Shareware Extravaganza - Disc 4/Shareware Extravaganza - Over 25,000 Programs (The Ultimate Shareware Company)(Disc 4 of 4)(1993).iso/cad/grperf.zip/GRPERF.DOC

Text File | 1990-12-07 | 11KB | 243 lines

                               GRPERF v1.1


           AutoCAD(tm) Video Graphics Performance Benchmark Suite


                Copyright 1990 Off-Broadway Business Systems

                           (All Rights Reserved)


               Portions Copyright 1990 A/E Automation Systems




         AutoCAD and DXF are registered trademarks of Autodesk,
Inc.




GRPERF is a benchmark suite for measuring the relative performance
of graphics display hardware and drivers used in conjunction with
AutoCAD(tm), Release 11 and Release 10.

The tests performed by GRPERF are very detailed and are designed to
measure the draw performance of various types of geometry, and to
identify certain types of geometry which a driver or display board
is more capable of redrawing faster or slower than other types.
This benchmark was developed primarily for evaluating the "raw"
performance of hardware and video drivers, and for those involved
in developing same.  As a user, you may not find the results of
GRPERF to be very meaningful in terms of the kind of work you do,
but it will prove to be useful in evaluating the relative
performance of different display boards and display- list
processing ADI drivers.

Redraw tests.

There are nine tests performed by GRPERF.  Each test is composed of
a vector image that is exactly 1 x 1 drawing units in size.  The
redraw time of this image is measured at nine different display
magnifications, starting with a magnification that displays the 1
x 1 test image at 25 screen pixels, which doubles with each
successive magnification, ending with an image that is 256x larger
(1600 screen pixels) than the initial one.

At each of the nine display magnifications five redraws are
performed on the test image, and the time required to perform each
redraw is recorded.  Note that this time includes the time required
to first clear the display, as this operation is itself intrinsic
to the REDRAW operation.

For each of the five redraw times that are obtained at each
magnification, the highest and lowest times are discarded and the
three remaining times are then averaged together to produce a
single redraw time for each of the nine test magnifications.

Note that all times are only accurate to +/- 0.1 seconds. Once the
test is complete, the average redraw time of each of the nine
magnifications is then averaged together again, to produce a single
redraw time for the entire test.

Note that the test is designed to show variations in redraw time at
various different points, starting at one where the test image is
entirely within the view, and proceeding to a point where the view
contains only a portion of the image. You will probably note a
change in the redraw time at the point where the test image exceeds
the extents of the view, requiring the test image to be clipped to
the edge of the view.  This can give you a rough indication of the
performance of a display controller or an ADI driver's "trivial"
acceptance, rejection and line-clipping algorithms.

Benchmark performance times.

GRPERF contains built-in benchmark times for each
test/magnification as well as average times for all magnifications,
for all tests.  These times reflect the use of GRPERF with AutoCAD
Release 10 on a 25 Mhz. 386 PC with a standard VGA card running at
640 x 480, using the stock Autodesk-supplied VGA driver, with the
right-side SCREEN menu enabled, and where all redraw activity was
limited entirely to cached 32-bit system RAM without requiring any
disk I/O.

The benchmark reference times scale the performance of the system
on which GRPERF is run to provide a performance indicator or
"index" relative to the benchmark system performance as described
above.  This relative performance factor indicates the redraw
performance of the system being tested relative to the benchmark
system, where a value of 1.0 is equal to the performance of the
benchmark system, and 2.0 is twice as fast as the benchmark system.

Note that comparing a higher-resolution display configuration or
one with the AutoCAD SCREEN menu disabled (effectively increasing
the number of pixels in the view) may not be very meaningful in
terms comparing those results with other results using a different
display resolution.  In order to make a fair comparison between any
two test results, the dimensions of the view in screen pixels must
be exactly the same for all tests whose results are compared.

The redraw tests should be performed on a system with enough RAM to
maintain the test drawing's vector display list ENTIRELY in RAM,
without paging any portion of it to disk.  If a disk access is
required during the redraw test, then the test results should be
considered invalid and entirely meaningless.

Running GRPERF.

To execute GRPERF, place all files from the archives into a
directory in the, AutoCAD library search path, and REMOVE any
ACAD.LSP file so that it will not be loaded Automatically by
AutoCAD.  DO NOT attempt to run GRPERF with an any ACAD.LSP file
present in a path which permits it to be automatically loaded!

Then, at the DOS PROMPT:

   C:\ACAD > GRPERF [<redraw alias>]

Where the optional parameter <redraw alias> is the command that
initiates an AutoCAD or ADI driver redraw.  If your ADI display
driver uses the AutoCAD REDRAW command, then there is no need to
specify this option.  You will then be placed in the AutoCAD
drawing editor.  At this point, you can select one of the nine
tests by pulling down the left-most item on the pull-down menu:

    ┌─────┐
    │Tests│  Results   Options
    ├─────┴─────────────────┐
    │ Orthogonal lines      │
    │ Orthogonal linetypes  │
    │ Orthogonal polylines  │
    │ Random lines          │
    │ Random polylines      │
    │ Orthogonal text       │
    │ Rotated text          │
    │ Solid flood           │
    │ Circles               │
    │ --------------------- │
    │ Run ALL tests         │
    └───────────────────────┘
After you select a test, the drawing editor will be exited and re-
initialized and the test geometry will be generated.  If you have
not provided a hardware description for the test (a description of
the system or configuration being tested, in 30 characters or less,
which is included on the graph and tabular results), then you will
be prompted to specify this once. You can also change this
description by selecting the appropriate item from the "options"
pull-down menu.  The test description is stored in the file
GRCONFIG.TXT, and it will be used until you explicitly change it,
or erase the file.

After each test is complete, GRPERF displays the detailed results
in tabular form.  These results will also be posted in the file
GRPERF.LOG automatically for future reference.

Graphing test results.

After performing a test, GRPERF can draw a bar-graph representation
of the tabular results.  You can do this by selecting "graph test
results" from the second pulldown menu titled "results".  Once the
graph is generated you can then save it as either an AutoCAD
drawing file, or a slide, by selecting the appropriate item from
the "results" pulldown menu.

Graph drawings and slides are saved in filenames named "TESTn"
where 'n' is a number from 1 to 9 indicating the test number.  The
test numbers are the same as the order in which the corresponding
tests appear in the pulldown menu.

Running all tests in batch mode.

You can also have GRPERF run all nine tests sequentially and
automatically by selecting "run all tests" from the "tests" pull-
down menu.  When running all tests in batch mode, GRPERF
automatically generates the graph of each result and saves it as a
drawing file named in accordance with the conventions noted above.
It also posts tabular results from each test in the file GRPERF.LOG

When running all tests in batch mode GRPERF will also compute the
average redraw time and relative performance factor (VGA @ 25 Mhz
= 1.0) for each test and generate a bar graph of all test results,
and save the graph in the drawing file SUMMARY.DWG.

Altering test data.

Please note that the drawing and DXF(tm) files included with GRPERF
are not to be modified in any way.  These are the drawings which
are used to generate the test geometry.  They are named TEST1.DXF
thru TEST9.DXF and GRPERF.DWG, GR1.DWG, and GR2.DWG.  These .DWG
and .DXF files should NEVER BE MODIFIED for any reason whatsoever!
If they are, then GRPERF will not run, and no results that is
produced if it does run is valid.

A word about benchmarks.

All benchmarks can be somewhat deceptive, and accurate
representation of the true performance of any single aspect of two
dissimilar systems is extremely difficult to measure.  In addition,
results do not take into consideration a specific type of drawing
or discipline which can have a tremendous influence on the
effective performance that is achieved while using AutoCAD for
serious work, as opposed to running benchmarks.

GRPERF is primarily intended to help evaluate the 'raw' performance
aspects of graphics hardware and drivers, and is not designed to be
very meaningful as in indicator of performance as it may relate to
an end user and/or his or her specific application.  The tests
results can be easily misinterpreted due to the possibility of
their being influenced by many other factors other than video
performance which is almost impossible to compensate for.

In short, results should be never be considered as an accurate
indicator or reference standard by which the performance of any two
or more systems can be measured.  The use of this benchmark should
be limited to comparing AutoCAD graphics performance of different
graphics hardware or ADI drivers installed in the same system with
all other variables being precisely equal.

If you would like to compare the AutoCAD vector drawing performance
of two or more different video boards running at the same
resolution and with the same ADI display driver, then you should
swap the boards into the same system and run the benchmark on each
of them the same system, making sure that all other variables have
been eliminated.

Do not run the benchmark on different systems with different
graphics cards or drivers and expect the test results to be an
accurate indication of the raw performance of the hardware or the
drivers in each system.


Robert Schutz
Off Broadway
TEL [phone]
FAX [phone]
Compuserve  74435,16
Tony Tanzillo
A/E Automation Systems
Compuserve 71241,2067