"Wolfgang Kern" <nowhere@[EMAIL PROTECTED]
> wrote in message
news:g6hu5b$vus$1@[EMAIL PROTECTED]
>
> "cr88192" wrote:
> ...
>
> [...code styles and performance]
>
>>> There might be optimising tools around, but I'm confident sure that I
>>> could do it shorter/faster/smarter in machine code if I had all info.
>>> By 'input code' you mean the GPU-commands (instruction/parameters) ?
>> I meant, in general.
>
>> input code I indirectly meant, the code being written by the
developers.
>> for example, if we write in assembler, then the efficiency is often
> limited
>> by said assembler code (if the coder isn't very skilled, or the system
>> architecture changes, the code is no longer so efficient...).
>
>> if we write in C, then optimizability and ****tability somewhat increase
> (in
>> many cases, the compiler can adjust to features or shortcommings of the
>> target, and the demands on the developer are lower).
>
>> if we target system libraries, rather than hard-coding everything, then
>> improvements made to the libraries silently benefit the apps that use
> them,
>
> I can't agree on this one.
> Hardware-drivers can be written totally independent from OS and of
> course every brand will need its own driver.
> You probably mean the function-links to the OS's API which usually
> come as a set of DLLs that show only part of hardware capabilities.
> And I see no way in windoze to overcome this detours.
>
> Unlike ATAPI, VESA and legacy ****ts, standards for 2D/3D Graphic
> are unfortunately still missing, so I heavy daubt that a compiler
> can optimize hardware related code without knowing all details.
>
it can optimize most user-written code fairly well, and compete fairly
well
with "mundanely written" assembler code, as was demonstrated long ago with
early FORTRAN compilers, and later with C.
most people who write assembler, just work on getting it done, and so
produce code that is on average slower than the compiler output.
more so, if later we ****t it to a new arch, the ****ted assembler is likely
to poorly utilize the arch, and in the process require huge amounts of
alteration. the C code ****ts often with little or no change (someone can
write some code, and have it work on x86, x86-64, and ARM, often with no
real modification).
or, even if the arch adds new features, and the programmers don't bother
to
update their code, but the compilers do update their code generation
processes.
more so, in the case of video cards, using the standardized APIs will
often
allow net faster code, since very often these API will include many
specialized optimizations which would not be used in frontend code (the
developer trying to write code that works on a variety of cards, ...).
now, these are not compiler optimizations, but library and OS-level
optimizations.
likewise, do we really expect each and every app to need to worry about
the
details of, for example, the ATA and SATA interfaces, how FAT32 and NTFS
work, ... or is it better, to leave most of this up to the OS's
drivers?...
> [...future CPU] ok
> [...multithreaded design] ok
> [...power]
>>> I think the fusion generator above our heads got enough power for
>>> all our needs and may last for 'some' time.
>
>> the problem is getting it in usably large amounts...
>
> 1..4 KiloWatt/square-meter sound a lot when I look at the roof surface
> of a small house. The problem is still how to store it, we could
> use anchient methods like lift a huge stone or what's already in use
> for decades: Hydroelectric plants use spare energy to pump water uphill.
>
here, I use some lead/acid batteries, and really the
capacity/volume/weight
tradeoffs are pretty poor, for example, 72 cubic inches only holds around
500VA (approx 500W).
lithium ion and NiMH batteries are a lot better, but still fairly low
density, and I suspect a reasonable-sized capacity for an android is
likely
to to be closer to around 850kVA (or, about a whole damn car full of
lead-acid batters, or a large backpack of NiMH batteries).
several orders of magnitude more power could be extracted from 1L of
gasoline (or ethanol, or whatever else).
for example, assuming an approx 25% conversion efficiency (poor quality
4-cycle and a generator) and butanol fuel (lower energy than gasoline), we
still get around 7.5MW, or around 9x the storage of a backpack full of
batteries.
if, say, the torso could store approx 6L of fuel, we would have around
45MW,
which could keep the thing running pretty good...
>> biofuels typically rely on solar anyways, but the volumes are more
>> limited
>> (more fuel means bigger facility, ...).
>
> We could take some knowledge on DNA and/or nature to optimise this.
>
possibly.
>> nuclear can be very productive and very cheap, but for whatever reason
> many
>> people are afraid of it (especially breeder reactors and likewise).
>
> We got a nuclear power plant in Austria since many years, but even it's
> kept and held in service because it was an expensive toy, it never were
> turned on. Our former Cancellor built it and asked the crowd after done,
> and the crowd cried: "NO!" :)
>
yes, apparently.
people are afraid enough even of stupid light-water reactors, and as far
as
I know, there has only been 1 major incident with one of these (3 mile
island), and this didn't even breach containment.
many people are terrified of breeder reactors (in a worst-case scenario,
one
of these could have an uncontrolled chain reaction and turn into a nuclear
explosive, which does not happen with a LWR, though a worst case for a LWR
is that it breaches containment and starts leaking radiation, very likely
in
the form of radioactive steam in the cooling tower...).
yet, Japan and China have had plenty of good success with these (actually,
I
think for a long time the US has been giving Japan crap about them using
breeder reactors).
but, why should people be so afraid anyways? after all, people drive cars,
which are a statistically FAR more powerful instrument of death than a
reactor would be...
>> China, India, and Japan use large amounts of nuclear power (uncluding
>> breeder reactors) with no real problems. even the soviets only had a
few
>> problems, and they were pioneering many of these technologies, while
> making
>> use of minimal safety and poor workman****p.
>
>> I still personally believe that nuclear is likely to be both the
safest,
>> cleanest, and most economical option, and we have enough nuclear fuel
(if
> we > include U-238, Plutonium, Thorium, ...) that there is no danger of
> running
>> out (only that many people in the US and Europe are afraid of anything
> that
>> isn't the ultra-rare U-235, which then they only burn a small amount of
>> it
>> and try to discard the rest, in a show of pure inefficiency...).
>
>> for example, of all companies, Hitachi is making liquid-Sodium cooled
>> breeder reactors, why should the US be so damn afraid?...
>
> Not only US-citicens are afraid, there is some danger: "the human
factor".
>
potentially.
but, if you mean terrorists getting plutonium, ... well then, they can
just
buy it from someone selling it (say, Russia, Iran, ...).
China has lots of it, but I doubt is all that likely to sell it to them
(if
need be, they are likely more inclined to use it for themselves...).
IMO, it is about like people being afraid of carrying a gun (fearing
someone
might steal it and turn on them) in a crowd of people with guns. it makes
little sense...
>
>>> Hydrogen could be the key, create it with solar-panels or similar
>>> natural 'tools' (plants/bacteria may know better than silicon).
>>> Then first burn it, feed motors or O-H cells with it, and after that
>>> reuse (we may even drink) the resulting waste.
>
>> burning oxygen and hydrogen tends to produce some small amount of
>> peroxide
>> in the waste as well, so drinking it directly may not be entirely safe
>> (distilation or passing it through an organic filter would probably
>> react-out the peroxide though), but then again, there are much less
safe
>> things around...
>
> Sure. I once suggested an OH-burner which also cleans air and waste
water
> by sort out all the dirt with gas-chromatic and mud filters driven by
the
> otherwise wasted side heat. The calculated effectivty was about 50%, so
> just above a coal/steam generator, but this was 30 years ago.
>
yes, ok.
> [...Batteries]
>>> Electrical cars are available and aren't too bad on milage/load.
>>> We just haven't enough reload/replace stations yet.
>>> Another idea is to have power-rails on highways...
>
>> yes, however, if we consider how much power usage an android is likely
to
>> have, and existing battery capacities, the batteries would likely be
> fairly
>> decent size (say, a decent chunk of the torso), and may only allow the
> thing > to run for a time measurable in hours (one would have to be sure
> to
> recharge > their android often).
>
> Little vacuum cleaner robots are available already, and it's rumoured
> that the next generation may find a power-outlet by themself ...
>
but, if we note an im****tant difference here:
these robots use wheels and lack appendages.
anthropromorphic robots (such as androids) would have much higher energy
usage.
reason: limbs and walking are hardly the most efficient means of movement.
this is easily evidenced if we give someone a bicycle, and notice how much
faster they can go and how much further they can go before getting tired,
vs
walking or running somewhere...
now, we could just make robots that are like rolling cylinders with wheels
and manipulator arms, but although these would need far less energy to
move,
they are hardly human-like...
>> I would think, ideally, the thing 'should' be be able to run a few days
>> or
>> weeks or more if needed without reflueling or recharging...
>
>> none the less, lithium ion and NiMH batteries are still a lot better
than
>> lead acid (which would likely be pitiful, largish lead-acid batteries
> unable > to even keep my damn desktop running for any reasonable amount
of
> time...).
>
> It may be just a cost reason why cars still must carry heavy lead ...
>
yes.
lead/acid batteries are a lot cheaper...
they are also a lot better at high-amperage short-burst usage (aka:
shallow
cycle), used for starting a car engine, which would cause a good deal of
damage to a a LiON or NiMH battery.
these are also rather different that the lower-density, heavier, but
long-lasting NiFe batteries often used in trains and some industrial
power-storage uses...
> [gas vs liquid]
>> hydrogen though, would need to be stored in pressurized tanks.
>> but, yes, a hydrogen-based fuel storage would offer a good deal of
energy
>> capacity (vs current litium-ion batteries), but as can be noted, would
> still > need either a small engine or fuel cells (likely to remain
> expensive), to
>> convert it back to electricity.
>
> I look one more time on how Nature decided to power us.
> Electricity is just used for sensing and control, while motions avoid
> rotating at all and are driven really effective by chemical power.
>
> Why not 'design' faster horses or grow big birds to ride on it ?
> and for androids, we could train apes.... oh, another movie yet :)
>
possibly, but apes would still need better treatment than androids, and
could not be as effectively used (un-augmented) for tasks requiring a good
deal of processing power.
>> also, as a detractor, electrolysis is not a very efficient way of
> producing
>> hydrogen (so, a better option may be needed). for example, I had read
> before > of someone pulling off higher-efficiency water cracking with
> microwaves, so
>> efficient production could be possible.
>
> I've heard that UV-light would be cheaper, cause we got 'some' natural
> sources.
>
maybe...
> __
> wolfgang
>
>
>
|
