unfortunately, that not working.
./ptsminer.exe -u x -m 28 -a gpuv4amd ( I have 3+ GB gpu memory)
Starting OpenCLMomentum V4 AMD Optimized
Device 00: Tahiti
Max work group size: 256
ERROR: -61, CL_INVALID_BUFFER_SIZE, if size is 0.Implementations may return CL_INVALID_BUFFER_SIZE if size is greater than the CL_DEVICE_MAX_MEM_ALLOC_SIZE value specified in the table of allowed values for param_name for clGetDeviceInfo for all devices in context.
assertion "_MY_ERR_X == CL_SUCCESS" failed: file "OpenCLObjects.cpp", line 379, function: OpenCLBuffer* OpenCLContext::createBuffer(size_t, cl_mem_flags, void*)
Aborted (core dumped)

with ./ptsminer.exe -u x -m 27 -a gpuv4amd
show around 17k cpm

Warning: found more candidate collisions than storage space available
[WORKER] collision found: 17088 <-> 16407 #28 @ 1390279937 by 0
[WORKER] collision found: 36544 <-> 35863 #82 @ 1390279937 by 0
[WORKER] collision found: 56000 <-> 55319 #96 @ 1390279937 by 0
[WORKER] collision found: 57399 <-> 58352 #126 @ 1390279937 by 0
[WORKER] collision found: 91992 <-> 91175 #142 @ 1390279937 by 0
[WORKER] collision found: 123888 <-> 122935 #154 @ 1390279937 by 0
[WORKER] collision found: 99160 <-> 98343 #164 @ 1390279937 by 0
[WORKER] collision found: 79552 <-> 78871 #196 @ 1390279937 by 0
[MASTER] submitted share -> REJECTED
[STATS] 2014-Jan-21 11:52:16 | 17280.0 c/m | 480.0 sh/m | VL: 0 (0.0%), RJ: 10 (100.0%), ST: 0 (0.0%)
[MASTER] submitted share -> REJECTED
[STATS] 2014-Jan-21 11:52:16 | 17280.0 c/m | 480.0 sh/m | VL: 0 (0.0%), RJ: 11 (100.0%), ST: 0 (0.0%)
[MASTER] submitted share -> REJECTED
too many rejects (3) in a row, forcing reconnect.
[STATS] 2014-Jan-21 11:52:16 | 17280.0 c/m | 480.0 sh/m | VL: 0 (0.0%), RJ: 12 (100.0%), ST: 0 (0.0%)
no connection to the server, reconnecting in 10 seconds
Warning: found more candidate collisions than storage space available

yeah, that's the problem with not having an AMD GPU at home to test Probably some memory sync problem, but it will be impossible to debug without a GPU of my own the 17K cpm is misleading. all the hashes are probably 0 because the sha512 is not woking...

Tanks anyway. I will find a way to debug it...

THANK YOU for compiling this for Windows. I've been banging my head against the wall trying to get cygwin to like it and I'm sure I'd have succeeded eventually, but now I can forsake that sad battle

Also, I'm glad to finally see open source code enabling GPU mining for ProtoShares using all modern GPUs on all platforms.

[Edit: I'm sending you a PTS donation.]

Stats:
~400c/p on Radeon HD 78 (using 512 mem. Crashes with 1024 mem).
Crashes on Radeon HD 580, not sure why.

I think I'll break down and migrate my rig to linux, as I suspect I'll get much better speeds via that platform . . .

Questions: does this include the hidden options for different algorithms, and if so how are they invoked? Also, what ocl binary file is it wanting to read (when run on Windows), and would it load faster with that, and if so, where can I get said file?

(And if so, do you use your powers for good or for awesome?)

It does have the hidden options. "-a gpuvN" where N can be 2, 3 or 4.

But i suspect we are talking about different things here. There is no ocl binary or anything like that on my miner. You probably downloaded another miner thinking it was mine?

Quote from: girino on January 17, 2014, 11:54:25 am

john the ripper has sha512 code specific for AMD, but i was unable to test it since i have no AMD card. I will create a "gpuv6" that uses it instead of the generic version, and see if we get closer to them. Any other tips and tricks for tuning opencl code on AMD cards are welcome.

thanks for add gpuv6! I will test asap after adding.

Just got it ready. 80% improvement at my low end AMD at work (from 28 to 45 cpm). The options is -a gpuv4amd since it does not work on nvidia. Please test it and tell me if it improves performance for you.

Yeah speed is a problem... With 2x280x i am getting speed of 1x280x with 1gh.

To pass 1000 cpm i needed to run it with -m1024 (witm m2048 i got lower cpm)

oops, i had a bad optimization leal into production code. Just reverted it. Please re-download and retest it. Should be way faster now.

WINDOWS BINARIES NOW AVAILABLE. See first post for links.

I just compiled using mingw, but did not test. Please report any errors.

Thanks.

i'm try.... but...;

Quote

$ make
g++  -O3 -march=native -c -o jhlib.o jhlib.cpp
In file included from global.h:9:0,
                 from jhlib.cpp:1:
win.h:52:0: attention : « FIONBIO » redéfini [enabled by default]
 #define FIONBIO 0
 ^
In file included from /usr/include/cygwin/socket.h:51:0,
                 from /usr/include/cygwin/in.h:21,
                 from /usr/include/netinet/in.h:14,
                 from /usr/include/arpa/inet.h:14,
                 from win.h:12,
                 from global.h:9,
                 from jhlib.cpp:1:
/usr/include/asm/socket.h:27:0: note: ceci est la localisation d'une précédente définition
 #define FIONBIO 0x8004667e /* To be compatible with termiost version */
 ^
In file included from global.h:9:0,
                 from jhlib.cpp:1:
win.h:77:0: attention : « __declspec » redéfini [enabled by default]
 #define __declspec(x) __##x
 ^
jhlib.cpp:1:0: note: ceci est la localisation d'une précédente définition
 #include"global.h"
 ^
In file included from global.h:9:0,
                 from jhlib.cpp:1:
win.h:83:6: erreur: #elif with no expression
 #elif
      ^
jhlib.cpp: In function ‘uint32 streamEx_substream_writeData(void*, void*, uint32)’:
jhlib.cpp:599:15: erreur: ‘__debugbreak’ was not declared in this scope
  __debugbreak(); // no write access for substreams?
               ^
jhlib.cpp: In function ‘void streamEx_substream_setSize(void*, uint32)’:
jhlib.cpp:611:15: erreur: ‘__debugbreak’ was not declared in this scope
  __debugbreak(); // not implemented
               ^
Makefile:25: recipe for target 'jhlib.o' failed
make: *** [jhlib.o] Error 1

I'll get this working in a few hours. Just need to grab my windows box.

Quote from: girino on January 20, 2014, 05:56:33 am

I had no intention of releasing my miner for ypool in the begining (the 70% harming the network, etc, etc), but since too many people asked, i am now releasing it. As always, the code is Open sourced and free. As I used jh00 jhProtominer as a base, and the minar and the pool support donation fees, i included a 2.5% developer fee in the code, that will only apply for those mining on ypool.net. If you mine on any other pool, please remember to donate.

Code is available at: https://github.com/girino/jhProtominer

A linux 64bit compiled binary is available at: https://mega.co.nz/#F!oVBUmDjR!atcpQhUQ7C123F-XFiub0w

Hope you enjoy!

2.5% developer fee + 5% pool fee = 7,5% ? 

Yup. I have no intention of driving more people to ypool. But you can still use it on other pools with no fee. Call it an incentive...

I had no intention of releasing my miner for ypool in the begining (the 70% harming the network, etc, etc), but since too many people asked, i am now releasing it. As always, the code is Open sourced and free. As I used jh00 jhProtominer as a base, and the minar and the pool support donation fees, i included a 2.5% developer fee in the code, that will only apply for those mining on ypool.net. If you mine on any other pool, please remember to donate.

Code is available at: https://github.com/girino/jhProtominer

BINARY FILES ARE AVAILABLE HERE: https://www.dropbox.com/sh/y3zsy1xe6sibusl/KtCKHkxpwR
(windows, Linux and OSX available).

EDIT: Windows compile now available. (Not tested, i still don't have my windows machine, cross-compiled on linux using mingw64). Link: https://www.dropbox.com/sh/y3zsy1xe6sibusl/KtCKHkxpwR

Edit again: The opencl files included in the previous zip file were outdated. please redownload for way better performance.

Edit 3: New optimizations in the sha512 code. Now 10% faster.

Edit 4: Per request, i made a Mac OSX version available.

Edit 5: Further optimized and customized sha512 code. At least 15% faster on NVidia. Binaries for Windows, Linux and OSX available.

Edit 6: Corrected bug in opencl that prevented it to work on AMD GPUs. Now works on every platform. Binaries for Windows, Linux and OSX available.

Hope you enjoy!