Running PF on an Ubuntu virtual machine on a Windows 10 host,
the full development for a reference image takes more than twice the time (2’:30") than on a native Windows 10 or a native Debian system (1’:0").
I thought first about a memory allocation issue but I’ve found a visible difference about the CPU usage.
On native systems, PF makes the best usage of available CPUs (i5 4 cores), close to 100% on windows task manager.
On the VM, it starts around a bit more of 50% and then drops to around 33%. This is coherent with the times I can see.
The VM settings are:
Does the host machine have more than 4 cores? If not, it is best to leave at least 1 core for the host.
When the host needs to use even just one core, then it could freeze all 4 cores for the guest VM, because in many cases (for VMWare) all prescribed cores need to be available at the same time before they are allocated to the guest.
So, it might be worth trying assigning just 3 or 2 cores to the guest, which oddly can sometimes improve the performance of the guest.
I’ve just installed VirtualBox + GuestAdditions with the same Ubuntu-64 16.10.
The time is about same, 2’:40" while strangely CPU looks better. Notice the same kind of drop (75% to 50%) after a while (at the same time there is a red glitch on the orange “caching” status):
The output image is 6000x4000 large. The treatment is:
When removing the layer “noise reduction” the CPU drop disappears and, as expected, the time is better, around 40". I have not the drop on Windows (and native Linux as well because I’ve the same execution time), but why “noise reduction” would use less CPUs than the other layers ?
Because it’s easy to test, how about with 2, for the same reasoning as before (if your host needs to use 2 cpus).
Do you have Intel-VT support enabled in BIOS? Depending on BIOS, this is sometimes found under a security menu. But I think this might be one of the differences between i5 and i7.
But after this, I’m afraid I am out of suggestions.
It is taxing for Windows to allocate more than half of the RAM to the VM. Try increasing the swap space in the VM instead. E.g., if you have 8 GB, allocate only 4 GB. Then increase your Ubuntu swap to 4 GB.
It might be a driver compatibility issue. WMware Tools has poor compatibility with the latest host hardware and Linux software. E.g, my GPU doesn’t support OpenGL 3.3 on VMWare.
There is a Windows version of PhotoFlow. @Carmelo_DrRaw recently built a new one.
Then you don’t need that much swap. 4 GB is just an arbitrary number. The point is that Windows and VMware appear to work better when the host retains at least half of the RAM and a couple of cores. It might not look like they are being used for anything on the host, but in my experience, things magically run smoother. Every system setup is different, so you will need to experiment.
Not strange at all. The VM should not be able to see the stats of the host, unless you enable it to do so with your hacking skills.
PS Latest Windows builds can be found here (currently photoflow-20170521.zip).
Thanks :-), I’m testing this solution too. Cuirently I’ve some issue with it mainly due to image size. And Digikam on windows doesn’t provide the “Open With” feature (so far).
At the end I’ve to find out a triplet OS-DAM-Editor (non-destructive) which works for me (to replace Lightroom standalone).
The Ubuntu VM seems the most open, handy (for me) and reliable but suffers lower performance, not using the full power of the machine.
Concerning the problem with large images on windows, I’ll try to create a 64-bits windows package during the next few days… probably this will be the good solution.
4 GB is just an arbitrary number. The point is that Windows and VMware appear to work better when the host retains at least half of the RAM and a couple of cores. It might not look like they are being used for anything on the host, but in my experience, things magically run smoother. Every system setup is different, so you will need to experiment.