diff --git a/README.md b/README.md index e526760..e2b90df 100755 --- a/README.md +++ b/README.md @@ -71,10 +71,11 @@ License: 8-bit assembly code: -* Z80 decompressors (size- and speed-optimized) written by [introspec](https://github.com/specke) +* Z80 decompressors (size- and speed-optimized) written by [introspec](https://github.com/specke) with optimizations by [uniabis](https://github.com/uniabis) * 6502 and 8088 size-optimized improvements by [Peter Ferrie](https://github.com/peterferrie) +* 6502 speed-optimized decompressor by [John Brandwood](https://github.com/jbrandwood) * 8088 speed-optimized decompressor by [Jim Leonard](https://github.com/mobygamer) -* 6809 decompressors (for the Tandy CoCo, Thomson MO5/MO6/TO7/TO8/TO9, Dragon 32/64..) +* 6809 decompressors (Tandy CoCo, Thomson MO/TO, Dragon 32/64..) optimized by [Doug Masten](https://github.com/dougmasten) External links: diff --git a/asm/6809/unlzsa1.s b/asm/6809/unlzsa1.s index ec49aef..5b55862 100644 --- a/asm/6809/unlzsa1.s +++ b/asm/6809/unlzsa1.s @@ -1,4 +1,4 @@ -; unlzsa1.s - 6809 decompression routine for raw LZSA1 - 117 bytes +; unlzsa1.s - 6809 decompression routine for raw LZSA1 - 111 bytes ; compress with lzsa -r ; ; in: x = start of compressed data @@ -23,7 +23,45 @@ ; misrepresented as being the original software. ; 3. This notice may not be removed or altered from any source distribution. -decompress_lzsa1 +decompress_lzsa1 equ lz1token + +lz1bigof ldb ,x+ ; O set: load long 16 bit (negative, signed) offset + lda ,x+ ; (little endian) +lz1gotof leau d,y ; put backreference start address in U (dst+offset) + + puls b ; restore token + + clra ; clear A (high part of match length) + andb #$0F ; isolate MMMM (embedded match length) + addb #$03 ; add MIN_MATCH_SIZE + cmpb #$12 ; MATCH_RUN_LEN? + bne lz1gotln ; no, we have the full match length, go copy + + addb ,x+ ; add extra match length byte + MIN_MATCH_SIZE + MATCH_RUN_LEN + bcc lz1gotln ; if no overflow, we have the full length + bne lz1midln + + ldb ,x+ ; load 16-bit len in D (low part in B, high in A) + lda ,x+ ; (little endian) + bne lz1gotln ; check if we hit EOD (16-bit length = 0) + tstb + bne lz1gotln ; go copy matched bytes if not + + rts ; done, bail + +lz1midln tfr b,a ; copy high part of len into A + ldb ,x+ ; grab low 8 bits of len in B + +lz1gotln pshs x ; save source compressed data pointer + tfr d,x ; copy match length to X + +lz1cpymt lda ,u+ ; copy matched byte + sta ,y+ + leax -1,x ; decrement X + bne lz1cpymt ; loop until all matched bytes are copied + + puls x ; restore source compressed data pointer + lz1token ldb ,x+ ; load next token into B: O|LLL|MMMM pshs b ; save it @@ -50,59 +88,18 @@ lz1declt lsrb ; shift literals count into place lsrb lsrb lsrb - + lz1gotlt tfr x,u tfr d,x ; transfer 16-bit count into X lz1cpylt lda ,u+ ; copy literal byte - sta ,y+ + sta ,y+ leax -1,x ; decrement X and update Z flag bne lz1cpylt ; loop until all literal bytes are copied tfr u,x lz1nolt ldb ,s ; get token again, don't pop it from the stack - - tstb ; test O bit (small or large offset) - bmi lz1bigof + bmi lz1bigof ; test O bit (small or large offset) ldb ,x+ ; O clear: load 8 bit (negative, signed) offset lda #$ff ; set high 8 bits bra lz1gotof - -lz1bigof ldb ,x+ ; O set: load long 16 bit (negative, signed) offset - lda ,x+ ; (little endian) -lz1gotof leau d,y ; put backreference start address in U (dst+offset) - - puls b ; restore token - - clra ; clear A (high part of match length) - andb #$0F ; isolate MMMM (embedded match length) - addb #$03 ; add MIN_MATCH_SIZE - cmpb #$12 ; MATCH_RUN_LEN? - bne lz1gotln ; no, we have the full match length, go copy - - ldb ,x+ ; grab extra match length byte - addb #$12 ; add MIN_MATCH_SIZE + MATCH_RUN_LEN - bcc lz1gotln ; if no overflow, we have the full length - bne lz1midln - - ldb ,x+ ; load 16-bit len in D (low part in B, high in A) - lda ,x+ ; (little endian) - bne lz1gotln ; check if we hit EOD (16-bit length = 0) - tstb - bne lz1gotln ; go copy matched bytes if not - - rts ; done, bail - -lz1midln tfr b,a ; copy high part of len into A - ldb ,x+ ; grab low 8 bits of len in B - -lz1gotln pshs x ; save source compressed data pointer - tfr d,x ; copy match length to X - -lz1cpymt lda ,u+ ; copy matched byte - sta ,y+ - leax -1,x ; decrement X - bne lz1cpymt ; loop until all matched bytes are copied - - puls x ; restore source compressed data pointer - bra lz1token ; go decode next token diff --git a/asm/6809/unlzsa1b.s b/asm/6809/unlzsa1b.s index f658872..eff2fba 100644 --- a/asm/6809/unlzsa1b.s +++ b/asm/6809/unlzsa1b.s @@ -1,4 +1,4 @@ -; unlzsa1b.s - 6809 backward decompression routine for raw LZSA1 - 116 bytes +; unlzsa1b.s - 6809 backward decompression routine for raw LZSA1 - 112 bytes ; compress with lzsa -r -b ; ; in: x = last byte of compressed data @@ -62,9 +62,7 @@ lz1cpylt lda ,-u ; copy literal byte tfr u,x lz1nolt ldb ,s ; get token again, don't pop it from the stack - - tstb ; test O bit (small or large offset) - bmi lz1bigof + bmi lz1bigof ; test O bit (small or large offset) ldb ,-x ; O clear: load 8 bit (negative, signed) offset lda #$ff ; set high 8 bits @@ -84,8 +82,7 @@ lz1gotof nega ; reverse sign of offset in D cmpb #$12 ; MATCH_RUN_LEN? bne lz1gotln ; no, we have the full match length, go copy - ldb ,-x ; grab extra match length byte - addb #$12 ; add MIN_MATCH_SIZE + MATCH_RUN_LEN + addb ,-x ; add extra match length byte + MIN_MATCH_SIZE + MATCH_RUN_LEN bcc lz1gotln ; if no overflow, we have the full length bne lz1midln diff --git a/asm/6809/unlzsa2.s b/asm/6809/unlzsa2.s index f99c453..e21d0cd 100644 --- a/asm/6809/unlzsa2.s +++ b/asm/6809/unlzsa2.s @@ -1,4 +1,4 @@ -; unlzsa2.s - 6809 decompression routine for raw LZSA2 - 187 bytes +; unlzsa2.s - 6809 decompression routine for raw LZSA2 - 183 bytes ; compress with lzsa -f2 -r ; ; in: x = start of compressed data @@ -40,8 +40,7 @@ lz2token ldb ,x+ ; load next token into B: XYZ|LL|MMM cmpb #$12 ; LITERALS_RUN_LEN_V2 + 15 ? bne lz2gotlt ; if not, we have the full literals count, go copy - ldb ,x+ ; load extra literals count byte - addb #$12 ; add LITERALS_RUN_LEN + 15 + addb ,x+ ; add extra literals count byte + LITERALS_RUN_LEN + 15 bcc lz2gotlt ; if no overflow, we got the complete count, copy ldb ,x+ ; load low 8 bits of little-endian literals count diff --git a/asm/6809/unlzsa2b.s b/asm/6809/unlzsa2b.s index 162f972..5defa09 100644 --- a/asm/6809/unlzsa2b.s +++ b/asm/6809/unlzsa2b.s @@ -1,4 +1,4 @@ -; unlzsa2.s - 6809 backward decompression routine for raw LZSA2 - 189 bytes +; unlzsa2b.s - 6809 backward decompression routine for raw LZSA2 - 187 bytes ; compress with lzsa -f2 -r -b ; ; in: x = last byte of compressed data @@ -42,8 +42,7 @@ lz2token ldb ,-x ; load next token into B: XYZ|LL|MMM cmpb #$12 ; LITERALS_RUN_LEN_V2 + 15 ? bne lz2gotlt ; if not, we have the full literals count, go copy - ldb ,-x ; load extra literals count byte - addb #$12 ; add LITERALS_RUN_LEN + 15 + addb ,-x ; add extra literals count byte + LITERALS_RUN_LEN + 15 bcc lz2gotlt ; if no overflow, we got the complete count, copy ldd ,--x ; load 16 bit count in D (low part in B, high in A) diff --git a/src/lzsa.c b/src/lzsa.c index 82aa902..29d4aa5 100755 --- a/src/lzsa.c +++ b/src/lzsa.c @@ -48,7 +48,7 @@ #define OPT_RAW_BACKWARD 8 #define OPT_STATS 16 -#define TOOL_VERSION "1.2.2" +#define TOOL_VERSION "1.3.0" /*---------------------------------------------------------------------------*/ diff --git a/src/shrink_block_v2.c b/src/shrink_block_v2.c index 32d76b9..67d9b81 100644 --- a/src/shrink_block_v2.c +++ b/src/shrink_block_v2.c @@ -1048,7 +1048,7 @@ static int lzsa_write_block_v2(lzsa_compressor *pCompressor, lzsa_match *pBestMa return -1; if (pCompressor->flags & LZSA_FLAG_RAW_BLOCK) - pOutData[nOutOffset++] = (nTokenLiteralsLen << 3) | 0x47; + pOutData[nOutOffset++] = (nTokenLiteralsLen << 3) | 0xe7; else pOutData[nOutOffset++] = (nTokenLiteralsLen << 3) | 0x00; nOutOffset = lzsa_write_literals_varlen_v2(pOutData, nOutOffset, nMaxOutDataSize, &nCurNibbleOffset, &nCurFreeNibbles, nNumLiterals); @@ -1081,7 +1081,6 @@ static int lzsa_write_block_v2(lzsa_compressor *pCompressor, lzsa_match *pBestMa if (nOutOffset >= nMaxOutDataSize) return -1; - pOutData[nOutOffset++] = 0; /* Match offset */ nOutOffset = lzsa_write_nibble_v2(pOutData, nOutOffset, nMaxOutDataSize, &nCurNibbleOffset, &nCurFreeNibbles, 15); /* Extended match length nibble */ if (nOutOffset < 0) return -1; @@ -1119,12 +1118,12 @@ static int lzsa_write_raw_uncompressed_block_v2(lzsa_compressor *pCompressor, co int nTokenLiteralsLen = (nNumLiterals >= LITERALS_RUN_LEN_V2) ? LITERALS_RUN_LEN_V2 : nNumLiterals; int nOutOffset = 0; - int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v2(nNumLiterals) + (nNumLiterals << 3) + 8 + 4 + 8; + int nCommandSize = 8 /* token */ + lzsa_get_literals_varlen_size_v2(nNumLiterals) + (nNumLiterals << 3) + 4 + 8; if ((nOutOffset + ((nCommandSize + 7) >> 3)) > nMaxOutDataSize) return -1; pCompressor->num_commands = 0; - pOutData[nOutOffset++] = (nTokenLiteralsLen << 3) | 0x47; + pOutData[nOutOffset++] = (nTokenLiteralsLen << 3) | 0xe7; nOutOffset = lzsa_write_literals_varlen_v2(pOutData, nOutOffset, nMaxOutDataSize, &nCurNibbleOffset, &nCurFreeNibbles, nNumLiterals); if (nOutOffset < 0) return -1; @@ -1137,8 +1136,6 @@ static int lzsa_write_raw_uncompressed_block_v2(lzsa_compressor *pCompressor, co /* Emit EOD marker for raw block */ - pOutData[nOutOffset++] = 0; /* Match offset */ - nOutOffset = lzsa_write_nibble_v2(pOutData, nOutOffset, nMaxOutDataSize, &nCurNibbleOffset, &nCurFreeNibbles, 15); /* Extended match length nibble */ if (nOutOffset < 0) return -1;