Re: FastMove ?

FastMove should never be slower than using the standard system move. In
fact, the system move routine uses virtually the same code as used by
FastMove (for processors that do not support MMX, SSE etc). It could be
that just adding FastMove to the project causes memory to be allocated from
a different location (not in level 1/2 cache etc), or causes another routine
to be sub-optimally aligned etc.

How much slower is it, and how are you measuring the move performance?

--
Regards,
John

--
The Fastcode Project: http://www.fastcodeproject.org/

"Sanyin" <prevodilac@xxxxxxxxxxx> wrote in message
news:48032fa2$1@xxxxxxxxxxxxxxxxxxxxxxxxx

"John O'Harrow" <john@xxxxxxxxxxxxxxxxxxxx> wrote in message
news:47f4feb6$1@xxxxxxxxxxxxxxxxxxxxxxxxx
Version 3.03 (March 2006) is still the latest version.

It can be found at http://fastcode.sourceforge.net/

--
Regards,
John

When using fastmove from pointer obtained by windows memory mapping API,
It is slower than regular move (d7)...
Any reason for this?



.

Relevant Pages

  • Re: FastMove ?
    ... FastMove (for processors that do not support MMX, SSE etc). ... routine to be sub-optimally aligned etc. ... How much slower is it, and how are you measuring the move performance? ...
    (borland.public.delphi.language.basm)
  • Re: FastCode Poll - Which Functions are Used
    ... Move slower. ... That's why i proposed the FastMove. ... The Fastcode library should simply have all the winner functions and nothing ... Fastmove code isn't efficient moving few amount of bytes ...
    (borland.public.delphi.language.basm)
  • Re: just for curiosity fastmove slower in some case
    ... Hi John, ... fastmove is 10% slower than without. ... Prev by Date: ...
    (borland.public.delphi.language.basm)
  • just for curiosity fastmove slower in some case
    ... just like inform that with the poker benchmark library, D2006 with fastmove is 10% slower than without. ... Maybe with small data pieces the plain AI32 is more efficient that using bigger algorithm? ... Anyway a typical delphi software average the 90% of the ram move calls are used in smallest pieces. ...
    (borland.public.delphi.language.basm)
  • Re: just for curiosity fastmove slower in some case
    ... Roby DP wrote: ... > Hi John, ... > fastmove is 10% slower than without. ... D2006 uses Fastmove by default doesn't it? ...
    (borland.public.delphi.language.basm)