Re: MASM Expert needed immediately

"f0dder" <f0dder_nospam@xxxxxxxxxxxxxxxx> écrivait
news:447845f3$0$60786$157c6196@xxxxxxxxxxxxxxxxxxxxx:

Betov wrote:
Phil Carmody <thefatphil_demunged@xxxxxxxxxxx> écrivait
news:87u07b6ad3.fsf@xxxxxxxxxxxxxxxxxxxx:

Funnily enough I do have an assembly related question (a week
ago I tried to optimise 20 lines of C code into 10 lines of
SSE2 asm, and made it 15% slower in so doing).

Two errors at a time:

1) Optimizing at Code Level is simply absurd.

2) SSE has never been a thing but selling propaganda,
as demonstrated by all and any attempt like yours.

Heh.

http://board.flatassembler.net/topic.php?t=5122 . Optimizing from x87
FPU to SSE2 code gives...

~185% speed of the original for Pentium4/Prescott
~137% speed of the original for AMD64x2,
~219% speed of the original on Merom (upcoming CPU from intel)

Not to mention that mandelbrot rendering is quite a nice example of
where threading is good - we achieved almost perfect parallelism on
the code (for two CPUs, anyway).

Zero interrest.


Betov.

< http://rosasm.org >




.

Relevant Pages

  • Re: MASM Expert needed immediately
    ... ago I tried to optimise 20 lines of C code into 10 lines of ... SSE2 asm, and made it 15% slower in so doing). ... ~219% speed of the original on Merom (upcoming CPU from intel) ... Betov:) ...
    (alt.lang.asm)
  • Re: MASM Expert needed immediately
    ... ago I tried to optimise 20 lines of C code into 10 lines of ... SSE2 asm, and made it 15% slower in so doing). ... ~219% speed of the original on Merom (upcoming CPU from intel) ...
    (alt.lang.asm)
  • Re: MASM Expert needed immediately
    ... ago I tried to optimise 20 lines of C code into 10 lines of ... SSE2 asm, and made it 15% slower in so doing). ... SSE has never been a thing but selling propaganda, ...
    (alt.lang.asm)