Re: Which assembler can handle the BIG stuff ?


"Frank Kotler" <fbkotler@xxxxxxxxxxx> schrieb im Newsbeitrag
news:UdydnUPkCrouoHnfRVn-3g@xxxxxxxxxxxxxx
| wolfgang kern wrote:
| > Thanks Annie,
| >
| > | > but I have no clue what 'X' you mean here.
| > | > Don't care ...
| > | > __ _____
| > | > wolfgang ((( `\
| > | _ _`\ )
| > | Frank is talking about the (^ ) )
| > | X-Window GUI on Linux, Wolfie. ~-( )
| > | _'((,,,)))
| > | As you know, the Linux GUI ,-' \_/ `\
| > | is not a part of the O.S. ( , |
| > | It's a separate application, `-.-'`-.-'/|_|
| > | somewhat like WinDoze 3.xx. \ / | |
| > | =()=: / ,' aa
| >
| > I see, it could be compared to my 2KB OS-core module
| > (native graphics/coloured text).
|
| You could say "compared" - unfavorably. "Contrasted" might be better. I
| expect your 2K has less functionallity than the full-blown "X windows
| system"... but not proportionally so.
|
| > But what I then not understand is the huge speed difference
| > for counting character occurence in a memory image... ?
|
| The problem is that with all the "graphics cruft" in memory, I'm
| swapping pages to disk. I've "only" got 128M of physical ram. Minimum
| requirement for X plus KDE (a fat "window manager") is 192M or something...
|
| > Let me estimate how long it would take on my old 500 Mhz AMD:
| >
| > 38 MB somewhere in RAM, lets point it by esi,
| > reserve a cleared result-field (table) for counts of interest
| > ie: 256*4 bytes and let edi point to
| >
| > mov esi,... ;source
| > mov edi,... ;result-array
| > mov ecx,... ;the size of it (non-zero)
| > loop:
| > movzx al,[esi]
|
| movzx ???
|
| > inc esi
| > ;cmp al,20h ;*only if a filtered range is desired
| > ;jc nocnt
| > ;cmp al,80h
| > ;jnc nocnt
| > inc dword[edi+eax*4]
|
| How's this compare to "add [...], 1"?
|
| > nocount: ;*
| > dec ecx
| > jnz loop ;or JNS/JNGE if endmark included in size
| > --- and that's it.
| > (* I doubt any filter will increase speed here,
| > so I'd just ignore unwanted results.)
|
| Right. This is our "inner loop", more or less. Anything we do here
| happens for each character, so "wasting" a bit of memory to count
| "ignored" characters is a speed increase, compared to any "special
| caseing", AFAICS.
|
| > 1024 bytes results (only a temporary needed buffer!)
| > plus just a handful code bytes.
| >
| > Takes (w/o filter) 8..10 cycles per iteration,
| > which I would count as 304..380 Mcycles,
| > which is less (~0.75) than a second on an 500 MHz machine.
|
| I'm only doing 300MHz, but - given some memory to work in - I'm seeing
| results in this ballpark.
|
| > I forgot to sort and display the result ?
|
| Yeah, but this is done once, not once per character, so it isn't going
| to add that much.
|
| > Ok, I'd use KESYS Fn50 to create a formatted, sorted table
| > which would take ~200 mS on an 1024*768, 8-bit color full screen.
|
| Well, I haven't got "Fn50", but I think I get the idea. I don't sort
| anything (yet). Sorting algorithms have "been done", but might be worth
| discussing. The "newer guys" (including myself) can learn something, and
| the "experts" can show off how smart they are... and *they* might learn
| something, too!
|
| > So all together is less than a second on my old 500 MHz.
|
| As I said, at 300MHz I'm not quite matching that, but getting close.
| I've got a reply to write to Alex, then maybe I'll post some more
| (crude) code. Probably in a brand new thread - as you observed, this
| one's getting cluttered.
|
| Best,
| Frank

.

Relevant Pages

  • REVIEWS: The X-Axis - 18 December 2005
    ... And for storytelling purposes, he's a good point of view character for the baffled reader, which helps to keep Cable's schemes more mysterious and suspicious. ... Their solution is to keep Quicksilver with one foot in the superhero world. ... Usually this sort of thing would lead to him failing to realise his goal but learning an Important Lesson about what he really wanted. ... WOLVERINE is the only one I'm buying, but from the descriptions, it's not immediately clear how that makes any difference. ...
    (rec.arts.comics.marvel.xbooks)
  • Re: looking for implementation of strtok
    ... contains a ',' character. ... we enter the ugly loop that simply implements strspn: ... loop sets spanp to delim, then loops while seting sc (the "span ... If there are no more tokens, we set *lastp to NULL (this is, I ...
    (comp.lang.c)
  • Re: And the WoW clone rain begins!
    ... help make raids a bit more accessible to more players. ... like that sort of thing. ... If your first character was ... yeah to me it's pretty much game over at max ...
    (comp.sys.ibm.pc.games.rpg)
  • Re: help with VBA
    ... Within your loop you determine that you're at ... and know that you need to sort everything on all of the rows from 4 to this ... "indraneel" wrote: ... The first thing to do is to name the ranges you want sorted, ...
    (microsoft.public.excel.worksheet.functions)
  • Re: Using "uniq" or something different to get rid of duplicate lines.
    ... Use character SEPARATOR as the field separator when finding the ... empty string between a non-blank character and a blank character. ... That is, given the input line ` foo bar', `sort' breaks it into ... To specify a null character as the field separator, ...
    (comp.unix.shell)