• Please review our updated Terms and Rules here

An OS built in 100% assembly language?

Power is nearing the time of being phased out. IBM paid Global Foundries $1.5 Billion to take over IBM's processor production lines. Global Foundries will continue to offer Power chips for 10 years. So with Power8 being fairly new, there will be one or two more generations of Power designs before the last Power chip is sold.

Intel can't slow R&D all that much. Not only are ARM designs closing on the lower parts of Intel's market which provide the volumes needed to keep fabs running at full output but Intel has to convince buyers that the new CPUs are enough of improvement to replace older CPUs. If systems only get replaced on failure which would average about every 10 years, Intel would lose about 80% of their sales.
 
It just got overtaken by x86 eventually, since Intel had a much larger market, and therefore a much larger R&D budget than Apple/IBM/Motorola.
Click to expand...

This isn't really true. IBM had, and has, a gargantuan R&D budget: they're famous for it. IBM just didn't really want to be in the consumer market and they even got out of the workstation market because they (correctly) perceived that the days of high-margin sales were gone. True to their predictions, workstations and PCs have since commenced on a race to the bottom (low-sales boutique systems and the R2D2 Mac Pro notwithstanding). If Apple had paid them, IBM would have delivered (see also Gekko, Broadway, Cell and Xenon), but IBM didn't have anything close to a laptop chip and they weren't going to work on it for free when it wasn't where IBM as a corporation wanted to be.

Power is nearing the time of being phased out. IBM paid Global Foundries $1.5 Billion to take over IBM's processor production lines. Global Foundries will continue to offer Power chips for 10 years. So with Power8 being fairly new, there will be one or two more generations of Power designs before the last Power chip is sold.
Click to expand...

"Power Systems represents IBM's growth strategy with key investments in PowerLinux and OpenPOWER." -- Errol Rasit, Gartner Research
http://www.gartner.com/newsroom/id/2997118

z-Machines are not a growth industry, and having exited the x86 server business (itself in a race to the bottom), Power Systems and service contracts for Power Systems is all IBM's got. I doubt very much they're putting themselves on the 10-year slowboat to extinction, and I'm quite sure GlobalFoundaries will be delighted to continue making chips for IBM on a new contract.
 
ClassicHasClass said:
This isn't really true. IBM had, and has, a gargantuan R&D budget: they're famous for it. IBM just didn't really want to be in the consumer market and they even got out of the workstation market because they (correctly) perceived that the days of high-margin sales were gone.
Click to expand...

That is what I meant obviously: They weren't interested in investing in the PowerPC specifically.
Which meant development slowed down in the late 90s, and eventually came to a halt after the G5. Which allowed x86 to overtake it by default.
Yes, IBM *could* have sunk more money into it, but they chose not to. In fact, IBM was never all that interested in the markets that Apple was in, which is why Motorola was the main supplier for Apple initially. IBM also made PowerPC chips in those days, but not aimed at Apple's range.

Just as Intel *could* have gone all out on Itanium, but instead chose, after the Pentium 4, which was originally meant to be the last x86 ever, to put the focus back on x86, and have Itanium as a second-class citizen.
 
Scali said:
I have rarely seen any 64-bit software that is actually faster on x86 rather than slower, which can not be explained by hitting the memory limits in 32-bit mode. In other words, 64-bit code that actually performs better at the instruction level.
Click to expand...
Higan/bsnes and Dolphin emulators are two examples. In fact, Dolphin dropped support for 32-bit x86, while the author of higan says he will continue to support it. But my machine- a Sandy Bridge Core i7- can't run games at full speed using the 32-bit build of higan (it can using the 64-bit build).

Scali said:
And you still have all these crazy optimization rules that make your head explode.
Click to expand...
And other RISC CPUs don't :p? Even MIPS by R3000 was unpleasant to think about its 8 stage pipeline. Well, MIPS code is unpleasant to write period (takes way too many instructions to get anything done), but I digress...
 
cr1901 said:
But my machine- a Sandy Bridge Core i7- can't run games at full speed using the 32-bit build of higan (it can using the 64-bit build).
Click to expand...
Higan must not be very well put together, then; I can run games at full speed in other emulators for the same systems on a Core 2 Duo laptop running 32-bit XP. Heck, time was I could do it on an 800MHz Celeron.
 
commodorejohn said:
Higan must not be very well put together, then;
Click to expand...

Indeed, there are plenty of cycle-exact emulators for eg C64 and Amiga, which should be at least as complex to emulate as a NES.
And they generally work fine on a modest Core2 Duo, or less.

So I wonder what kind of horrible code is hiding inside Higan then :)
 
Scali said:
Indeed, there are plenty of cycle-exact emulators for eg C64 and Amiga, which should be at least as complex to emulate as a NES.
And they generally work fine on a modest Core2 Duo, or less.

So I wonder what kind of horrible code is hiding inside Higan then :)
Click to expand...
Maybe it does a dedicated emulation of every single transistor? :rolleyes:
 
You must log in or register to reply here.
Back
Top