7 replies to this topic

[Pointer2APointer]

I checked Intel's site, and they have no development manual, and/or references explaining the opcodes, practices, or overview of the Intel Celeron 64-bit processor, which is the one I have.

Do I request one?

Would I have to pay?

Would I get an electronic format, physical manual, or how?

I don't know what to do at this point, but all I know is that I need more specifics on the register types, features, opcodes, ISA, etc. for the processor for reference.

Edited by Pointer2APointer, 24 September 2012 - 02:42 PM.

[frob]

The instruction set is all consolidated in the general architecture manuals available here.

All the chip-specific stuff is included in the manuals, they don't create a new separate manual for each chip since nearly everything is identical for each chip. Differences between processors is noted as they apply.


The architecture manuals are much more detailed than a game programmer is likely to need. Mostly it is information for compiler vendors and OS developers. Any reason you are asking for it?

[Ravyne]

Also, keep in mind that Celeron is just a marketting name, just like Pentium, i7 or what have you. It can span several different micro-architectures over 10s of years. There are 64-bit celeron cores going all the way back to the Prescott core, which was a 64bit evolution of the Pentium 4's netburst core.

What you really care about is the micro-architecture. For the most part, a Celeron processor is identical to contemporary Pentium, Core, or 'i'-somethings of its time, just with fewer-cores, slower speeds, less cache, and lacking some of the higher-end, niche features (things like virtualization support, etc).

[Pointer2APointer]

Mostly it is information for compiler vendors and OS developers. Any reason you are asking for it?

I intend to pursue OS development.

Thanks for clearing up the specifics on micro-architecture and brand names.

I honestly thought that they might use different ISA for different processor brands.

[Ravyne]

You'll have to start at the beginning, quite literally.

Even modern x86 processors boot into 16bit mode, and you then have to write a program that transitions first into 32bit and then 64bit mode. Even getting just that far is non-trivial.

I'd not worry a bit about modern processors just yet -- try writing a simple 16bit x86 OS first -- Bring it up to use just the first 64k for Prog and Data segments, then add support for having Prog and Data occupy different RAM addresses, then add in support to page in additional memory. You'd end up with a basic dos-style OS, but that's plenty to keep you busy.

[Pointer2APointer]

Wow, thanks for all of that. You just cleared even more things up for me.

Despite the difficulty, I planned on eventually (not anytime soon) accomplishing a 64-bit OS(not like Windows, Mac, or Unix-like interfaces, but I planned on having a graphical user-interface at some point in the future).

The thing is that I have a 64-bit processor, and I wasn't so sure about which mode corresponded to each individual architecture, ISA of Intel, or other processor families, when booted up. It's kind of unadvanced, to me at least, that they don't draw these things out for hobbyists without all of the more "boring" details of the processor's functioning being laid out in hundreds of pages beforehand.

But I honestly didn't know for a fact that you could write a bootable, usable 16-bit operating system on a 64-bit processor.

[Ravyne]

I'm glad you're learning something, but I warn you now that if you're already worried about a sea of "boring details", then you should quit now, or change your attitude and dig in for the long haul -- those details are there for a reason, and they are critical to anything you're doing at the operating system-level. And if you think just understanding the details at that level is boring, just wait until you have to debug something at that level -- which, by the way, means that you're probably dealing with kernel dumps, because you may not even have even console output over a serial connection working yet, much less video, much-much less a workable debugger, and much-much-much less a debugger as nice as you might be used to in Visual Studio or even GDB. What's more, you probably can't initially trust even a single function in the entire system because, well, *you* wrote it yourself, or at least ported it.

Then, you'll deal with hardware stuff like control registers being undocumented or documented incorrectly or incompletely -- for example, the hardware might only initialize if the registers are touched in the correct order, and this might not be documented -- it wouldn't be the first time. Then, even *hardware* has bugs (so called "Errata") where perhaps a specific feature doesn't work like it did in last year's processor, or a certain sequence of instructions causes the processor to do something unexpected like fire an IRQ, cause an exception, or go into a hard lockup. Again, it wouldn't be the first time it's happened. There's usually an entire appendix dedicated to errata in just about any hardware manual.

This is not to say that you should't go for it, just realize exactly what it is that you're about to commit to. As a veteran PC programmer for more than 10 years, my first real experience with hardware-level programming (on an ARM microprocessor development board) consumed an entire evening just to get my development tools set up (and that was with 30 pages of step-by-step directions), and another evening (or was it two?) pouring over the reference manual to bring up the hardware (setting the system clock, , filling in the interrupt table, initializing timers, etc), just so that I could write two-dozen lines of my "application" that lit up a single LED 30 times a minute.

It is fun, despite the work, and there certainly is a deep feeling of achievement the first time you make an LED light up completely on your own, but it's not something where everyone will enjoy the journey, or find the pay-off worthwhile, or that everyone is even cut out for.

[Pointer2APointer]

I didn't mean that the details were "boring" in the sense that I didn't like or understand why they're there and what they do, what they can be used for, etc.

I just meant that they'd generally be seen more boring from the perspective of someone comparing stack dumps, machine code, interrupt controllers/devices, and the like with something more interesting to a specific cause or interest, such as 4 GB RAM, 1 GB video memory, 3.40GHz processor, multi-core, etc.

People would normally see the latter of the last sentence more intriguing than the real low-level stuff, but I don't think the processor's native architecture, registers, memory, and controls are boring - but they are really difficult to actually get a good grasp of understaning on.

And the difficult part to me is that Assembly confuses me, because I can't really translate every directive call, move, or shift to a possible correlated machine instruction while excluding the ideal of interconnected hardware and control, and how each instruction would exactly "work", let's say.

I don't find most typical programming languages (at least the ones I use, like C++, C, Java, C#, etc.) to be difficult at all once you use them enough.

The problem I have is understanding how the languages from the OS and API perspective work "under the hood" to see how every dependency, instruction, kernel memory management and interrupt controlling, and graphics control, would be handled in any fathomed, linear or straightforward way.

It just, to me, seems like the low-level controls are beyond my ideas to manage and incorporate as like higher-level code isn't as hard to.

But if you've been at this for over ten years, and you've made an LED light up like that, I'm guessing you know a thing or two about hardware through low-level controls, kernel operation, and maybe even native, modern Assembly instruction transitions and drivers.

I get all of the basics, and even some of the "under the hood" stuff, but understanding how it all goes together, and how it all could go together, still seems like a nightmare.