[...] But my application-coded debugging brain kept looking at abstractions like they would provide all the answers. I rationally knew that the abstractions wouldn’t help, but my instincts hadn’t gotten the message.
That feels like the wrong takeaway for me. Assembly still runs on abstractions: You're ignoring the CPU microcode, the physical interaction with memory modules, etc. If the CPU communicates with other devices, this has more similarities with network calls and calling the "high level APIs" of those devices. For user space assembly, the entire kernel is abstracted away and system calls are essentially "stdlib functions".
So I think it has a different execution model, something like "everything is addressable byte strings and operates on addressable byte strings". But you can get that execution model occasionally in high-level languages as well, e.g. in file handling or networking code. (Or in entire languages built around it like brainfuck)
So I think assembly is just located a few levels lower in the abstraction pile, but it's still abstractions all the way down...
Yes and no. There's no way to "get" to these. Arguably assembly is an abstraction on top of codes (hexcodes or binary if you want to see it that way), but the assembly instructions are the lowest level we get to access. For as a programmer you don't get to access the microcodes emulating an amd64 architecture and you cannot decide to use these microcodes directly.
Otherwise it's just electricity. Then it's just electrons.
So it's not false that it's all abstractions but it doesn't help much to view it that way.
There is a lot going on at the hardware level that is going on and hidden from the view of assembly. Hardware is not magic, there are a ton of design decisions that go into every architecture, most of which isn't immediately obvious by looking at the ISA.
Embedded CPU assembly is what I do most often, for the last 40 years, and there aren't really any abstractions at all - not even microcode. You have a few KB or ROM and maybe a few KB of RAM, ALU, registers, peripherals, and that's it - no APIs, no kernel, no system calls, no stdlib. Just the instructions you burn into the ROM.
Many common hardware features like out of order instruction issuing, register renaming, and even largely caching and segmentation, are largely or entirely hidden at the assembly level.
You can totally convert your assembly into machine code by hand. There is no lower level. Not all cpus have microcode.
Abstractions do exist (disagreeing with the single other post in here) and they also exist in most flavours of assembly, because assembly itself is still an abstraction for machine code. A very thin one, sure, but assemblers will generally provide a fair amount of syntactic sugar on top, if you want to make use of it.
Protip: your functions should be padded with instructions that'll trap if you miss a return.
Galaxy brained protip: instead of a trap, use return instructions as padding, that way it will just work correctly!
Some compilers insert trap instructions when aligning the start of functions, mainly because the empty space has to be filled with something, and it's better to use a trapping instruction if for some reason this unreachable code is ever jumped to. But if you have to do it manually, it doesn't really help, since it's easier to forget than the return.
The application programmers ended up with a dump and had no idea on how to diagnose the problem. I once walked up to a programmer puzzled by a dump and solved it in about three minutes (practice helps) putting out of joint the nose of the senior programmer who was being consulted.
The compiler library handled this situation just fine, but the guru who wrote the Assembler program insisted that programmers had to be more careful.
I wrote another routine to intercept the calls and use the subroutine library when needed.
The flags are another abstraction that might not mean what it says. The 6502 N flag and BPL/BMI instructions really just test bit 7 and aren't concerned with whether the value is really negative/positive.
I haven't done serious work in C in quite some time. I wonder if modern compilers are better at reporting that sort of thing.
If you want a better understanding of the architecture, reading the documentation from the hardware vendor will serve you better.
If you want your code to be faster, almost certainly there will be better ways to go about it. C++ is plenty fast in 99% of the situations. So much so that it is what hardware vendors use to write the vast majority of their high-performance libraries.
If you are just curious and are doing it for fun, sure, go ahead and gnaw your way in. Before you do so, why not have a look at how hand-written assembly is used in the rare niches where it can still be found? Chances are that you will find C/C++ with a few assembly intrinsics thrown in more often than long whole chunks of code in plain assembly. Contain that assembly into little functions that you can call from your main code.
For bonus brownie points, here is a piece of trivia: the language is called assembly and the tool that translates it into executable machine code is called the assembler.
For bonus brownie points, here is a piece of trivia: the language is called assembly and the tool that translates it into executable machine code is called the assembler.
> In Sum# > Abstractions. They don’t exist in assembler. Memory is read from registers and the stack and written to registers and the stack.
Abstractions do not exist periodi. They are patterns, but these patterns aren’t isolated from each other. This is how a hacker is born, through this deconstruction.
It’s just like the fact that electrons and protons don’t really exist. but the patterns in energy gradients are consistent enough to give them names and model their relationship. There are still points where these models fail (QM and GR at plank scale, or just the classical-quantum boundaries). It’s gradients all the way down, and even that is an abstraction layer.
Equipped with this understanding you can make an exploit like Rowhammer.
Now, there are abstractions (which exist in your brain, whatever the language) and tools to represent abstractions (in ASM you've got macros and JSR/RET; both pretty leaky).
My point is that we settle with what we see for convenience/utility and base our models on that. We build real things on top of these models. Then the result meets reality. If only that transition were so simple.
When an effect jumps unexpectedly between layers of abstraction we call it an abstraction leak. As you mentioned. The correct response is to re-examine these leaks and make other frameworks to cover the edge cases, not to blame the world.
Hackers actively seek these “leaks” by suspending assumptions that arise out of the abstractions that humans tend to rely on.
I’m not surprised that my OP got downvoted. It can be very upsetting when one’s conceptual frameworks are challenged without prescription. No one even mentioned the specific example that I referenced. Well, if they can’t parse it, they don’t deserve it. Keeps me in the market.
The general conceptual model of "asm" is simple.
Some instruction sets and architectures are hideous, though.
> merely being a human compiler is not particularly enlightening nor useful.
I don't think I can agree with that. At least it teaches you what the compiler is doing. And abiding by conventions (HLL-esque control flow, but also things like "put the return value in r0" and "put constant pools after the function") can definitely make it easier to make sense of the code. (Although you might share a constant pool across a module or something, if the instructions reach far enough.)
Not to say that you can't do interesting things, and can't ever beat the compiler. One of the things I most enjoyed discovering, in mid-00s era THUMB (i.e. 16-bit ARM) code, is that the compiler was implementing switch statements with tables of 32-bit constants that it would load into an indirect jump. I didn't get around to it, but I figured I could mechanically replace these with a computed jump into a "table" of 16-bit unconditional branches (except for very long functions, but this helped bring the branch distances under thresholds).
Ya totally I can also keep 32 registers, a memory file, and stack pointer all in my head at once ...fellow human... (In 2026 I might actually be an LLM in which I really can keep all that context in my "head"!)
Contrast this with trying to figure out all the nested implicit actions that a single line of some HLL like C++ will do.
I disagree with that. My abstractions are pure in my mind (well, I hope even if it sounds a bit pretentious). I try to get the best out of the tools I have at hand to represent them. I'm perfectly fine with leaky abstractions, mismatches, inconvenent languages, etc. I live with that. But I certainly don't actively seek these leaks. Quite the opposite :-) (well instead I'm pursuing another goal like performance, in which case, I blow abstractions away). Oh, and in case you wonder, I did write tons of assembly on 8086/80386 and 6502 and now I'm full on on rust, julia and python. I know what an abstraction is :-)
But I thnk we globally agree nonetheless
I have no interest in continuing this pointless internet interaction.