26 lines
1.4 KiB
Markdown
26 lines
1.4 KiB
Markdown
|
# Fast Compiliing C++ Library
|
||
|
|
-----------------------------------------
|
||
|
|
|
||
|
|
In my prior attempts at developing MACE what I discovered is that compile times
|
||
|
|
would explode to unreasonable levels that hinder the rate of development more
|
||
|
|
than what can be saved by reduced typing. So I began a quest to get C++ to compile
|
||
|
|
as quickly as Java or C# and the result is this library.
|
||
|
|
|
||
|
|
One of the major drawbacks to templates is that they place everything in header and
|
||
|
|
must be compiled with every run and generate a lot of object code. With Link Time Optimization,
|
||
|
|
the benefit of inline methods mostly disapears, leaving only static vs dynamic polymorphism.
|
||
|
|
|
||
|
|
For the vast majority of applications, a virtual method call is not the bottleneck and the
|
||
|
|
increased compile times costs more than is otherwise justified; therefore, the Fast Compiling C++
|
||
|
|
library opts for virtual interfaces to handle reflection instead of template interfaces. One could
|
||
|
|
argue that both types of reflection could be useful.
|
||
|
|
|
||
|
|
Another source of slowness was the standard template library itself. Most standard template library
|
||
|
|
classes cannot be forward declared and import thousands of lines of code into every compilation unit.
|
||
|
|
|
||
|
|
Another source of slowness is the need to include headers simply because the 'size' of the object must
|
||
|
|
be known. A new utility class allows you to 'forward declare' the size required for certain types which
|
||
|
|
allows you to remove their inclusion from the header file.
|
||
|
|
|
||
|
|
|