Image size on VAX

[Fekko Stubbe]

I have a problem with an executable-size on VAX.

When I compile/link the program (some 55 files) on alpha the image size
is about 3500 blocks.

The same sources compiled on vax gives me an image of 24000 blocks.
Looking at the image I see there are a lot of zero pages (upto 16000 contiguous pages).
I wonder why the linker does not create a new demand zero image section.
When I add the DZRO_MIN=1 to the linker options file, the size does not change.
Adding ISD_MAX=96 to the linker options file does not make any difference.

The number of image sections is 27, so the must be room enough for more.

PROCSECTCT is set to 128

[Hein van den Heuvel]

Try linking with an option file and the option:

UNSUPPORTED = 1 ! force demand zero pages

Groetjes,
Hein.

[Fekko Stubbe]

Hein,
Good shot, but it did not work, the image is still some 24000 blocks

[Hein van den Heuvel]

Along with remembering the UNSUPPORTED option I seem to vaguely remember something about their being a 16 bit limit in the maximum number of Dzero pages. There may be a threshold at 65K.
That would mean that when you need it most, you don't get it!
Just for an experiment can you reduce the need under 65K pages and see if it gets squished?

Can you divide the space up in clusters?

Ultimately you may be better of using some GETVM / MALLOC / CRMPSC scheme to just grab that memory on the fly.

Hein.

[Fekko Stubbe]

After increasing ISD_MAX to 256 the image came down to some 2800 blocks.
There seems to be a discrepancy between the map (telling me in the image synopsys)
Number of image sections: 222

And the anal/image looking for "image section"
pipe ana/ima dix_vax.exe |search sys$input "image section"/noout/stat
...
Records matched 155.

In the old (big image) the map said
Number of image sections : 95
and ana/ima said "27"

The 95 is close to the default isd_max of 96.

So summarizing : adding the ISD_MAX=256 solved the problem.

[Fekko Stubbe]

Increasing isd_max in the linker options file solved the problem