[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Gm2] some comments on -fm2-whole-program
From: |
Gaius Mulley |
Subject: |
[Gm2] some comments on -fm2-whole-program |
Date: |
Sun, 17 Aug 2014 13:44:47 -0500 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/23.4 (gnu/linux) |
Hi,
I thought it might be interesting to post some preliminary results with
the -fm2-whole-program option. I've compiled the gm2/example/map
program with "-O3 -fm2-whole-program" alternatively with "-O3". In this
application (a maze generation program) which uses arrays heavily the
-fm2-whole-program almost doubles the speed of the final executable with
only about 7% increase size of code on the x86_64 platform.
time ./map-whole-program > t2
real 0m1.515s
user 0m1.500s
sys 0m0.000s
time ./a.out > t1
real 0m2.836s
user 0m2.824s
sys 0m0.004s
size a.out
text data bss dec hex filename
164522 920 492016 657458 a0832 a.out
size map-whole-program
text data bss dec hex filename
175292 912 492112 668316 a329c map-whole-program
The option is still alpha, as I'm currently hunting down some
linking bugs, however the results are very pleasing. I'll
add the m2-whole-program option to the examples/map/Makefile.in
when it is fully operational.
It is also interesting to note that the small amount of effort
required to implement this switch vs. the increase in performance. A
slight word of caution though - this flag is unlikely to work with
projects which mix the language dialects or mix pim/iso library
modules.
regards,
Gaius
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [Gm2] some comments on -fm2-whole-program,
Gaius Mulley <=