Compiler flags and memory usage

Antonio Trande anto.trande at
Mon Jun 16 18:02:25 UTC 2014

On 06/16/2014 12:12 AM, Kevin Kofler wrote:
> Antonio Trande wrote:
>> To pass these problems, I had to decrease '-O' optimization flag
>> gradually to '-O1' or '-O0', latest attempt is the removal of '-pipe'
>> from RPM_OPT_FLAGS.
>> What's best choice in these cases?
> Definitely not -O0, that generates really crappy (slow and large) code. I'd 
> also not recommend -O1. If you must change something, remove -pipe, that 
> only makes builds slower (given sufficient RAM for the piping), but does not 
> affect the quality of the generated code.

On 06/16/2014 12:37 AM, Reindl Harald wrote:
> Am 16.06.2014 00:12, schrieb Kevin Kofler:
> i would add here
> %_smp_mflags -j1
> if it comes to reduce memory due compile

Thanks for your tips.

On 06/16/2014 03:54 AM, John Reiser wrote:>> coin-or-lemon is still in
f20 updates-testing repos
> Please mention such facts early and prominently.  It saves time
> for those who try to help you.


>> So my RAM is not adequate.
> It *can* work, but your machine is not configured properly for large
> -----
>> $ cat /proc/swaps
>> Filename				Type		Size	Used	Priority
>> $     ### /proc/swaps is empty.  There is no swap space.
> -----

Honestly, I already had a swap partition but I never verified if it was
available and used.

> Get some swap space, either a disk partition, or a file of at least 6GB:
>     dd if=/dev/zero of=my_swap_space bs=32k count=200k
>     mkswap my_swap_space
>     swapon my_swap_space

Okay, I have made ready a swap partition mounted at boot time by
changing /etc/fstab; 10GB should be sufficient. This is the swap after a
first *complete* compilation of SeqAn (at last):

$ cat /proc/swaps
Filename     Type		Size	Used	Priority
/dev/sda2    partition	       10485756	188752	-1

My hardware needs more than 1 hour just for pair_align.cpp with "make
-j1" in .spec. In any case, now I can complete SeqAn building on my PC.

Thank you very much.

Antonio Trande

GPG Key: D400D6C4

More information about the devel mailing list