Vasily,

On 02.02.2015 14:44, Vasily wrote:
Hi Dirk,

Do you have any methods on how do one finds those places to improve subst() 
performance?

all I've ever done is to profile SCons with cProfile, and then looked at the 
output.

I mean, how can I understand which subst's are used, and which of those could 
benefit from simplifying most (like FastCPP tool does
with suffixes)?

There is no silver bullet here. In the case of these C++ benchmarks it's clear that the CXXCOM action gets used over and over again, so it's a low hanging fruit for optimization. But as you see, I didn't pre-expand the linker command...it gets called only 200 times for the whole build, so it's not worth the effort.

In general, one also has to take into account the fact that these benchmarks use very rudimentary C++ files/headers. The compile time per file is kept small, which means that timing the full build is measuring the "overhead" that the used build system adds to the actual compile/link time.
As soon as your sources get larger and more complicated, the impact of the 
mentioned optimizations will get less visible.
They reduce the overhead, but that doesn't automatically result in a 20% faster 
build every time. ;)

Best regards,

Dirk

_______________________________________________
Scons-dev mailing list
[email protected]
https://pairlist2.pair.net/mailman/listinfo/scons-dev

Reply via email to