I have to admit I don't know where this is going. Does whatever we're talking about have to be an advantage all the time? To expect that just seems like a pessimistic point of view. At least assembly on the Mill is more legible than the average dialect these days.
At least it has to be not a disadvantage.Originally Posted by whiteflags
But I'm mostly just curious as this is the first time I've heard of using "C++ as assembly," and of course, getting a significant speed boost of x86.
Question about Rust; does it really have a standard ABI?
Note: I have found info that implies it supports multiple ABI; but, no info that implies it support a standard ABI as like C does.
C supports a standard ABI that is used by multiple Compiler Vendors.
Tim S.
"...a computer is a stupid machine with the ability to do incredibly smart things, while computer programmers are smart people with the ability to do incredibly stupid things. They are,in short, a perfect match.." Bill Bryson
C does not have a standard ABI AFAIK. It only happens that due to its simplicity, it often just works.
No. It may be on their roadmap, but for now it's very much like the C ABI.
That's a good thing. ABIs should be an implementation detail.
Incorrect! It's much more complicated than one might naively think. The reason it "just works", is because each platform has their own standard C ABI.
O_oC supports a standard ABI that is used by multiple Compiler Vendors.
Elysia is correct.
There is no C standard ABI. Sometimes things will work out because the language is simpler. (You don't, for example, need mangling to account for overloading.) The ABI you are probably familiar with is one of a few common ABI mostly designed by hardware engineers. (A variation of the "Itanium" ABI is very common these days even in x8664 land.) Even then, the ABI you are familiar with only has so much conformance. (On x86, for example, different compilers use different "internal padding" for variables depending on the options. You might, same example, compile with some variation of "use double precision for all mathematics" changing the size of all structures with a `float' member. Depending on other compilers, you'll never be able to directly interface the compiled code because some compilers do not have options for aligning `double' to 96 bits while that is the standard for other compilers.) One you reach beyond default everything, the ABI gets less and less "standard".
Soma
“Salem Was Wrong!” -- Pedant Necromancer
“Four isn't random!” -- Gibbering Mouther
That is a bad thing. ABIs should be standardized so that any two compilers can just plug and play. Of course, it may be that it needs to be on a platform level.
:facepalm:
When a language aspect is standardized by the platform rather than the language specification itself, that makes it an implementation detail, just as I said.
No, it doesn't.
A standard may still mandate what a platform ABI may and may not do.
Besides, many things such as name mangling can be part of the standard ABI, while lower level things such as calling convention may be part of the platform ABI.
O_oBesides, many things such as name mangling can be part of the standard ABI, while lower level things such as calling convention may be part of the platform ABI.
If the standard mandated a particular mangling scheme, most vendors would not follow that part of the standard.
*shrug*
Face it, you aren't going to see a standard ABI which will be widely adopted for years and years if ever.
For a vendor, "backwards compatibility" would be more important than rigid adherence to the standard because clients expect a level of compatibility.
Soma
“Salem Was Wrong!” -- Pedant Necromancer
“Four isn't random!” -- Gibbering Mouther
That's why it's important to have in place from day #1 so no legacy code will be able to "get in the way." That gives the highest probability of there ever existing a common ABI.
You just don't get it, do you? The 'B' in ABI means 'binary', as in, machine code. It cannot be fully standardized by the language specification without making it unportable (to anything not covered by the spec, obviously).
I do get it, and I never said it had to be fully specified by the standard.
So... Are we gonna all just ignore the fact that MILL architecture CPUs sounds absolutely astonishingly amazing? It's free performance boosts without code rewrites. I started watching the presentation on branching and I'm really starting to get into this. I mean, forget writing in assembly. Learning assembly in this day and age just seems silly unless you're either getting paid for it or you're getting paid a lot for it.
The speaker is also a perfect caricature of a computer scientist, imo. And I absolutely loved his joke, "Now, some of you younger folks may not recognize it but this is 'assembly language'."
it doesn't
both of those examples expand to simple codes when compiled, just a few repeated instructions in the binary
Last edited by iopq; 06-22-2014 at 02:54 PM.
