On Thu, Mar 5, 2015 at 12:05 PM, Sven Van Caekenberghe <sven@stfx.eu> wrote:
Phil,
On 05 Mar 2015, at 11:44, phil@highoctane.be wrote:
Images are starting slow, that's my experience. Here, a Pharo3 on Windows with quite a bunch of things in it takes six seconds to load on first launch and three after that.
And that's on a i7 4770K clocked 3.85GHz with SSD drives.
It takes less time to revive a VMWare VM :-(
Maybe it is Windows ? Or certain extra stuff that you have in your image that does work at startup ? Because you should see sub second startup times.
(This is on the smallest Digital Ocean instance, 1 virtual CPU, 512 Mb RAM, SSD)
root@stfx:~/experiments/pharo4# curl get.pharo.org/40+vm | bash % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 2885 100 2885 0 0 17974 0 --:--:-- --:--:-- --:--:-- 18031 Downloading the latest 40 Image: http://files.pharo.org/image/40/latest.zip Pharo.image Downloading the latest pharoVM: http://files.pharo.org/vm/pharo/linux/stable.zip pharo-vm/pharo Downloading PharoV10.sources: http://files.pharo.org/sources//PharoV10.sources.zip Downloading PharoV20.sources: http://files.pharo.org/sources//PharoV20.sources.zip Downloading PharoV30.sources: http://files.pharo.org/sources//PharoV30.sources.zip Creating starter scripts pharo and pharo-ui
root@stfx:~/experiments/pharo4# ./pharo Pharo.image printVersion [version] 4.0 #40535
root@stfx:~/experiments/pharo4# ls -lah Pharo.image -rw-rw-r-- 1 root root 22M Mar 4 14:51 Pharo.image
root@stfx:~/experiments/pharo4# time ./pharo Pharo.image eval '100 factorial' 93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000
real 0m0.386s user 0m0.184s sys 0m0.028s
root@stfx:~/experiments/pharo4# time ./pharo Pharo.image eval '100 factorial' 93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000
real 0m0.422s user 0m0.192s sys 0m0.032s
root@stfx:~/experiments/pharo4# time ./pharo Pharo.image eval '100 factorial' 93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000
real 0m0.863s user 0m0.404s sys 0m0.064s
It is the latest (smaller) image, but that should not make much difference.
On my CentOS Linux, headless tests are fast as well. Granted, my Windows image was super large. 300Megs or something. I cleaned the Monticello things with flush cache, which removed 41 megs. I can't test the CLI version on Windows, as I do get a window opening. There are a ton of windows and morphs open in there and I've run a couple of experiments. Now, the Linux image was large too. 200Megs or so. And the window showed immediately, then it took a second and a half or so to show the images back. There is a black background before. Maybe the system is processing its startUp list, unhibernating forms, I don't know what's going on but it takes that 1.5 secs. The image gets large because of Monticello caches, and TWM (Tiling Window Manager) keeping a handle on browsers etc over time. So, I cleaned the caches, closed TWM, gc'ed a time or two and then got the image back to 90 megs. Then the image opens in like 0.5 sec. It is hard to measure as what I want is the time between issuing the pharo-ui XYZ.image and the fact that I can actually see the UI. How would one do that? Sorry for the rant but generally speaking, the Pharo UI seems slower than the Squeak 4.5 ui. Or the EToys UI for that matter. I am using those for little educational programs and they feel faster. I'd like to know why. Both systems have morphic. Is it because of aliased fonts? Squeak 4.5 also has that. Is there any way to monitor what's going on during image startup? Phil
Sven
-- Sven Van Caekenberghe Proudly supporting Pharo http://pharo.org http://association.pharo.org http://consortium.pharo.org