How long does it take you to compile wine?
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip... Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ). Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
On Sun, Jun 1, 2008 at 2:44 AM, Dan Kegel <dank(a)kegel.com> wrote:
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip...
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
What are your arguments to make? -- James Hawkins
On Sun, Jun 1, 2008 at 12:47 AM, James Hawkins <truiken(a)gmail.com> wrote:
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
What are your arguments to make?
Details are on buried that page, but it's "-j3" for the dual core / dual CPU systems. I found a nice graph of Wine's codebase size over time: http://www.ohloh.net/projects/wine/analyses/latest Looks like it's just about exactly doubled since that last measurement in late 2002. I suppose my e7200 is a budget performance system; a dual Xeon back then was pretty special. Probably have to compare it to something heftier today? - Dan
On Sunday 01 June 2008 10:16:19 Dan Kegel wrote:
I suppose my e7200 is a budget performance system; a dual Xeon back then was pretty special. Probably have to compare it to something heftier today?
It's 4m55s on a 2 way dual core Xeon machine (4 cores, 2 Ghz, 4MB cache each). This is on a hot disk cache, i.e. I ran $ make distclean && ./configure && time make -j5 twice and took the second measurement. Compiler is gcc 4.3.0. -Hans
On Sun, Jun 1, 2008 at 2:16 AM, Dan Kegel <dank(a)kegel.com> wrote:
I found a nice graph of Wine's codebase size over time: http://www.ohloh.net/projects/wine/analyses/latest Looks like it's just about exactly doubled since that last measurement in late 2002.
Anyone else notice that link shows we have some license inconsistencies? There could be a legitimate reason for that, I didn't have time to look. Anyway, it's showing 5 files with LGPL v2.1 (huh? shouldn't all the files be showing as LGPL v2.1?), 3 files BSD (X11?), and 2 files GPL. -Brian
On Sun, Jun 1, 2008 at 1:16 AM, Dan Kegel <dank(a)kegel.com> wrote:
http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
Details are on buried that page, but it's "-j3" for the dual core / dual CPU systems.
I found a nice graph of Wine's codebase size over time: http://www.ohloh.net/projects/wine/analyses/latest Looks like it's just about exactly doubled since that last measurement in late 2002.
http://techgage.com/article/intel_core_2_duo_e7200_-_the_new_budget_supersta... has some recent wine build benchmarks: Time to build wine-0.9.59: quad core, -j 5: QX9770: 182s QX9650: 193s QX6850: 200s Q9450: 216s Q6600: 251s dual core, -j3: e8400: 363s e7200: 437s e6750: 422s e6600: 472s Golly. Maybe I should have gotten a quad core :-) BTW, I just compared the e7200 and e8400 on the 'valgrind wine test suite with -j2' task. Results: e8400 is 18% faster (83 minutes vs. 99 minutes). I'd love to try this on a quad core... maybe I'll find one at work and give it a shot.
Dan Kegel wrote:
On Sun, Jun 1, 2008 at 1:16 AM, Dan Kegel <dank(a)kegel.com> wrote:
http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-) Details are on buried that page, but it's "-j3" for the dual core / dual CPU systems.
I found a nice graph of Wine's codebase size over time: http://www.ohloh.net/projects/wine/analyses/latest Looks like it's just about exactly doubled since that last measurement in late 2002.
http://techgage.com/article/intel_core_2_duo_e7200_-_the_new_budget_supersta... has some recent wine build benchmarks:
Time to build wine-0.9.59: quad core, -j 5: QX9770: 182s QX9650: 193s QX6850: 200s Q9450: 216s I have tested this on F9 with gcc-4.3.0 and Wine-1.0 and get for the Q9450 305 seconds. The compiler advances kill performance faster than CPU vendors can add it... Though to be fair gcc-4.3.0 does more checks than gcc-4.1.x by default with -Wall; the bounds check come to mind.
bye michael
Q6600: 251s
dual core, -j3: e8400: 363s e7200: 437s e6750: 422s e6600: 472s
Golly. Maybe I should have gotten a quad core :-)
BTW, I just compared the e7200 and e8400 on the 'valgrind wine test suite with -j2' task. Results: e8400 is 18% faster (83 minutes vs. 99 minutes). I'd love to try this on a quad core... maybe I'll find one at work and give it a shot.
-- Michael Stefaniuc Tel.: +49-711-96437-199 Consulting Communications Engineer Fax.: +49-711-96437-111 -------------------------------------------------------------------- Reg. Adresse: Red Hat GmbH, Otto-Hahn-Strasse 20, 85609 Dornach bei Muenchen Handelsregister: Amtsgericht Muenchen HRB 153243 Geschäftsführer: Brendan Lane, Charlie Peters, Michael Cunningham, Werner Knoblich
2008/6/1 Dan Kegel <dank(a)kegel.com>:
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip...
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
Have you checked how much Wine has grown in size since then and factored that in as well? Five years seems to be before CodeWeavers took the torch (I don't know when they started contributing *big time* to Wine) with the MSI, DirectX and other improvements. A way to get a better idea would be to get the number of lines in the make output and have a lines/second comparison. This is because Wine could be twice as big as it was five years ago. - Reece
Hi Dan, it's 31 minutes on AMD Athlon(tm) 64 Processor 3700+, 2GB RAM. Cheers Vit Dan Kegel wrote:
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip...
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
On Sun, 1 Jun 2008, Vít Hrachový wrote:
Hi Dan, it's 31 minutes on AMD Athlon(tm) 64 Processor 3700+, 2GB RAM.
He he, I think I hold the record. Wall clock | CPU | RAM | gcc | Machine ------------+------------------+-------+-------+-------------------- 145m38.115s | P3 600MHz | 192MB | 4.2.4 | Debian Laptop ------------+------------------+-------+-------+-------------------- 39m30.058s | Pentium M 1.5GHz | 768MB | 4.2.3 | Ubuntu Laptop (1) ------------+------------------+-------+-------+-------------------- 53m12.450s | Athlon XP 2600+ | 1GB | 4.2.4 | Debian Desktop (2) | (2.1GHz) | | | 12m34.218s | | | | 100% ccache hitrate ------------+------------------+-------+-------+-------------------- 50m58.424s | Athlon XP 1800+ | 1GB | 4.2.4 | Debian Desktop | (1.53GHz) | | | ------------+------------------+-------+-------+-------------------- 29m30.277s | Distcc | | | (3) ------------+------------------+-------+-------+-------------------- (1) This one is missing Esd, NAS, Jack and Valgrind (memcheck.h) support. However in other tests this does not seem to make much difference. It's also the only one with a different distribution and gcc version which makes comparisons tricky. (2) This is actually my main development machine. And damn, it's slow! I hope to retire it soon but it still hurts. I guess part of the reason for it being so slow is that it's never really idle. There's always half a dozen Firefox windows around, gkrellm, xmms, a bunch of background processes like MythTV, MySQL, software RAID 1, 90% full disks, etc. (3) This is with 1 job on the Athlon 2600+ and 2 on the Athlon 1800+. I tried getting Wine 2002 to compile but even in my Debian 3.1 chroot and with gcc 2.95 it wouldn't. So we don't have a good point of comparison. -- Francois Gouget <fgouget(a)free.fr> http://fgouget.free.fr/ If it stinks, it's chemistry. If it moves, it's biology. If it does not work, It's computer science.
Dan Kegel a écrit :
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip...
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth :-)
I used to compiled it over 30minutes... now I barely wait... 6mn26s make -j5, Q6600 @ 2.4GHz gcc version also matters (4.2.3 for me) A+ -- Eric Pouech "The problem with designing something completely foolproof is to underestimate the ingenuity of a complete idiot." (Douglas Adams)
On Sun, Jun 1, 2008 at 4:50 AM, Eric Pouech <eric.pouech(a)orange.fr> wrote:
I used to compiled it over 30minutes... now I barely wait... 6mn26s
make -j5, Q6600 @ 2.4GHz
Ouch! I almost chose a Q6600 for this machine! Well, I suppose I should test it with the motherboard graphics turned off, that saps a lot of performance.
gcc version also matters (4.2.3 for me)
I'm using same version. What graphics card, btw? - Dan
On Sun, Jun 1, 2008 at 10:20 AM, Dan Kegel <dank(a)kegel.com> wrote:
make -j5, Q6600 @ 2.4GHz
for comparison I have an older 2.4 ghz quad CPU single core build box with gcc 4.1.x and Fiesty. make -j4 real 16m34.006s user 56m54.569s sys 3m26.561s -- Steven Edwards "There is one thing stronger than all the armies in the world, and that is an idea whose time has come." - Victor Hugo
real 5m22.625s user 17m16.993s sys 1m1.336s This is on Kubuntu 8.04 w/GCC 4.2.3. The system uses a Q6600 overclocked to 2.9Ghz, 4GB of DDR2 RAM(a)1066Mhz, and a 9600GT video card. I had Amarok and some other stuff going in the background, but nothing that would make a huge diff. -J
Dan Kegel <dank <at> kegel.com> writes:
Way back five years ago, we all compared notes on how long it took to build Wine on our systems. http://www.winehq.org/site/?issue=149#Compile%20Time%20Comparisons%20/%20Tip...
Today I measured it on a new computer (with a spiffy new Intel dual core e7200 processor with 2GB of RAM, http://kegel.com/new-computer-2008.html ).
Result: ... it takes me a bit longer today on this machine (11 minutes) to build as it did then on a dual cpu hot rod (8 minutes). I guess Moore's Law isn't quite keeping up with Wine's code growth
Depends on the machines, on my new PC I just built it is... rather fast and very useful for regression testing. Doing a configure, build and install from a completely clean setup, without ccache on my machine takes 3 minutes :) Quite nice. Ben H
participants (12)
-
Ben Hodgetts -
Brian Vincent -
Dan Kegel -
Eric Pouech -
Evil Jay -
Francois Gouget -
Hans Leidekker -
James Hawkins -
Michael Stefaniuc -
Reece Dunn -
Steven Edwards -
Vít Hrachový