[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[bigjme]

Yeh i use to get some like that, just cancel it and start again

[djm34]

Yeh i use to get some like that, just cancel it and start again

still compiling... (I did restart though...)  the compiler really doesn't like what I put in whirlpool   (obviously it is where it is stuck... and I will have to recompile as it will just run over one thread...)

[bigjme]

Hmm, there was in issue like that in ccminer for 64 bit.
Which version are you compiling?

[djm34]

Hmm, there was in issue like that in ccminer for 64 bit.
Which version are you compiling?

right now, I am compiling in 32bit (the pre X13 release, without killer groestl... for the wrong reason I must precise)

[bigjme]

Strange. If you watch the compile directory size does it keep increasing?

[cbuchner1]

lol, I have entered compilation of doom.... for some reason cicc.exe is compiling since half an hour and use almost 2Gb of ram

too much loop unrolling causes that, also for some algorithms going below 80 registers per thread also causes these issues.

[djm34]

the compiler doesn't like MAKE_ULONGLONG,

[bigjme]

Do the intergers need to be that long?

[djm34]

Do the intergers need to be that long?

whirlpool uses tables of uint64_t, the shared memory doesn't seem to take more than uint32_t...
(and I don't know enough to modify a lot the algo)

[bigjme]

Is there a way to split it into 2 and merge it later?

[djm34]

Is there a way to split it into 2 and merge it later?

I did that yesterday, ended up with new uint32_t tables of 512 numbers  (instead of 256 uint64_t)... it divided the computing time the hashrate by two...
right know, the shared memory seems to split them automatically, just needs to merge them.

[bigjme]

like this

Code:

uint32 a = 0xff00ff00;
uint32 b = 0x00ff00ff;

ulong64 c = ((ulong64)a) << 32 | b;

[djm34]

still compiling... each mistake cost a lot at the moment... (it wasn't that bad yesterday...)

[bigjme]

A compile over 10 minutes is a bad one i think

[djm34]

it works   (good luck compiling though...)
(in my first attemp the two uint32 wasn't in the right order...)

gpu usage stable around 78 (not good but better than yesterday)
speed: 2.9Mhash/s (2 instance gives between 3-3.2MHash/s) with the killer groestl version.

I will release as is (meaning it runs with -a x13... adding new option, is clearly the part I hate... and I don't know well ccminer interface)

(less than 5minutes of change over 3 hours of compilation   )

edit: I have no idea on how to make it work on linux...

[Amph]

it works   (good luck compiling though...)
(in my first attemp the two uint32 wasn't in the right order...)

gpu usage stable around 78 (not good but better than yesterday)
speed: 2.9Mhash/s (2 instance gives between 3-3.2MHash/s) with the killer groestl version.

I will release as is (meaning it runs with -a x13... adding new option, is clearly the part I hate... and I don't know well ccminer interface)

(less than 5minutes of change over 3 hours of compilation   )

you just added the two .c/.cu file and changed two line of code?

[djm34]

it works   (good luck compiling though...)
(in my first attemp the two uint32 wasn't in the right order...)

gpu usage stable around 78 (not good but better than yesterday)
speed: 2.9Mhash/s (2 instance gives between 3-3.2MHash/s) with the killer groestl version.

I will release as is (meaning it runs with -a x13... adding new option, is clearly the part I hate... and I don't know well ccminer interface)

(less than 5minutes of change over 3 hours of compilation   )

you just added the two .c/.cu file and changed two line of code?

If you can find the two cu anywhere let me know (I haven't used the piece of code I "traded", as it would have required a lot more work actually to insert it into ccminer)... you still have to write them (was two days ago... though...)
5min (over the last few changes...   once I have found how get something in and out of the shared memory)



as it may takes a little while before I get the source out, I put here the windows binaries (it still runs with -x13, though I will try to fix that).
It is compiled from the killer groestl version with only one executable:

https://mega.co.nz/#!IUsXGBoZ!fMJQK_BG2x6GAyhX_oaD_cmurLD7-4oB4B2Cys5Qh-o

[Neo.op]

it works   (good luck compiling though...)
(in my first attemp the two uint32 wasn't in the right order...)

gpu usage stable around 78 (not good but better than yesterday)
speed: 2.9Mhash/s (2 instance gives between 3-3.2MHash/s) with the killer groestl version.

I will release as is (meaning it runs with -a x13... adding new option, is clearly the part I hate... and I don't know well ccminer interface)

(less than 5minutes of change over 3 hours of compilation   )

edit: I have no idea on how to make it work on linux...

where are the codes located? I'll give it a try on linux and see if it compiles through.

[djm34]

it works  (good luck compiling though...)
(in my first attemp the two uint32 wasn't in the right order...)

gpu usage stable around 78 (not good but better than yesterday)
speed: 2.9Mhash/s (2 instance gives between 3-3.2MHash/s) with the killer groestl version.

I will release as is (meaning it runs with -a x13... adding new option, is clearly the part I hate... and I don't know well ccminer interface)

(less than 5minutes of change over 3 hours of compilation  )

edit: I have no idea on how to make it work on linux...

where are the codes located? I'll give it a try on linux and see if it compiles through.

sorry not available yet

[bigjme]

djm do you want me to see how long it takes to compile on my windows system in case it was your pc doing it slow?
I wont release it as i have no interest in it but i am willing to check it for you