Board index » delphi » Re: Who is Delphi competing against?

Re: Who is Delphi competing against?


2008-05-21 12:05:03 AM
delphi167
Adem writes:
Quote
Rudy Velthuis [TeamB] writes:

>IOW, people can already suggest improvements based on what they can
>see.

True. But, that adds several barriers.

-- It isn't easy to read (and comprehend) the output at FPU/CPU
window.
I find it pretty easy.
Quote
-- Suggestions would be too abstract to be meaningful
Not at all. One can say: hey, every time this and that is done, there
is an unnecessary opcode such and such. After all, the compiler and
linker produce machine code, not readable source.
--
Rudy Velthuis [TeamB] www.teamb.com
"Imagine if every Thursday your shoes exploded if you tied them
the usual way. This happens to us all the time with computers,
and nobody thinks of complaining." -- Jeff Raskin.
 
 

Re: Who is Delphi competing against?

Anders Isaksson writes:
Quote
Suggestions would not be abstract at all. They would concentrate on
the generated code, and how it could look when opitimized - no need
for every suggester to understand the compiler.
Exactly.
--
Rudy Velthuis [TeamB] www.teamb.com
"One of the symptoms of an approaching nervous breakdown is the
belief that one's work is terribly important."
-- Bertrand Russell (1872-1970)
 

Re: Who is Delphi competing against?

Alan Garny writes:
Quote
"willr" <XXXX@XXXXX.COM>writes
news:4832e274$XXXX@XXXXX.COM...
>Alan Garny writes:
>>>We are doing mapping projects, sample data collection and processing and
>>>geophysical modeling -- including mag data processing and datum and
>>>projection conversions -- yadda yadda and so forth.
>>>
>>>Delphi for the overall application is pretty good.
>>>
>>>The scientific math is a touch weak tis true. :-(
>>Why not the mathematical part in C/C++ and compile it to a DLL using the
>>Intel C++ compiler? that is certainly what I'd do. In fact, I have
>>done that for a part of my project which uses a third party C numerical
>>library (for solving ODE systems).
>I am looking more at the Complex Number functions and array processing
>than anything. I don't think C++ is much better. I could be wrong.

I am not really advocating C++ over Delphi, but rather the Intel compilers
(C++ and Fortran).

Alan


Duh...
Of course -- plus Linpac -- or whatever the latest incarnation is -- I
actually did a time estimate on moving the Array Processing to the Linux
Server for just this reason -- since Interbase is there now I can do
UDF's or whatever to process the arrays.
I was looking at the Sun Fortran Compiler -- 64 bit -- parallel
processing...
I think I am back on track here.
--
Will R
PMC Consulting
 

Re: Who is Delphi competing against?

Adem writes:
Quote
-- It isn't easy to read (and comprehend) the output at FPU/CPU
window.
Somebody who can read compiler source would find it easy.
--
Dave Nottage [TeamB]
 

Re: Who is Delphi competing against?

Anders Isaksson writes:
Quote
Adem writes:

>-- It isn't easy to read (and comprehend) the output at FPU/CPU
>window.

But it is definitely easier than reading (and comprehending) the
complete compiler source!
You wouldn't always need to read and comprehend the whole source. If
you had the source, you could step into it and then trace your problem
back from that point.
Quote
>-- Suggestions would be too abstract to be meaningful --they
>could not refer or offer alterations/additions to the code lines.

Suggestions would not be abstract at all. They would concentrate on
the generated code, and how it could look when opitimized - no need
for every suggester to understand the compiler.
I can see that you [plus 2 TeamB heavies :) ] basically agree on this,
but from where I look at it, the generated coder isn't Pascal (here, I
am assuming compiler source is Pascal). Which means, anyone who isn't
assembeler guru will be reduced to naggers..
Quote
The most difficult thing when getting suggestions about changes in
software is that too few people actually tell what the problem is, and
too many suggests a solution/implementation instead, without having
all the facts.
But, is this really the shared experience with large open source
projects?
When you have access to the source, chances are, you're more likley to
describe the problem --or whereabouts of it.
If you have no access, all you can do is to file an abstract bug report
--along with a piece of code to reproduce it, that is if you're lucky
to be able to come up with something short and suitable--, and expect
some poor time constrained soul on the other end of the screen will
first encounter you bug report, then find it
interesting/comprehensible/important enough to work on it; and finally
have the time to actually do it..
Such a long chain that one feels rather lethargic to even think about
filing a bug report..
 

Re: Who is Delphi competing against?

It would be great, and a good selling point to the scientific/engineering
community, if CG would license the Intel Performance Primitives and perhaps
the Math Kernal Library and include them and their proper header conversions
in the package. Then Delphi would be in my opinion untouchable for ease of
use and power.
Dan Sox
"Alan Garny" <XXXX@XXXXX.COM>writes
Quote
"willr" <XXXX@XXXXX.COM>writes
news:4832e274$XXXX@XXXXX.COM...
>Alan Garny writes:
>>>We are doing mapping projects, sample data collection and processing
>>>and geophysical modeling -- including mag data processing and datum and
>>>projection conversions -- yadda yadda and so forth.
>>>
>>>Delphi for the overall application is pretty good.
>>>
>>>The scientific math is a touch weak tis true. :-(
>>
>>Why not the mathematical part in C/C++ and compile it to a DLL using the
>>Intel C++ compiler? that is certainly what I'd do. In fact, I have
>>done that for a part of my project which uses a third party C numerical
>>library (for solving ODE systems).
>
>I am looking more at the Complex Number functions and array processing
>than anything. I don't think C++ is much better. I could be wrong.

I am not really advocating C++ over Delphi, but rather the Intel compilers
(C++ and Fortran).

Alan

 

Re: Who is Delphi competing against?

DJSox writes:
Quote
It would be great, and a good selling point to the scientific/engineering
community, if CG would license the Intel Performance Primitives and perhaps
the Math Kernal Library
Just an idea: Why don't jou propose to the Basm/Fastcode guys to develop
such a library from scratch? They love such challenges.
--
Arthur Hoornweg
(In order to reply per e-mail, please just remove the ".net"
from my e-mail address. Leave the rest of the address intact
including the "antispam" part. I had to take this measure to
counteract unsollicited mail.)
 

Re: Who is Delphi competing against?

Rudy Velthuis [TeamB] writes:
Quote
The source of the RTL and VCL is open to see for every Delphi Pro+
user. The source of the compiler isn't, but it only takes a few glances
in a CPU/FPU window to see what kind of code it generates in which
circumstances, and that is the only thing that counts.

But that does not give the true picture as to how the compiler was coded.
 

Re: Who is Delphi competing against?

Adem writes:
Quote

I can see that you [plus 2 TeamB heavies :) ] basically agree on this,
but from where I look at it, the generated coder isn't Pascal (here, I
am assuming compiler source is Pascal). Which means, anyone who isn't
assembeler guru will be reduced to naggers..

I doubt the compile is coded in Pascal.
 

Re: Who is Delphi competing against?

"Arthur Hoornweg" <XXXX@XXXXX.COM>writes
Quote
DJSox writes:
>It would be great, and a good selling point to the scientific/engineering
>community, if CG would license the Intel Performance Primitives and
>perhaps the Math Kernal Library

Just an idea: Why don't jou propose to the Basm/Fastcode guys to develop
such a library from scratch? They love such challenges.
Remember: the Intel guys know (or should know) their processors inside out
and they have been working on their compiler and libraries for years. To try
and 'compete' against them would be very time consuming and potentially
meangingless: Intel will always be ahead (as is Microsoft with .NET).
So, why try to reinvent the wheel when you can just buy it from its original
inventor?!
Alan
 

Re: Who is Delphi competing against?

Adem writes:
Quote
You wouldn't always need to read and comprehend the whole source. If
you had the source, you could step into it and then trace your problem
back from that point.
Do you have *any* idea about the complexity of a compiler the size of
Delphi's? Do you have *any* idea how many steps you would have to go
before coming to the interesting parts? How do you know what the
interesting part is? A problem seen in the final codegen end could be
caused by something done during the initial parse tree generation.
Quote
here, I am assuming compiler source is Pascal
AFAIK, the compiler is mostly written in C - google for "Danny Thorpe
Delphi Compiler" to make sure.
Quote
>The most difficult thing when getting suggestions about changes in
>software is that too few people actually tell what the problem is,
>and too many suggests a solution/implementation instead, without
>having all the facts.

But, is this really the shared experience with large open source
projects?
Probably, yes. At least it is my experience, both in commercial
software development and (as a bystander) in OS projects. People are
much more interested in telling the devs "what to do" instead of
elaborating on "why should it be done".
Quote
When you have access to the source, chances are, you're more likley to
describe the problem --or whereabouts of it.
In general, I don't think so. Of course, some people are able to grok
large code bases quickly, and give good advise (or direct code fixes),
but for most of us a sufficiently large code base makes it almost
impossible to get the insight needed to actually be useful (talking
about *large* code bases here).
--
Anders Isaksson, Sweden
BlockCAD: web.telia.com/~u16122508/proglego.htm
Gallery: web.telia.com/~u16122508/gallery/index.htm
 

Re: Who is Delphi competing against?

Yogi Yang 007 writes:
Quote
Rudy Velthuis [TeamB] writes:
>The source of the RTL and VCL is open to see for every Delphi Pro+
>user. The source of the compiler isn't, but it only takes a few
>glances in a CPU/FPU window to see what kind of code it generates
>in which circumstances, and that is the only thing that counts.
>
But that does not give the true picture as to how the compiler was
coded.
Yes, indeed. But we don't need it anyway.
--
Rudy Velthuis [TeamB] www.teamb.com
"Ketchup left overnight on dinner plates has a longer half-life
than radioactive waste." -- Wes Smith.
 

Re: Who is Delphi competing against?

Anders Isaksson writes:
Quote
Yogi Yang 007 writes:

>But that does not give the true picture as to how the compiler was
>coded.

But you don't need that picture. You won't be able to fix a compiler
bug anyway (at least most of us wouldn't), so why burden you brain
with trying to understand the compiler source? Spend the time on
getting good test cases instead, that is worth more, I think.
Indeed.
--
Rudy Velthuis [TeamB] www.teamb.com
"Plato was a bore."
-- Friedrich Nietzsche (1844-1900)
 

Re: Who is Delphi competing against?

Anders Isaksson writes:
Quote
Do you have any idea about the complexity of a compiler the size of
Delphi's? Do you have any idea how many steps you would have to go
before coming to the interesting parts? How do you know what the
interesting part is?
Not really.
All I have done, in the past, is to look at the size of the GCC
compiler. And, yes, it was huge.
But, then, same goes fo all sorts of current open source projects
--starting from seemingly simple text editors all the way to complex
files systems and fully-fledged OSes..
Quote
A problem seen in the final codegen end could be
caused by something done during the initial parse tree generation.
But, isn't this the case for almost all SW..
Quote
>here, I am assuming compiler source is Pascal

AFAIK, the compiler is mostly written in C - google for "Danny Thorpe
Delphi Compiler" to make sure.
I did, but couldn't find anything definitive.
I do, however, seem to remember --very vaguely-- that Delphi there was
some talk of Delphi becoming self-compiling. Though I couldn't find
where and when that was said either, it must have stuck in my memory to
assume that Delphi's compiler (from some point on) has been Pascal.
Quote
>But, is this really the shared experience with large open source
>projects?

Probably, yes. At least it is my experience, both in commercial
software development and (as a bystander) in OS projects. People are
much more interested in telling the devs "what to do" instead of
elaborating on "why should it be done".
And, the devs, usually tell people to "read the source" and get back to
them with a more tangible arguement..
Quote
>When you have access to the source, chances are, you're more likley
>to describe the problem --or whereabouts of it.

In general, I don't think so. Of course, some people are able to grok
large code bases quickly, and give good advise (or direct code fixes),
but for most of us a sufficiently large code base makes it almost
impossible to get the insight needed to actually be useful (talking
about large code bases here).
This is, of course, true. But, not having the source, there's little
chance that those who are able to grok large code bases will be able to
contribute.
 

Re: Who is Delphi competing against?

Anders Isaksson writes:
Quote
Yogi Yang 007 writes:

>But that does not give the true picture as to how the compiler was
>coded.

But you don't need that picture. You won't be able to fix a compiler
bug anyway (at least most of us wouldn't), so why burden you brain
with trying to understand the compiler source? Spend the time on
getting good test cases instead, that is worth more, I think.
This sounded (to me, at least) very much like telling a woman not to
bother her pretty head with things like (such as) who to vote for :P