Board index » cppbuilder » Re: bad code generation

Re: bad code generation


2006-07-20 07:36:20 AM
cppbuilder65
At 01:21:36, 20.07.2006, Ed Mulroy wrote:
Quote
I do not know what FileZilla does or why it is needed.
It is an ftp program, and a good one as well.
And it does work with FileZilla. I entered the data as described, and was
able to see and download everything.
For many donwloads, FileZilla is better than IE, believe me. <g>
--
Rudy Velthuis [TeamB] rvelthuis.de/
"We are not retreating - we are advancing in another Direction."
-- General Douglas MacArthur (1880-1964)
 
 

Re:Re: bad code generation

At 01:32:45, 20.07.2006, Brion L. Webster wrote:
Quote
IE handles case-senstive things a little more loosely than other
programs.
Actually, it is the OS of the FTP host which decides if case sensitivity
is a problem or not.
--
Rudy Velthuis [TeamB] rvelthuis.de/
"All our knowledge merely helps us to die a more painful death than
animals that know nothing." -- Maurice Maeterlinck.
 

Re:Re: bad code generation

"Jonathan Benedicto" < XXXX@XXXXX.COM >wrote in message
Quote
Dennis Jones wrote:
>Unfortunately, 2.0.7 is even worse -- it just hangs. At least 2.0.6
>compiled (albeit very slowly).

I found and fixed that bug. Please could you download
www.jomitech.com/downloads/mtbcc32CompileExecutor.zip and extract
into C:\Program Files\JomiTech\TwineCompiler.
Okay, this version is better, but it still hangs occasionally. Out of four
attemps, it hung twice: once in the IDE and once on the command line. In
both cases I had to kill the executor tasks and execute the build again.
My guess is you have a deadlock or a race condition somewhere.
- Dennis
 

{smallsort}

Re:Re: bad code generation

Dennis Jones wrote:
Quote
Okay, this version is better, but it still hangs occasionally. Out of
four
attemps, it hung twice: once in the IDE and once on the command line. In
both cases I had to kill the executor tasks and execute the build again.

My guess is you have a deadlock or a race condition somewhere.
This is really strange.
After it hangs again, please kill the executors and send me mtbcc32.log from
BCB\Bin, and all the mtbcc32CompileExecutor*.log files from C:\Program
Files\JomiTech\TwineCompiler.
Jonathan
 

Re:Re: bad code generation

PS. Please could you send it to support at jomitech dot com.
Jonathan
 

Re:Re: bad code generation

Quote
it is universal (and the best IMO) linker supporting bcc, bcc32, msc,
msvc, gcc (cygwin/mingw), masm, tasm, watcom, icl, dm, you_name_it

as regarding to bcb AFAIK it supports almost all ilink options and
offers many more. it produces target executables in order of
magnitudes faster, smaller up to 20% and several times smaller tds
files
How can it be used with BCB(5/2006)? Can it directly replace the
current linker or must a third-party build tool (ie Twine, BB) make use
of it?
Thanks
David
 

Re:Re: bad code generation

On Thu, 20 Jul 2006 10:28:36 +0100, David Perkins < XXXX@XXXXX.COM >
wrote:
Quote
How can it be used with BCB(5/2006)? Can it directly replace the
it depends on what meaning you put in "directly" :) if you build your
projects from within IDE without 3rd party tools you probably won't be
able to use ulink. however this means you have no any problems with
bcb speed so you won't need it anyway :)
otherwise you can specify what linker to use using command-line tools,
see bcb5 online help or Ed Mulroy's howtos
Quote
current linker or must a third-party build tool (ie Twine, BB) make use
I never used Twine/BB so cannot comment on that, but bcc32pch supports
alternate linker
I personally mostly use bcc32pch Execute Makefile window with
Additional make parameter: -DLINKER=ulink
--
Vladimir Ulchenko aka vavan
 

Re:Re: bad code generation

Thanks for that, I'll check it out when time allows...
 

Re:Re: bad code generation

Quote
For many donwloads, FileZilla is better than IE,
believe me. <g>
Ok, I'll look around for it. It might be interesting to play with.
. Ed
Quote
Rudy Velthuis wrote in message
news:xn0eoxoy2ma1y9c00m-velthuis@www.teamb.com...

>I do not know what FileZilla does or why it is needed.

It is an ftp program, and a good one as well.

And it does work with FileZilla. I entered the data as described,
and was able to see and download everything.

For many donwloads, FileZilla is better than IE, believe me. <g>
 

Re:Re: bad code generation

In article <xn0eoxru8mdye7v000-velthuis@www.teamb.com>,
"Rudy Velthuis [TeamB]" < XXXX@XXXXX.COM >wrote:
Quote
Actually, it is the OS of the FTP host which decides if case sensitivity
is a problem or not.
Are more particularly, the filesystem used by the OS. Some OSs can
handle more than one type of filesystem, and some filesystems come in
both case sensitive and non case sensitive versions.
--
-David
Quis custodiet custodes ipsos?
 

Re:Re: bad code generation

David Perkins wrote:
Quote
How can it be used with BCB(5/2006)? Can it directly replace the current
linker or must a third-party build tool (ie Twine, BB) make use of it?
For BDS yes, go to the BDS\Bin dir, and rename ilink32.exe to
old_ilink32.exe, and then paste ulink into this dir, and rename it to
ilink32.exe. This should do it.
HTH
Jonathan
-----------------------------------------------------------------
TwineCompiler - Blazingly fast C++ compiles on ANY machine.
www.jomitech.com/
-----------------------------------------------------------------
 

Re:Re: bad code generation

J Alexander a écrit :
Quote
2) Declaring functions __fastcall can result in bad generated code. I got access
violation crashes after one template class library was peppered with
__fastcall's "to speed it up." Don't use __fastcall, it hardly improves
performance, but can generate bad code.
If the implementation of this function is not compatible with the
__fastcall calling convention; for sure it will crash. This case often
happens when you try to import functions from a DLL.
--
Florent Ouchet
 

Re:Re: bad code generation

"Florent Ouchet" < XXXX@XXXXX.COM >wrote in message
Quote
J Alexander a écrit :
>2) Declaring functions __fastcall can result in bad generated code. I got
>access violation crashes after one template class library was peppered with
>__fastcall's "to speed it up." Don't use __fastcall, it hardly improves
>performance, but can generate bad code.

If the implementation of this function is not compatible with the __fastcall
calling convention; for sure it will crash. This case often happens when you
try to import functions from a DLL.
Yeah, that's the first thing one would think about, but the guess is incorrect.
There wasn't DLL or anything with different calling conventions. The function
was part of a static library, compiled together with the main application with
the same compiler settings. Knowing that BDS Make may miss dependencies,
everything was painfully rebuilt.
I suspect on the call site function parameters were put into registers but
function implementation still got them from the stack. Examining function
parameters in the de{*word*81} on the call site they were correct, inside the
function they were complete mess. But it's my guess, I don't have time for
experiments, all __fascalls are gone now and so are those problems.
Interestingly, there were many other template functions with __fastcall from the
same library called before that function and they did the right thing.
J
 

Re:Re: bad code generation

At 16:19:55, 20.07.2006, David Dean wrote:
Quote
>Actually, it is the OS of the FTP host which decides if case
>sensitivity is a problem or not.

Are more particularly, the filesystem used by the OS.
Correct.
--
Rudy Velthuis [TeamB] rvelthuis.de/
"Should array indices start at 0 or 1? My compromise of 0.5 was
rejected without, I thought, proper consideration."
-- Stan Kelly-Bootle
 

Re:Re: bad code generation

At 14:53:21, 20.07.2006, Ed Mulroy wrote:
Quote
>For many donwloads, FileZilla is better than IE,
>believe me. <g>

Ok, I'll look around for it. It might be interesting to play with.
I can really recommend it.
--
Rudy Velthuis [TeamB] rvelthuis.de/
"I failed to make the chess team because of my height." -- Woody Allen