On 4/26/09 9:57 PM, "Ruslan Zasukhin" <[email protected]> wrote:
Hi REAL,
Hi Geoff,
I want once again point on PROBLEM I see in work with Arrays from plugins.
Also I wonder if the same problem lives inside of RB.
PROBLEM is very bad performance.
1) if we use REALobjectArray then we have only way extract items of variants
using
REALGetPropValue()
but as Apple Shark show easy this means that RB calls PARSER yylex
on EACH such call.
it is obvious that this is SLOW.
I THINK that much better way will be add into PLUGIN SDK
few methods to GET/SET different types of VARIANT values:
GetInt16() SetInt16()
...
GetDouble() SetDouble
...
As result we will get VERY FAST to variant values.
Again, right now REALObjectArray is very slow.
It is slower of REALstringArray in about 10 times (!!!)
Although expected to be 4-5 times FASTER of REALstringArray,
Because no string conversions when work with numbers...
Big Performance speed up possible here :-)
2) REALstringArray -- here also I see odd things.
In particular when I have run test under Shark / Malloc watching,
It show that internally REALBasic do wow conversions
of Doubles and Ints into strings.
Why this happens? I just cannot understand it,
Because in RB code I have really create array of STRINGs
And did assign STRINGs ...
-------------------------------------------------------------------------
> This is from shark.
>
> Work of VARIANT is very slow.
> Tested with RB 2009, but I have see this yet in RB 2005 I believe.
>
>
> 0.0% 80.2% V4RB.rbx_0.dylib
REALGetPropValue(REALobjectStruct*, char const*, long*)
> 0.4% 80.0% rbframework.dylib REALGetPropValueInt
> 0.6% 69.0% rbframework.dylib REALLoadObjectMethod
> 0.1% 62.3% rbframework.dylib
BuildClassEvent(ObjectDefinition*, REALevent*)
> 0.9% 61.2% rbframework.dylib REALSetPropValueCFStringRef
> 0.1% 50.4% rbframework.dylib yylex()
> 2.9% 50.1% rbframework.dylib yylex_ex()
> 1.1% 15.8% rbframework.dylib string::ConstructFromBuffer(char
const*, unsigned long)
> 1.5% 12.7% rbframework.dylib string::AllocateBuffer(unsigned
long, void* (*)(unsigned long))
> 0.7% 11.0% libSystem.B.dylib malloc
> 0.6% 10.2% libSystem.B.dylib malloc_zone_malloc
> 2.9% 9.7% libSystem.B.dylib szone_malloc
> 3.6% 4.8% libSystem.B.dylib tiny_malloc_from_free_list
> 1.3% 1.3% libSystem.B.dylib tiny_free_list_add_ptr
> 1.5% 1.5% libSystem.B.dylib __spin_lock
> 0.2% 0.2% libSystem.B.dylib dyld_stub__spin_lock
> 0.1% 0.1% libSystem.B.dylib spin_unlock
> 0.1% 0.1% libSystem.B.dylib dyld_stub__spin_unlock
> 0.1% 0.1% libSystem.B.dylib spin_lock
> 0.0% 0.0% libSystem.B.dylib tiny_free_list_add_ptr
> 0.1% 0.1% libSystem.B.dylib dyld_stub_malloc_zone_malloc
> 0.0% 0.0% libSystem.B.dylib szone_malloc
> 0.1% 0.1% rbframework.dylib dyld_stub_malloc
> 0.3% 2.0% rbframework.dylib umemcpy(void*, void const*,
unsigned long)
> 0.0% 0.0% libSystem.B.dylib __memcpy
> 0.7% 14.9% rbframework.dylib stringStorage::RemoveReference()
--
Best regards,
Ruslan Zasukhin
VP Engineering and New Technology
Paradigma Software, Inc
Valentina - Joining Worlds of Information
http://www.paradigmasoft.com
[I feel the need: the need for speed]
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives:
<http://support.realsoftware.com/listarchives/lists.html>
