Index
> DOS > DB -- dump bytes (hexdump) |
| Author |
|
code that I whipped up recently.
|
rugxulo 28 Aug 2016, 02:42
Well, not that anybody else cares, but here's a slightly newer version I've been fiddling with. No major speedups, though. Further studying (and testing) is required.
Code: ; ---------------------------------------------------------------------- ; DB.ASM -- dump raw bytes as hex from a file (using FASM) ; ; public domain, free for any use, nenies proprajho ; ; rugxulo _AT_ gmail ; ; ; Saturday, August 20, 2016 ; ---------------------------------------------------------------------- ;USE_186=1 ;USE_386=1 ;USE_AAM=1 ;USE_ALIGN=1 ;USE_BLOAT=1 ;USE_MMX=1 ;USE_XLATB=1 ARGC=80h CLOSE=3Eh CR=13 DOS=21h GOODBYE=20h LF=10 LFCR=0A0Dh LINEMAX=16 OPENREAD=3D00h READBYTES=3Fh WRITESTR=9 macro ALIGNED { if defined USE_ALIGN if defined ALIGNMENT align ALIGNMENT else align 16 end if end if } macro NIB2ASC { cmp al,10 sbb al,105 das } org 100h ; DOS .COM ; section .text ; begin main Komenco: mov si,ARGC lodsb xor bx,bx xchg al,bl mov byte [si+bx],bh .open: cmp byte [si],CR jz Fino lea dx,[si+1] mov ax,OPENREAD int DOS jc Fino mov word [handle],ax ALIGNED .read: mov ah,READBYTES mov bx,word [handle] mov cx,LINEMAX mov dx,buf int DOS jc .close mov word [bytesread],ax test ax,ax jz .close .show: call writeline if defined USE_386 movzx eax,word [bytesread] add dword [ofs],eax else mov ax,word [bytesread] add word [ofs],ax adc word [ofs+2],0 end if jmp .read .close: call setupofs mov di,line+8 call writeline.writestr mov ah,CLOSE mov bx,word [handle] int DOS Fino: if defined USE_MMX emms end if int GOODBYE ; end main ALIGNED ; proc setupofs: if defined USE_MMX mov eax,[ofs] bswap eax movd mm0,eax .convert: punpcklbw mm0,mm0 movq mm1,mm0 psrlw mm0,4 pand mm0,qword [mask] pand mm1,qword [mask] psrlw mm0,8 por mm0,mm1 paddb mm0,[zeros] movq mm2,mm0 pcmpgtb mm2,[nines] pand mm2,qword [sevens] paddb mm0,mm2 movq qword [line],mm0 else mov cx,4 mov si,ofs mov di,line+6 .ofsbyte: rept 4 { lodsb call split mov [di],ax sub di,2 } end if ret ; endp ALIGNED ; proc writeline: call setupofs .setupbytes: mov cx,word [bytesread] mov si,buf mov di,line.bytes ALIGNED .load: lodsb call split mov [di],ax add di,3 loop .load dec di .writestr: mov word [di],LFCR mov byte [di+2],'$' mov ah,WRITESTR mov dx,line int DOS ret ; endp ALIGNED ; proc split: if defined USE_AAM display "AAM 16 is unsupported by NEC V20/V30" aam 16 else mov ah,al and al,15 if defined USE_186 shr ah,4 else rept 4 { shr ah,1 } end if end if .byte2hexasc: if defined USE_XLATB push bx mov bx,hexarray xlatb xchg ah,al xlatb pop bx else NIB2ASC mov dl,al mov al,ah NIB2ASC mov ah,dl end if .bye: ret ; endp ; section .data if defined USE_BLOAT if defined BLOAT db BLOAT dup(0) else db 4*1024 dup(0) end if end if ALIGNED line db 'HHHHHHHH: ' .bytes db 'XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX' .end db CR,LF,'$' if defined USE_MMX ALIGNED zeros dq '00000000' nines dq '99999999' sevens db 8 dup(7) mask dw 4 dup(0Fh shl 8) end if if defined USE_XLATB ALIGNED hexarray db '0123456789ABCDEF' end if ALIGNED ofs dd 0 ; section .bss ALIGNED bytesread rw 1 ALIGNED handle rw 1 ALIGNED buf rb LINEMAX ; EOF |
|||
28 Aug 2016, 02:42 |
|
|
bitRAKE 28 Aug 2016, 19:05
_________________ ¯\(°_o)/¯ “languages are not safe - uses can be” Bjarne Stroustrup |
|||
|
28 Aug 2016, 19:05 |
|
|
rugxulo 04 Sep 2016, 01:07
bitRAKE wrote: http://wm.ite.pl/articles/convert-to-hex.html While I didn't study this in-depth and while it's still vaguely interesting, it doesn't show anything uniquely amazing. Maybe I'll have to time/test some of it myself, but at first glance nothing obvious pops out. (Besides, I already reluctantly wrote some crude MMX, and it didn't help at all. Alignment seems to make a bigger difference, sadly.) Anyways, I also briefly looked at Hugi Compo 17 (2002), but even that didn't show any obvious hints. (The "obvious" hint now is that slow hard disk seeking is the primary bottleneck, hence I need a better buffer algorithm. Except for one or two weirdo compilers, the ANSI C equivalent is always lightning fast, thanks to buffered I/O.) Anyways, I started to vainly wonder if I could shrink the code size. So I wrangled it down to 175 bytes. I also wanted a short (one-page) source version, so I converted it to Octasm. Not counting blanks, it's only 23 lines. Then I wondered if I could do similar terseness with a FASM version. I almost just cobbled together my own tr utility just to be able to put multiple instructions on one line, but in the end I found I could use a macro to help out. So 39 lines (not counting blanks) isn't bad. (It also adds up to only 71 instructions.) Yeah, pointless but still fun! 
Code: org 100h ; 175 bytes MAX=16 b equ byte w equ word DOS equ int 21h AAM16 equ aam 16 macro do [instr] { forward instr } ;macro AAM16 { do <mov ah,al>,<and al,15>,<shr ah,4> } Komenco: do <mov si,80h>,lodsb,<xor bx,bx>,<xchg al,bl>,<mov [si+bx],bh> open: do <cmp b[si],13>,jz Fino,<lea dx,[si+1]>,<mov ax,3D00h>,DOS do jc Fino,<mov [handle],ax> read: do <mov ah,3Fh>,<mov bx,[handle]>,<mov cx,MAX>,<mov dx,buf>,DOS do jc close,<mov [count],ax>,<test ax,ax>,jz close show: do call writeln,<mov ax,[count]>,<add w[ofs],ax>,<adc w[ofs+2],0>,jmp read close: do call setupofs,<mov di,line+8>,call writestr,<mov ah,3Eh> do <mov bx,[handle]>,DOS Fino: int 20h setupofs: do <mov si,ofs>,<mov di,line+6>,<mov w[di+2],': '>,<mov cx,4> @@: do lodsb,call split,stosw,<sub di,4>,loop @b,ret writeln: do call setupofs,<mov cx,[count]>,<mov si,buf>,<mov di,line+10> @@: do lodsb,call split,stosw,<mov al,' '>,stosb,loop @b,dec di writestr: do <mov w[di],0A0Dh>,<mov b[di+2],'$'>,<mov ah,9>,<mov dx,line>,DOS,ret split: AAM16 do <cmp al,10>,<sbb al,105>,das,<mov dl,al>,<mov al,ah> do <cmp al,10>,<sbb al,105>,das,<mov ah,dl>,ret do ofs dd 0,count dw ?,handle dw ?,buf db MAX dup(?) line db 10+(MAX*3)-1+3 dup(?) |
|||
|
04 Sep 2016, 01:07 |
|
< Last Thread | Next Thread > |
Forum Rules:
|