Watch, Follow, &
Connect with Us

For forums, blogs and more please visit our
Developer Tools Community.


Welcome, Guest
Guest Settings
Help

Thread: Performance Comparison Delphi 2010-XE6 (COMPLETE)



Permlink Replies: 42 - Last Post: Jul 2, 2014 6:01 AM Last Post By: Nick Hodges
Thomas Grubb

Posts: 61
Registered: 2/27/01
Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 8, 2014 3:27 PM
Hi All,

After almost a month, 12 blog posts, and over 12,000 words, I have finally completed my series of blog posts on comparing performance from Delphi 2010 to XE6. I wrote 3 blog posts this weekend to complete it. Surprisingly, at least to me and according to these specific series of tests, the numbers backed up Embarcadero's claim of Quality, Performance, Stability; at least so far as performance is concerned. Delphi XE6 did neither as well as I hoped nor as bad as I feared. It was a solid performer and competitive for all platforms. Compared to its immediate successor (Delphi XE5), it is a huge advance in performance.

The series:
Introduction
http://blogs.riversoftavg.com/index.php/2014/05/12/performance-comparison-from-delphi-2010-to-delphi-xe6-introduction/

Hello World Compilation Tests (Speed and Size)
http://blogs.riversoftavg.com/index.php/2014/05/12/performance-comparison-from-delphi-2010-to-xe6-part-2/

Hello World Execution Tests, VCL (Win32 and Win64) and FMX (Win32)
http://blogs.riversoftavg.com/index.php/2014/05/19/performance-comparison-from-delphi-2010-to-delphi-xe6-part-3/

Hello World Execution Tests (FMX Win64, OSX)
http://blogs.riversoftavg.com/index.php/2014/05/26/performance-comparison-from-delphi-2010-to-delphi-xe6-part-4-hello-world-win64-and-osx/

Hello World Mobile Execution Tests (iOS and Android)
http://blogs.riversoftavg.com/index.php/2014/05/26/performance-comparison-from-delphi-2010-to-delphi-xe6-part-5-hello-world-mobile/

Inference Engine Component Suite Execution Tests
http://blogs.riversoftavg.com/index.php/2014/06/01/performance-comparison-from-delphi-2010-to-delphi-xe6-part-6-iecs/
http://blogs.riversoftavg.com/index.php/2014/06/01/performance-comparison-from-delphi-2010-to-delphi-xe6-part-6a-iecs-advanced-console/
http://blogs.riversoftavg.com/index.php/2014/06/01/performance-comparison-from-delphi-2010-to-delphi-xe6-part-6b-iecs-basic-console/
http://blogs.riversoftavg.com/index.php/2014/06/01/performance-comparison-from-delphi-2010-to-delphi-xe6-part-6c-iecs-mobile-app/

RiverSoftAVG SVG Component Library Drawing Tests
http://blogs.riversoftavg.com/index.php/2014/06/08/performance-comparison-from-delphi-2010-to-delphi-xe6-part-7-rscl/
http://blogs.riversoftavg.com/index.php/2014/06/08/performance-comparison-from-delphi-2010-to-delphi-xe6-part-7b-rscl-primitives/

Conclusion
http://blogs.riversoftavg.com/index.php/2014/06/08/performance-comparison-from-delphi-2010-to-delphi-xe6-conclusion/

Regards,
Tom
R Horbury-Smith

Posts: 9
Registered: 12/29/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 8, 2014 7:48 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Nice to see a talented practitioner taking the time and making the
effort to put the marketing crap in perspective with some empirical
facts.

A very big THANK YOU Thomas


Thomas Grubb wrote:

Hi All,

After almost a month, 12 blog posts, and over 12,000 words, I have
finally completed my series of blog posts on comparing performance
from Delphi 2010 to XE6. ...

--
Robert Horbury-Smith
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 4:50 AM   in response to: R Horbury-Smith in response to: R Horbury-Smith
Thanks Robert. I appreciate the kind words. It was a lot of work. I came away with a restrained appreciation for XE6. Embarcadero truly did a lot of work for XE6; it just doesn't jump out at you from the numbers and is instead much more subtle.

Tom
Alex Belo

Posts: 626
Registered: 10/8/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 8, 2014 10:14 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas Grubb wrote:

comparing performance

It would be nice also to compare "native vs. managed" code in small
apps with the same visual functionality (like filling combo/memo,
drawing pictures).

--
Alex
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 4:53 AM   in response to: Alex Belo in response to: Alex Belo
Alex Belo wrote:
Thomas Grubb wrote:

comparing performance

It would be nice also to compare "native vs. managed" code in small
apps with the same visual functionality (like filling combo/memo,
drawing pictures).

Hi Alex,
Do you mean using the native iOS and Android wrappers on sourceforge? That is a good idea, and I may look at that eventually. For now, though, I have to say I am burned out with this and need to get back to developing components.
Regards,
Tom
Alex Belo

Posts: 626
Registered: 10/8/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 11:54 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas Grubb wrote:

compare "native vs. managed" code

Do you mean

I mean comparing Delphi native (mobile) CPU program performance against
interpreted languages. Is native code really faster or/and battery
friendly?

--
Alex
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 4:10 AM   in response to: Alex Belo in response to: Alex Belo
Ah! Understood. It is a good idea. Someone else would have to do that as I have never done any mobile development until Delphi. With my limited time, it would be too much to learn for just this test.

Regards,
Tom
Bob Devine

Posts: 107
Registered: 8/16/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 2:31 AM   in response to: Thomas Grubb in response to: Thomas Grubb
"It is probable that the FMX library bogs down as the number of
primitives goes up"

This certainly appears to be the case as evidenced by the poor TGrid
performance. TGrid performance on OSX (XE6) is particularly bad - again
supported by your results.

Cheers, Bob
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 4:57 AM   in response to: Bob Devine in response to: Bob Devine
Bob Devine wrote:
"It is probable that the FMX library bogs down as the number of
primitives goes up"

This certainly appears to be the case as evidenced by the poor TGrid
performance. TGrid performance on OSX (XE6) is particularly bad - again
supported by your results.
Hi Bob,
Thanks for the kind words. I was very interested when I read your experiences with TGrid. If I ever do this again, I need to include the TGrid. It sounds like the stuff of nightmares :-)
Regards,
Tom
Richard Stevens

Posts: 52
Registered: 5/1/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 7:13 AM   in response to: Thomas Grubb in response to: Thomas Grubb
Thank you very much. Very interesting and quite encouraging conclusions about XE6, even if I would have preferred, "OMFG XE6 is just AMAZING".

Richard
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 5:15 PM   in response to: Richard Stevens in response to: Richard Stevens
Richard Stevens wrote:
Thank you very much. Very interesting and quite encouraging conclusions about XE6, even if I would have preferred, "OMFG XE6 is just AMAZING".
LOL, that is what I hoped would happen too. It was actually interesting how my opinion of XE6 changed over the course of this testing. I started with the opinion that XE6 seemed informally faster and it was going to rock in the benchmarks. Then, I couldn't get any of the numbers to really back it up and I got very disgusted with it. Now, I have come to a mature realization that it improved, but not as much as I would like (I feel like I am describing a relationship... "I had a crush on her, then that wore off and we fought a lot, but now our relationship is in a good place" :-) )

Tom
Arnaud BOUCHEZ

Posts: 143
Registered: 2/17/02
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 2:11 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thanks Tom for sharing.
This set of benchmarks is informative... but I'm always doubtful about the question asked itself.

IMHO there is no "compiler magic", something like marketing sirens of saying "the new compiler will generate faster code, because it is newer".

To be honest, the Delphi compiler did not improve much since Delphi 6/7/2007.
The RTL did slow down in fact...
The compiler is NOT the true bottleneck of any business application, even if it does make a difference for floating-point arithmetic, due to new hardware support.

I've just published some benchmarks about Delphi 6, 7, 2007, XE4 and XE6, under Win32 and Win64, and my experiment is that those compilers generates more or less the same code...
See http://blog.synopse.info/post/2014/06/09/Performance-comparison-from-Delphi-6%2C-7%2C-2007%2C-XE4-and-XE6

To be honest, XE6 generates more or less the same asm than Delphi 2007.
Since function inlining, there was no noticeable optimization, for business code.
The RTL is slowest in XE6.
Win64 RTL is much slower than Win32 RTL, for most of it.

The compiler is not the main point.
Algorithms matter!
For instance, FastMM4 is what gives the bigger speed improvement between raw Delphi 7 and Delphi XE6.
But using FastMM4 on Delphi 7 or 2007 will let our 16,500,000 individual regression tests perform faster than the latest Delphi XE6!

What is nice in those blog posts is to see that FMX and VCL are not yet in the same speed range...
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 5:09 PM   in response to: Arnaud BOUCHEZ in response to: Arnaud BOUCHEZ
Hi Arnaud,
I completely agree. I truly didn't expect much differences between the compilers, except possibly with the new compilers as there might have been optimizations done as they mature to the level of the Win32 compiler (certainly they have gotten more stable, if you want to blow up the early Delphi ARM compilers, use the property X IMPLEMENTS ISomething construct). If I had expected big differences in the compilers, I would have called the series "Performance Comparison of Delphi compilers 2010-XE6" :-)

Since each version of Delphi is both its compilers and its architecture, and FMX has been changing radically each version (and the RTL as you pointed out), I was curious to see how the differences would manifest.
Regards,
Tom
Janez Atmapuri ...

Posts: 240
Registered: 2/8/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 4:52 AM   in response to: Arnaud BOUCHEZ in response to: Arnaud BOUCHEZ
Dear Arnaud,

That is a nice comparison, but saying that same code runs equally fast or
slow is not something new. The code styling is what needs to be changed.
Once you change that, you are in for the "change". The fact that some 15
years after vectorization was introduced in desktop CPUs, the databases are
still unable to make use of it, is just stunning. The hardware architecture
has changed dramatically, but the code remained largely the same. I think it
is a combination of several factors:

1.) Programmers don’t have experience.
2.) Tools like Delphi completely ignored vectorization together with .NET
framework.
3.) The amount of codebase is just way out too big to rewrite everything
from scratch.

What this world needs is a DB company to make a 10x speed jump over Oracle,
mySQL and the likes and move the old school in to the closet by giving
everybody something to run for.

Atmapuri

"Arnaud BOUCHEZ" wrote in message news:679739 at forums dot embarcadero dot com...

Thanks Tom for sharing.
This set of benchmarks is informative... but I'm always doubtful about the
question asked itself.

IMHO there is no "compiler magic", something like marketing sirens of saying
"the new compiler will generate faster code, because it is newer".

To be honest, the Delphi compiler did not improve much since Delphi
6/7/2007.
The RTL did slow down in fact...
The compiler is NOT the true bottleneck of any business application, even if
it does make a difference for floating-point arithmetic, due to new hardware
support.

I've just published some benchmarks about Delphi 6, 7, 2007, XE4 and XE6,
under Win32 and Win64, and my experiment is that those compilers generates
more or less the same code...
See
http://blog.synopse.info/post/2014/06/09/Performance-comparison-from-Delphi-6%2C-7%2C-2007%2C-XE4-and-XE6

To be honest, XE6 generates more or less the same asm than Delphi 2007.
Since function inlining, there was no noticeable optimization, for business
code.
The RTL is slowest in XE6.
Win64 RTL is much slower than Win32 RTL, for most of it.

The compiler is not the main point.
Algorithms matter!
For instance, FastMM4 is what gives the bigger speed improvement between raw
Delphi 7 and Delphi XE6.
But using FastMM4 on Delphi 7 or 2007 will let our 16,500,000 individual
regression tests perform faster than the latest Delphi XE6!

What is nice in those blog posts is to see that FMX and VCL are not yet in
the same speed range...

Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 3:03 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas,

| Compared to its immediate successor (Delphi XE5), it is a huge
| advance in performance.

Great, even FANTASTIC, job!

But,... I looked at your pages and only see an XE6 "advance in
performance" over XE5 in the Mobile/FMX environment. Am I overlooking
something?

--

Q

06/09/2014 15:01:52

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Eli M

Posts: 1,346
Registered: 11/9/13
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 3:56 PM   in response to: Quentin Correll in response to: Quentin Correll
XE6 got 2 wins, 3 second places, and 2 third places. But it also supports all 4 platforms which is the only reason I'm using Delphi at all at the moment. XE3 and XE4 performed best on some specific platforms and if performance on that specific platform was paramount in a specific project it might be worth looking into compile that project on that specific version of Delphi. However, if you're targeting all 4 platforms in a cross platform manner XE6 is the clear winner.
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 7:28 PM   in response to: Eli M in response to: Eli M
Eli,

| However, if you're targeting all 4 platforms in a cross platform
| manner XE6 is the clear winner.

I only deal in Windows. From Thomas' testing it looks like I should
have stuck with IE4. <g>

--

Q

06/09/2014 19:27:52

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 5:20 PM   in response to: Quentin Correll in response to: Quentin Correll
Hi Quentin,
No, you are not missing anything. As far as VCL, any version of Delphi works well (though I do like the new features they have added in recent versions: styles, BaaS, FireDAC, etc). It is a little disappointing that after so many years Borland/Inprise/CodeGear/Embarcadero have not add more optimizations to the compilers though. The big differences are in FMX and cross-platform.

Tom
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 9, 2014 7:31 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas,

Thanks again!

You did an OUTSTANDING job!!!!!!!!!!!!

Would it be possible to put everything into a PDF? I would like to
have your work available in a single-access on my desktop environment
document.

--

Q

06/09/2014 19:29:15

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Steve Thackery

Posts: 151
Registered: 4/29/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 3:10 AM   in response to: Quentin Correll in response to: Quentin Correll
Quentin Correll wrote:

You did an OUTSTANDING job!!!!!!!!!!!!

Seconded!

Would it be possible to put everything into a PDF? I would like to
have your work available in a single-access on my desktop environment
document.

Seconded again.

--
SteveT
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 4:07 AM   in response to: Steve Thackery in response to: Steve Thackery
Thank you both. I can look into making it all one document.
Tom
Steve Thackery

Posts: 151
Registered: 4/29/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 5:11 AM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas Grubb wrote:

Thank you both. I can look into making it all one document.
Tom

Great. Please be sure to let us all know! At the moment you are the
closest thing we've got to a hero, so you should expect all sorts of
other unreasonable requests to come your way (y'know, like God must
feel when he listens to our prayers).

:-)

--
SteveT
Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 15, 2014 5:26 PM   in response to: Steve Thackery in response to: Steve Thackery
LOL. I don't know what I did that was heroic, but if I'm the closest thing we've got, we are in trouble :-)

Anyway, I put the PDF up on the blog.

Thank you for the kind words.
Regards,
Tom
Steve Thackery

Posts: 151
Registered: 4/29/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 2:03 AM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas Grubb wrote:

Anyway, I put the PDF up on the blog.

Hey, Tom, could you post the link again, please?

--
SteveT

Thomas Grubb

Posts: 61
Registered: 2/27/01
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 4:50 AM   in response to: Steve Thackery in response to: Steve Thackery
Steve Thackery wrote:

Hey, Tom, could you post the link again, please?
http://blogs.riversoftavg.com/index.php/2014/06/16/performance-comparison-from-delphi-2010-to-xe6-pdf/
Steve Thackery

Posts: 151
Registered: 4/29/06
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 7:59 AM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas Grubb wrote:

Steve Thackery wrote:

Hey, Tom, could you post the link again, please?
http://blogs.riversoftavg.com/index.php/2014/06/16/performance-comparison-from-delphi-2010-to-xe6-pdf/

Thank you!

--
SteveT
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 2:25 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas,

| Anyway, I put the PDF up on the blog.

Snagged it!

Many thanks!

--

Q

06/16/2014 14:25:21

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 29, 2014 6:05 PM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas,

Finally got around to reading and studying your PDF in detail.

It's now painfully obvious to me that since I don't do any mobile work
at all that I should have stuck with XE2! <sigh>

If it wasn't such a huge PITA to install all the 3rd-party components I
use I would immediately retro-grade myself from XE5 to XE2.

And I may skip XE6, seriously. XE6 which you show as the "Winner" only
has an insignificant-piddlely edge over XE5 in the non-mobile
environment. And both produce horribly large executables!

I am REALLY PO'D at their executable sizes!!! <GRUMBLE> I have one
small app (VERY SMALL, self-contained, only 430 lines, including
comments and blank lines, and uses NO other routines!) that produces a
Debug executable size of over 28MB!!! That's absolutely LUDICROUS!!!
Its Release mode size is over 8MB! <shaking head> The more I think
about it the more I am giving serious consideration to returning to
XE2!!!!!!!

TA!

--

Q

06/29/2014 17:44:04

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]
Robert Dawson

Posts: 211
Registered: 7/28/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 29, 2014 6:30 PM   in response to: Quentin Correll in response to: Quentin Correll
"Quentin Correll" wrote

Debug executable size of over 28MB!!! That's absolutely LUDICROUS!!!

Perhaps. But the oldest development VM I use has 3GB of memory--the size of
the debug executable hardly seems an issue. IOW, it might be ludicrous, but
it doesn't seem particularly significant. At least to me.

FMPOV, the Delphi VCL community seems ever more polarized into two camps:

(1) those who want Delphi to be a better C++ (small exes, high speed,
anything fancy in optional libs., and
(2) those who want Delphi to be a better C#/ (decent containers, rooted type system, built-in MM (ARC or GC), better reflection/abstraction/generics/aspects and all the other pixie dust that modern languages increasingly have.

Don't fixate on the languages or curly braces; they're just meant to capture
the difference between those who are primarily business logic developers,
and those who are primarily technical, close-to-the-metal developers. There
are both using Delphi at present.

The first camp appears to me more sensitive to legacy breakage--perhaps
because the latter see more to gain from the changes they advocate for the
kinds of code they write.

Cpt. Obvious comment: anything done to make camp2 happier tends to anger camp1.

bobD
Nick Hodges

Posts: 2,414
Registered: 9/22/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 6:18 AM   in response to: Robert Dawson in response to: Robert Dawson
Robert Dawson wrote:

Cpt. Obvious comment: anything done to make camp2 happier tends to
anger camp1.

In my view, Camp 2 is the future of popular computing and Camp 1 is a
shrinking demographic.

--
Nick
Delphi Programming is Fun
Robert Dawson

Posts: 211
Registered: 7/28/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 7:04 PM   in response to: Nick Hodges in response to: Nick Hodges
"Nick Hodges" wrote

In my view, Camp 2 is the future of popular computing and Camp 1 is a
shrinking demographic.

I tend to see both as having validity--since I have code that falls on both
sides. But I recognize them as having divergent needs, and at times,
directly opposed needs.

And they've been pulling on the language (in different directions) for a
long time. Remember the bru-ha-ha when the item notify code was added to
TList?

bobD
Nick Hodges

Posts: 2,414
Registered: 9/22/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jul 2, 2014 6:01 AM   in response to: Robert Dawson in response to: Robert Dawson
Robert Dawson wrote:

I tend to see both as having validity--since I have code that falls
on both sides. But I recognize them as having divergent needs, and at
times, directly opposed needs.

Oh, please don't misunderstand me -- both have validity. It's just
that one is a shrinking group and one is a growing group in my view.

And they've been pulling on the language (in different directions)
for a long time. Remember the bru-ha-ha when the item notify code was
added to TList?

Yep -- it's a titanic struggle. ;-)

--
Nick
Delphi Programming is Fun
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 11:36 AM   in response to: Robert Dawson in response to: Robert Dawson
Robert,

| IOW, it might be ludicrous, but it doesn't seem particularly
| significant. At least to me.

My POV is that it's just that type of thinking that leads to such poor
quality software.

--

Q

06/30/2014 10:51:50

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Robert Dawson

Posts: 211
Registered: 7/28/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 12:11 PM   in response to: Quentin Correll in response to: Quentin Correll
Quentin Correll wrote:

My POV is that it's just that type of thinking that leads to such poor
quality software.

No software is perfect, and time is limited. So I guess I'd say that given all the things I'd like to see fixed/changed in Delphi, debug executable size isn't in my top 10. It just doesn't chart at all.

Is there something specific you can't do, or can't do as well, because of this issue?

bobD
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 3:03 PM   in response to: Robert Dawson in response to: Robert Dawson
Robert,

| Is there something specific you can't do, or can't do as well,
| because of this issue?

I have a couple of apps that use large amounts of RAM (Probability math
computations on a large, continuously growing, database.) which run on
older, lower memory available, laptops. Every byte is valuable. It's
necessary to use RAM instead of the HD, some calculations are presently
taking up to a couple of hours to run as it is.

--

Q

06/30/2014 14:59:15

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]

Robert Dawson

Posts: 211
Registered: 7/28/00
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 30, 2014 6:53 PM   in response to: Quentin Correll in response to: Quentin Correll
"Quentin Correll" wrote
older, lower memory available, laptops. Every byte is valuable. It's
necessary to use RAM instead of the HD, some calculations are presently
taking up to a couple of hours to run as it is.

But do you deploy debug builds on those machines?

bobD
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jul 1, 2014 8:47 AM   in response to: Robert Dawson in response to: Robert Dawson
Robert,

| But do you deploy debug builds on those machines?

No, just MadExcept.

--

Q

07/01/2014 08:47:15

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]
Dalija Prasnikar

Posts: 2,325
Registered: 11/9/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jul 1, 2014 3:17 AM   in response to: Robert Dawson in response to: Robert Dawson
Robert Dawson wrote:
"Quentin Correll" wrote

Debug executable size of over 28MB!!! That's absolutely LUDICROUS!!!

Perhaps. But the oldest development VM I use has 3GB of memory--the size of
the debug executable hardly seems an issue. IOW, it might be ludicrous, but
it doesn't seem particularly significant. At least to me.

FMPOV, the Delphi VCL community seems ever more polarized into two camps:

(1) those who want Delphi to be a better C++ (small exes, high speed,
anything fancy in optional libs., and
(2) those who want Delphi to be a better C#/ (decent containers, rooted type system, built-in MM (ARC or GC), better reflection/abstraction/generics/aspects and all the other pixie dust that modern languages increasingly have.

Don't fixate on the languages or curly braces; they're just meant to capture
the difference between those who are primarily business logic developers,
and those who are primarily technical, close-to-the-metal developers. There
are both using Delphi at present.

The first camp appears to me more sensitive to legacy breakage--perhaps
because the latter see more to gain from the changes they advocate for the
kinds of code they write.

I belong in both camps. Actually, I don't think there should not be two camps.
You can have things done for camp 2, that don't obliterate camp 1. You just
have to do things systematically, carefully and do them properly. Of course,
there will always be some small price to pay, but if it is small enough all will
be happy campers.

What we have now, is that new features are coming in too slowly, with too much
recklessness and both camps are unhappy. For instance, generics are huge problem
for exe size, but it is not due to nature of generics as feature, but poor compiler
implementation that cannot fold generic code.

RTTI is nice feature, but does it really have to be fully imposed on RTL/VCL/FMX
core (especially private and protected members). It does more harm than good there.

Live bindings are also another example of half baked feature that is both huge, slow
and that would be much better if it would be backed up by compiler.

But making all those things right takes time (and skill), and that is something EMBT
obviously lacks.

The most problematic part for camp 1, would probably be ARC, because it can bring
significant performance issues in server and multithreaded scenarios. Anything else
should be beneficial, if done with care. And camp 2 would also profit from faster code
and smaller exes.

Cpt. Obvious comment: anything done to make camp2 happier tends to anger camp1.

Especially, if you do it half arsed.

Dalija Prasnikar
John Treder

Posts: 349
Registered: 8/2/02
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE) [Edit]
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 29, 2014 8:53 PM   in response to: Quentin Correll in response to: Quentin Correll
Quentin Correll wrote:

Thomas,

Finally got around to reading and studying your PDF in detail.

It's now painfully obvious to me that since I don't do any mobile work
at all that I should have stuck with XE2! <sigh>

If it wasn't such a huge PITA to install all the 3rd-party components I
use I would immediately retro-grade myself from XE5 to XE2.

And I may skip XE6, seriously. XE6 which you show as the "Winner" only
has an insignificant-piddlely edge over XE5 in the non-mobile
environment. And both produce horribly large executables!

I am REALLY PO'D at their executable sizes!!! <GRUMBLE> I have one
small app (VERY SMALL, self-contained, only 430 lines, including
comments and blank lines, and uses NO other routines!) that produces a
Debug executable size of over 28MB!!! That's absolutely LUDICROUS!!!
Its Release mode size is over 8MB! <shaking head> The more I think
about it the more I am giving serious consideration to returning to
XE2!!!!!!!

TA!

3rd party components? NexusDB? (Yeah, it's great!) You have "Use Debug DCUs" checked in the Debug build, right?
Try a release build and shut off MadExcept; you might be slightly mollified. :-)

--
Ol' Tred
Quentin Correll


Posts: 2,412
Registered: 12/1/99
Re: Performance Comparison Delphi 2010-XE6 (COMPLETE)
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 10, 2014 10:18 AM   in response to: Thomas Grubb in response to: Thomas Grubb
Thomas,

What Steve typed.

--

Q

06/10/2014 10:18:25

1.19.1.372 [Q'sBrokenToolBar] [Running on TQ]
Arthur Hoornweg

Posts: 414
Registered: 6/2/98
Hello Earth
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 1:25 AM   in response to: Thomas Grubb in response to: Thomas Grubb
The Space Shuttle's primary flight system needs only one megabyte memory. A Delphi "Hello World" needs 2-8 ???? This is getting ridiculous!

Source:http://www.nasa.gov/mission_pages/shuttle/flyout/flyfeature_shuttlecomputers.html
Gilbert Padilla

Posts: 315
Registered: 3/8/04
Re: Hello Earth
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 7:28 AM   in response to: Arthur Hoornweg in response to: Arthur Hoornweg
Mike Margerum

Posts: 590
Registered: 12/1/99
Re: Hello Earth
Click to report abuse...   Click to reply to this thread Reply
  Posted: Jun 16, 2014 9:47 AM   in response to: Gilbert Padilla in response to: Gilbert Padilla
On 6/16/2014 10:28 AM, Gilbert Padilla wrote:
Apparently they didn't use that same rigorous testing for their website.
Got an error about a script running too long.
Legend
Helpful Answer (5 pts)
Correct Answer (10 pts)

Server Response from: ETNAJIVE02