🚀 go-pugleaf

In article <1992Apr3.092216.14918@zooid.guild.org>, ross@zooid.guild.org (Ross Ridge) writes:
>> 
>> SORROW%oak.decnet@pine.circa.ufl.edu writes:
>> >Pertaining to REP STOSD, on my video card ( Diamond SpeedStart/+ w/1MB
>> >and Tseng Et4000ax ), I get roughly a 20% speed increase over 
>> >REP STOSW.  Althought I shouldn't in theory, in practice it works
>> >beautifully.
>> 
>> ross@zooid.guild.org (Ross Ridge) writes:
>> >I'll have to give it a try myself.  I was suprised to get a 100%
>> >speed increase using REP STOSW over REP STOSB.  Everything I had
>> >read up to that point said I shouldn't.
>> 
>> dme21270@uxa.cso.uiuc.edu (David M. Erazmus) writes:
>> >If you're using a 16-bit video card on an AT bus then you SHOULD get a 100%
>> >speed increase due to the fact that you're moving twice as much memory in
>> >parallel as with the 8-bit instruction.
>> 
>> The problem is that not all 16-bit VGA cards are created equal, up until
>> about a year ago no VGA card supported 16-bit writes to graphics memory.
>> Well at least that's what an article in DJJ said.  

Well, all VGA cards support 16-bit CPU read/writes.  The video card just separates the access into 2 8-bit accesses.  What you can't do is do video to video moves because of the way 16-bit moves are brocken into 2 operations.  Any word access that requires a read(modify)write works rather weird.  Typically the high-byte (the one loaded last) is written to both the low- and high-byte of the destination word.  

If your just reading or writing to the video buffer (but not both in the same instruction) the 16-bit access should work properly for EGA and VGA cards.
 
   cs911477@ariel.yorku.ca
   James Bennett