The World of Computer Software/World_Of_Computer_Software-02-386-Vol-2of3.iso/b/bench12.zip/BENCH12.DOC

Text File | 1992-12-22 | 9KB | 212 lines

Doc for BENCH12.SCR 12/21/92

How to run this benchmark.---------------------------------------------------

First, completely configure AutoCAD for the driver environment
to be tested.  Use either the driver's configuration program
and/or the AutoCAD video configuration to set this up.  Ensure
that 32-bit display space operation is enabled in the driver, if
it has this option.

NOTE:  This benchmark will not work with any driver that only
supports 16-bit display space, as these drivers are not true ADI
4.2 drivers and do not provide you, as an end-user, with all the
features of AutoCAD R12.

After the driver has been configured, return to DOS.

All of the Autodesk supplied drivers will perform this
benchmark.  Additionally, the ACAD.DWG prototype drawing used
for this trial should be the default drawing shipped by
Autodesk.

Starting from the directory where ACAD.EXE is located, enter
AutoCAD.  You might want to use the STATUS command to note both
the amount of available memory and the amount of memory used by
the driver without a drawing loaded.

Use the SCRIPT command to invoke BENCH12.SCR.  This benchmark
will take anywhere from two minutes up to an hour, depending on
your system speed, amount of memory and the driver used. On a
486/66 DX/2 with 16Mb of memory, Autodesk's standard VGA driver
takes just under five minutes, and requires about 13.5Mb of
memory.

Remember, the amount of memory you have will determine how much
disk access is required during this benchmark. Disk access speed
is tremendously slower than RAM access speed, so more RAM equals
more overall speed.



The files used.--------------------------------------------------

This script file tests driver performance using 32-bit display
list with Release 12.  It must be run from the ACAD directory
(or the main directory where ACAD.EXE is located).  To find the
following files (SAMPLE\PSSHADOW.DWG, SAMPLE\HOUSEPLN.DWG,
SUPPORT\CHROMA.DWG), this benchmark assumes you have followed
the standard AutoCAD installation.

You must also have read/write access to the directory where this
script was started from.  The script file creates a temporary
file, $$$$TEMP.EPS, for the postscript input/output functions.

NOTE: The only place any of the files are referenced is at the
beginning of the script file. This has been done so that you may
easily change path names to enable this benchmark to function on
a network or customized system.

You should have at least 14Mb of free memory for AutoCAD to use
when running this benchmark.  The Autodesk STANDARD VGA 640x480
16 color driver uses about 13.5Mb to run this script. Other
drivers, especially display list, will require more.

The amount of extra memory required by other display drivers
will make a difference in the performance of this benchmark. As
mentioned before, the major differences in speed are due to the
radically different access times encountered while retrieving
information from either the hard disk or the RAM.

NOTE: Retrieving information from the RAM will be faster than
retrieving information from the hard disk.


What this benchmark measures.----------------------------------

This benchmark was developed at request of the many users, users
groups, graphics board manufacturers, and reviewers who needed a
benchmark that demonstrated actual AutoCAD performance as
affected by display list drivers.

Traditional driver benchmarks only show the benefits of faster
zoom and pan operations, but they ignore the side-effects that a
display list driver may bring to the system, even during the
most normal editing operations.

This new benchmark is meant to demonstrate how a display list
driver can affect an AutoCAD system in a working environment --
it is not simply a test structured to show off the few benefits
available from many display list drivers. If overall AutoCAD
performance is not enhanced, then a display list driver is not
providing a benefit for your system.

The main objective of this benchmark is to show the overall,
performance gains or losses when using a display list driver.
Many of the functions performed in this benchmark typically
cause display list drivers to increase their memory usage
dramatically.

With this in mind, the BENCH12 benchmark may expose poor
implementations of display list editing and memory management.
Additionally, the display list operations typically enhanced by
benchmarks (i.e. zooming and panning) are also performed to show
traditional display list gains.

The second objective of this benchmark was to avoid generating a
regen.  Autodesk is promoting the 32-bit display space built
into AutoCAD R12 and its dramatic reduction of regens. This
script will not generate a regen (after the initial Zoom
Extents) on any driver that is fully ADI 4.2 compatible and
supports 32-bit display space.

The third objective of this benchmark was to test the driver
with 32-bit display list objects. This was done to ascertain if
the driver being tested would perform correctly while scaling
and clipping. Without proper handling of these objects, many
display list drivers exhibit erratic behavior up to and
including random system crashes.


About the Script.----------------------------------------------

This script file first inserts PSSHADOW.DWG, uses PSOUT to
create a temporary .EPS file, and imports it with PSIN to get
the shaded image. This portion is not timed since it is
virtually 100 percent CPU and hard disk related.

The drawing is generated by moving the shaded image, importing
the HOUSEPLN.DWG drawing, importing the HOUSEPLN.DWG drawing at
a reduced scale, importing the CHROMA drawing, and importing the
CHROMA drawing at a reduced scale.  The total time for this
section is reported at the end.

A ZOOM E(xtents) is performed (this causes a regen) and the time
is recorded.

All inserted blocks are EXPLODEd and a donut is created in the
corner.  This time is also reported at the end.

A series of ZOOM commands are used to test the classic display
list benefits with the time being reported at the end.
Accelerated boards should benefit from this section and should
have lower times than non-accelerated boards.  About 28 bits of
display list space (out of the 31 available) are used to prevent
regens at very high resolutions and other aspect ratio
considerations.  The donut created in the Zoom E(xtents) section
is used to check polygon clipping.

A number of edits are performed (i.e. ROTATE, UNDO, PEDIT,
PLINE, CIRCLE, DONUT, MOVE and ERASE).  This section contains
redraws after most edits, simulating a typical user cleaning up
the pick blips.  This time is reported at the end.

After the above test sequence, a small drawing is created
tallying the results.

A STATUS command is also executed to show the amount of virtual
memory used by AutoCAD and the driver. This is important because
some drivers require an excessive amount of RAM during editing
sessions. This excessive use of RAM means more hard disk
accessing and slower overall system performance.


What to look for.-----------------------------------------------

If a regen occurs, most likely the driver does not implement the
full 32-bit display space and is an ADI 4.1 driver (Release 11)
pretending to be an ADI 4.2 driver.  These drivers will most
likely stop in the middle of the script, because AutoCAD was
unable to pick an object due to the reduced display space forced
by the driver.

During this benchmark, pay attention to the redraws as they may
be showing previous edits.  These previous edits may appear as
temporary, or "flashing" ghosts.  If so, then the driver is not
performing real-time display list editing.

When objects are moved or erased, you may see "left-over-bits"
or pixels.  A driver that does not properly match highlighted
lines with normal lines will leave these annoying artifacts.

Additionally, if the "Virtual Memory Allocated to Program" (use
STATUS command) grows more then a couple of megabytes from when
the drawing was created (post-ZOOM E command), then the driver
is not releasing or re-using erased display list memory.  Some
drivers may cause the system to access the hard disk excessively
due to the amount of memory required.

If your system hangs, locks, reboots, gets an error message (i.e
Divide By Zero, Abnormal Program Termination, Fatal Error,
etc...) then it is the driver that has caused the problem.

First, run the benchmark with the Autodesk supplied VGA driver
to get a feel for what the display should look like during
execution. Then, run the benchmark with each of the drivers you
are testing to see if they are operating correctly.


The future.---------------------------------------------------

Some of the areas that this script does not address are:
Multiple Viewports, Paper Space, AVE Render, Grid, Command Area
text speed and dialog-box speed. Overall, however, this
benchmark has proven to be a reasonable attempt to reproduce a
working environment that is using the enhanced features provided
by AutoCAD R12.

If you have any questions or comments concerning this benchmark,
please FAX them to us at [phone].

-- end of BENCH12.DOC --