High compilation time with gnat

High compilation time with gnat

[Joachim Schr�er]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 656 bytes --]

Hello,
we have a package spec which uses an enormous compilation time. One of the
specs it withes has about 40_000 lines of source. When I remember correctly
the gnat parser parses all specs in the closure of a given compuation unit.
This sum has a size of about 60_000 loc.
The compilation time seems to rise very nonlinear with the size of single
compilation units. The system is mainly paging to and from harddisc.
Is there a compilation switch for gnat to increase the amount of main memory
it uses or is there any other measure one may take?
Rewriting the source is no option, cause it is a delivery.

Thanks for any help
    Joachim Schr�er

Re: High compilation time with gnat

[Dale Stanbrough]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 832 bytes --]

Joachim Schr�er wrote:

> Hello,
> we have a package spec which uses an enormous compilation time. One of the
> specs it withes has about 40_000 lines of source. When I remember correctly
> the gnat parser parses all specs in the closure of a given compuation unit.
> This sum has a size of about 60_000 loc.
> The compilation time seems to rise very nonlinear with the size of single
> compilation units. The system is mainly paging to and from harddisc.
> Is there a compilation switch for gnat to increase the amount of main memory
> it uses or is there any other measure one may take?
> Rewriting the source is no option, cause it is a delivery.


How much RAM does your computer have? If you have insufficient
the performance will be bound by the hard drive transfer speeds.

Dale

-- 
dstanbro@spam.o.matic.bigpond.net.au

Re: High compilation time with gnat

[Georg Bauhaus]

"Joachim Schr�er" <joachim.schroeer@dornier.eads.net> wrote:
: Hello,
: we have a package spec which uses an enormous compilation time.

There has been a discussion of long compilation time which
is/was due to GNAT's way of doing things, in a few cases.
(It had to do with Matrices, and persons named dewar and brunot
have been involved, iirc.)

Re: High compilation time with gnat

[Volkert]

"Joachim Schr�er" <joachim.schroeer@dornier.eads.net> wrote in message news:<bq1ou1$1tcm0d$1@ID-76083.news.uni-berlin.de>...
> Hello,
> we have a package spec which uses an enormous compilation time. One of the
> specs it withes has about 40_000 lines of source. When I remember correctly
> the gnat parser parses all specs in the closure of a given compuation unit.
> This sum has a size of about 60_000 loc.
> The compilation time seems to rise very nonlinear with the size of single
> compilation units. The system is mainly paging to and from harddisc.
> Is there a compilation switch for gnat to increase the amount of main memory
> it uses or is there any other measure one may take?
> Rewriting the source is no option, cause it is a delivery.
> 
> Thanks for any help
>     Joachim Schr�er

You haven`t a service contract with ACT ???? 
I would ask them ...

Volkert

Re: High compilation time with gnat

[Craig Carey]

On 28 Nov 2003 00:11:47 -0800, Volkert wrote:

>On Wed, 26 Nov 2003 09:40:19 +0100, "Joachim Schr�er" wrote:
...
>> This sum has a size of about 60_000 loc.
>> The compilation time seems to rise very nonlinear with the size of single
>> compilation units. The system is mainly paging to and from harddisc.

Any child packages ?.

>> Is there a compilation switch for gnat to increase the amount of main memory
...

>You haven`t a service contract with ACT ???? 
>I would ask them ...
>

Can't you ask on behalf of the first author?.
If ACT does not reply then you could cancel your own contract.
Then still the number of clients would alter by 1.

This is comp.lang.ada and 2 line responses giving seemingly bad advice
lacking the needed argument, is not ideal.


On Wed, 26 Nov 2003 12:00:31 GMT, Dale Stanbrough wrote:
>How much RAM does your computer have? If you have insufficient
>the performance will be bound by the hard drive transfer speeds.


I suppose that GNAT compiler doubles the size of its tables which
could be just about optimal. It seems that more RAN chips may be
appropriate if the CPU is slowed by attending to paging.

There is this gnatmake option (which may not speed things up):

  -gnatTnnn :  All compiler tables start at nnn times usual starting size

(Possibly ASIS (Sigada) could aid in the rearranging of any large Ada 95
program.

I had a look and I failed to find the message of Mr Brunot and Dewar
when using the groups.google.com search engine.


Craig Carey

Re: High compilation time with gnat

[Georg Bauhaus]

Craig Carey <research@ijs.co.nz> wrote:
: I had a look and I failed to find the message of Mr Brunot and Dewar
: when using the groups.google.com search engine.

The memory is fading, I haven't been too successful myself, sorry.
However, I have noted (or at least I think I have) that nested
packges have resulted in compilation times in excess of what the
the same packages as child packages require (the packages contain
a few instantiations of Charles containers).

Another source of long compilation time is displayed in

   procedure PT is
      X : constant := 1.0 + 2.0 ** (-1000000);
      Y : constant := 1.0 + 2.0 ** (-1000001);
      Z : constant := 2.0 ** (1000001) * (X - Y);
   begin
      Put_Line (Float'Image (Z));
   end PT;

which was posted with smaller powers by Laurent Guerby on
2000-09-10.



-- Georg

Re: High compilation time with gnat

[Alexandre E. Kopilovitch]

Georg Bauhaus <sb463ba@l1-hrz.uni-duisburg.de> wrote:

> Craig Carey <research@ijs.co.nz> wrote:
> : I had a look and I failed to find the message of Mr Brunot and Dewar
> : when using the groups.google.com search engine.
>
> The memory is fading, I haven't been too successful myself, sorry.

Perhaps it was not in c.l.a, but in GNAT-chat mailing list (when it was open
for general public). At least I vaguely remember quite long (and somehow heated)
exchange in that list between N. Brunot and Robert Dewar, at the end of which
the GNAT-chat mailing list was restricted to ACT customers only.



Alexander Kopilovitch                      aek@vib.usr.pu.ru
Saint-Petersburg
Russia