nanogui: Thread: Re: Herc in Microwin

[hercules ... snip]

: You said that hline and vline need to be rewriten. I am volintering for the 
: task (unless someonee has allready done it).
: 
	Great!

: I looked at the code and at the moment it just uses a loop to draw the line 
: pixel by pixel. How does it need to be improved??
: 
	Are you familiar with the Herc memory/screen layout?  Instead of
looping for each bit, the C code needs to set up to 8 pixels at a time with a
single MOV instruction.  This entails pre-calculating the bits in the screen memory
and writing it once.

Greg

>: I looked at the code and at the moment it just uses a loop to draw the line 
>: pixel by pixel. How does it need to be improved??
>: 
>	Are you familiar with the Herc memory/screen layout?  Instead of
>looping for each bit, the C code needs to set up to 8 pixels at a time with a
>single MOV instruction.  This entails pre-calculating the bits in the screen memory
>and writing it once.

 Oh, wait a second. I've got some very, very old C-code somewhere on my very old 386SX 16MHz PC somewhere. Nah, I wouldn't bother considering it's ten years old and I was fairly new to C at the time... :-)

 Well, if you're unfamiliar with the hercules screen-memory layout it might provide some guidance. Here's a bit of code for TurboC 2.0: (please don't comment the code-quality)

#define US unsigned
#define BYTE US char

#define HERCADDR(column,row)  \
	((BYTE far *) ( 0xB0000000L |  (US long) (0x2000u*(row & 3u) + 90u*(row>>2u) + (column>>3u)) ))

void vert_line(US x1, US x2, US y, const BYTE color) {
	BYTE far *addr1;
	BYTE far *addr2;
	BYTE byte1, byte2;

	byte1 =  128u >> (x1 & 7u);
	byte1 |= byte1-1;

	byte2 =  128u >> (x2 & 7u);
	byte2--;
	byte2 ^= 0xFF;

	addr1 = HERCADDR(x1, y);
	addr2 = HERCADDR(x2, y);

	if ( color ) {
		if ( FP_OFF(addr2) == FP_OFF(addr1) ) {
			*addr1 |= byte1 & byte2;
		} else {
			*addr1 |= byte1;
			*addr2 |= byte2;
			addr1++;
			while ( addr1 < addr2 )
				*addr1++ = 0xFF;
		}
	} else {
		if ( FP_OFF(addr2) == FP_OFF(addr1) ) {
			*addr1 &= (byte1 & byte2) ^ 0xFF;
		} else {
			*addr1 &= byte1 ^ 0xFF;
			*addr2 &= byte2 ^ 0xFF;
			addr1++;
			while ( addr1 < addr2 )
				*addr1++ = 0;
		}
	}
}

 Hmmm. Wasn't hercules a four color graphics card? Maybe I decided a two color model was more "speed optimized".


//Björn.

On Wed, 8 Sep 1999, Bjorn Eriksson wrote:
> Hmmm. Wasn't hercules a four color graphics card? Maybe I decided a
> two color model was more "speed optimized".

Hercules was mono. CGA was four colour.

--------------- Linux- the choice of a GNU generation. --------------
: Alex Holden (M1CJD)- Caver, Programmer, Land Rover nut, Radio Ham :
-------------------- http://www.linuxhacker.org/ --------------------

On Wednesday, September 08, 1999 11:44 AM, Bjorn Eriksson 
####@####.#### wrote:
: >: I looked at the code and at the moment it just uses a loop to draw the 
line
: >: pixel by pixel. How does it need to be improved??
: >:
: >	Are you familiar with the Herc memory/screen layout?  Instead of
: >looping for each bit, the C code needs to set up to 8 pixels at a time 
with a
: >single MOV instruction.  This entails pre-calculating the bits in the 
screen memory
: >and writing it once.
:
:  Oh, wait a second. I've got some very, very old C-code somewhere on my 
very old 386SX 16MHz PC somewhere. Nah, I wouldn't bother considering it's 
ten years old and I was fairly new to C at the time... :-)
:
:  Well, if you're unfamiliar with the hercules screen-memory layout it 
might provide some guidance. Here's a bit of code for TurboC 2.0: (please 
don't comment the code-quality)
: [snip]
	Totally cool!  This is exactly the code that Thomas Stuart needs to speed 
up
the Herc support.

	Thomas - go ahead and use Bjorn's code for the horizontal line draw 
speedup.
Let me know if you have questions.  BTW, Bjorn's code is incorrectly called 
"vert_line"
it really draws a horizontal line.  We won't get any speedups on the 
vertical
line drawing.

Greg