On 23/11/2020 19:16, ASSI wrote:
> Hamish McIntyre-Bhatty via Cygwin-apps writes:
>> That is strange if it's so much faster on your system.
> I've checked again to be sure:
>
> Compile 41 minutes, install 7 (32bit) / 4 (64bit) minutes, packaging 5
> minutes.

Okay, wow that is a LOT quicker. I haven't timed mine precisely buts
it's something like:

Compile 2 hrs, install 40 minutes (64bit), 120 minutes (32bit),
packaging probably like 5 minutes.

>> Fortunately I have loads of memory so it's not too much of an issue
>> for me. Most of the time on 32-bit Cygwin is spent stripping the
>> executables, which is _way_ slower than on 64-bit Cygwin, by a factor
>> of probably 3-4.
> What's the disk based on?  I've recently switched to NVMe and while it's
> not doing wonders compared to the SATA-SSD I've had before it has helped
> to shave some time off the builds, generally in the order of 10…25% (for
> gcc that means about half an hour so that's very welcome even if it
> doesn't sound much).  Running a VM probably means you're running on a
> filesystem image rather than a dedicated disk and that probably means an
> extra slowdown anyway.  NVMe would enable you to make that overhead go
> away, but it needs to be supported through the whole hypervisor / VM
> stack to be effective.

I have an NVME disk, but all my VMs (including this one) are on a SATA
disk, and it's a Samsung QVO 1TB one that was cheaper but also slower
(sustained write slows down to 80 MB/s eventually). Read is quick
though. IIRC it's presented as SATA to the VM too.

I'd better have a look at my configuration. Maybe do some disk and CPU
benchmarks. This is in virtualbox so perhaps it's yet another problem
with that, in the long list of problems I seem to have with virtualbox.
The VM has 16GB of RAM dedicated to it, how does that compare to your
system?

Hamish