flat assembler
> Examples and Tutorials > Simple pattern matching procedure. |
| Author |
|
|
Here is a simple pattern match program, similar to the DOS/Linux glob matching. Please, if someone wants, make some tests. Also, suggestions about size optimizations are welcome.
Code: ;*************************************************************************************** ; ; Provides pattern matching of the .pattern in the .string . ; Returns CF=1 if the pattern MATCH ; CF=0 if the pattern DOES NOT MATCH ; ; In the pattern, "*" char match 0 or more characters and "?" char match ; exactly 1 character. If the characters "*" and "?" have to be directly matched ; they can be escaped with "\". "\\" escapes the backslash itself. If the backslash ; is located at the end of the pattern, it is treated as a normal character. ; ; If a character between "@" and "_" is escaped, a controll code is generated equal ; to the ascii code of the char - $40. ; ;*************************************************************************************** proc StrMatchPattern, .pattern, .string begin pushad mov esi, .pattern mov edi, .string xor ecx, ecx .main_loop mov ah, edi ; from the string inc edi .main_loop2 mov al, esi ; from the pattern inc esi test al, al jz .eo_pattern cmp al, '*' jne .not_star mov ecx, esi ; the address after the star. mov ebx, edi jmp .main_loop2 .not_star test ah, ah jz .not_match ; the string ended, but the pattern not. cmp al, '?' je .main_loop cmp al, '\' jne .compare cmp byte esi , 0 je .compare mov al, esi inc esi cmp al, $40 jb .compare cmp al, $60 jae .compare sub al, $40 ; the control codes escaped. .compare cmp al, ah je .main_loop jecxz .not_match .char_ne mov edi, ebx mov esi, ecx inc ebx jmp .main_loop .eo_pattern test ah, ah jz .match jecxz .not_match cmp byte ecx , 0 jne .char_ne .match stc popad return .not_match clc popad return endp _________________ Tox ID: 48C0321ADDB2FE5F644BB5E3D58B0D58C35E5BCBC81D7CD333633FEDF1047914A534256478D9 |
|||
07 Nov 2013, 11:15 |
|
|
JohnFound wrote: Here is a simple pattern match program, similar to the DOS/Linux glob matching. Please, if someone wants, make some tests. Also, suggestions about size optimizations are welcome.  |
|||
|
07 Nov 2013, 11:20 |
|
|
It is FreshLib syntax. The procedure header can be easily edited for whatever macro library you use.
|
|||
|
07 Nov 2013, 11:27 |
|
|
JohnFound wrote: It is FreshLib syntax. The procedure header can be easily edited for whatever macro library you use.
Another question. Why don't use 'lodsb' here: Code: .main_loop2
mov al, esi ; from the pattern
inc esi |
|||
|
07 Nov 2013, 12:25 |
|
|
MHajduk wrote: Another question. Why don't use 'lodsb' here: It better corresponds with the next instruction pair "mov ah, [edi]". As long as both make the same (load char from the string), I considered they have to look the same. It makes the code more readable. _________________ Tox ID: 48C0321ADDB2FE5F644BB5E3D58B0D58C35E5BCBC81D7CD333633FEDF1047914A534256478D9 |
|||
|
07 Nov 2013, 13:03 |
|
|
Well, to substitute these two instructions:
Code: mov ah, edi ; from the string inc edi Code: xchg esi, edi xchg ah, al lodsb xchg ah, al xchg esi, edi |
|||
|
07 Nov 2013, 13:16 |
|
|
Stupid post removed.
Last edited by HaHaAnonymous on 28 Feb 2015, 19:10; edited 1 time in total |
|||
|
07 Nov 2013, 13:17 |
|
|
If you're not too concerned about alignment and would prefer to save space, use 'align 1' in front of your procedures. This will save, on average, a couple bytes per procedure.
|
|||
|
09 Nov 2013, 22:32 |
|
|
ejamesr wrote: If you're not too concerned about alignment and would prefer to save space, use 'align 1' in front of your procedures. This will save, on average, a couple bytes per procedure. Well, "align 1" is actually NOP. The assembler always align on 1. The alignment of the procedures in FreshLib is optional, so the user of the library can choose the alignment with "options.AlignCode". _________________ Tox ID: 48C0321ADDB2FE5F644BB5E3D58B0D58C35E5BCBC81D7CD333633FEDF1047914A534256478D9 |
|||
|
10 Nov 2013, 05:10 |
|
|
'align 1' does nothing; it will not insert a NOP into the code. But if it replaces a larger-sized 'align' statement, it can eliminate NOPs.
My previous comment assumed you used 'align 4' or even 'align 16' for your procedures. If not, please disregard my comment. Since I'm always more interested in speed than size (using assembly pretty much guarantees my code will be smaller than when using a different language), I periodically consult the latest version of the "Intel® 64 and IA-32 Architectures Optimization Reference Manual" (July 2013). On page 3-9: Quote: Assembly/Compiler Coding Rule 12. (M impact, H generality) All branch targets should be 16-byte aligned. Quote: Alignment of code can be an issue for the Pentium M, Intel Core Duo and Intel Core 2 Duo processors. |
|||
|
11 Nov 2013, 21:55 |
|
< Last Thread | Next Thread > |
Forum Rules:
|