forums.silverfrost.com

[Chemie]

Hello
I'm transferring now all my programs to 64bit, they will be faster and I can allocate memory up to 12 GB unter 16GB RAM.
I have only problems with browse_for_folder@, it always produce an arror. I get an Silverfrost exception report, RPC server is unavailable within KERNELBASE.dll.
On the other hand, I can't read with %ss any prarameters from the .INI file and also, I can't write to the .INI file.
What I have made wrong?

Thanks for Help

Chemie

[PaulLaidler]

In both cases a short sample program would help to identify reasons for the failure. browse_for_folder@ should be OK. There might be problems with %ss concerning protected folders. Post some samples and I will check them out.

[Chemie]

Hi Paul,
here are the two sample programs:
browse_for_folder@:

winapp
program suchefolder
use mswin
implicit none
character*80 title,path
logical folder
title='Suche Folder'
call browse_for_folder@(title,path)
! folder=browse_for_folder@(title,path)
end

In both cases of call there will be the exception report.

%ss:

winapp
program sstest
use mswin
implicit none
integer*4 ssctrl,testvalue,a,openctrl
testvalue=0
ssctrl=-1
a=winio@('%ss[SSTest/SSTest]&',ssctrl)
a=winio@('%ww[invisible]&')
a=winio@('%nlTestvalue: %rd&',testvalue)
a=winio@('%lw',openctrl)
call sleep@(0.01)
openctrl=0
call window_update@(openctrl)
a=winio@('%ca[Result]&')
a=winio@('%nlResult: %wd&',testvalue)
a=winio@('%2nl%cn%9`bt[OK]')
end

The File SSTEST.INI contains:

[SSTest]
Testvalue=1000

When I build with win32, testvalue is 1000, with x64 testvalue is 0.
Thank you for looking.
Chemie

[PaulLaidler]

I can confirm that both of these need fixing for 64 bits and I made a note of this.

[silicondale]

I'm getting the same problem. It opens the browse-for-folder dialog window but fails when I click on almost anything - though did actually create a new folder when I simply clicked on 'Make new folder' (in the default location)

[PaulLaidler]

The %ss issue has now been fixed.

For me, browse_for_folder@ does not even get started (Exception: RPC server not available).

[Chemie]

[PaulLaidler]

64 bit browse_for_folder@ has now been fixed for the next release.

Use /debug and try launching SDBG64.exe using your exe and step through the program or click on "continue".

[Chemie]

[JohnCampbell]

The problem appears to be with the result of ichar for characters with the 8th bit set. /64 gives the wrong result. Looks as if ichar is giving integer*1 result then equated to I*4. See the following code for /32 and /64

[Chemie]

Hello

Sure, it works with iand(ichar(c),255) instead of ichar(c) but unfortunately, I have do change all my programs.

But, sorry, I have found 2 more problems:

winapp

program timetest
use mswin
implicit none

integer*4 a
real*8 zeitanf,zeitend

zeitanf=HIGH_RES_CLOCK@(.TRUE.)
call sleep@(1.0)
zeitend=HIGH_RES_CLOCK@(.FALSE.)

a=winio@('%ca[Meldung]&')
a=winio@('%cnCPU-Zeit!&')
a=winio@('%2nlDie CPU-Zeit betr�gt: %wf sec&',zeitend-zeitanf)
a=winio@('%2nl%cn%9`bt[OK]')

end
gives absurd values for the cputime, always different, when I start once more. Win32 works normal.

The other problem is the reading of jpeg-files. get_dib_size@ and get_dib-block@ are not working with some (in my case photos from my digitalcamera, I can't send here an example) jpegfiles. I don't know, what is the difference between my photos and other "normal" jpegfiles. The error occurs also with Win32.

Chemie

[Chemie]

The diffrence between my digitalcamera jpegs and normal jpegs is that there are missing the JFIF-notation. When I insert at position 3 the sequenz

FF E0 00 10 4A 46 49 46 00 01 01 01 00 B4 00 B4 00 FF E1

then I have a normal jpeg.

Chemie

[JohnCampbell]

Chemie,

HIGH_RES_CLOCK@ is VERY badly named, as it does not work in /64 and works very badly in 32 bit.

Preferable /64 routines for elapsed time are:
SYSTEM_CLOCK intrinsic,
STDCALL QUERYPERFORMANCECOUNTER 'QueryPerformanceCounter' (REF):LOGICAL*4, or
INTEGER(KIND=4) FUNCTION RDTSC_VAL@()

The clock tick rate is an issue. For these last 2 timers is found by:
STDCALL QUERYPERFORMANCEFREQUENCY 'QueryPerformanceFrequency' (REF):LOGICAL*4
From my recent testing, RDTSC_RATE@ = QueryPerformanceFrequency * 1024.

All these timers appear to be based on RDTSC

My recommendation is to use SYSTEM_CLOCK (ease of use), or RDTSC_VAL@() (most accurate)

Note: The function cpu_clock@ is not available for 64 bit applications and has been replaced by rdtsc_val@...
INTEGER(KIND=4) FUNCTION RDTSC_VAL@()

[JohnCampbell]

As a more extensive comparison of preferred elapsed time timers, all based on RDTSC and working for /32 or /64, the following example is a approach I have adopted and works on all recent pc's I have.

[JohnCampbell]