mersenneforum.org - mfakto: an OpenCL program for Mersenne prefactoring

mersenneforum.org (https://www.mersenneforum.org/index.php)

- GPU Computing (https://www.mersenneforum.org/forumdisplay.php?f=92)

- - mfakto: an OpenCL program for Mersenne prefactoring (https://www.mersenneforum.org/showthread.php?t=15646)

[QUOTE=LaurV;495387]I tried this in the past but it didn't really work, if you use 'auto' it will default to a very slow, partially functional, 'gcn'. You may be able to squeeze a fraction of GHzDays if you tune, it does not worth the effort.

Edit: Whoops, didn't see the other thread.[/QUOTE]
That's odd, I got 17.8 GhzD/day out of the HD620 (igp) built in to my i7-7500U laptop, with type auto and mostly default mfakto.ini. What intel hardware were you seeing fractional ghzd/day on?

[QUOTE=ET_;495337]I started using mfakto witn an INTEL HD4400 GPU. Not as speedy as a "real" Nvidia or AMD, but worth at leasat 5 more GHz-days/day.

as I never used mfakto before, I noticed some tuning parameters on the INI file I never encountered on mfaktc for Nvidia:

[code]
# 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=AUTO

GPUType=AUTO
[/code]What should I use with my Intel environment, INTEL or AUTO?

And how can I get the best from it? Should I tweak [COLOR=Red]GPUSievePrimes[/COLOR], [COLOR=red]GPUSieveSize[/COLOR] or [COLOR=red]GPUSieveProcessSize[/COLOR] to have the best performances on my INTEL GPU?

Luigi[/QUOTE]
My setup: HD620 on i7-7500U equipped laptop, Win10 x64, latest mfakto, nearly default mfakto.ini. I changed checkpointdelay to 600 and that seemed to reduce the wait period of the gpu while checkpoint files were being written to an acceptable compromise, raising overall throughput a bit. This was with type=auto. It cranked out about 17.8ghzd/day. As I recall I tried some other tuning but it didn't seem to matter. It had some impact on prime95 because it's part of the same chip and heat budget as the cpu cores. As I recall there was a net gain in total throughput but it was small.
Note that installing the NVIDIA SDK or AMD SDK or VS or something took out something that opencl function on the hd620 relied on, stopping mfakto from continuing to work on the HD620, so it would grab the cpu instead. Haven't bothered to investigate or resolve that yet.

If you want to explore tuning it further, there might be something useful in past posts regarding HD4xxx or Intel in general. From my mfakto forum thread notes, posts that might be useful:
674 HD4000 tune parameters
1122 getting hd 4xxx recognized etc
1124 intel opencl driver download link
1177 0.15pre2 with change list
1232 page 112 0.15pre4 summary of testing results, including hd4600 working at 18-19ghzd/day
1236 hd4600 performance profiling results (up to 24ghzd/day depending on p, bits, kernel)
1248 0.15pre5 and a few new features listed
1249 more hd 4600 profiling results
1299 page 119 mfakto on hd4600; sieve on gpu disabled or driver crashes, marginal effect on prime95; 7-14 ghzd/day
1301 hd4600 with cpu sieving
1322 (page 121) some tuning recommendations from bdot

Have fun!

[QUOTE=kriesel;495403]My setup: HD620 on i7-7500U equipped laptop, Win10 x64, latest mfakto, nearly default mfakto.ini. I changed checkpointdelay to 600 and that seemed to reduce the wait period of the gpu while checkpoint files were being written to an acceptable compromise, raising overall throughput a bit. This was with type=auto. It cranked out about 17.8ghzd/day. As I recall I tried some other tuning but it didn't seem to matter. It had some impact on prime95 because it's part of the same chip and heat budget as the cpu cores. As I recall there was a net gain in total throughput but it was small.
Note that installing the NVIDIA SDK or AMD SDK or VS or something took out something that opencl function on the hd620 relied on, stopping mfakto from continuing to work on the HD620, so it would grab the cpu instead. Haven't bothered to investigate or resolve that yet.

If you want to explore tuning it further, there might be something useful in past posts regarding HD4xxx or Intel in general. From my mfakto forum thread notes, posts that might be useful:
674 HD4000 tune parameters
1122 getting hd 4xxx recognized etc
1124 intel opencl driver download link
1177 0.15pre2 with change list
1232 page 112 0.15pre4 summary of testing results, including hd4600 working at 18-19ghzd/day
1236 hd4600 performance profiling results (up to 24ghzd/day depending on p, bits, kernel)
1248 0.15pre5 and a few new features listed
1249 more hd 4600 profiling results
1299 page 119 mfakto on hd4600; sieve on gpu disabled or driver crashes, marginal effect on prime95; 7-14 ghzd/day
1301 hd4600 with cpu sieving
1322 (page 121) some tuning recommendations from bdot

Have fun![/QUOTE]

:tu:

Has anyone here compiled mfakto for a Mac?

Anyone know whether mfakto will work on newer integrated GPUs, such as the HD Graphics 530?

[QUOTE=ixfd64;504230]Anyone know whether mfakto will work on newer integrated GPUs, such as the HD Graphics 530?[/QUOTE]

It surely works on Intel 4xx, but the CPU penalty is noticeable.

I got a laptop with an HD Graphics 620 integrated GPU. mfakto correctly recognizes the device but gets stuck at the self-test:

[CODE]OpenCL device info
name Intel(R) HD Graphics 620 (Intel(R) Corporation)
device (driver) version OpenCL 2.1 (21.20.16.4678)
maximum threads per block 256
maximum threads per grid 16777216
number of multiprocessors 24 (24 compute elements)
clock rate 1100MHz

Automatic parameters
threads per grid 0
optimizing kernels for INTEL

Loading binary kernel file mfakto_Kernels.elf
Compiling kernels.
GPUSievePrimes (adjusted) 81206
GPUsieve minimum exponent 1037054
Started a simple selftest ...
######### testcase 1/30 (M1031831[63-64]) #########[/CODE]

What am I doing wrong?

[QUOTE=ixfd64;504234]I got a laptop with an HD Graphics 620 integrated GPU. mfakto correctly recognizes the device but gets stuck at the self-test:

[CODE]OpenCL device info
name Intel(R) HD Graphics 620 (Intel(R) Corporation)
device (driver) version OpenCL 2.1 (21.20.16.4678)
maximum threads per block 256
maximum threads per grid 16777216
number of multiprocessors 24 (24 compute elements)
clock rate 1100MHz

Automatic parameters
threads per grid 0
optimizing kernels for INTEL

Loading binary kernel file mfakto_Kernels.elf
Compiling kernels.
GPUSievePrimes (adjusted) 81206
GPUsieve minimum exponent 1037054
Started a simple selftest ...
######### testcase 1/30 (M1031831[63-64]) #########[/CODE]What am I doing wrong?[/QUOTE]
It's hard to tell. That's all the output you get? How long did you wait? Remember, the HD620 is only about 18GhzD/day, so it will take quite a while, 10 to 50 times what a current separate gpu takes, to get anything done. I'm running mfakto on an HD620 sometimes, with an essentially stock mfakto.ini. It does cut into the prime95 throughput. You could try running gpu-z or some other gpu monitoring tool to see if the igp is working or idle. [url]https://www.mersenneforum.org/showpost.php?p=488474&postcount=6[/url]

Suddenly sluggish GPU = dying GPU?

I have an AMD Radeon HD 7770 that's been running TF like a champ for almost six years. In the last few days, when MFAKTO 0.14 for Windows is running the mouse pointer has started to lag annoyingly behind my mouse movements and program windows are hesitating noticeably when maximizing or minimizing them. As a result, the PC is getting harder and harder to use for other purposes while TF'ing. GPU load is at 99% according to GPU-Z.

This lag/hesitation never used to happen, but it seems to be getting worse by the day. Could it be a sign that the graphics card is nearing the end of the line?

Only other possibility I can think of is the driver (24.20.13019.1008), which was updated about four weeks ago. Are newer drivers [B]less[/B] TF-capable than older ones?

GPU temperature is about normal for this computer (76C) and the PC case was dusted with compressed air about three weeks ago.

Thanks for any insights/hypotheses/sheer speculation.

[QUOTE=Rodrigo;504453]I have an AMD Radeon HD 7770 that's been running TF like a champ for almost six years. In the last few days, when MFAKTO 0.14 for Windows is running the mouse pointer has started to lag annoyingly behind my mouse movements and program windows are hesitating noticeably when maximizing or minimizing them. As a result, the PC is getting harder and harder to use for other purposes while TF'ing. GPU load is at 99% according to GPU-Z.

This lag/hesitation never used to happen, but it seems to be getting worse by the day. Could it be a sign that the graphics card is nearing the end of the line?

Only other possibility I can think of is the driver (24.20.13019.1008), which was updated about four weeks ago. Are newer drivers [B]less[/B] TF-capable than older ones?

GPU temperature is about normal for this computer (76C) and the PC case was dusted with compressed air about three weeks ago.

Thanks for any insights/hypotheses/sheer speculation.[/QUOTE]
The system is not a new one. "six years".

How I might approach this:

[B]Using Task Manager[/B]:
Check for free and available memory. Does it look low?
Check for running processes. Are there things running you don't expect? Can you terminate some, change their start status from automatic to manual?
Check "Show processes from all users" on the processes tab.
Any with unexpectedly high memory usage?

Check for disk paging and utilization. Does it seem high? Which processes are using the most paging? I've seen web browsers grow to multiple GB of ram and cause high paging over time. Stop/restart the browser app takes care of it for a while. Reducing number of open tabs helps. Some web pages seem particularly memory hungry and are better closed when not in active use. Maybe a bad memory leak in their scripting??
Does the six year old system still have the same amount of ram installed as when new, and can it be doubled? Automatic updates for applications like browsers may have changed the memory usage of a fixed list of applications.
Are you also running prime95 on this system? P-1 stage 2 and ECM stage 2 want a lot of memory. More so as exponents assigned go up with time. What was it running years ago vs. now?

What's using the most cpu? On versions of Windows that are no longer supported with updates, automatic update availability checking could still be enabled, and using significant cpu, making the system sluggish. It's a system service that can absorb a full core for days on end, until that process is stopped in Task Manager.

A disk that's becoming unreliable, or that has become too full or fragmented, or connected to an unreliable interface, can become sluggish.
[B]Using Event Viewer[/B]:
Check for disk retries and errors. If you find numerous disk retries or errors it might be about to fail, so promptly make sure you have a recent backup on other media.
Look also for other events, that were not present or not as common a year or more ago but are now.
[B]Using Explorer[/B]:
Local Disk, C:, Properties (how full?)
then Tools (how fragmented? Is defragmented scheduled to run automatically at low use times?)
[B]Using Add/remove programs[/B]:
Anything you don't want to keep, and could be thrown overboard to free disk space and make backups smaller and quicker?

See also [URL]https://www.mersenneforum.org/showthread.php?t=23415[/URL]

Something like the above is what I use on my 8 year old main use laptop with 8GB (maximum & installed) ram when it gets less responsive. Right now it has ~1GB "available" and essentially 0 free while running P-1 with 3.6GB and is pretty responsive.

Mfakto:
The latest version is V0.15pre6. I suggest upgrading, after the issue is sorted out.

Good luck, and let us know what you find.