6502/lzsa
1
0
Fork 0
mirror of https://github.com/emmanuel-marty/lzsa.git synced 2025-01-03 04:33:15 +00:00
Code Issues Projects Releases Wiki Activity

Faster LZSA1 z80 decompression

Browse Source
This commit is contained in:
Emmanuel Marty 2019-08-27 13:16:20 +02:00 committed by GitHub
parent ef259e6867
commit 9de7e930e9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 284 additions and 275 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
asm/z80/unlzsa1_fast_v1.asm
View File

@ -1,5 +1,6 @@
;
; Speed-optimized LZSA1 decompressor by spke (v.1 03.1-22/08/2019, 107 bytes);
; Speed-optimized LZSA1 decompressor by spke (v.2 03/04/2019-27/08/2019; 111 bytes);
; with improvements by uniabis (30/07/2019,22/08/2019; -3 bytes, +3% speed).
;
; The data must be compressed using the command line compressor by Emmanuel Marty
; The compression is done as follows:
@ -59,6 +60,10 @@
ex de,hl : add hl,de
ENDM
MACRO COPY_MATCH
ldi : ldi : ldir
ENDM
MACRO BLOCKCOPY
ldir
ENDM
@ -74,6 +79,10 @@
ld a,d : sbc h : ld h,a ; 4*4+3*4 = 28t / 7 bytes
ENDM
MACRO COPY_MATCH
ldd : ldd : lddr
ENDM
MACRO BLOCKCOPY
lddr
ENDM
@ -83,8 +92,8 @@
@DecompressLZSA1:
ld b,0 : jr ReadToken
NoLiterals: xor (hl) : NEXT_HL
push de : ld e,(hl) : jp m,LongOffset
NoLiterals: xor (hl)
push de : NEXT_HL : ld e,(hl) : jp m,LongOffset
; short matches have length 0+3..14+3
ShortOffset: ld d,#FF : add 3 : cp 15+3 : jr nc,LongerMatch
@ -93,7 +102,7 @@ ShortOffset: ld d,#FF : add 3 : cp 15+3 : jr nc,LongerMatch
CopyMatch: ld c,a
.UseC NEXT_HL : ex (sp),hl ; BC = len, DE = offset, HL = dest, SP ->[dest,src]
ADD_OFFSET ; BC = len, DE = dest, HL = dest-offset, SP->[src]
BLOCKCOPY : pop hl ; BC = 0, DE = dest, HL = src
COPY_MATCH : pop hl ; BC = 0, DE = dest, HL = src
ReadToken: ; first a byte token "O|LLL|MMMM" is read from the stream,
; where LLL is the number of literals and MMMM is
@ -103,8 +112,8 @@ ReadToken: ; first a byte token "O|LLL|MMMM" is read from the stream,
cp #70 : jr z,MoreLiterals ; LLL=7 means 7+ literals...
rrca : rrca : rrca : rrca ; LLL<7 means 0..6 literals...
ld c,a : ld a,(hl) : NEXT_HL
BLOCKCOPY
ld c,a : ld a,(hl)
NEXT_HL : BLOCKCOPY
; next we read the first byte of the offset
push de : ld e,(hl)
@ -121,7 +130,7 @@ LongerMatch: NEXT_HL : add (hl) : jr nc,CopyMatch
; the codes are designed to overflow;
; the overflow value 1 means read 1 extra byte
; and overflow value 0 means read 2 extra bytes
.code1 NEXT_HL : ld b,a : ld c,(hl) : jr nz,CopyMatch.UseC
.code1 ld b,a : NEXT_HL : ld c,(hl) : jr nz,CopyMatch.UseC
.code0 NEXT_HL : ld b,(hl)
; the two-byte match length equal to zero
@ -130,17 +139,17 @@ LongerMatch: NEXT_HL : add (hl) : jr nc,CopyMatch
pop de : ret
MoreLiterals: ; there are three possible situations here
xor (hl) : NEXT_HL : exa
ld a,7 : add (hl) : NEXT_HL : jr c,ManyLiterals
xor (hl) : exa
ld a,7 : NEXT_HL : add (hl) : jr c,ManyLiterals
CopyLiterals: ld c,a
.UseC BLOCKCOPY
.UseC NEXT_HL : BLOCKCOPY
push de : ld e,(hl)
exa : jp p,ShortOffset : jr LongOffset
ManyLiterals:
.code1 ld b,a : ld c,(hl) : NEXT_HL : jr nz,CopyLiterals.UseC
.code0 ld b,(hl) : NEXT_HL : jr CopyLiterals.UseC
.code1 ld b,a : NEXT_HL : ld c,(hl) : jr nz,CopyLiterals.UseC
.code0 NEXT_HL : ld b,(hl) : jr CopyLiterals.UseC

8
asm/z80/unlzsa1_small_v1.asm
View File

@ -1,5 +1,5 @@
;
; Size-optimized LZSA1 decompressor by spke (v.1 23/04/2019, 67 bytes);
; Size-optimized LZSA1 decompressor by spke (v.1 23/04/2019, 68 bytes);
; with improvements by uniabis (30/07/2019, -1 byte, +3% speed).
;
; The data must be compressed using the command line compressor by Emmanuel Marty
@ -86,7 +86,7 @@
; first a byte token "O|LLL|MMMM" is read from the stream,
; where LLL is the number of literals and MMMM is
; a length of the match that follows after the literals
ReadToken: ld a,(hl) : NEXT_HL : push af
ReadToken: ld a,(hl) : exa : ld a,(hl) : NEXT_HL
and #70 : jr z,NoLiterals
rrca : rrca : rrca : rrca ; LLL<7 means 0..6 literals...
@ -95,10 +95,10 @@ ReadToken: ld a,(hl) : NEXT_HL : push af
ld c,a : BLOCKCOPY
; next we read the low byte of the -offset
NoLiterals: pop af : push de : ld e,(hl) : NEXT_HL : ld d,#FF
NoLiterals: push de : ld e,(hl) : NEXT_HL : ld d,#FF
; the top bit of token is set if
; the offset contains the high byte as well
or a : jp p,ShortOffset
exa : or a : jp p,ShortOffset
LongOffset: ld d,(hl) : NEXT_HL