diff --git a/examples/graphbench.pas b/examples/graphbench.pas
index 327e72e..9abbfba 100644
--- a/examples/graphbench.pas
+++ b/examples/graphbench.pas
@@ -1,5 +1,17 @@
 program graphbench;
+uses sprites;
+
 var starttime,endtime:DateTime;
+    spriteData:SpritePixels;
+
+procedure readSpriteData(filename:string);
+var f:file;
+begin
+	open(f,filename,ModeReadOnly);
+	seek(f,8); (* skip file header *)
+	read(f,spriteData);
+	close(f);
+end;
 
 procedure startBench(name:string);
 begin
@@ -13,7 +25,7 @@ var secDelta, minDelta, hourDelta:integer;
   begin
 	if i < 10 then
 		write('0');
-		write(i);
+	write(i);
   end;
 begin
 	endTime := GetTime;
@@ -49,6 +61,20 @@ begin
 	randint := r;
 end;
 
+procedure drawsprites(count:integer);
+var i,col,x,y:integer;
+begin
+	col := 1;
+	for i := 1 to count do
+	begin
+		x := randint(350);
+		y := randint(350);
+		PutSprite(x,y,spriteData);
+		col := col + 1;
+		if col > 15 then col := 1;
+	end;
+end;
+
 procedure drawlines(count:integer);
 var i,col,x1,y1,x2,y2:integer;
 begin
@@ -80,13 +106,20 @@ begin
 end;
 
 begin
+	readSpriteData('rocket.sprt');
+
 	InitGraphics;
-	startBench('200K points');
+	startBench('points 200K');
 	drawpoints(200000);
 	endBench;
 
 	InitGraphics;
-	startBench('10K lines');
+	startBench('lines 10K');
 	drawlines(10000);
 	endBench;
+
+	InitGraphics;
+	startBench('sprites 50K');
+	drawsprites(50000);
+	endBench;
 end.
diff --git a/examples/sprites.s b/examples/sprites.s
index 6962eda..ab2e580 100644
--- a/examples/sprites.s
+++ b/examples/sprites.s
@@ -6,28 +6,13 @@
         .EQU FB_WA      $904
         .EQU FB_IO      $908
         .EQU FB_PS      $90C
-
-; calculate mask for a word of pixels
-; args: word of pixels with four bits per pixel
-; returns: value that masks out all pixels that are set
-CALC_MASK:
-	LOADC $F	; pixel mask
-C_M_L0:
-	SWAP		; swap mask and pixels value
-	AND.S1.X2Y	; isolate one pixel, keep args
-	CBRANCH.Z C_M_L1 ; if pixel is zero, dont set mask bits
-	OVER		; copy current mask
-	OR		; or into pixels value
-C_M_L1:
-	SWAP 		; swap back, ToS is now mask bits
-	SHL 2		; shift mask for next pixel to the left
-	SHL 2
-
-	DUP
-	CBRANCH.NZ C_M_L0 ; if mask is zero, we are done
-	DROP		; remove mask bits
-	NOT		; invert result	
-	RET
+	.EQU FB_PD	$910
+	.EQU FB_CTL	$914
+	.EQU FB_SHIFTER $918
+	.EQU FB_SHIFTCOUNT $91C
+	.EQU FB_SHIFTERM $920
+	.EQU FB_SHIFTERSP $924
+	.EQU FB_MASKGEN $928
 
 ; calculate vmem address from coordinates
 ; args: x,y
@@ -67,13 +52,19 @@ CALC_VMEM_ADDR:
 	.EQU PS_SHIFT_C		20
 	.EQU PS_SPILL		24
 	.EQU PS_STRIPE_C	28
-	.EQU PS_FS 32
+	.EQU PS_BPSAVE		32
+	.EQU PS_FS 36
 PUTSPRITE:
 	FPADJ -PS_FS
 	STORE PS_SPRITE_DATA
 	STORE PS_Y
 	STORE PS_X
 
+	LOADREG BP
+	STORE PS_BPSAVE
+	LOADC 0
+	STOREREG BP
+
 	; calculate vmem address
 	LOAD PS_X
 	LOAD PS_Y
@@ -81,11 +72,6 @@ PUTSPRITE:
 	CALL	
 	STORE PS_VMEM_ADDR
 
-	LOAD PS_X	; shift count = x mod 8
-	LOADC 7
-	AND
-	STORE PS_SHIFT_C
-
 	LOADC SPRITE_HEIGHT
 	STORE PS_SPRITE_LINES
 
@@ -93,12 +79,10 @@ PUTSPRITE:
 PS_LOOP1:
 	; set read and write address
 	; in the vga controller
-	LOADC FB_RA ; read address register
 	LOAD PS_VMEM_ADDR
-	STOREI 4 ; use autoincrement to get to the next register
-	LOAD PS_VMEM_ADDR
-	STOREI
-	DROP
+	DUP
+	STORE.B FB_RA
+	STORE.B FB_WA
 
 	LOAD PS_SPRITE_DATA	; address of sprite data
 	DUP
@@ -106,61 +90,19 @@ PS_LOOP1:
 	STORE PS_SPRITE_DATA	; and store it again
 	LOADI			; load word from orig. address
 
+	; ------- one word of sprite pixels on stack
 
-	LOADC 0
-	STORE PS_SPILL
+	STORE.B FB_SHIFTER
+	LOAD PS_X
+	STORE.B FB_SHIFTCOUNT
 
-	; loop to shift pixel data to right
-	LOAD PS_SHIFT_C		; load shift count
-PS_LOOP2:
-	DUP			; test it for zero
-	CBRANCH.Z PS_LOOP2_X
+	LOAD.B FB_SHIFTERM	; get shifted mask
+	LOAD.B FB_IO		; and background pixel data
+	AND			; remove foreground pixels
 
-	SWAP			; swap count with pixels
-
-	; save the pixel that is shifted out
-	LOADC $F		; mask the four bits
-	AND.S0			; keep original value on stack
-	BROT			; and move them to MSB
-	BROT
-	BROT
-	SHL 2
-	SHL 2			; shift by 28 in total
-
-	LOAD PS_SPILL		; load spill bits
-	SHR			; shift by four to make space
-	SHR
-	SHR
-	SHR
-	OR			; or with orig value
-	STORE PS_SPILL		; store new value
-
-	SHR			; shift pixels right
-	SHR			; four bits per pixel
-	SHR
-	SHR
-
-	SWAP			; swap back, count now ToS
-	DEC 1
-	BRANCH PS_LOOP2
-PS_LOOP2_X:
-	DROP			; remove shift count, shifted pixels now in ToS
-
-	DUP
-	LOADCP CALC_MASK	; calculate sprite mask for this word
-	CALL
-
-	LOADCP FB_IO		; address of the i/o register
-	LOADI			; read word from video mem
-
-	AND			; and word with mask
-	
-	OR			; OR sprite data with original pixels
-
-	LOADCP FB_IO
-	SWAP
-	STOREI			; store result into i/o reg
-	DROP
+	LOAD.B FB_SHIFTER	; get shifted pixels
+	OR			; combine with background
+	STORE.B FB_IO		; store into vmem
 
 	; set counter for remaining stripes
 	LOADC SPRITE_STRIPES - 1
@@ -170,8 +112,8 @@ PS_LOOP2_X:
 	; process spilled bits and next vertical stripe of sprite data
 	;
 PS_NEXT_STRIPE:
-	; put spill bits on stack for later
-	LOAD PS_SPILL
+	;use spill bits from first column
+	LOAD.B FB_SHIFTERSP
 
 	LOAD PS_SPRITE_DATA	; address of sprite data
 	DUP
@@ -179,65 +121,20 @@ PS_NEXT_STRIPE:
 	STORE PS_SPRITE_DATA	; and store it again
 	LOADI			; load word from orig. address
 
-	; reset spill bits
-	LOADC 0
-	STORE PS_SPILL
-
-	; last spill bits are on ToS now
-
-	; shift pixel data to right
-	LOAD PS_SHIFT_C		; load shift count
-PS_LOOP3:			; test it for zero
+	STORE.B FB_SHIFTER	; store into shifter
+	LOAD PS_X
+	STORE.B FB_SHIFTCOUNT	; shift stuff
+	LOAD.B FB_SHIFTER	; get shifted pixels
+	OR			; combine with spill bits (see above)
 	DUP
-	CBRANCH.Z PS_LOOP3_X
+	STORE.B FB_MASKGEN	; store to mask reg to get new mask
 
-	SWAP			; swap count with pixels
+	LOAD.B FB_MASKGEN	; get mask for spill bits + shifted pixels
+	LOAD.B FB_IO		; get vmem data
+	AND			; remove foreground pixels from bg
 
-	; save the pixel that is shifted out
-	LOADC $F		; mask the four bits
-	AND.S0			; keep original value on stack
-	BROT			; and move them to MSB
-	BROT
-	BROT
-	SHL 2
-	SHL 2			; shift by 28 in total
-
-	LOAD PS_SPILL		; load spill bits
-	SHR			; shift by four to make space
-	SHR
-	SHR
-	SHR
-	OR			; or with orig value
-	STORE PS_SPILL		; store new value
-
-	SHR			; shift pixels right
-	SHR			; four bits per pixel
-	SHR
-	SHR
-
-	SWAP			; swap back, count now ToS
-	DEC 1
-	BRANCH PS_LOOP3
-PS_LOOP3_X:
-	DROP			; remove shift count, shifted pixels now in ToS
-
-	OR			; or together with spill bits
-
-	DUP
-	LOADCP CALC_MASK	; calculate sprite mask
-	CALL
-
-	LOADCP FB_IO		; load original pixels
-	LOADI
-
-	AND			; and with mask
-
-	OR			; or together with original pixels
-
-	LOADCP FB_IO
-	SWAP
-	STOREI
-	DROP
+	OR			; combine with shifted pixels
+	STORE.B FB_IO		; write to vmem
 
 	LOAD PS_STRIPE_C	; decrement stripe count
 	DEC 1
@@ -246,22 +143,18 @@ PS_LOOP3_X:
 	CBRANCH.NZ PS_NEXT_STRIPE ; if non-zero, next stripe
 
 	; write spilled bits of the last stripe into next vmem word
-	LOAD PS_SPILL		; get spill bits
+	LOAD.B FB_SHIFTERSP 	; get spill bits
 	DUP
-	LOADCP CALC_MASK	; calculate sprite mask for spill bits
-	CALL
+	STORE.B FB_MASKGEN
+	LOAD.B FB_MASKGEN	; get sprite mask for spill bits
 
-	LOADCP FB_IO
-	LOADI			; load next vmem word
+	LOAD.B FB_IO		; load next vmem word
 	AND			; apply sprite mask
 	
 	OR			; OR in spill bits
 
-	LOADCP FB_IO
-	SWAP			; swap pixels and addr
-	STOREI			; write back
-	DROP
-		
+	STORE.B FB_IO		; write to vmem
+
 	LOAD PS_SPRITE_LINES	; decrement lines count
 	DEC 1
 	DUP
@@ -275,7 +168,10 @@ PS_LOOP3_X:
 	BRANCH PS_LOOP1
 PS_L_XT:
 	DROP
-			
+
+	LOAD PS_BPSAVE
+	STOREREG BP
+
 	FPADJ PS_FS
 	RET