I just checked in after a while... Is there a precompiled 64-bit linux binary with airsquirrels' changes (I searched the forum but didn't see one)? If not, I can build one myself, but I thought I'd ask first.

Thanks!

16.3.2 gave me a 40 point a day boost. From about 580 to ~620 on an R9 390X@1100/6000

This is an OpenCL question, but not an mfacto question. I have OpenCL code that gives me an error when I call clBuildProgram. Apparently the problem is with calls to "mul_hi". These calls work on my MacBook Pro, but not on my iMac. If I remove them, clBuildProgram doesn't give an error but my program doesn't return the right results. Does anyone here have any ideas? I've done a lot of searching, but haven't had any luck figuring out what I'm missing.

This could be anything, from bad hardware over outdated drivers to code errors. But usually, clBuildProgram returns some error code, that me help restricting the search space.

Furthermore, if the program does not just die but return some compile error, you can call clGetProgramBuildInfo to get the build log.

I would suggest you also use clGetProgramBuildInfo on the working machine, maybe it shows some warnings that help you further.

Hi all,

My GTX560Ti has recently gone south, and replacement won´t happen anytime soon, due to "budget constraints". So an old ATI Radeon 4350 I had taken out of service was called up.
I´m running Win10 64 bits, and the card was happily recognized, installed, and is running perfeclty.
Then I decided to give mfakto a go. The trouble is the program will just start and immediately exit with the following message:


Runtime options
Inifile mfakto.ini
Verbosity 1
SieveOnGPU yes
MoreClasses yes
GPUSievePrimes 82486
GPUSieveProcessSize 24Ki bits
GPUSieveSize 96Mi bits
FlushInterval 8
WorkFile worktodo.txt
ResultsFile results.txt
Checkpoints enabled
CheckpointDelay 300s
Stages enabled
StopAfterFactor class
PrintMode compact
V5UserID none
ComputerID none
TimeStampInResults yes
VectorSize 2
GPUType VLIW5
SmallExp no
UseBinfile mfakto_Kernels.elf
Compiletime options
[I]Select device - Error -1001 (Unknown errorcode (not an OpenCL error)): clGetPlatformIDs(num)
ERROR: init_CL(3, 0) failed[/I]]

I´m usiing mfakto 0.14p, but the same happened with 0.15pre6.

The driver is up to date.

I remember that some years ago I briefly used this very card for mfakto (can´t recall which version) under windows 7 (or 8 ?...) and it worked really well. It was slow, but hey.

Any ideas?

Thx a lot.

Hi,

what exactly does this mean:
[QUOTE=lycorn;433936]
The driver is up to date.
[/QUOTE]

Did you leave it to Windows and it said the driver was up to date, or did you install the latest AMD driver yourself?

The failing call ist the first OpenCL call mfakto is doing. My guess would be that some MS driver is feeding your GPU but does not support OpenCL. Try running clinfo to learn about your OpenCL-supporting devices. Should be in C:\Windows\System32\clinfo.exe if the driver installed OK. Otherwise you can try installing the AMD APP SDK, which would bring the tool in
C:\Program Files (x86)\AMD APP SDK\3.0\bin\x86_64\clinfo.exe
C:\Program Files (x86)\AMD APP SDK\3.0\bin\x86\clinfo.exe


My second guess would be that you really installed the latest AMD drivers, but they no longer support an HD 4xxx. You would need to go back a few generations of AMD drivers, and I'm not sure if there's overlap in support for your GPU and support for Win10.

Good luck.

R9 390 issues with 0.14p
 

Been playing around with the previously linked 0.14p Win7 64-bit version on my [URL="http://www.newegg.com/Product/Product.aspx?Item=N82E16814150729&cm_re=r9_390-_-14-150-729-_-Product"]XFX R9 390[/URL]. It continues to fail, including with the -st flag. I have updated AMD drivers, etc. Not sure what's wrong.

[CODE]
mfakto 0.14-MGW (64bit build)


Runtime options
Inifile mfakto.ini
Verbosity 1
SieveOnGPU yes
MoreClasses yes
GPUSievePrimes 82486
GPUSieveProcessSize 24Ki bits
GPUSieveSize 96Mi bits
FlushInterval 8
WorkFile worktodo.txt
ResultsFile results.txt
Checkpoints enabled
CheckpointDelay 300s
Stages enabled
StopAfterFactor class
PrintMode compact
V5UserID none
ComputerID none
TimeStampInResults yes
VectorSize 2
GPUType AUTO
SmallExp no
UseBinfile mfakto_Kernels.elf
Compiletime options
Select device - Get device info - Loading binary kernel file mfakto_Kernels.elf
Compiling kernels.

OpenCL device info
name Hawaii (Advanced Micro Devices, Inc.)
device (driver) version OpenCL 2.0 AMD-APP (2079.5) (2079.5 (VM))
maximum threads per block 256
maximum threads per grid 16777216
number of multiprocessors 40 (2560 compute elements)
clock rate 1015MHz

Automatic parameters
threads per grid 256
optimizing kernels for GCN

Started a simple selftest ...
ERROR: selftest failed for M50752613 (cl_barrett15_88_gs)
no factor found
ERROR: selftest failed for M50752613 (cl_barrett32_87_gs)
no factor found
ERROR: selftest failed for M50752613 (cl_barrett32_77_gs)
no factor found
Selftest statistics
number of tests 108
successful tests 105
no factor found 3

selftest FAILED![/CODE]

Followup question, does this automatically report the results to the GIMPS server? Or do I have to upload them myself, like I had to download the worktodo.txt?

There are a few different automation programs/scripts.

I created [URL="https://github.com/Kunde21/MersenneManager"]these programs[/URL] to solve my own automation problems.

Later I found [URL="https://github.com/teknohog/primetools"]these Python scripts[/URL] and [URL="http://www.mersenneforum.org/misfit/"]this management program[/URL] that have similar functionality.

You do it by yourself, or use a tool (like Misfit, for windoze, or some py script for *nix). They can organize all the work for you, taking assignments, reporting results, etc. Set it and forget it.

opportunistic GPUSieve
 

I've been running mfakto (and mfaktc) on a number of workstations, but when GPUSieve=1 the workstation graphics become heavily latented.


What other tasks than sieving does mfakto perform?


Would it be possible to implement something similar to the CUDALucas Polite setting, or otherwise opportunistic GPU sieving, so that when GPU resources are underutilised mfakto switches to GPU sieving.

I can't address mfakto. However, in mfaktc, adjusting "GPUSieveSize" will reduce the GPU load and improve responsiveness. Normally, I run this value at the maximum of 128. I have found that dropping it to the range of 30-40 has a good effect for other tasks on the computer.