Re: Delphi 2008 native?

"Jolyon Smith" <jsmith@xxxxxxxxxxxxxxxxxxxxx> wrote in message news:MPG.2225fc6d492829ef989f4b@xxxxxxxxxxxxxxxxxxxxxxxxx
In article <47ba0d01@xxxxxxxxxxxxxxxxxxxxxx>, says...

I don't think we need a conspiracy

No, just Economics 101.

Companies need sales. To increase sales companies need to provides
motivation to their customers to keep buying new stuff. If their
customers don't NEED that stuff then those companies have to come up
with clever ways to make them THINK they do.


I don't think MS staff wander around thinking, "What will
slow the PC?"

Really? I think they do, but not in those terms precisely. They are
thinking "What will make our customers want to buy a new/upgraded OS
from us?" and at the same time Intel are thinking the same thing.

The combined effect is no different from if Microsoft and Intel *did*
sit down and collude.

Yes, but I was responding to a poster who claimed an *actual conspiracy* as in criminal.

Every tech business is working out how to extract $$ out of wallets. That has driven the economy since the industrial revolution. That is why CPUs get faster. Do you want a planned economy? I can do more with Windows or Mac or Linux now than 13 years ago. Personally, I do want my PC to do more.


I'm not a .NET enthusiast, but I can see why MS introduced
it as a move toward safer, faster programming.

That may be what they SAID it was for. Bottom line - it was to get some
more sales.

We needed something. VB was dying. I don't think native code, including Delphi was suitable for the VB people. In many ways, .NET is a successor to the spirit of VB.

"Pointer" languages like
C/C+ and Pascal make assumptions about the computer architecture

Fairly safe assumptions given that those assumptions continue to be
valid and aren't in any real danger of losing that status in the
foreseeable future.

Well, your "future" has been here for decades - Harvard architecture CPUs with separate code and data memories for a 2X memory bandwidth improvement. I have been writing code for these for 15 years- and it is clumsy with C/Pascal because you have multiple types of incompatible pointers. Now that we are starting to hit the speed limits of silicon, we need to consider different architectures. C/Pascal use highish level syntax which compiles very directly down to assembler. Is that all we'll ever look forward to? I don't feel confident to make assumptions and predict the "forseeable future". Most tech predictions are proved wrong, why not yours?

An important difference between .NET and Pascal is that
.NET potentially can move objects to new memory locations *during*
execution

Um, that's got nothing to do with Pascal - Chrome is Pascal and produces
programs that perform very happily in a .NET environment.

Does the Chrome language have real x86 pointers? Not if its .NET. So its not the classic C/Pascal pointer language I was referring to.

And my native code has it's memory moved around the whole time - in and
out of the pagefile specifically.
:)

The issue here is memory fragmentation - and all x86 processors are vulnerable to it when running C/Unix type memory handling. Once your runtime can relocate an object, fragmentation can be prevented. In the case of the .NET speed penalty, you might argue that the cure is worse than the problem - but it was an attempt!

so that memory fragmentation is avoided

Ah. You mean the additional indirection between the code and the data
it is working with. How does that benefit the code? The developer?
The user?

Its not the pointer lookups which slow .NET. Its the whole Managed deal where everything is sheperded at runtime. You may not want Managed - but it has its benefits: cross platform where .NET and Java run on many devices, CPU independence, security, a clean break from the aging WinAPI. Imagine being an OS company, kowing all too well the limitations of current OS thinking, and wanting to move. Tell me they got it wrong, but don't tell me they did not need to move.

Or is it perhaps just about selling the "benefits" of change in order to
justify $$ spend on that change?

Maybe?

You can say that of your TV, automobile, house. All changed to make money for companies. Is it improvement or are we just chasing change? As I get older, all the change does disturb me. Well, the next generation can take over.

In my opinion, hardware and software is moving forward, including MS products.


Also, programs can run
on wildly different architectures, possibly as a step toward future machines
we have not yet dreamt of.

Build a dream machine THEN devise the best tools for it. Trying to
shape todays tools for machines about which we know nothing is only ever
going to result in compromised tools for today with no real certainty
that they will even be suitable in the future.

Yes, but MS had to act. 1995 thru 2000 were times when business software was a mess of VB hacks and flakey C++ and MS knew that was wrong. For all its limitations, .NET was a decent stab at the future. And it is not locked in to a particular architecture, so it can survive even radical OS change. Remember that MS cannot determine the future - it had to bet on technologies which gave it a chance of being part of that future.

It would work poorly as a desktop machine with any modern OS. Just try
loading the latest desktop Linux - unusably slow!

The question has to be "Why?"

Why is it so slow?

AND

Why load the latest version if an old version is doing the job it needs
to do?

The old versions were not doing the job to my satisfaction. I don't want to see Win95, NT4 or an early Linux or Mac ever again. And they all ran like dogs on hardware of the day. Now I see a lot less of the hourglass. I open many programs and don't fear the machine going down.

I looked in on my 14 year old son last night - he was playing a 3D game - it
looked like a movie, with leaves on the trees, explosions, hundreds of
soldiers all running around, lag-free response to the controls, LAN enabled.

And would you say more or less successful (in delivering an
entertainment experience) than DOOM?

I've not seen DOOM. This was World of Warcraft where the soldiers and vehicles have their own intelligence and get high level direction from the player. I saw a pitched battle with a large number of entities fighting - bombing from above, grenade launchers, flamethrowers all viewed from above. He could zoom down through the battle and see a single soldier in good detail while all this was going on. Pure hardware grunt.

That same game on my development machine is slow and blurry.

So play Doom/Quake on that PC - my guess is you will be having just as
much fun, and quite possibly MORE fun with those games.

There are familiar arguments to be trotted about about game-play vs eye-
candy. Economics 101 again - it comes down to "shifting boxes".

When games can be sold by how pretty they look, you get pretty, vacuous
games. When there are limits to how pretty a game can be, the game has
to deliver on the game-play.

The eye candy is part of the illusion. My son definitely values it. PC games are not like chess - this is a living world to the player. This game gets him engrossed for day after day. Each race and unit has different strengths, weaknesses. You can configure the combatants and terrain to give you an easy or hard battle. Its not for me, but boys and adults are playing this stuff. I suppose its another form escapism, like watching a movie or reading a book - humans place value on breaking from reality.

My point is
that PCs are going to become more than Word, Excel, Firefox - we are looking
for better interfaces, instant response

"Better interfaces". "Better" being measured how? Intuitive,
functional and efficient should be the measure. I suspect you still
mean "prettier" though.

Instant response - funny thing is, that has always been a demand, and
developers have always managed to deliver on it.

I'd like a PC to be an intelligent entity, an assistant or co-worker. I'd like to talk to it in human terms. I'd like it to have a wide ranging knowledge. Developers have rarely delivered "instant response" to my satisfaction.


Users today are running 3GHz workstations talking over 100Mb LANs to
RDBMS servers running multiple multi-GHz CPUs.

This is a world removed from the days when I cut my teeth in Windows
Client/Server, when workstations were 33MHz 386's, the LAN was 10Mb if
you were lucky, and the server was barely more powerful than the
workstation.

Workstations are 100x faster (in the CPU) the LAN is 10 (or sometimes
also 100) x faster, and servers are massively more powerful.

Are customers sub-second response targets now sub-milli-second targets?

No, they are still sub-second, because despite all this extra grunt
being thrown at the same old problems, stuff still just doesn't seem to
get any faster.

I don't know what software you drive, but I get exposed to CAD programs of awsome power. And these have continuously improved. The problems tackled are far more ambitious - not the "same old problems".

I know things have changed vastly over our lifetimes. But of itself, that amount of change does not tell us that what we now have is all we need. Those faster LANs, faster PCs and bigger screens are now showing medical x-rays to any PC in a hostpital, so avoiding lost x-ray film, repeat x-ray exposure, doctor guesswork, thus reducing mortalities. The more you provide, the more can be done. Now that LANs reach video speeds, new apps can come. Fields like genetics, engineering and research thrive on processing power. The world is not just databases, spreadsheets and email.


speech recognition.

You've been watching too much Star Trek.

:)

This isn't a technical problem, it's an interaction and contextual
problem. Recognising speech is fairly easy. Interpreting it and
knowing when it is being directed in a fashion that demands recognition
and interpretation and determining the context for that interpretation I
suspect will prove to be beyond the capabilities of computation
certainly in mine or my granchildrens lifetime.

Sure its hard. But I'm not satisfied. I welcome the move forward. And you need more processor grunt to move research forward. The PCs we have now are only possible because billion dollar semiconductor factories produce large numbers of chips to a mass market. That market has an appetite for more.

Just a gentle dig, but are you in danger of sounding like an old fart who dislikes the "newfangled stuff"? The good old days?

Roger Lascelles

.