Previous: cadlinc-graphics-workstation-24-apr-82- Up: ../grinfo.html Next: dec-vt240-24-sep-85-


DEC-GIGI [22-Aug-81]

 The DEC GIGI terminal has 8 colors  or 8 grey shades and a screen  with
 resolution 768 horizontally by  240 vertically.  Price  with a black  &
 white monitor is about  $2800, and with a  color monitor, about  $3500.
 Through the courtesy of the local DIGITAL office, I was able to make  a
 test plot  on  the GIGI  to  verify  the correctness  of  the  <PLOT79>
 interface, and it might be of interest to record my experience.

 The plot chosen was  the 3-color plot DEM66  which shows a  hidden-line
 view on which contours are superimposed;  the surface is enclosed in  a
 box with labelled 3D axes, and  the contours are also displayed on  the
 top and bottom surfaces.  CPU time for construction of the plot on  the
 DEC-20/60 was 77 sec.  The plot was transmitted from our DEC-20/60  via
 a 1200-baud  dialup  connection,  and  took  approximately  fifty  (50)
 minutes to complete.  The plot file contained 19497 vectors represented
 in 329195 characters.  This gives an average transmission rate of about
 1100 baud.  Editing  with EMACS  to remove  NULs and  CR LF  characters
 reduced the  size to  245039 characters  (74.4% of  original).   During
 editing, I noticed the presence of duplicate vectors (due to coordinate
 roundoff into  the same  pixel); a  small utility  program showed  that
 there were 1545 duplicate vectors (7.9% of original).  Compacting could
 thus have  reduced the  file to  about 66%  of its  original size,  and
 elimination of  successive  points  on the  same  straight  line  would
 further reduce it by a undetermined amount.  The GIGI will not keep  up
 with transmission above  4800 baud,  so a reasonable  time estimate  at
 that speed (with file  compaction) might be  about 8 minutes.   Halving
 the grid size in the array could further reduce this to perhaps about 2
 minutes.  Clearly, such action would be necessary to make the  plotting
 time acceptable.

 The vector format of the GIGI looks something like

 !P[nnn,nnn]V[nnn,nnn]
 !V[nnn,nnn]

 for a  move then  draw, and  for a  subsequent draw,  respectively.   A
 similar verbosity is found in the Tektronix 4025 intelligent  terminal.
 Eleven characters (plus two more for CR LF) are on average sent for one
 visible vector.  The data format used  in the Tektronix 4010 series  is
 more  compact,  but  not  human-readable,  since  it  packs  a   10-bit
 coordinate  into  the  low-order  five  bits  of  two  characters;  the
 high-order two bits in each character  are used as flag bits, since  it
 is possible to omit unchanged  chunks.  It is therefore interesting  to
 compare with the plot file produced using the Tektronix 4010  interface
 and subsequently processed by the  TKVECS utility.  The 4010 file  took
 124932 characters (38%  of GIGI  file).  It contained  83296 NULs  (for
 delay time  padding at  9600-baud  transmission rates),  leaving  41636
 non-NUL characters, and had 11681 draw commands and 3443 move commands.
 There were 682  duplicate commands and  5578 points on  the same  line.
 Elimination of the  NULs indicates  that a vector  on average  requires
 about 2.75 characters, about 25% of the number required by GIGI or  the
 Tektronix 4025.  Thus, despite the awkward format of the Tektronix 4010
 vector file,  there  is a  substantial  storage and  transmission  time
 reduction.

 During plotting on the GIGI, I noticed that when a colored line crossed
 a white line, there appeared to be color bleeding into several adjacent
 pixels on the white line.  The  GIGI User's Manual revealed the  reason
 -- there are only 3  color bits and 1  blinking bit in graphics  memory
 for 12 adjacent  pixels on  a raster line.   This resulted  in the  red
 contour lines broadening into a 12-pixel wide band on the blue surface;
 when the lines run diagonally, the width on the screen is 2 to 3 mm.  I
 personally find this  quite objectionable, since  it seriously  reduces
 the resolution when color is used.