--- trunk/mkinitrd-magellan/isolinux/rllpack.inc	2007/09/01 22:45:15	532
+++ trunk/mkinitrd-magellan/isolinux/rllpack.inc	2010/08/19 09:50:43	1133
@@ -1,7 +1,7 @@
-; -*- fundamental -*-
-; -----------------------------------------------------------------------
-;   
-;   Copyright 2004 H. Peter Anvin - All Rights Reserved
+; -*- fundamental -*- ---------------------------------------------------
+;
+;   Copyright 2007-2009 H. Peter Anvin - All Rights Reserved
+;   Copyright 2009 Intel Corporation; author: H. Peter Anvin
 ;
 ;   This program is free software; you can redistribute it and/or modify
 ;   it under the terms of the GNU General Public License as published by
@@ -10,7 +10,6 @@
 ;   (at your option) any later version; incorporated herein by reference.
 ;
 ; -----------------------------------------------------------------------
-; $Id: rllpack.inc,v 1.1 2007-09-01 22:44:05 niro Exp $
 
 ;
 ; rllpack.inc
@@ -20,114 +19,146 @@
 ;
 ; Format of leading byte
 ; 1-128		= x verbatim bytes follow
-; 129-255 	= (x-126) times subsequent byte
+; 129-223	= (x-126) times subsequent byte
+; 224-255	= (x-224)*256+(next byte) times the following byte
 ; 0		= end of data
 ;
+; These structures are stored *in reverse order* in high memory.
+; High memory pointers point to one byte beyond the end.
+;
 
 		section .text
 
 ;
 ; rllpack:
-;	Pack CX bytes from DS:SI into ES:DI
-;	Returns updated SI, DI and CX = number of bytes output
+;	Pack ECX bytes from ESI into EDI.
+;	Returns updated ESI and EDI.
 ;
 rllpack:
-		push ax
-		push bx
-		push cx
-		push bp
-		push di
+		push word .pmentry
+		call simple_pm_call
+		ret
+
+		bits 32
+.pmentry:
+		push ecx
+		push ebx
+		push edx
 .startseq:
-		xor ax,ax		; Zero byte
-		xor bx,bx		; Run length zero
-		mov bp,di		; Pointer to header byte
-		stosb			; Store header byte (might be zero)
-		jcxz .done_null
+		xor eax,eax		; Zero byte
+		xor ebx,ebx		; Run length zero
+		dec edi
+		mov edx,edi		; Pointer to header byte
+		mov [edi],al		; Create header byte
+		jcxz .done		; If done, this was the terminator
 .stdbyte:
 		lodsb
-		stosb
-		dec cx
+		dec edi
+		mov [edi],al
+		dec ecx
 		cmp ah,al
 		je .same
 .diff:
 		mov ah,al
-		xor bx,bx
+		xor ebx,ebx
 .plainbyte:
-		inc bx
-		inc byte [es:bp]
-		jcxz .done
+		inc ebx
+		inc byte [edx]
+		jcxz .startseq
 		jns .stdbyte
 		jmp .startseq
 .same:
 		cmp bl,2
 		jb .plainbyte
 		; 3 bytes or more in a row, time to convert sequence
-		sub byte [es:bp],bl
+		sub [edx],bl
 		jnz .normal
-		dec di			; We killed a whole stretch, remove start byte
+		inc edi			; We killed a whole stretch,
+					; drop start byte
 .normal:
-		inc bx	
-		sub di,bx
-		mov bp,di
-		mov al,bl
-		add al,126
-		stosb
-		mov al,ah
-		stosb
+		inc ebx
+		add edi,ebx		; Remove the stored run bytes
 .getrun:
-		jcxz .done
-		cmp bl,255-126
-		jae .startseq
+		jcxz .nomatch
 		lodsb
 		cmp al,ah
 		jne .nomatch
-		inc bx
-		inc byte [es:bp]
-		dec cx
+		cmp bx,(256-224)*256-1	; Maximum run size
+		jae .nomatch
+		inc ebx
+		dec ecx
 		jmp .getrun
 .nomatch:
-		dec si
+		cmp bx,224-126
+		jae .twobyte
+.onebyte:
+		add bl,126
+		dec edi
+		mov [edi],bl
+		jmp .storebyte
+.twobyte:
+		add bh,224
+		sub edi,2
+		mov [edi],bx
+.storebyte:
+		dec edi
+		mov [edi],ah
+		dec esi			; Reload subsequent byte
 		jmp .startseq
 .done:
-		xor al,al
-		stosb
-.done_null:
-		pop dx
-		sub dx,di
-		neg dx
-		pop bp
-		pop cx
-		pop bx
-		pop ax
+		pop edx
+		pop ebx
+		pop ecx
 		ret
+
+		bits 16
 ;
 ; rllunpack:
-;	Unpack bytes from DS:SI into ES:DI
-;	On return SI, DI are updated and CX contains number of bytes output
-;	
+;	Unpack bytes from ESI into EDI
+;	On return ESI, EDI are updated and
+;	ECX contains number of bytes output.
+;
 rllunpack:
-		push ax
-		push di
-		xor cx,cx
+		push word .pmentry
+		call simple_pm_call
+		ret
+
+		bits 32
+.pmentry:
+		push edi
+		xor ecx,ecx
 .header:
-		lodsb
-		and al,al
-		jz .done
-		cmp al,129
+		dec esi
+		mov cl,[esi]
+		jcxz .done
+		cmp cl,129
 		jae .isrun
 		; Not a run
-		mov cl,al
-		rep movsb
+.copy:
+		dec esi
+		mov al,[esi]
+		stosb
+		loop .copy
 		jmp .header
 .isrun:
-		sub al,126
-		mov cl,al
-		lodsb
+		cmp cl,224
+		jae .longrun
+		sub cl,126
+.dorun:
+		dec esi
+		mov al,[esi]
 		rep stosb
 		jmp .header
+.longrun:
+		sub cl,224
+		mov ch,cl
+		dec esi
+		mov cl,[esi]
+		jmp .dorun
 .done:
-		pop cx
-		sub cx,di
-		neg cx
-		pop ax
+		pop ecx
+		sub ecx,edi
+		neg ecx
 		ret
+
+		bits 16