SvgaLib/doc/man6/accel.6

.TH accel 6 "29 July 1997" "Svgalib (>= 1.2.11)" "Svgalib User Manual"
.SH NAME
accel \- tests the new style svgalib accelerator interface
.SH SYNOPSIS

.B accel

.SH DESCRIPTION
Test new-style accelerated functions (As of this writing: Ark, Cirrus, Chips &
Technologies cards, and Mach32 only).
For other cards the demo will not work (well it will complain about missing accelerator
support). Don't worry about this.

During the development of the Mach32 new style driver for 1.2.12, this demo was massively
extended to check the Mach32 functions.

Upon startup it lists all supported SVGA modes and asks you to enter a number
identifying the mode to be tested. The supported subfunctions of
.BR vga_accel (3)
in this mode are listed and the demo instructs to press
.B <Return>
to start the demos.

If supported, all drawing operations are performed in the background.

Then the following tests are performed:
.SS Positioning tests
These tests were originally intended to check that the accelerator commands work on the
proper screen locations. The screen shows 12 (4 x 3) smaller areas with red crosses in the
corners. When everything is ok, the drawings should reach right in the corners of the crosses.

A given card may not support all operations listed here. In that case the resp. test area
just shows the red crosses. For tests performed, the name of the test is printed below the
area. The tests are (from left to right, top to bottom):

.IP 1.
A green box is drawn with
.BR vga_accel(ACCEL_FILLBOX) .
.IP 2.
A cross of green lines is drawn with
.BR vga_accel(ACCEL_DRAWLINE) .
.IP 3.
A linux pixmap just fitting into the crosses is drawn with
.BR vga_accel(ACCEL_PUTIMAGE) .
.IP 4.
A pixmap just fitting into the crosses is drawn into the red crosses (by
.BR vgagl (5)
which may or may not use the accelerator). The pixmap is then copied to a few lines/columns
below. Green crosses mark the intended destination position.
.IP 5.
Works like 3. but copies to an area above the origin. The accelerator must ensure that the
overlapping areas are handled non corrupting. Thus, watch that the copy operation is properly
performed.
.IP 6.
A green triangle is drawn above the top/left to bottom/right diagonal by use of
.BR vga_accel(ACCEL_DRAWHLINELIST) .
.IP 7.
Certain bitmaps are copied to the screen. In the corners you'll see the digits
.RS
.IP 0
top/left, green on red.
.IP 1
top/right, red on green.
.IP 2
bottom/left, black on white.
.IP 3
bottom/right, white on black. Note that some black border, not the digit will be aligned 
to the red crosses.
.RE
.IP
Finally, a yellow wizard image is drawn into the center.

The bit ordering for bitmaps is a bit weird. Please check that the digits are not mirrored or
flipped.
.IP 8.
This time bitmap transparency is tested by drawing wizard images onto the aforementioned
linux pixmap (left to right, top to bottom) in yellow, red, green, and cyan. The background
of the yellow wizard is masked out by a black border bitmap.
Note that the wizard will not reach into the red corners because the bitmap has some
(transparent) border.
.IP 9.
The text below this box is copied as a monochrome bitmap from the screen into the corners
listed under 7. in the same colors.
.IP 10.
Two green rectangles with an edge cut out from the bottom is drawn using
.BR vga_accel(ACCEL_POLYLINE) . There is an intended bug which draws the center of the
top line twice. If supported, the lower rectangle is drawn in cyan and with the xor
raster operation s.t. the buggy point is not drawn thus leaving a pin hole.
.IP 11.
.BR vga_accel(ACCEL_POLYHLINE) is used to draw some green lines which makes this area look
like a green box with a cut out, black M-style shape.
.IP 12.
A weird green polygon is filled in red with
.BR vga_accel(ACCEL_POLYFILLMODE)
using the techniques given in
.BR vga_accel (3).
This needs some offscreen memory. If VGA memory is tight in that resolution the test cannot
be performed.
.PP
After this screen, you'll have to hit
.B <Return>
to continue.
.SS Raster operations
Again, red cross bordered areas are drawn on the screen, this time for each of the supported
raster operations. For
.BR ROP_AND " and " ROP_XOR
the areas are filled in white first.

Three overlapping boxes
.BR A ", " B ", " C
are drawn such that you see the following areas.

.RB AAAAAA dddd BBBBBB
.br
.RB AAAAAA dddd BBBBBB
.br
.RB AAAAAA dddd BBBBBB
.br
.RB AAAA ee gggg ff BBBB
.br
.RB AAAA ee gggg ff BBBB
.br
.RB AAAA ee gggg ff BBBB
.br
.RB "    CCCCCCCC"
.br
.RB "    CCCCCCCC"
.br
.RB "    CCCCCCCC"

The pictures should show:
.IP 1.
Replace mode. A, B, C are red, green, blue. They just overlap, yielding
d - green and e, f, g - blue.
.IP 2.
The colors mix using
.B ROP_OR
(and a nice color table). The overlapping areas become the additive color mix:
d - yellow, e - magenta, f - cyan, and g - white.
.IP 3.
.B ROP_AND
is used. The background is filled white first, s.t. there is something in video memory
to and with non trivially. We have A, B, C in cyan, magenta, yellow and d, e, f, g
in blue, green, red, black.
.IP 4.
.B ROP_XOR
is used and the background filled white first too. A, B, C are red, green, blue again, but
the overlapping areas d, e, f, g  become blue, green, red, white.
.IP 5.
.B ROP_INV
is used, s.t. A, B, C are all white and d, e, f, g  become black, black, black, white.
Note that this is not done by using
.B ROP_XOR
and drawing A, B, C in white. Instead A, B, C are drawn in the usual
 red, green, blue. However, the accelerator just inverts the memory contens.
.PP
If the accelerator supports raster operations for
.B ACCEL_DRAWHLINELIST
actually disks (well, ellipses) are drawn instead of boxes.
.PP
After this screen, you'll have to hit
.B <Return>
to continue.

.SS Replace QuixDemo
If
.B ACCEL_DRAWLINE
is supported, a Quix like bouncing series of lines in varying colors is drawn. The lines
are removed from the screen by overdrawing them in black, thus erasing the dots and text
on the background.

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS XOR Mode QuixDemo
As before, but this time all lines are drawn in
.B ROP_XOR
mode (if
.B ACCEL_DRAWLINE
supports raster operations). Thus the background will not be destroyed this time.

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS FillBox Demo
The screen is
.BR ACCEL_FILLBOX
filled with a series of boxes of increasing colors.
In truei/high color modes you'll probably only see a series of varying blue tones (because these
are at the beginning of the color table and there are soo many of them).

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS ScreenCopy Demo
Some random dots are drawn on the screen and thirds of the screen contents are moved
around using
.BR ACCEL_SCREENCOPY .

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS Scroll Demo
Some random dots are drawn on the screen and moved one line up with
.BR ACCEL_SCREENCOPY .
In offscreen memory a new line is prepared which will be cleared by
.BR ACCEL_FILLBOX and move in from below. This test requires some offscreen and will not
be performed if video memory is very tight.

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS FillBox with DrawHLineList Demo
Like the FillBox test, but no box fill is done but the screen is filled with a list of
horizontal lines drawn with
.BR ACCEL_DRAWHLINELIST .

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS FillBox XOR Mode Demo
Like the FillBox test, but the XOR raster operation is used.

The test lasts about 5 seconds and some statistics are printed to stdout.

.SS PutBitmap Demo
The screen is filled with bitmasks consisting of tiny vertical lines alternating in red and
blue.

The test lasts about 5 seconds and some statistics are printed to stdout.

.SH SOME DATAPOINTS
Here is a list of speed listings for some cards. Please keep in mind that also the
calling overhead for the program is measured. This seems to be esp. true for the
QuixDemo.

.SS "Results on a Cirrus GD5434-E with 2Mb:"
.TP
.B "640x480x256 60 Hz"
.br
FillBox: 200.3 Mpixels/s (200.3 Mbytes/s)
.br
ScreenCopy: 51.0 Mpixels/s (51.0 Mbytes/s)
.br
Scroll Demo: 50.5 Mpixels/s (50.5 Mbytes/s)
.br
FillBox XOR: 83.2 Mpixels/s (83.2 Mbytes/s)
.TP
.B "320x200x256 70 Hz"
.br
FillBox: 200.1 Mpixels/s (200.1 Mbytes/s)
.br
ScreenCopy: 52.3 Mpixels/s (52.3 Mbytes/s)
.br
Scroll Demo: 51.2 Mpixels/s (51.2 Mbytes/s)
.br
FillBox XOR: 87.1 Mpixels/s (87.1 Mbytes/s)
.TP
.B "640x480x32K 60 Hz"
.br
FillBox: 90.9 Mpixels/s (181.8 Mbytes/s)
.br
ScreenCopy: 23.1 Mpixels/s (46.3 Mbytes/s)
.br
Scroll Demo: 23.0 Mpixels/s (46.1 Mbytes/s)
.br
FillBox XOR: 37.2 Mpixels/s (74.5 Mbytes/s)
.TP
.B "640x480x16M (32-bit) 60 Hz"
.br
FillBox: 35.5 Mpixels/s (142.3 Mbytes/s)
.br
ScreenCopy: 9.3 Mpixels/s (37.3 Mbytes/s)
.br
Scroll Demo: 9.2 Mpixels/s (37.1 Mbytes/s)
.br
FillBox XOR: 14.6 Mpixels/s (58.6 Mbytes/s)

.SS "On a Cirrus Logic 5426 VLB (50 MHz MCLK):"
.TP
.B "640x480x256 60 Hz"
.br
FillBox: 32.8 Mpixels/s (32.8 Mbytes/s)
.br
ScreenCopy: 16.4 Mpixels/s (16.4 Mbytes/s)
.br
Scroll Demo: 16.3 Mpixels/s (16.3 Mbytes/s)
.br
FillBox XOR: 16.5 Mpixels/s (16.5 Mbytes/s)
.TP
.B "640x480x32K 60 Hz"
.br
FillBox: 12.2 Mpixels/s (24.4 Mbytes/s)
.br
ScreenCopy: 6.1 Mpixels/s (12.2 Mbytes/s)
.br
Scroll Demo: 6.0 Mpixels/s (12.1 Mbytes/s)
.br
FillBox XOR: 6.1 Mpixels/s (12.2 Mbytes/s)

.SS "Tweaked to 60 MHz MCLK:"
.TP
.B "640x480x256 60 Hz"
.br
FillBox: 42.1 Mpixels/s (42.1 Mbytes/s)
.br
ScreenCopy: 21.0 Mpixels/s (21.0 Mbytes/s)
.br
Scroll Demo: 20.9 Mpixels/s (20.9 Mbytes/s)
.br
FillBox XOR: 21.1 Mpixels/s (21.1 Mbytes/s)
.TP
.B "640x480x32K 60 Hz"
.br
FillBox: 16.7 Mpixels/s (33.5 Mbytes/s)
.br
ScreenCopy: 8.3 Mpixels/s (16.7 Mbytes/s)
.br
Scroll Demo: 8.3 Mpixels/s (16.7 Mbytes/s)
.br
FillBox XOR: 8.3 Mpixels/s (16.7 Mbytes/s)
.SS Results on a Mach32 EISA with 2Mb VRAM:
.TP
.B 1280x1024x256 60 Hz
Replace QuixDemo: 12.1 Klines/s (6.7 Mpixels/s or 6.7 Mbytes/s)
.br
Xor QuixDemo: 9.9 Klines/s (5.1 Mpixels/s or 5.1 Mbytes/s)
.br
FillBox: 75.4 Mpixels/s (75.4 Mbytes/s)
.br
ScreenCopy: 26.4 Mpixels/s (26.4 Mbytes/s)
.br
Scroll Demo: 28.7 Mpixels/s (28.7 Mbytes/s)
.br
FillBox with DrawHlineList: 73.1 Mpixels/s (73.1 Mbytes/s)
.br
FillBox XOR: 37.9 Mpixels/s (37.9 Mbytes/s)
.br
PutBitmap: 15.6 Mpixels/s (15.6 Mbytes/s)

.TP
.B 1024x768x64K 72Hz
Replace QuixDemo: 12.3 Klines/s (5.2 Mpixels/s or 10.5 Mbytes/s)
.br
Xor QuixDemo: 9.0 Klines/s (5.1 Mpixels/s or 10.3 Mbytes/s)
.br
FillBox: 37.6 Mpixels/s (75.2 Mbytes/s)
.br
ScreenCopy: 13.2 Mpixels/s (26.4 Mbytes/s)
.br
Scroll Demo: 13.2 Mpixels/s (26.4 Mbytes/s)
.br
FillBox with DrawHlineList: 37.0 Mpixels/s (74.0 Mbytes/s)
.br
FillBox XOR: 18.9 Mpixels/s (37.8 Mbytes/s)
.br
PutBitmap: 15.2 Mpixels/s (30.5 Mbytes/s)

.PP
You're encouraged to send in more data.
This demo is part of svgalib and can be found in the
.I demos/
subdirectory of the original svgalib distribution. However, it is not installed in the system
by default, s.t. it is unclear where you can find it if your svgalib was installed by some
linux distribution. Even then, when you have the demo on your system, you probably won't have
the sources s.t. it is only of limited use for you.

In case of any such problem, simply get an svgalib distribution from the net. You even
don't need to install it. Just
.B make
in the
.I demos/
subdirecty. As of this writing,
.I svgalib-1.2.12.tar.gz
is the latest version and can be retrieved by ftp from
.IR "sunsite.unc.edu" " at " "/pub/Linux/libs/graphics"
and
.IR "tsx-11.mit.edu" " at " "/pub/linux/sources/libs"
which will most probably be mirrored by a site close to you.

.SH SEE ALSO

.BR svgalib (7),
.BR vgagl (7),
.BR libvga.config (5),
.BR vga_accel (3),
.BR threed (6),
.BR bg_test (6),
.BR eventtest (6),
.BR forktest (6),
.BR fun (6),
.BR keytest (6),
.BR mousetest (6),
.BR scrolltest (6),
.BR speedtest (6),
.BR spin (6),
.BR testaccel (6),
.BR testgl (6),
.BR testlinear (6),
.BR vgatest (6),
.BR plane (6),
.BR wrapdemo (6)

.SH AUTHOR

This manual page was edited by Michael Weller <eowmob@exp-math.uni-essen.de>. The
demo and most of its documentation is due to
Harm Hanemaayer <H.Hanemaayer@inter.nl.net>.