qemu-system-arm on my amd64 laptop seems quite a bit slower than building on a cortex a9 imx.
I don't have numbers for qemu but on real hardware a few weeks ago: bbb with nfs root, make build 25 hrs 31 min cubox with usb flash root, make build 9 hrs 21 min At least part of the bbb slowdown is likely due to the cpsw hardware/driver. On Mon, Jun 29, 2015 at 10:47:38PM -0500, Edwin Amsler wrote: > Building on top of Qemu with dpb on my 2013 Macbook Pro with SSD doesn?t seem > fantastic at the moment either. The single core-ness on OpenBSD might be a > little to blame but the system usage is ~50% while building. Might be the > virtio drivers? > > I?ll see about snapshotting the filesystem 8x and running a few more VMs with > one acting as an NFS server. > > On Jun 15, 2015, at 9:59 PM, andrew fabbro <and...@fabbro.org> wrote: > > > On Sat, May 23, 2015 at 9:30 PM, Philip Cheney <phi...@packwidth.net> wrote: > > > >> So far, the only issue I've had building anything on arm is the speed > >> difference. The limited hardware really does take its toll: limited RAM, > >> extremely slow IO, heat issues with continuous high load on fanless > >> systems. Building the entire ports tree in one go is not representative of > >> most of this platform's use-cases, though. > >> > > > > In the last 72 hours, I've built 81 packages for 5.7, compiling pretty much > > continuously on my Beagle Bone Black. I haven't seen any crashes, errors, > > or problems (other than a couple of my own making/learning opportunities). > > But yes, it's very slow - python took something like 20 hours to build > > using the internal flash. > > > > But hey, that's why we have batch features in the shell. > > > > I wonder if this will be like painting the Golden Gate Bridge - if you > > start with the 5.7 release, by the time all the packages are built, 5.8 > > will be out :-) > >
