Topic: [ANN]: cpuminer-opt v3.8.8.1, open source optimized multi-algo CPU miner - page 166. (Read 444040 times)

Quote from: Wolf0 on May 27, 2016, 06:14:05 PM

Your HOdl is broke - on linux, and probably on every platform. You put a WHOLE lot of wasteful C++ garbage in, butchering my HOdl code - also breaking the compile.

What version? That wastefull c++ code is to support non AES_NI.

I'm currently using 3.2.3 and I made no changes to hodl in 3.2.4 or 3.2.5.

Just compiled and ran 3.2.5, no porblem.

I tried the version linked in the OP - and you know, you can do a non-AES-NI version really simply with a lookup table... or a few, for speed.

I'l download it and give it a try.

Edit: downloaded, compiled and runs ok on hodl.

GCC/G++ version 6.1.1?

gcc (Ubuntu 4.8.4-2ubuntu1~14.04.1) 4.8.4

Oh lord that's old. Probably why.

The new compoiler should be backward compatible, or have a compatibility switch. If not it's pretty inconsiderate
of them to break existing code.

There are a lot of implicit function declarations in my code, did you set the option to allow them? I'm still
undecided on whether to leave them ro make it easier to add new algos, or require every algo's registration
function be defined in algo-gate-api.h.

It actually dies in C++ headers somewhere. I was going to fix it, but then I looked at what was causing it and realized that all that inefficient C++ wasn't actually needed at all, so pointed it out.

So the problem is not in your code but in the non AES_NI version. Is the error easy to fix, I'll do it if you send me the info.

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: joblo on May 27, 2016, 06:11:08 PM

Quote from: Wolf0 on May 27, 2016, 06:08:48 PM

Quote from: joblo on May 27, 2016, 05:58:59 PM

Quote from: Wolf0 on May 27, 2016, 06:08:48 PM

Your HOdl is broke - on linux, and probably on every platform. You put a WHOLE lot of wasteful C++ garbage in, butchering my HOdl code - also breaking the compile.

What version? That wastefull c++ code is to support non AES_NI.

I'm currently using 3.2.3 and I made no changes to hodl in 3.2.4 or 3.2.5.

Just compiled and ran 3.2.5, no porblem.

I tried the version linked in the OP - and you know, you can do a non-AES-NI version really simply with a lookup table... or a few, for speed.

I'l download it and give it a try.

Edit: downloaded, compiled and runs ok on hodl.

GCC/G++ version 6.1.1?

gcc (Ubuntu 4.8.4-2ubuntu1~14.04.1) 4.8.4

Oh lord that's old. Probably why.

The new compoiler should be backward compatible, or have a compatibility switch. If not it's pretty inconsiderate
of them to break existing code.

There are a lot of implicit function declarations in my code, did you set the option to allow them? I'm still
undecided on whether to leave them ro make it easier to add new algos, or require every algo's registration
function be defined in algo-gate-api.h.

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: joblo on May 27, 2016, 05:58:59 PM

Your HOdl is broke - on linux, and probably on every platform. You put a WHOLE lot of wasteful C++ garbage in, butchering my HOdl code - also breaking the compile.

What version? That wastefull c++ code is to support non AES_NI.

I'm currently using 3.2.3 and I made no changes to hodl in 3.2.4 or 3.2.5.

Just compiled and ran 3.2.5, no porblem.

I tried the version linked in the OP - and you know, you can do a non-AES-NI version really simply with a lookup table... or a few, for speed.

I'l download it and give it a try.

Edit: downloaded, compiled and runs ok on hodl.

GCC/G++ version 6.1.1?

gcc (Ubuntu 4.8.4-2ubuntu1~14.04.1) 4.8.4

joblo

legendary

Activity: 1470

Merit: 1114

Your HOdl is broke - on linux, and probably on every platform. You put a WHOLE lot of wasteful C++ garbage in, butchering my HOdl code - also breaking the compile.

What version? That wastefull c++ code is to support non AES_NI.

I'm currently using 3.2.3 and I made no changes to hodl in 3.2.4 or 3.2.5.

Just compiled and ran 3.2.5, no porblem.

I tried the version linked in the OP - and you know, you can do a non-AES-NI version really simply with a lookup table... or a few, for speed.

I'l download it and give it a try.

Edit: downloaded, compiled and runs ok on hodl.

joblo

legendary

Activity: 1470

Merit: 1114

Your HOdl is broke - on linux, and probably on every platform. You put a WHOLE lot of wasteful C++ garbage in, butchering my HOdl code - also breaking the compile.

What version? That wastefull c++ code is to support non AES_NI.

I'm currently using 3.2.3 and I made no changes to hodl in 3.2.4 or 3.2.5.

Just compiled and ran 3.2.5, no porblem.

arransiv

sr. member

Activity: 420

Merit: 250

I will be mining with this miner soon. I will try different algos and different coins and post my feedback here. I can't throw big hash.

Wolf0

member

Activity: 81

Merit: 1002

It was only the wind.

Quote from: joblo on May 27, 2016, 01:12:22 AM

Why not publish the compiled version?

If you mean Windows it's because it won't compile in Windows. If you mean Linux you get
better performance compiling yourself.

What algo are you working on now?

ctgiant

sr. member

Activity: 298

Merit: 250

Quote from: ctgiant on May 27, 2016, 01:04:32 AM

Quote from: joblo on May 26, 2016, 09:37:37 PM

Quote from: ctgiant on May 27, 2016, 01:04:32 AM

Why not publish the compiled version?

If you mean Windows it's because it won't compile in Windows. If you mean Linux you get
better performance compiling yourself.

What algo are you working on now?

None at the moment, hmq1725 was last adddition and lyra2v2 the last optimization.

I got your miner working on windows..had to remove the hodl algo and make some code changes, but it is working on my sandybridge cpu...I've only tested the hmq1725 algo so I am not sure if I broke anything while getting it to compile, but it is working well. In order to get it to work for both windows and linux I would need clean up my changes a bit. I also couldn't get it compiled statically, so it needs to be run with the dll's in tow.

I just thought you might be interested to know that it can compile on windows. If you want to discuss let me know and I'll get back to you.

That's great news, yes please discuss, I have many questions.

If HMW1725 works then many other algos such as x11, quark etc will work.
Hodl is a special case and so are a few other unique algos but I think I'll be able to work it out.
Can't wait to see the source.

I only made small changes other than not using hodl. I set up a github so that you could see what the changes were. Here is the commit to my repo with the changes detailed: https://github.com/ctgiant/cpuminer-opt/commit/1aa77613bc835d0693eef29084d888fa920002f0

I should also note, just so you know, it was built with mingw-w64-x86_64.

mingw, even better, no messing around with vstudio. Thanks.

No problem...if anyone wants to test it I posted the executable and DLL's here: https://github.com/ctgiant/cpuminer-opt/releases/download/v3.2.5/cpuminer-opt.zip

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: joblo on May 26, 2016, 09:37:37 PM

Quote from: joblo on May 26, 2016, 09:37:37 PM

Why not publish the compiled version?

If you mean Windows it's because it won't compile in Windows. If you mean Linux you get
better performance compiling yourself.

What algo are you working on now?

None at the moment, hmq1725 was last adddition and lyra2v2 the last optimization.

I got your miner working on windows..had to remove the hodl algo and make some code changes, but it is working on my sandybridge cpu...I've only tested the hmq1725 algo so I am not sure if I broke anything while getting it to compile, but it is working well. In order to get it to work for both windows and linux I would need clean up my changes a bit. I also couldn't get it compiled statically, so it needs to be run with the dll's in tow.

I just thought you might be interested to know that it can compile on windows. If you want to discuss let me know and I'll get back to you.

That's great news, yes please discuss, I have many questions.

If HMW1725 works then many other algos such as x11, quark etc will work.
Hodl is a special case and so are a few other unique algos but I think I'll be able to work it out.
Can't wait to see the source.

I only made small changes other than not using hodl. I set up a github so that you could see what the changes were. Here is the commit to my repo with the changes detailed: https://github.com/ctgiant/cpuminer-opt/commit/1aa77613bc835d0693eef29084d888fa920002f0

I should also note, just so you know, it was built with mingw-w64-x86_64.

mingw, even better, no messing around with vstudio. Thanks.

ctgiant

sr. member

Activity: 298

Merit: 250

Quote from: Epsylon3 on May 26, 2016, 11:28:12 PM

Why not publish the compiled version?

If you mean Windows it's because it won't compile in Windows. If you mean Linux you get
better performance compiling yourself.

What algo are you working on now?

None at the moment, hmq1725 was last adddition and lyra2v2 the last optimization.

I got your miner working on windows..had to remove the hodl algo and make some code changes, but it is working on my sandybridge cpu...I've only tested the hmq1725 algo so I am not sure if I broke anything while getting it to compile, but it is working well. In order to get it to work for both windows and linux I would need clean up my changes a bit. I also couldn't get it compiled statically, so it needs to be run with the dll's in tow.

I just thought you might be interested to know that it can compile on windows. If you want to discuss let me know and I'll get back to you.

That's great news, yes please discuss, I have many questions.

If HMW1725 works then many other algos such as x11, quark etc will work.
Hodl is a special case and so are a few other unique algos but I think I'll be able to work it out.
Can't wait to see the source.

I only made small changes other than not using hodl. I set up a github so that you could see what the changes were. Here is the commit to my repo with the changes detailed: https://github.com/ctgiant/cpuminer-opt/commit/1aa77613bc835d0693eef29084d888fa920002f0

I should also note, just so you know, it was built with mingw-w64-x86_64.

Epsylon3

legendary

Activity: 1484

Merit: 1082

ccminer/cpuminer developer

the important is the new stratum methods (in util.c) only one is required for the bench (get_stats) and then sysutils.c code

joblo

legendary

Activity: 1470

Merit: 1114

nice initiative the git hmage... and surprising it build directly Wink

i m very busy and missed several versions...

it could be nice to import my last cpuminer-multi changes to do some "public" benchs on yiimp

i found also a nice way to get the cpu name from the cpuid instruction... should work on both linux and windows

i tried to find a way to get the internal temperature like that on windows, but no way... seems to require driver (ring0) instructions

So It's the hmage fork that's working? I'll have to give it a spin.

It's probably good that you missed several versions, it was a little messy for a while. It's stabilized quite a
bit so now is a good time.

I see you've made some recent changes to the windows branch. Is that what I should use? Are the changes
mostly large blocks of code or is it line by line? In other words how hard will it be to manually
merge?

I already have cpuid working thanks to a contributor, don't know about windows but I don't see why not.

Epsylon3

legendary

Activity: 1484

Merit: 1082

ccminer/cpuminer developer

nice initiative the git hmage... and surprising it build directly Wink

i m very busy and missed several versions...

it could be nice to import my last cpuminer-multi changes to do some "public" benchs on yiimp

i found also a nice way to get the cpu name from the cpuid instruction... should work on both linux and windows

i tried to find a way to get the internal temperature like that on windows, but no way... seems to require driver (ring0) instructions

joblo

legendary

Activity: 1470

Merit: 1114

Why not publish the compiled version?

If you mean Windows it's because it won't compile in Windows. If you mean Linux you get
better performance compiling yourself.

What algo are you working on now?

None at the moment, hmq1725 was last adddition and lyra2v2 the last optimization.

I got your miner working on windows..had to remove the hodl algo and make some code changes, but it is working on my sandybridge cpu...I've only tested the hmq1725 algo so I am not sure if I broke anything while getting it to compile, but it is working well. In order to get it to work for both windows and linux I would need clean up my changes a bit. I also couldn't get it compiled statically, so it needs to be run with the dll's in tow.

I just thought you might be interested to know that it can compile on windows. If you want to discuss let me know and I'll get back to you.

That's great news, yes please discuss, I have many questions.

If HMW1725 works then many other algos such as x11, quark etc will work.
Hodl is a special case and so are a few other unique algos but I think I'll be able to work it out.
Can't wait to see the source.

ctgiant

sr. member

Activity: 298

Merit: 250