the problem is due to the suprisingly slow cga console.

these timings are based on rdtsc() subtractions around the
named areas for the simple test of cat'ing /lib/pci to the
console.  they are huge:

cycles
printing chars		482510340
scrolling			135746656968
total			137112900000

by introducing a frame buffer to avoid reading from the
cga console for scrolling (a guess based on problems with
graphics performance), we get about a 10x improvement:

printing normal chars	1080381568
scrolling			12046340760
total			13610262120

by guessing that any string >40 bytes is likely to induce scrolling,
we can redraw the whole screen once we're done.  this gives us
100/1000x improvment on our hot spots, but just 7x in run time.

printing chars		33186956
scrolling			24111480
total			1854594800

this looks like about all we can do.

Reference: /n/atom/patch/applied/cgafb
Date: Thu Feb 13 04:01:19 CET 2014
Signed-off-by: quanstro@quanstro.net

--- /sys/src/nix/k10/cga.c	Thu Feb 13 04:00:30 2014
+++ /sys/src/nix/k10/cga.c	Thu Feb 13 04:00:31 2014
@@ -36,6 +36,8 @@
 	Lock;
 	int	pos;
 	int	initdone;
+	uchar	blankline[Width];
+	uchar	fb[Width*Height];
 	uchar	*cga;
 } cga;
 
@@ -97,7 +99,13 @@
 }
 
 static void
-cgaputr(Rune c)
+cgaredraw(void)
+{
+	memmove(cga.cga, cga.fb, Width*Height-Postlen*2);
+}
+
+static void
+cgaputr(Rune c, int fb)
 {
 	int i;
 	uchar *p;
@@ -109,12 +117,12 @@
 	else if(c == '\t'){
 		i = 8 - ((cga.pos/2)&7);
 		while(i-- > 0)
-			cgaputr(' ');
+			cgaputr(' ', fb);
 	}
 	else if(c == '\b'){
 		if(cga.pos >= 2)
 			cga.pos -= 2;
-		cgaputr(' ');
+		cgaputr(' ', fb);
 		cga.pos -= 2;
 	}
 	else{
@@ -129,36 +137,42 @@
 					break;
 				}
 			}
-		cga.cga[cga.pos++] = c;
-		cga.cga[cga.pos++] = Attr;
+		cga.fb[cga.pos++] = c;
+		cga.fb[cga.pos++] = Attr;
+		if(!fb){
+			cga.cga[cga.pos-2] = c;
+			cga.cga[cga.pos-1] = Attr;
+		}
 	}
 	if(cga.pos >= (Width*Height)-Postlen*2){
-		memmove(cga.cga, &cga.cga[Width], Width*(Height-1));
-		p = &cga.cga[Width*(Height-1)-Postlen*2];
-		for(i = 0; i < Width/2; i++){
-			*p++ = ' ';
-			*p++ = Attr;
-		}
+		memmove(cga.fb, &cga.fb[Width], Width*(Height-1));
+		p = &cga.fb[Width*(Height-1)-Postlen*2];
+		memmove(p, cga.blankline, Width);
 		cga.pos -= Width;
+		if(!fb)
+			memmove(cga.cga, cga.fb, Width*Height-Postlen*2);
 	}
-	cgacursor();
 }
 
 void
 cgaconsputs(char* s, int n)
 {
-	int i;
+	int i, fb;
 	Rune r;
 
 	if(sys->novga || cga.initdone == 0)
 		return;
 	ilock(&cga);
+	fb = n>=Width/4;
 	while(n > 0 && fullrune(s, n)){
 		i = chartorune(&r, s);
 		n -= i;
 		s += i;
-		cgaputr(r);
+		cgaputr(r, fb);
 	}
+	if(fb)
+		cgaredraw();
+	cgacursor();
 	iunlock(&cga);
 }
 
@@ -182,6 +196,11 @@
 void
 cgainit(void)
 {
+	char *s;
+	uchar *p, *e;
+
+	if((s = getconf("*novga")) != nil)
+		sys->novga = atoi(s);
 	if(sys->novga)
 		return;
 	ilock(&cga);
@@ -189,6 +208,12 @@
 	cga.pos |= cgaregr(0x0f);
 	cga.pos *= 2;
 	cga.cga = CGA;
+	e = cga.blankline + Width;
+	for(p = cga.blankline; p < e; ){
+		*p++ = ' ';
+		*p++ = Attr;
+	}
+memmove(cga.fb, cga.cga, Width*Height);
 	cga.initdone = 1;
 	iunlock(&cga);
 }