mfakto: an OpenCL program for Mersenne prefactoring

[kriesel]

Originally Posted by apocalypse

Has anyone successfully gotten this working and have any tips about what I need to do beyond installing the drivers?

Thanks!

Ubuntu 18.04.3 LTS
4.15.0-66-generic
Radeon R9 290X

Not much of a linux/AMD user myself, but I suggest you check that you have a working install of OpenCL for the AMD gpu, using some other opencl utility.

[ixfd64]

I just pushed another update to mfakto. This update includes the following changes:

1. Fixed an issue where verbosity was not passed to the valid_assignment() function. If you were trial factoring using the -tf option, the verbosity level would remain at -1 unless -v was the first parameter.
2. mfakto now detects the FirePro D-series GPUs used in late 2013 Mac Pros. Ideally we want to detect the micro-architecture based on the feature set rather than the device name, so I made a note of it in the to-do file.
3. Improved the README and the INI file documentation.

Thanks again to kracker for reviewing my code. Additional feedback is always appreciated.

[LaurV]

Originally Posted by ixfd64

mfakto now detects the FirePro D-series GPUs

Benchmarks to James please!
I was a bit disappointed to see that the benchmark page is missing almost ALL of the 4-digits series of AMD cards..

[James Heinrich]

But I have estimated performance for all the FirePro cards:
https://www.mersenne.ca/mfaktc.php?filter=firepro

What I really need is a benchmark from a 5700 XT...

[LaurV]

Well, by 4 digits I meant the RX xxxx that everybody is selling now for 3-4-5 hundred bucks. My guess is they are not so good as the AMD's flagship cards, but much better than the 5xx or their sisters.

[James Heinrich]

Originally Posted by LaurV

Well, by 4 digits I meant the RX xxxx that everybody is selling now for 3-4-5 hundred bucks. My guess is they are not so good as the AMD's flagship cards, but much better than the 5xx or their sisters.

The only RX xxxx cards I'm aware of is the Radeon RX 5700 (and the 5700 XT).
And as I said, I don't have any benchmarks for any RX 5700 series card, so I don't know how they perform with mfakto. I've asked, but it seems nobody on the forums has one (improbable, but nobody has answered my pleas). So they have a null entry on the performance charts for now until I get a benchmark.

[ixfd64]

I just fixed another issue: mfakto would delete the kernel file after compilation if a device number greater than 1 is specified using the -d option. In the original code, mfakto would always try to load the binary data from the first slot regardless of the device number. It should now search the slots until it finds a valid one.

[Uncwilly]

After searching until I was blue in the face, I have given up and am going to ask basically the same question again.

I was handed a new machine. It has an Intel integrated processor. Specifically an Intel UHD Graphics 630. A while back someone pointed to a version of makto that I could run on my laptop's integrated graphics processor. I think it was on James's Mersenne.ca
I was told it wasn't worth it. But it is beating the CPU on GHz-days/day. Anyone know where I could find this version. I don't have admin rights on it, so I can't run GPU-Z on it.

[James Heinrich]

Perhaps this one?
https://download.mersenne.ca/mfakto/mfakto-0.12-hd4000

edit: Although based on what Ken says below, you're likely fine with the normal latest-version (0.15-pre6):
https://download.mersenne.ca/mfakto/mfakto-0.15pre6

[kriesel]

Originally Posted by Uncwilly

After searching until I was blue in the face, I have given up and am going to ask basically the same question again.

I was handed a new machine. It has an Intel integrated processor. Specifically an Intel UHD Graphics 630. A while back someone pointed to a version of makto that I could run on my laptop's integrated graphics processor. I think it was on James's Mersenne.ca
I was told it wasn't worth it. But it is beating the CPU on GHz-days/day. Anyone know where I could find this version. I don't have admin rights on it, so I can't run GPU-Z on it.

Mfacto v0.15pre6 does a fine job of getting about 18GHD/day out of it. IF the Intel OpenCl driver is in place. I have a UHD630 running now, in the same system as mfaktc 0.21 keeps the GTX1050Ti busy. There is a hit on prime95 performance though, because the Intel cpu and Intel IGP are sharing a combined TDP power limit, and also memory. It also requires patience.

[Sun Nov 10 17:14:39 2019]
UID: kriesel/peregrine-uhd630, no factor for M94188673 from 2^75 to 2^76 [mfakto 0.15pre6-Win cl_barrett32_76_gs_2]

took 4 days nonstop.
The mfakto.ini says

# Different GPUs may have their best performance with different kernels
# Here, you can give a hint to mfakto on how to optimize the kernels.
#
# Possible values:
# GPUType=AUTO             try to auto-detect, if that does not work: let me know
# GPUType=GCN              Tahiti et al. (HD77xx-HD79xx), also assumed for unknown devices.
# GPUType=VLIW4            Cayman (HD69xx)
# GPUType=VLIW5            most other AMD GPUs (HD4xxx, HD5xxx, HD62xx-HD68xx)
# GPUType=APU              all APUs (C-30 - C-60, E-240 - E-450, A2-3200 - A8-3870K) not sure if the "small" APUs would work better as VLIW5.
# GPUType=CPU              all CPUs (when GPU not found, or forced to CPU)
# GPUType=NVIDIA           reserved for Nvidia-OpenCL. Currently mapped to "CPU" and not yet functional on Nvidia Hardware.
# GPUType=INTEL            reserved for Intel-OpenCL (e.g. HD4000). Not yet functional.
#
# Default: GPUType=INTEL

 GPUType=INTEL

HD4000, HD620, or UHD630 are functional, as are some others that have at least the HD4000's capability.
I think the version James referred to was when the Intel support was added.

[Uncwilly]

Originally Posted by kriesel

Mfacto v0.15pre6 does a fine job of getting about 18GHD/day out of it. IF the Intel OpenCl driver is in place.

I found an OpenCl.dll or 3 on the machine.
I dl/ed and unzipped the version in question from James's site.
I tried to run it in "PowerShell" and get the following at the output (I trimmed the good.)

Compiling kernels.

        BUILD OUTPUT

Unrecognized build options: -O3
        END OF BUILD OUTPUT
Error -11 (Build program failure): clBuildProgram
ERROR: load_kernels(0) failed