Re: Loading single word to a xmm register

"ldb" <ldb_nospam@xxxxxxxxxxx> écrivait news:1141325493.315218.245130
@i40g2000cwc.googlegroups.com:


You just miss the point that i do not care of speed.
And you know the reason why?... _I have it_!

No you don't.

???!!!...

If you aren't using SSE instructions, your code is slower
then mine.

??? As long as i have written an Assembler that compiles
1 Mega/Second on a Celron 1.3, i suppose that you have
written one that Compiles 1.5 Megas/Second. Didn't you?


If

I love this "If"...

I write a program to do jpeg compression of a video
stream, and I use SSE instructions, I will likely be able to attain
twice the framerate you can...

You are right on this point. There are special cases when
Code-Level Optimizations are _necessary_. Such cases are
extreemely few, and your example is a very traditional good
one.

Mind you, dispiting what i say, if i had, say, to implement
a Reverb effect in some sound engine, i would _of course_
optimize the Routine at the Code Level intensively, and there
is no contradition, with this. The general rule, for all the
everydays works is that Code Level Optimization is nothing
but ridiculous, and the demonstration that the Program
prganization is absurd. There are simply _exeptions_ to
the general rule.


and I'd write most of my code in a HLL.

Guess why this is not a surprise to me... :)


Being "fast enough" depends entirely on the application. If you aren't
using SSE instructions for numeric loops, your code will always be
slower.

Feel free to play with this if you like, instead of
understanding the reason why you need it, but your
ideas are very typical of an HLLers, who does not
understand what he is doing, what can never be the
case of an Assembly Programmer.


Betov.

< http://rosasm.org >


.

Relevant Pages

  • Re: Embedding assembler in a language
    ... be used to improve the performance of the main language instead. ... This is also the reason why I see ... for handcoded assembler, although there are a few rare exceptions. ... I decided to provide a solution in the Seed7 runtime library. ...
    (comp.lang.misc)
  • Re: yet another Forth: vmForth
    ... and MidLevel and HLL all Integrated in the Kernel. ... VM assembler and simulator from within the simulator for testing :-) ... higher level Forth constructs, like jz/jnz, lit, execute, call and jump. ... Compiles VM assembler mnemonics into the VM memory. ...
    (comp.lang.forth)
  • yet another Forth: vmForth
    ... higher level Forth constructs, like jz/jnz, lit, execute, call and jump. ... Compiles VM assembler mnemonics into the VM memory. ... (xt is address of type flag) ...
    (comp.lang.forth)
  • Dont be overawed by a C compiler.
    ... This is a classic x86 assembler newsgroup where assembler issues are ... a C compiler can do it better. ... measurably faster on current Intel and AMD hardware, there is no reason ... and while a lot of it is junk, a reasonable amount of the code ...
    (comp.lang.asm.x86)
  • Re: from elsewhere, an assembler
    ... compilation. ... reasoning mostly that assembler now exists in a state of great decline ... otherwise, from my POV, there is not much reason to use ... is used for the whole opcode). ...
    (alt.lang.asm)