Using all cores during compile

jwong · Post by **jwong** » Thu Oct 13, 2011 2:26 pm

Is there anyway to use all 4 cores on my quad core during compilation? It appears I am only using 1 core. I am hoping there is a flag I can pass to the makefile.

Alternatively, I could probably break the makefile in four different makefiles for compiling and a separate makefile for linking. From there I could execute the compile makefiles seperately and finally the link makefile when the others are done.

Has anyone tried this? I'd really like to speed up compilations, especially when doing regression testing.

ryan woodsmall · Post by **ryan woodsmall** » Thu Oct 13, 2011 2:58 pm

Is there anyway to use all 4 cores on my quad core during compilation? It appears I am only using 1 core. I am hoping there is a flag I can pass to the makefile.

Look at the "-j" option to the make command. It takes a numeric argument that signifies how many simultaneous jobs should be running.

ccache is another useful project that can considerably speed up compilation by caching previous compiler output:

http://ccache.samba.org/

Post by **André H.** » Thu Oct 13, 2011 4:58 pm

beside ccache there is also distcc, but on a quadcore a "make -j4" is all you need

FrÃ©dÃ©ric Delanoy · Thu Oct 13, 2011 5:49 pm

On Thu, Oct 13, 2011 at 21:26, jwong <[email protected]> wrote:

Is there anyway to use all 4 cores on my quad core during compilation? It appears I am only using 1 core. I am hoping there is a flag I can pass to the makefile.

Alternatively, I could probably break the makefile in four different makefiles for compiling and a separate makefile for linking. From there I could execute the compile makefiles seperately and finally the link makefile when the others are done.

Has anyone tried this? I'd really like to speed up compilations, especially when doing regression testing.

You can also pass --disable-tests to ./configure
Also "-O0" in CFLAGS can help

blaiseg07 · Post by **blaiseg07** » Thu Oct 13, 2011 9:39 pm

jwong wrote: Has anyone tried this? I'd really like to speed up compilations, especially when doing regression testing.

make -j5(one more than you physically have) and --disable-tests in the configure.

handy dandy bash script I use...adjust to your needs.

Code: Select all

#! /bin/sh

CC="gcc-4.6 -m32" LDFLAGS="-L/lib32 -L/usr/lib32 -Wl,-rpath,/lib32 -Wl,-rpath,/usr/lib32" ./configure -v --prefix=/usr --without-capi --without-cups --without-jack --without-sane --with-x --without-openal --disable-tests --without-nas --without-v4l --without-gphoto --with-opencl
make depend && make -j4
echo {your password} | sudo -S make install

FrÃ©dÃ©ric Delanoy · Fri Oct 14, 2011 8:45 am

On Fri, Oct 14, 2011 at 04:39, blaiseg07 <[email protected]> wrote:

jwong wrote:
Has anyone tried this? I'd really like to speed up compilations, especially when doing regression testing.
make -j5(one more than you physically have) and --disable-tests in the configure.

handy dandy bash script I use...adjust to your needs.

Code:
#! /bin/sh

CC="gcc-4.6 -m32" LDFLAGS="-L/lib32 -L/usr/lib32 -Wl,-rpath,/lib32 -Wl,-rpath,/usr/lib32" ./configure -v --prefix=/usr --without-capi --without-cups --without-jack --without-sane --with-x --without-openal --disable-tests --without-nas --without-v4l --without-gphoto --with-opencl
make depend && make -j4
echo {your password} | sudo -S make install

Using ccache gcc instead of gcc would help even more

jwong · Post by **jwong** » Fri Oct 21, 2011 11:47 am

Thanks everyone. Using the -j flag has dropped compiles significantly.