (source code only)
http://www.multiupload.nl/Y9YA3NV84F

-----------------------------------------------------
OLD Test Version: v5.0.1.0 - CHEAT TEST
http://www.multiupload.nl/05HQV1EELM
that is CMA-5.0.1.0-Beta-WIN-x64-Haswell.rar (1.33 MB)
has 6 windows 64bit versions.. sse2,sse4,avx2 (cheat test and normal versions) plus source code.

Test Version: v5.0.0.8 - cheat test
http://www.multiupload.nl/BB2173QQA7
Includes 4 builds + Source code (2 versions for 2 cpu's each) one with cheat mode on and the other with cheat mode off.
Working on Intel 4770k and AMD Phenom x4 9950 BE (Windows 7 x64 Tested)
*Should work on any CPU that supports SSE4 or any that support AVX2 on the other build.

Download Binaries:
v5.0.0.6 = http://www.multiupload.nl/SD35TVN47S
v5.0.0.1 = http://www.multiupload.nl/T4Q34B6BNR

Souce Code:
v5.0.0.6 = http://www.multiupload.nl/H987PPNYAK
v5.0.0.1 = http://www.multiupload.nl/RNB3XLRPIF

Anyway here is the info and i hope people like it

                                                  
                CPU Miner Advanced v5 - JackpotCoin Edition By Spoetnik

    Code base - Pooler's miner = https://bitcointalk.org/index.php?topic=55038.0
      Original release - Jackpot Team - 04/25/2014 - See Readme for more info.
             Original Github ID: 8e5ae495d90eeffa53d338c23ae2705f6eb69b89

Donations are appreciated i have spent lots of time coding and have little hardware
              If anyone can motivate me with some JPC coins thanks a lot
                  Jackpot address = JdPDpV9cLguK4vxgcfno8az4UWxXYrmcpv

Changelog: v5.0.2.0 - 05/20/2014 - Code Name: Rantology
- Added a Code Name identifier to the startup info to make versions more recognizable.
- Cheat Mode testing is done. Use cmd-line --cheat to try it out.
- Added Visual Studio Project files and changed source code to support this.
- Added New Window Title info including current hash rates (5 second avg)
- Moved Help info code into it's own source code file, use cmd-line --help
- Fixed Bug in console logging color code that cuased a race condition.
- Changed benchmark mode to show highest speed in Blue and normal summary in Green.
- Changed benchmark mode to show combined hash-rate from all threads in summary.
- Reworked scanhash() code, now cheat mode test uses it's own seperate function.
- Reworked and optimized / tweaked Jackpot algo hashing code. (default algo)
- Added expanded version info when running command-line --version.
- Updated and compiled all the libraries, such as LibCurl (general optimizations)
- Updated and expanded version.h header file code.
- Maybe more, sorry if i forgot to mention it

Changelog: v5.0.0.8 - 05/05/2014
- cheat mode test (from UncleBob's quark miner mod)

Changelog: v5.0.0.6 - 05/05/2014
- all documents are updated (found with source code)
- removed old LibCurl compilation error fix from autogen.sh and Makefile.am
- fixed misaligned help info console formatting.
* changed --retry-pause from 30 to 15 default.
* tweaked max64 algo modifier *again
* command-line -H, --hashdebug (enable hash debug output) is new on this mod.
- added used url message to console output (previously removed)
- ported my scantime setting fixs to cpuminer 2.3.3 updated code used.
- added algo name to console window title info.
- added .bat file to cleanup generated files on windows.
- removed 1 second delay between threads creation added previously.

Changelog: v5.0.0.5 - 05/04/2014
~ proper and full update of base code cpuminer 2.3.3
- The --url option is now mandatory
- Do not switch to Stratum when using an HTTP proxy
- Fix scheduling policy change on Linux (clbr)
- Fix CPU affinity on FreeBSD (ache)
- Compatibility fixes for various platforms like Solaris 8 and old OS X.
- A man page for minerd is now available

Changelog: v5.0.0.1 - 05/03/2014
- default algo is Jackpot coin.
- added color, see notes.
- added share submission benchmark timer function (Windows only)
- made various minor spelling changes for console messages.
- changed share submission notification from (yay!!!)/(booooo) to (good)/(bad)
- added a REAL WORKING scantime setting (fixed MAJOR big) see notes.
- added benchmark function, see notes.
- added additional debug info.
- added program name and version info to console window title.
- added scantime config setting info to console window title.
- added console message stating why a share is rejected.

* = forgot to mention before

Notes:
- Limited testing has been done. Please report any bugs.

- Use --help command line for more info.

- Algos supported are Jackpot, Quark, Scrypt and SHA256.

- New Timer Mode feature (windows only)
  This times in seconds / minutes how long it takes to reach x shares submitted.
  Only Good shares are counted.
  Use --timer 10 for example to time 10 submitted shares (or blocks found solo mining)
  If x shares is reached a message colored Purple will show the time elapsed.
  When the amount of shares is reached the timer is turned off and does nothing.
 
- New Benchmark Mode feature.
  Use Command-Line --benchmark to run the hashing test.
  This will test out your hashing speed and report back a summary on intervals.
  Benchmarking is done in offline mode no connection to pools or daemons needed.
  Example: minerd -t 6 --benchmark
 
- Color coding is as follows.
  Red = Error
  Green = Normal Share.
  Cyan = Fastest submitted Share.
  Purple = Special messages.
  Blue = New Block info.
 
- Default color mode is Bright.
  Use --color normal for normal colors
  Use --color bright for bright colors
  Use --color none for no colors (text will be grey)
  Or use -w instead of --color
  The Command-Line is only needed to set dull colors or none.
 
- Scantime command line bug fix / mod notes.
  The option has never worked on ANY cpu miner and i fixed it.
  The only cicumstances it would ever have any effect is when solo mining.
  I fixed this last year and checked numerous github repo's
  and every single cpu miner mod and cudaminer mod suffered from the same bug.
  what this means is that if you set --scantime on a Pool it does nothing at all.
  Teh code clearly shows how teh scan time command line option is only used
  when Long Polling or Stratum is NOT found and it then has ALWAYS defaulted to 60
  60 is a hard coded value on every cpu miner and cudaminer version i have ever seen.
  #define LP_SCANTIME   60 is written in cpuminer.c in all repo's
 
  So it's working now have fun and experiment with it

[Spoetnik]

[Spoetnik]

added download link to first post and source code is coming quick after i pack it up with WinRar..

i already noticed a bug lol
when i ran the --help command-line the new info i put in was not aligned correctly in the console so sorry about that people.

edit:
source code uploaded now.. it's exactly as it was compiled.

i want to do some more but i'll be taking a bit of a break for now i did all this pretty fast ..i need a smoke afterwards lol
was it as good for you as it was for me ?

[jpcsupport]

added download link to first post and source code is coming quick after i pack it up with WinRar..

i already noticed a bug lol
when i ran the --help command-line the new info i put in was not aligned correctly in the console so sorry about that people.

edit:
source code uploaded now.. it's exactly as it was compiled.

i want to do some more but i'll be taking a bit of a break for now i did all this pretty fast ..i need a smoke afterwards lol
was it as good for you as it was for me ?

It is really great idea and improvement for everybody who need a CPUMiner Back born.
And, thank for sharing your great work for the other ~

[Spoetnik]

thanks and gl with the coin and i will do more work soon for sure on the miner

[Spoetnik]

I fixed up the help formatting and here it is for reference..
Also changed the default IP/URL port for JPC solo mining.
Although as i just previewed this comment i noticed i changed the default retry time from 30 to 15 seconds and didn't update the help info lol
which would be the command -r or --retry-pause

Note: if you run help on Linux the new timer function is not seen or usable at all, it is not compiled in with Linux.
and you will also see 2 more command-lines for Linux only when running help on Nix..

  -S, --syslog          Use system log for output messages
  -B, --background   Run the miner in the background

[hero18688]

I can not run this under win8.1 x64.It stuck from beginning.  e3 1230v3 with haswell kernel
http://imgur.com/NdajDaw

But it can run under win7 x64. i7 4770non-k with haswell kernel

[Spoetnik]

your terminology is not used correctly eh..
the kernel used is called Jackpot coin.
all the build is that i uploaded is a compiler optimized version tailored specifically for the 4770k
so that means ANY cpu that supports the same instruction set as the 4770k should work and all others it will just crash when starting it up.

so this is the cpu it does not work on then right ?
http://ark.intel.com/products/75054/intel-xeon-processor-e3-1230-v3-8m-cache-3_30-ghz

i don't know why it would not work for you if that is your cpu because both have the same features / instruction set support..

you can see both our cpu features here..

mine
http://www.cpu-world.com/sspec/SR/SR147.html

yours
http://www.cpu-world.com/sspec/SR/SR153.html

both support these on that page specs..

AES
AVX
AVX2
MMX
SSE
SSE2
SSE3
SSE4
SSSE3

what that means is the compiler will use the highest and best options.
and since both of ours are the same spec wise i don't know..

You say you can't get it running on windows 8 ? did it work on Win 7 or something ?

anyway i should have more builds to upload shortly i finally got 32bit compilations figured out.
unfortunately this is the problem with using an optimized version.. it will run faster usually but you have to know what you got and what you need.
if you don't know you in for a headache lol

[Spoetnik]

well i screwed stuff up making the 32bit version which is slow as hell..
i have to re-install MINGW etc

even though i can see it is configured for x64 now it still is making 32bit .exe's grrrrrrr  

pretty sure i hosed my install trying to flip back and forth cross compiling and copying folders around etc

the only advice i have gotten from anyone on this stuff (even after asking millions of times) was from two tutorials i found
i got last year that only mentioned x64 and i had used that as s tarting point..
and recently i found a revised version of it on Google (no idea who made them)

since it's a bitch to get help with this stuff i will re-post the tutorials here so that some other guy holding clumps of hair can read them lol
i will post the newer one only because it basically repeats the older one so..
hopefully it will help some people.

SOURCE: http://pastebin.com/p2cRgHxd

Detailed Windows build instructions, using MinGW (64-bit):
   Install MinGW and the MSYS Developer Tool Kit (http://www.mingw.org/)
      * Choose C, C++ and MSys on install and select to have it update its libraries
      * Install into C:\MinGW
      * Add C:\MinGW\bin and c:\MinGW\msys\1.0 to your path
   Download MinGW64 from http://sourceforge.net/projects/mingw-w64/files/Toolchains%20targetting%20Win64/Automated%20Builds/
      * Choose mingw-w64-bin_i686-mingw_20111220.zip
      * Extract ZIP to C:\MinGW64
      * Add C:\MinGW64\bin to your path before C:\MinGW\bin
   Install libcurl devel (http://curl.haxx.se/download.html)
      * download curl-7.30.0.tar.gz from http://curl.haxx.se/download.html and put it in C:\deps\
      * launch an MSYS shell and enter the following commands (the configure step will take a long time!)
         cd /c/deps
         tar -xvzf curl-7.30.0.tar.gz
         cd curl-7.30.0
         ./configure --host=x86_64-w64-mingw32 –-prefix=/c/mingw64
         make
         make install
         cp /c/deps/curl-7.30.0/docs/libcurl/libcurl.m4 /c/mingw/share/aclocal/libcurl.m4
   Install pthreads
      * download pthreads-20100604.zip from http://sourceforge.net/projects/mingw-w64/files/External%20binary%20packages%20%28Win64%20hosted%29/pthreads/ and put it in C:\deps\
      * unzip the file to c:\deps\
      * In the mingw64 subdirectory is pthreads-w64.zip - extract the contents to C:\MinGW64
   Configure mingw path
      * create C:\MinGW\msys\1.0\etc\fstab with following configuration:
        c:/mingw      /mingw
   In the MSYS shell, navigate to the CPUminer source code direcctory
      * You will likely get higher hashrates by forcing the compiler to build the executable for your
        specific CPU architecture.  This is done by adding "-march=<value>" into the CFLAGS.  Those
        values can be found at http://gcc.gnu.org/onlinedocs/gcc/i386-and-x86-64-Options.html#i386-and-x86-64-Options
           common choices for intel are : core2, corei7, corei7-avx
           common choices for AMD are : athlon-fx
                       or -msse2,-msse3,-msse4,-mavx ...
      * Execute the following (replacing the value of -march with the value for your CPU type)
         ./autogen.sh
         ./configure --host=x86_64-w64-mingw32 CFLAGS="-O3 -march=core2"
         make
   Strip minerd.exe
      * Execute the following:
         x86_64-w64-mingw32-strip minerd.exe
   Combine the executable with the dependencies
      * copy minerd.exe, C:\MinGW64\bin\libcurl-4.dll, and C:\MinGW64\bin\pthreadGC2-w64.dll to the same directory

For Win32 target hosts
   Prepare compilations
      * place mstcpip.h and _mingw_unicode.h in MinGW\include (they already exist in the MinGW installation tree)
   Configure Curl compiling with:
      * ./configure --host=i686-pc-mingw32 CPPFLAGS="-DWIN32 -DWIN32_LEAN_AND_MEAN" –-prefix=/c/mingw
   Clean previous compilations and execute same steps as Win64 targets
      * make clean
        make
        make install
        if not already done, as the file never changes: cp /c/deps/curl-7.30.0/docs/libcurl/libcurl.m4 /c/mingw/share/aclocal/libcurl.m4
   Add pthreads win32 package:
      * through the MinGW Installation manager
   Configure CPUMiner compiling with:
      * ./autogen.sh
        ./configure --host=i686-pc-mingw32 CPPFLAGS="-DWIN32 -D_WIN32 -DWIN32_LEAN_AND_MEAN" CFLAGS="-O2 -march=core2"
        (Note: only O2 optimization to avoid crash. core2 architecture ideal for Xeon processor)
   Clean previous compilations and Compile CPUMiner
      * make clean
        make
   Strip minerd.exe
      * mingw32-strip minerd.exe
   Combine the executable with the dependencies
      * copy minerd.exe, C:\MinGW\bin\libcurl-4.dll, C:\MinGW\bin\pthreadGC2.dll and C:\MinGW\bin\libgcc_s_dw2-1.dll to the same directory

[Spoetnik]

added new version built for Haswell cpu's again..

edit:
changed a preview pic with the new version showing some minor changes. (first one shown)

[ig0tik3d]

strange, i dont see in source code any optimizations or changes from original cpu miner))

[SolidStateSurvivor]

Any of these Spoetnink miners that works well on an AMD Phenom 9850 BE?

[Spoetnik]

strange, i dont see in source code any optimizations or changes from original cpu miner))

thanks for your help and your snotty attitude..

strange i don't see any of your bullshit you copied and pasted from other devs
show me ONE lien of code your wrote sweet heart lol

[Spoetnik]

Any of these Spoetnink miners that works well on an AMD Phenom 9850 BE?

uhhhmmm ya only reason i didn't post one yet is i am not sure how fast i can get it..
i got my 9950 going from 80 to 240kh/s just by compiler optimizations
i keep forgetting about it i have a bunch of pm;s from another guy about other cpu's so.. been super busy lol
buuut i will make it next and upload a version that will run on my 9950 and that i assume should work good for you too

lastly, i uploaded an Intel 4770k Haswell class version again and the source code is inside the rar archive too
i ported a so called cheat mode taken last year from UncleBob's quark cpu miner mod..

http://www.multiupload.nl/4GVXU06HVT

what that does is basically it will drop working on a hash if Groestl is going to get run twice.
doing that makes hashing 4x faster but i also added UncleBobs's divider on hashing times, so it will show more like 2x faster not 4x faster lol
how effective this is i am not sure i tried experimenting with it lot son Quark ages ago and i was not share it actually made much difference earnings wise..
i got the idea from chubner saying something about this and i recalled the same thing already done last year.. by me
So i turned my 500kh/s cpu into 2.1mh/s but i cut the speed it reports in half so it shows about 1.1mh/s in console.
I also did some work on some of the other hashing code which did speed it up a bit.. init contexts out f loop for example.

i wasn't so busy spending 75% of my time fucking around with the MingW compiler flags etc i could get some coding done
i will try and add a couple more builds i have been requested done really soon ..an yeah i said i that before but i mean it this time, honest lol

[Spoetnik]

Any of these Spoetnink miners that works well on an AMD Phenom 9850 BE?

yes see first post

they are very different sizes because i ran STRIP the MingW compile command on one and not the other..


Here is the MingW commands i used if anyone wants to know..

This one for my Intel 4770k

cd /c/code/jpc-cpuminer-master-v2
./autogen.sh
./configure --host=x86_64-w64-mingw32 CFLAGS="-O3 -mavx2"
make
x86_64-w64-mingw32-strip minerd.exe
cp /c/Code/jpc-cpuminer-master-v2/minerd.exe "/c/Miners/Jackpotcoin CPUMiner"

And then this for my AMD x4 9950. The strip command would make my AMD cpu crash but not the Intel machine which i used to compile on..
so i omitted it on this build and if anyone gets a prompt about libgcc_s_dw2-1.dll let me know it's a MingW run time dll.

cd /c/code/jpc-cpuminer-master-v2
./autogen.sh
./configure --host=x86_64-w64-mingw32 CFLAGS="-O3 -msse4"
make
cp /c/Code/jpc-cpuminer-master-v2/minerd.exe "/c/Miners/Jackpotcoin CPUMiner"

[Spoetnik]

i don't think the cheat mode does any good.. long story short it will report your speed on pool web sites slower.
EVEN of it says your hashing faster in the miner. (Because you ARE hashing faster by using shortcuts)

i tried a bunch of variations and they all give the same result.

a more interesting part is no one anywhere has figured out what it didn't work on quark waaaay back last year (the same trick) (but i did then)
people who tried it found that it submitted like roughly 40% shares invalid to the pools etc..
why ? the answer is real simple.. the code everyone uses on all cpu miner mods is yet *Again broken in the post critical parts of all LOL
i seem to keep finding bugs that a million github repo's and other mods like cudaminer etc have..

if anyone writes a shortcut to Groestl what is going to happen is close to half of all shares submitted will say "error over target" in stratum responses.
see if you can spot why it's FIXED below
oh and ya ll might wanna update those millions of github repo's lol

static void jackpothash(void *state, const void *input) {

   uint32_t hash[16];
   jackpothash_context_holder ctx;

   memcpy(&ctx, &base_contexts, sizeof(base_contexts));

   sph_keccak512 (&ctx.ctx_keccak, input, 80);
   sph_keccak512_close(&ctx.ctx_keccak, hash);


#ifdef CHEAT
   int z=0;
#endif
   for (int r = 0; r < 3; r++) {
      if (hash[0]&0x1) {
         //if (r==2) return;
#ifdef CHEAT
         /* give up and try another nonce */
         if (z==1) return;
#endif
         sph_groestl512 (&ctx.ctx_groestl, hash, 64);
         sph_groestl512_close(&ctx.ctx_groestl, hash);
#ifdef CHEAT
         z=1;
#endif
      }
      else {
         sph_skein512 (&ctx.ctx_skein, hash, 64);
         sph_skein512_close(&ctx.ctx_skein, hash);
      }

      if (hash[0]&0x1) {
         sph_blake512 (&ctx.ctx_blake, hash, 64);
         sph_blake512_close(&ctx.ctx_blake, hash);
      }
      else {
         sph_jh512 (&ctx.ctx_jh, hash, 64);
         sph_jh512_close(&ctx.ctx_jh, hash);
      }
   }
   memcpy(state, hash, 32);
}

inline bool test(const uint32_t *hash, const uint32_t *p)
{
   if ( ((hash[7]<=p[7]) && (hash[7]!=0x0)) ){
      return true;
   }
   return false;
}

int scanhash_jackpot(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
   uint32_t max_nonce, unsigned long *hashes_done)
{
   uint32_t n = pdata[19] - 1;
   const uint32_t first_nonce = pdata[19];
   const uint32_t Htarg = ptarget[7];

   uint32_t hash64[8] __attribute__((aligned(32)));
   uint32_t endiandata[32];

   int kk=0;
   for (; kk < 32; kk++){be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]);};

   if (ptarget[7]==0) {
      do {
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (((hash64[7]&0xFFFFFFFF)==0) && (test(hash64, ptarget)==true)) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   
   }
   else if (ptarget[7]<=0xF)
   {
      do {
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (((hash64[7]&0xFFFFFFF0)==0) && (test(hash64, ptarget)==true)) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   
   }
   else if (ptarget[7]<=0xFF)
   {
      do {
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (((hash64[7]&0xFFFFFF00)==0) && (test(hash64, ptarget)==true)) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   
   }
   else if (ptarget[7]<=0xFFF)
   {
      do {//jackpot
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (((hash64[7]&0xFFFFF000)==0) && (test(hash64, ptarget)==true)) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   

   }
   else if (ptarget[7]<=0xFFFF)
   {
      do {
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (((hash64[7]&0xFFFF0000)==0) && (test(hash64, ptarget)==true)) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   

   }
   else
   {
      do {
         pdata[19] = ++n;
         be32enc(&endiandata[19], n);
         jackpothash(hash64, &endiandata);
         if (test(hash64, ptarget)==true) {
            *hashes_done = n - first_nonce + 1;
            return true;
         }
      } while (n < max_nonce && !work_restart[thr_id].restart);   
   }


   *hashes_done = n - first_nonce + 1;
   pdata[19] = n;
   return 0;
}

[SolidStateSurvivor]

i don't think the cheat mode does any good.. long story short it will report your speed on pool web sites slower.
EVEN of it says your hashing faster in the miner. (Because you ARE hashing faster by using shortcuts)

i tried a bunch of variations and they all give the same result.

a more interesting part is no one anywhere has figured out what it didn't work on quark waaaay back last year (the same trick) (but i did then)
people who tried it found that it submitted like roughly 40% shares invalid to the pools etc..
why ? the answer is real simple.. the code everyone uses on all cpu miner mods is yet *Again broken in the post critical parts of all LOL
i seem to keep finding bugs that a million github repo's and other mods like cudaminer etc have..

if anyone writes a shortcut to Groestl what is going to happen is close to half of all shares submitted will say "error over target" in stratum responses.
see if you can spot why it's FIXED below
oh and ya ll might wanna update those millions of github repo's lol

So with other words, don't use the CHEAT versions?

[SolidStateSurvivor]

Of the 4 blocks I have found on the pool(s), 2 of them have been found with the Phenom 9850 BE, with your minerd_SSE4 miner. The 7950 1 and none by 7970 or the other CPUs/APUs.

[goodluck0319]

I got i7 4770k which miner I shall use? avx or see4?

[Spoetnik]

I got i7 4770k which miner I shall use? avx or see4?

the AVX one.. i compiled it to use MingW GCC AVX2 instruction set optimizations (as stated earlier)
AVX+ is great and big improvement over SSE versions..

check this out..
Intel’s Haswell is an unprecedented threat to Nvidia, AMD

So that says some of us can do AVX2 we can go a bit faster hashing

i have to stress i used AVX2 not AVX (version 1)
so only Haswell class chips and higher will support that build/.exe

and to the other guy i have no clue why SSE4 would be faster than AVX2 it doesn't make any sense..
i am guessing the one did not run for you and the other did ?
i run a AMD x4 9950 and Intel 4770k for testing my other cpu's machines are dead (until recently i had a Core 2 Duo for testing with)
Intel avg = 487
AMD avg = 471
per thread i am seeing about maybe 3 or 4 kh/s on my AVX2 build vs the SEE4 build.. which above shows on my 6 threads about 17kh/s more total.
Not a lot but impressive by not changing any code at all and just a compiler flag when building it lol
To do that test i ran the --benchmark command-line with the threads setting and watched it for a while..

Last night i rewrote the cpu hashing code a lot to see if i can get some better speeds and some versions i tried i was getting as high as 768kh/s reported on the pool
Some using various shortcuts and some not.. normal non cheating builds i am making now are about the same speed
and the so called cheat test versions i am not sure it's REALLY hard to tell because i have to rely on what the pools report for numbers..

Side Note:
I just spent more time trying to see if i can get ccminer for nvidia cards to run on my Fermi compute 2.0 card but i am not sure what i am doing i am no cuda guy
i can get it built and running but i am getting a problem with the hashing.. i tried to do an alt to validating the branches and much more.. so far i'm failing lol