• Hxal <hxal freenode.d.channel> Dec 22 2007
• "Vladimir Panteleev" <thecybershadow gmail.com> Dec 22 2007
• "Janice Caron" <caron800 googlemail.com> Dec 22 2007
• "Vladimir Panteleev" <thecybershadow gmail.com> Dec 22 2007

Hxal <hxal freenode.d.channel> writes:

Leandro Lucarella Wrote:

 According to phobos changeset 535[1] I guess we have a new 'manifest'
 keyword to define manifest (to be redundant) constants. I really think he
 should went with alias or macro or with other better *existing* keyword,
 but is definitely better than enum.
 
 [1] http://www.dsource.org/projects/phobos/changeset/535


I wonder if now that we have "manifest" meaning "taking no storage space",
it could be made to work for functions too (inlining every call to the
function).

"Vladimir Panteleev" <thecybershadow gmail.com> writes:

On Sat, 22 Dec 2007 15:21:55 +0200, Hxal <hxal freenode.d.channel> wrote:

 Leandro Lucarella Wrote:

 According to phobos changeset 535[1] I guess we have a new 'manifest'
 keyword to define manifest (to be redundant) constants. I really think he
 should went with alias or macro or with other better *existing* keyword,
 but is definitely better than enum.

 [1] http://www.dsource.org/projects/phobos/changeset/535


 I wonder if now that we have "manifest" meaning "taking no storage space",
 it could be made to work for functions too (inlining every call to the
function).


That would be great, but I think macros will do that better.
I think it's noteworthy that some languages like Delphi have an "inline"
keyword for functions (and it works across units/modules).

-- 
Best regards,
 Vladimir                          mailto:thecybershadow gmail.com

"Janice Caron" <caron800 googlemail.com> writes:

On 12/22/07, Hxal <hxal freenode.d.channel> wrote:
 I wonder if now that we have "manifest" meaning "taking no storage space",
 it could be made to work for functions too (inlining every call to the
function).


Why would you want to do that? I don't see why the programmer should
need to know or care what gets inlined and what doesn't.

"Vladimir Panteleev" <thecybershadow gmail.com> writes:

On Sat, 22 Dec 2007 19:08:37 +0200, Janice Caron <caron800 googlemail.com>
wrote:

 On 12/22/07, Hxal <hxal freenode.d.channel> wrote:
 I wonder if now that we have "manifest" meaning "taking no storage space",
 it could be made to work for functions too (inlining every call to the
function).


 Why would you want to do that? I don't see why the programmer should
 need to know or care what gets inlined and what doesn't.


1) The compiler can't know which code is in more need of performance - this
gives more control over performance to the user. Also, inlining doesn't just
imply copying the machine code - the function call is substituted at the AST
level, which means that it allows optimizations to happen across function
boundaries.

2) Sometimes it's very useful for security. Reverse-engineering a 500kb
function of mostly inlined cryphographic and mathematical operations is hell,
and doesn't have any substantial negative effects (the worst part is the huge
stack frame).

-- 
Best regards,
 Vladimir                          mailto:thecybershadow gmail.com