[cmake-developers] CMake master slowdown in generation step

[Stephen Kelly]

Brad King wrote:

> On 04/04/2014 11:57 AM, Stephen Kelly wrote:
>> I've pushed another commit which gets the time for the 2000 file testcase
>> from about 20s to about 4s for both Ninja and Makefiles. It fails a few
>> unit tests though and I don't have time right now to find out why. 4s is
>> still a bit much, so there's probably more opportunity.
> 
> So far your optimizations seem to be in areas that could have been
> optimized before commit 8cd113ad1d715cc9ce865956870cd462d3659089
> caused this major slowdown.

Optimizations that were possible before a refactoring become necessary 
afterward. That is not surprising. 

I don't understand the point you're making here.

> That one is:
> 
>  cmTarget: Store strings instead of cmSourceFile* to represent SOURCES
> 
> I suspect the performance drop is due to carrying around lots of
> strings instead of raw pointers.

The carrying doesn't seem to have been the problem. The problem seemed to be 
requiring matching/mapping of strings back to cmSourceFile objects.

> You'll have to find another
> way to represent that.  Otherwise this performance drop is too
> high a cost to pay for the target-sources feature and we will
> have to revert that until another solution is found.

There is no cause for alarm. 

Things are going according to plan. Get the feature right, and optimize 
later:

 http://thread.gmane.org/gmane.comp.programming.tools.cmake.devel/3615/focus=5250

I agree of course that the performance is master currently is not 
acceptable.

Re-cmake with v2.8.12.2:
 $ time cmake ..
 real    0m0.712s
 user    0m0.692s
 sys     0m0.020s
 $ time cmake ..
 real    0m0.709s
 user    0m0.697s
 sys     0m0.012s
 $ time cmake ..
 real    0m0.718s
 user    0m0.716s
 sys     0m0.000s

 $ time cmake .. -GNinja
 real    0m0.692s
 user    0m0.674s
 sys     0m0.020s 
 $ time cmake .. -GNinja
 real    0m0.730s
 user    0m0.695s
 sys     0m0.036s
 $ time cmake .. -GNinja
 real    0m0.693s
 user    0m0.678s
 sys     0m0.016s


Re-cmake with optimize-source-file-processing:
 $ time cmake ..
 real    0m0.733s
 user    0m0.689s
 sys     0m0.046s 
 $ time cmake ..
 real    0m0.747s
 user    0m0.724s
 sys     0m0.024s 
 $ time cmake ..
 real    0m0.757s
 user    0m0.743s
 sys     0m0.012s

 $ time cmake .. -GNinja
 real    0m0.352s
 user    0m0.328s
 sys     0m0.024s
 $ time cmake .. -GNinja
 real    0m0.363s
 user    0m0.359s
 sys     0m0.004s
 $ time cmake .. -GNinja
 real    0m0.355s
 user    0m0.348s
 sys     0m0.008s


Thanks,

Steve.