Re: Gnat on OpenVMS

["Thierry Lelegard" <lelegard@club-internet.fr>]

> We also ported our application from OpenVMS+DECAda to Digital UNIX +
> DECAda to Digital UNIX + gnat. We found that the performance was
> degraded by a factor of 2 or 3 between DECAda and gnat. Have you sen
> anything of the sort with your port ?

Yes, roughly. However, I must admit that our tests are not
completed yet. Before testing the speed, we must get a
functional application, which we did not achieve yet...

Among nine large applications, only one works with gnat
at the present time. The source code is the same at the
working code with DEC Ada (with Ada 83/95 incompatibilites
fixed, at least those that the compiler can see, maybe
not all subtle behavior differences).

You observed a performance degradation by a factor of 2 or 3.
I suppose that this can be easily explained: the size of the
code is larger by the same factor.

For instance, we observed the following sizes:
- EXE generated with DEC Ada, no debug info: 9,000 blocks
- EXE generated with GNAT, no debug info: 19,000 blocks
- EXE generated with GNAT, with debug info: 32,000 blocks

I was forced to look at the generated code because some
programs experienced strange Program_Error (not elab pb) and
GDB failed when we tried to debug it. So, since I was crazy
enough in the past to work with Alpha assembler, I decided
to find the pb by reading the .S file. The code is uncredibly
unefficient, even with /optim=all: no visible instruction
rescheduling (critical issue on Alpha), "strange" usage
of registers and memory, redundant branches, etc. I have
often looked at GEM-generated code in the past, optimized
gnat code looks like non-optimized gem code.

Anyway, all of these issues may be improved in the future,
although the effort to improve the efficiency of a code
generator is huge.

Our immediate concern is to get a working environment
(I mean compiler/debugger, not the applications) and
ACT is currently working on this (we already got a new
version of the compiler, so they really work on it!).

Daniel, did you use GDB on OpenVMS? Did you run into
specific problems? 

I am not in love with the OpenVMS debugger, I just appreciate
its functionalities. If GDB can achieve the same features,
even with a different user interface (I don't care),
I will be delighted to use it.

Note: please answer to lelegard@canal-plus.fr, not the
"reply-to" address of this note.
________________________________________________________
Thierry Lelegard, Paris, France
E-mail: lelegard@club-internet.fr