Index
> Main > flat assembler 1.67.26-27Goto page 1, 2 Next |
| Author |
|
|
Tomasz Grysztar 27 Jan 2008, 18:39
I've decided to release the 1.67.26 version now, even though I haven't yet finished the SSE4 support I wanted for this version - my computer broke (see the other thread) and this caused a slight pause in development (but have a new machine up and running great, thanks to all the people that helped me!).
So 1.67.26 has a lot of SSE4 instructions already supported, but some of them will be supported in the next release (1.67.27), I hope to get it finished soon. I also did play with DOS interface a bit, I fixed some bugs and also managed to free some space under the 64k code limit, so that I still don't have to get rid of the unREAL mode from the DOS interface. |
|||
27 Jan 2008, 18:39 |
|
|
Tomasz Grysztar 27 Jan 2008, 19:34
revolution wrote: I have never understood the reason for using unreal mode. I can't see what advantage it has over PM. Well, on some old machines it was faster than PM. And this implementation of it (the one that fasm uses) is also a kind of my 'pet', piece of code I really like and was proud of when first invented it. |
|||
|
27 Jan 2008, 19:34 |
|
|
rugxulo 27 Jan 2008, 20:06
Tomasz Grysztar wrote: I've decided to release the 1.67.26 version now, even though I haven't yet finished the SSE4 support I wanted for this version - my computer broke (see the other thread) and this caused a slight pause in development (but have a new machine up and running great, thanks to all the people that helped me!). So the curse has been lifted? 
Quote:
Do any actual machines (for sale) even support it yet? If not, I don't see the huge need to hurry here (unless you like working under BOCHS, heh). Tomasz wrote:
Heh, I guess you're human after all.  But it's still not fixed. Really. I just now tried three different OSes on three different machines (even tried both unreal and DPMI in pure DOS), and none of 'em worked. Either silently starts/exits (WinXP), GPFs/exits the DOS window (WinME), prints garbage when assembling (unreal in pure DR-DOS), or causes an exception (DPMI under DR-DOS). But FASMD still seems to work fine.
And I even saw/found a bug in one of my own dinky programs (due to my optimization frenzy of late, heh). So that's good to know (fix), at least. 
Tomasz wrote:
Like I said before, there's plenty that could be improved size-wise (and that's just my barely-educated guess without looking at all). So, don't give up that dream just yet.
revolution wrote:
No external memory managers (or DPMI servers) needed. So, it's at least more independent this way. |
|||
|
27 Jan 2008, 20:06 |
|
|
Tomasz Grysztar 27 Jan 2008, 20:13
rugxulo wrote: Heh, I guess you're human after all. Sorry, forgot to test it after some final touches I did. Of course it's not working, even for me. Hang on a second, I'm correcting it. |
|||
|
27 Jan 2008, 20:13 |
|
|
Tomasz Grysztar 27 Jan 2008, 20:22
OK, please try it now.
|
|||
|
27 Jan 2008, 20:22 |
|
|
revolution 27 Jan 2008, 20:28
rugxulo wrote: No external memory managers (or DPMI servers) needed. So, it's at least more independent this way. |
|||
|
27 Jan 2008, 20:28 |
|
|
rugxulo 27 Jan 2008, 20:44
Tomasz Grysztar wrote: OK, please try it now. Just telling you really quickly before someone downloads it: you forgot the dir structure: Code: FAS16726.ZIP FASM.EXE 42,558 (75,192) Jan,27,2008 09:16:12p [Deflat] FASM.TXT 50,382 (196,211) Aug,06,2007 12:16:38a [Deflat] FASMD.EXE 52,786 (93,083) Jan,27,2008 09:16:12p [Deflat] FASMD.TXT 2,416 (6,370) Dec,29,2006 12:17:56a [Deflat] LICENSE.TXT 949 (1,819) Feb,10,2007 02:24:32p [Deflat] WHATSNEW.TXT 4,042 (11,104) Jan,27,2008 06:57:02p [Deflat] FASM.ASM 1,775 (6,553) Jan,27,2008 09:14:40p [Deflat] ASSEMBLE.INC 6,005 (38,606) Jan,19,2008 07:54:58p [Deflat] ERRORS.INC 892 (3,975) Dec,28,2007 12:55:50a [Deflat] MESSAGES.INC 797 (2,395) Dec,02,2007 12:39:56a [Deflat] EXPRESSI.INC 8,712 (52,639) Dec,05,2007 11:41:02a [Deflat] FORMATS.INC 12,795 (79,627) Dec,30,2007 07:03:54p [Deflat] MODES.INC 3,040 (10,540) Jan,27,2008 06:32:44p [Deflat] PARSER.INC 4,519 (26,335) Aug,25,2007 10:14:16a [Deflat] PREPROCE.INC 8,608 (51,657) Dec,08,2007 07:17:08p [Deflat] SYSTEM.INC 2,111 (8,102) Jan,27,2008 09:14:36p [Deflat] SYSDPMI.INC 722 (2,103) Jan,27,2008 06:26:40p [Deflat] TABLES.INC 7,519 (48,546) Jan,19,2008 08:39:46p [Deflat] VARIABLE.INC 833 (2,356) Jan,19,2008 08:32:26p [Deflat] VERSION.INC 955 (1,847) Jan,19,2008 03:42:22p [Deflat] X86_64.INC 13,496 (139,206) Jan,19,2008 08:44:46p [Deflat] MEMORY.INC 1,008 (5,931) Aug,05,2007 07:59:26p [Deflat] NAVIGATE.INC 2,075 (10,606) Aug,05,2007 07:59:30p [Deflat] EDIT.INC 3,586 (20,153) Aug,05,2007 07:59:14p [Deflat] BLOCKS.INC 2,165 (10,779) Dec,01,2007 11:56:32a [Deflat] SEARCH.INC 2,039 (10,963) Aug,05,2007 07:59:34p [Deflat] UNDO.INC 851 (4,795) Aug,05,2007 07:59:38p [Deflat] FASMD.ASM 16,501 (82,698) Dec,03,2007 09:47:28p [Deflat] COMDEMO.ASM 165 (233) May,07,2006 08:09:18p [Deflat] COMDEMO.COM 22 (22) Dec,03,2007 09:15:20p [Stored] EXEDEMO.ASM 131 (178) Feb,12,2005 01:59:58a [Deflat] EXEDEMO.EXE 52 (59) Dec,03,2007 09:15:22p [Deflat] DEVICE.ASM 400 (868) Feb,12,2005 01:59:58a [Deflat] DEVICE.SYS 88 (88) Dec,03,2007 09:15:20p [Stored] LIFE.ASM 886 (3,103) Oct,15,2005 05:29:52p [Deflat] LIFE.COM 365 (434) Dec,03,2007 09:15:22p [Deflat] MANDEL.ASM 445 (1,029) Sep,20,2007 11:01:12p [Deflat] MANDEL.COM 177 (183) Dec,03,2007 09:15:22p [Deflat] MULTISEG.ASM 223 (399) Oct,15,2005 05:29:38p [Deflat] MULTISEG.EXE 68 (101) Dec,03,2007 09:15:22p [Deflat] USEDPMI.ASM 611 (1,536) Dec,08,2006 11:38:26p [Deflat] USEDPMI.EXE 229 (261) Dec,03,2007 09:15:24p [Deflat] 42 items: 0 dirs + 42 files 257,999 bytes (1,012,685 unzipped) Created by ZIP 4.0 : DOS, OS/2, WinNT FAT Still doesn't work on XP (silently starts/exits), but DOSBox seems okay. So lemme test the WinME and DR-DOS cpus. |
|||
|
27 Jan 2008, 20:44 |
|
|
LocoDelAssembly 27 Jan 2008, 20:51
Quote:
Intel provides a library that installs a SEH handler so you can start doing SSE4 programs now and the exception handler will emulate them. (I posted a link somewhere in one of the peter's posts). |
|||
|
27 Jan 2008, 20:51 |
|
|
rugxulo 27 Jan 2008, 20:53
Nope, under WinME the DOS version still crashes/exits the DOS window.
revolution wrote:
Advantage to using PM being more compatible than unreal mode? Because DPMI is more compatible even still. I haven't tested (though I could) or really read up about old versions of Windows (e.g. 3.x), but something like VCPI doesn't work except in Standard mode. But I think DPMI works both (or at least in Enhanced). But whatever. It's obviously a minefield beyond my understanding. |
|||
|
27 Jan 2008, 20:53 |
|
|
rugxulo 27 Jan 2008, 21:12
P.S. Privalov, I wouldn't stress yourself out over this (today, at least). I mean, it's not like 1.67.23 is broken, so I'm still good, heh. (Oh darn, no SSE4, boo freakin' hoo. I still like plain ol' 8086!) But, just to hopefully share some helpful info:
Code: D:\TONY\*.* . (DIR) .. (DIR) fasm exe 75,192 attr6 asm 13,470 [ DR-DOS ] Sun 1-27-2008>fasm [ DR-DOS ] Sun 1-27-2008>fasm JemmEx: exception 0D occured at CS:EIP=04E5:0000000A, ERRC=00000000 SS:ESP=1850:00008000 EBP=000B17D6 EFL=00033246 CR0=80000011 CR2=00000000 EAX=10000471 EBX=00BC0000 ECX=00001850 EDX=FF80045B ESI=000D0000 EDI=00028000 DS=04E5 ES=04E5 FS=D208 GS=D208 [CS:IP]=67 A3 99 26 01 00 67 89 Press ESC to abort program [ DR-DOS ] Sun 1-27-2008>vv D:\TONY\*.* fasm exe 75,192 Jan,27,2008 09:16:12pm A... attr6 asm 13,470 Aug,27,2007 06:32:28pm A... [ DR-DOS ] Sun 1-27-2008>scrndump a:\fasm_err.txt Unreal mode (in DOS) seems to hang without a filename specified. But otherwise it seems to print like a billion whitespaces (continually scrolling the screen until I hit Ctrl-C). Just FYI. (Maybe you are cursed?!  ) |
|||
|
27 Jan 2008, 21:12 |
|
|
rugxulo 27 Jan 2008, 21:36
LocoDelAssembly wrote:
Not that I need it, but anyways, the closest link of yours I found via a quick search was this: http://softwarecommunity.intel.com/Wiki/DevelopforCoreprocessor/289.htm |
|||
|
27 Jan 2008, 21:36 |
|
|
LocoDelAssembly 27 Jan 2008, 21:48
http://board.flatassembler.net/topic.php?p=65013#65013 (the first post wasn't of peter
) |
|||
|
27 Jan 2008, 21:48 |
|
|
Tomasz Grysztar 27 Jan 2008, 22:04
rugxulo wrote: (Maybe you are cursed?! Sorry for the mess, it's just that after moving everything to a new machine much of my development/testing tools are broken (some things don't work well under xp64, you know), now I need to clean up. Give some time, I will try to upload a corrected package today still. |
|||
|
27 Jan 2008, 22:04 |
|
|
rugxulo 27 Jan 2008, 22:15
All I know about XP 64-bit is that it's (reputedly) 10% faster, some lack of drivers, and no V86 mode (i.e. no NTVDM / DOS stuff). It's so annoying (but slightly understandable) how "new" always breaks old. Oh well, can't have everything I guess.
P.S. Vista 32-bit is actually worse than XP 32-bit: only 32 MB DPMI limit (which I know I've mentioned a thousand times ... hopefully SP1 will come out "within a month" with that fix!). All the world's a VAX? No, but if it ain't Windows or Linux, people hate it: Quote:
Dare I say it ... thanks (big time) for the DOS support. Last edited by rugxulo on 13 Feb 2008, 09:02; edited 1 time in total |
|||
|
27 Jan 2008, 22:15 |
|
|
Tomasz Grysztar 27 Jan 2008, 23:34
OK, please try it now. I think I have cleaned up the mess a bit.
|
|||
|
27 Jan 2008, 23:34 |
|
|
rugxulo 28 Jan 2008, 21:39
Lightly tested it (without using "-d"), and it seems to work okay. WinXP, WinME, MS-DOS 6.22 all ran it fine (even unreal mode).
However, the problem comes into play when using "-dBLAH" without a value. It seems to not like that. (At least, unreal mode FASM prints a billion whitespaces until I hit Ctrl-C ... DPMI version probably would crash, dunno, didn't bother testing). "-dBLAH=" or "-dBLAH=1" seem to work. IMHO, "-dBLAH" should be the same as "-dBLAH=1", and I assume you'll probably agree. Of course, besides that minor nit (which hardly anyone utilizes yet, do they?), it's much more stable now.
EDIT: Thanks again!  Last edited by rugxulo on 30 Jan 2008, 03:54; edited 1 time in total |
|||
|
28 Jan 2008, 21:39 |
|
|
Tomasz Grysztar 28 Jan 2008, 21:45
rugxulo wrote: IMHO, "-dBLAH" should be the same as "-dBLAH=1", and I assume you'll probably agree. I'm not so sure about. The "-d" is the same as "define", that's why you can do "-dBLAH=" and it will do exactly the same as "define BLAH", that is - define symbolic constant with an empty value. And since symbolic don't really have much to do with numbers, it wouldn't make much sense to give it such a default value. However you're right about the "-dBLAH" bug, I'm going to fix it in 1.67.27. |
|||
|
28 Jan 2008, 21:45 |
|
|
rugxulo 28 Jan 2008, 22:30
Tomasz Grysztar wrote:
That's all I was really implying. You're of course free to "define BLAH" without value as default. No complaints here. |
|||
|
28 Jan 2008, 22:30 |
|
|
Kevin_Zheng 02 Feb 2008, 13:25
Hi, Tomaszy:
I recommed that you maybe combine all of release versions onto one package. For example, we can build below code tree as below figure: \fasm\ fasmd.exe ==> fasm for dos fasmw.exe ==> fasm for windows ide fasm ==> fasm for linux fasm.o ==> fasm for unix fasm.exe ==> fasm for windows console source\ all of versions inlude\ all of headers examples\dos\dos demo progames \windows\windows demo programs \linux\linux demo programs \unix\unix demo programs And other, PROC and ENDP are very excellent macro defines for FASM; but it only used for windows enviroment. Why don't use it for all os OS enviroment? I think that it maybe help other assembly users to transfer FASM world. I've built one macros to implement this idea and it seems okay. Please see the extended macro defines (proc and endp) in attahced source code package, I've used those macros "proc" and "endp" to build three demo programs for DOS, windows and Linux-64; it seems okay, now. Kenneth Zheng Jan 02, 2008
_________________ Pure Assembly Language Funs |
|||||||||||
|
02 Feb 2008, 13:25 |
|
| Goto page 1, 2 Next < Last Thread | Next Thread > |
Forum Rules:
|