From: sms@antinode.org (Steven M. Schweda)

> GnuPG 1.4.9 exists now. Looks like a minor bug-fix release. I made
> no new VMS-specific changes:
>
> http://antinode.org/dec/sw/gnupg.html


That page now leads to version 1.4.9b, where I _have_ made some new
VMS-specific changes. I believe that I've located the major file open
operations, and have added the usual access callback function to adjust
some RMS parameters to bigger values (unless the user overrides using
SET RMS_DEFAULT). The callback function also sets the SQO (sequential
access only) bit for output files, which should make the program a
better citizen on a disk with highwater marking enabled. I also changed
the file locking code for the random seed file (normally,
"/SYS$LOGIN/gnupg/random_seed"). I believe that this will stop the
annoying accumulation of multiple versions of this file.

Using a 2.5GB test file (four copies of a VMS V7.3-2 installation
CD-ROM image) on a zx2000 (IA64) workstation (more idle, and less
CPU-bound, than anything else here), I got some detectable speed
improvements. The (compressed and) encrypted file is about 1GB.
Typical SHOW PROCESS /ACCOUNTING results follow.

Encrypt:

1.4.9a Accounting information:
Buffered I/O count: 33807 Peak working set size: 8944
Direct I/O count: 608383 Peak virtual size: 182512
Page faults: 491 Mounted volumes: 0
Images activated: 2
Elapsed CPU time: 0 00:54:41.24
Connect time: 0 01:13:58.02

1.4.9b Accounting information:
Buffered I/O count: 865 Peak working set size: 9392
Direct I/O count: 136807 Peak virtual size: 182832
Page faults: 519 Mounted volumes: 0
Images activated: 2
Elapsed CPU time: 0 00:49:01.18
Connect time: 0 00:51:31.75

Decrypt:

1.4.9a Accounting information:
Buffered I/O count: 80047 Peak working set size: 8160
Direct I/O count: 383534 Peak virtual size: 182064
Page faults: 442 Mounted volumes: 0
Images activated: 2
Elapsed CPU time: 0 00:05:35.05
Connect time: 0 00:55:03.79

1.4.9b Accounting information:
Buffered I/O count: 1016 Peak working set size: 8624
Direct I/O count: 57116 Peak virtual size: 182672
Page faults: 468 Mounted volumes: 0
Images activated: 2
Elapsed CPU time: 0 00:04:53.99
Connect time: 0 00:08:57.77

With two points, a line always fits, but it looks as if the reduction of
"Connect time" in this case is roughly proportional to the size of the
output file.

I haven't tested this kit much more than usual, but simple
encrypt-decrypt operations seem to have survived, so what could go
wrong? Complaints are always welcome.

------------------------------------------------------------------------

Steven M. Schweda sms@antinode-org
382 South Warwick Street (+1) 651-699-9818
Saint Paul MN 55105-2547