Defining variables in c++ inline assembly - c++

I am searching for a way to define variables in c++ inline assembly. I found an interesting way to do it. But it confuses me, how this can work.
__asm
{
push ebp
mov ebp, esp
add esp, 4
mov [ebp - 4], 2
mov esp, ebp
pop ebp
}
I see this code as - Push base pointer address to the stack, move stack pointer address into base pointer's (Stack logically should collapse here, because this is common epilogue function of cleaning the stack). Then we move 4 to the esp address (Not even the value) And then remove that 4 from esp. So we get back to the same esp address. Strange fact for me is that, it even compiles, and it works. But when I try to test it by outputting the value
uint32_t output;
__asm
{
push ebp
mov ebp, esp
add esp, 4
mov [ebp - 4], 2
mov output,[ebp-4]
mov esp, ebp
pop ebp
}
std::cout << output;
It does not compile, showing "Operand size conflict", which seems weird to me, because I use 32 bit integer and register is also 32 bit. When using [ebp-4] without [], it gives garbage values, as expected.
So, maybe someone could explain how this works without giving error :)
And one additional question, why does db does not work in inline assembly?

It doesn't work, that doesn't define a C++ variable.
It just messes with the stack to reserve some new storage below the stack frame created by the compiler. And you modify EBP so compiler-generated addressing modes that use EBP will be broken.1
If you want to define or declare a C++ variable, do it with C++ syntax like int tmp.
asm doesn't really have variables. It has registers and memory. Keep track of where values are using comments. If you want to use some extra stack space from MSVC inline asm, I think that's safe, but don't modify EBP if you also want to reference C++ local variables.
Footnote 1:
That would be the case if your code assembled at all, which it won't because mov output,[ebp-4] has 2 explicit memory operands. MSVC inline asm can't allocate C++ variables in register.
Also mov [ebp - 4], 2 has ambiguous operand-size: neither operand has a size associated with it because neither is a register. Maybe you want mov dword ptr [ebp - 4], 2

Related

Decoding assembly from MSVC 32-bit release (homework). What does no-op do?

Hi heads up this is a homework. I'm given an assembly generated by MSVC 32-bit Release with optimizations on, and I'm supposed to decode it back into C++. I've included the top of the function to the line I'm having problems with. The comments are mine, which I'm wrote while trying to understand this.
Note: Code is supposedly generated from C++. Not traditional ASM.
Note 2: There is one area of undefined behavior in the code.
Here are the lines I'm stuck with
TheFunction: ; TheFunction(int* a, int s);
0F2D4670 push ebp ; Push/clear/save ebp
0F2D4671 mov ebp,esp ; ebp now points to top of stack
0F2D4673 push ecx ; Push/clear/save ecx
0F2D4674 push ebx ; Push/clear/save ebs
0F2D4675 push esi ; Push/clear/save esi
0F2D4676 mov ebx,edx ; ebx = int s
0F2D4678 mov esi,1 ; esi = 1
0F2D467D push edi ; calling convention ; Push/clear/save edi
0F2D467E mov edi,dword ptr [a (0F2D95E8h)] ; edi = a[0]
0F2D4684 cmp ebx,esi ; if(s < 1)
0F2D4686 jl SomeFunction+3Ch (0F2D46ACh) ; Jump to return
0F2D4688 nop dword ptr [eax+eax] ; !! <-- No op involving dereferencing? What does this do?
0F2D4690 mov eax,dword ptr [edi+esi*4-4] ; !! <-- edi is *a, while esi is 1. There is no address
here!
..... More code but I've figured these out ....
I've more or less got the gist of the function. Its a function that takes a pointer to an int, with an underlying array, and a size. It then goes through each element in the array from last to first, adding to each subsequent one and printing it out. However, I still haven't got the details down and need help
Two questions, both at the end of the code snippet. What does no op on a dereference pointer do, and am I reading the last line in that its attempting to dereference something not in memory?
The nop dword ptr [eax+eax] instruciton does nothing. It doesn't even access the memory location given by the operand. It literally performs no operation.
It's just there so the next instruction is aligned to a 16-byte boundary. You'll notice that next instruction address is 0F2D4690 which ends with 0 which means it's 16-byte aligned. This can improve the performance of loops. Somewhere there will be an instruction that jumps back to 0F2D4690 as part of a loop. This particular form of a NOP instruction is used because it encodes a single NOP instruction in 8 bytes.
There is no corresponding C++ code for this instruction. You shouldn't try to represent it in your C++ code, just ignore it.
Also note that your comment for mov edi,dword ptr [a (0F2D95E8h)] is incorrect. Instead of being edi = a[0] it's simply edi = a. The variable a isn't a parameter at all, instead it's a global (or file level static) variable located at memory location 0F2D95E8h. This instruction just loads the value from memory.

MSVC Assembly function arguments C++ vs _asm

I have a function which takes 3 arguments, dest, src0, src1, each a pointer to data of size 12. I made two versions. One is written in C and optimized by the compiler, the other one is fully written in _asm. So yeah. 3 arguments? I naturally do something like:
mov ecx, [src0]
mov edx, [src1]
mov eax, [dest]
I am a bit confused by the compiler, as it saw fit to add the following:
_src0$ = -8 ; size = 4
_dest$ = -4 ; size = 4
_src1$ = 8 ; size = 4
?vm_vec_add_scalar_asm##YAXPAUvec3d##PBU1#1#Z PROC ; vm_vec_add_scalar_asm
; _dest$ = ecx
; _src0$ = edx
; 20 : {
sub esp, 8
mov DWORD PTR _src0$[esp+8], edx
mov DWORD PTR _dest$[esp+8], ecx
; 21 : _asm
; 22 : {
; 23 : mov ecx, [src0]
mov ecx, DWORD PTR _src0$[esp+8]
; 24 : mov edx, [src1]
mov edx, DWORD PTR _src1$[esp+4]
; 25 : mov eax, [dest]
mov eax, DWORD PTR _dest$[esp+8]
Function body etc.
add esp, 8
ret 0
What does the _src0$[esp+8] etc. even means? Why does it do all this stuff before my code? Why does it try to [apparently]stack anything so badly?
In comparison, the C++ version has only the following before its body, which is pretty similar:
_src1$ = 8 ; size = 4
?vm_vec_add##YAXPAUvec3d##PBU1#1#Z PROC ; vm_vec_add
; _dest$ = ecx
; _src0$ = edx
mov eax, DWORD PTR _src1$[esp-4]
Why is this little sufficient?
The answer of Mats Petersson explained __fastcall. But I guess that is not exactly what you're asking ...
Actually _src0$[esp+8] just means [_src0$ + esp + 8], and _src0$ is defined above:
_src0$ = -8 ; size = 4
So, the whole expression _src0$[esp+8] is nothing but [esp] ...
To see why it does all these stuff, you should probably first understand what Mats Petersson said in his post, the __fastcall, or more generally, what is a calling convention. See the link in his post for detailed informations.
Assuming that you have understood __fastcall, now let's see what happens to your codes. The compiler is using __fastcall. Your callee function is f(dst, src0, src1), which requires 3 parameters, so according to the calling convention, when a caller calls f, it does the following:
Move dst to ecx and src0 to edx
Push src1 onto the stack
Push the 4 bytes return address onto the stack
Go to the starting address of the function f
And the callee f, when its code begins, then knows where the parameters are: dst and src0 are in the registers ecx and edx, respectively; esp is pointing to the 4 bytes return address, but the 4 bytes below it (i.e. DWORD PTR[esp+4]) is exactly src1.
So, in your "C++ version", the function f just does what it should do:
mov eax, DWORD PTR _src1$[esp-4]
Here _src1$ = 8, so _src1$[esp-4] is exactly [esp+4]. See, it just retrieves the parameter src1 and stores it in eax.
There is however a tricky point here. In the code of f, if you want to use the parameter src1 multiple times, you can certainly do that, because it's always stored in the stack, right below the return address; but what if you want to use dst and src0 multiple times? They are in the registers, and can be destroyed at any time.
So in that case, the compiler should do the following: right after entering the function f, it should remember the current values of ecx and edx (by pushing them onto the stack). These 8 bytes are the so-called "shadow space". It is not done in your "C++ version", probably because the compiler knows for sure that these two parameters will not be used multiple times, or that it can handle it properly some other way.
Now, what happens to your _asm version? The problem here is that you are using inline assembly. The compiler then loses its control to the registers, and it cannot assume that the registers ecx and edx are safe in your _asm block (they are actually not, since you used them in the _asm block). Thus it is forced to save them at the beginning of the function.
The saving goes as follows: it first raises esp by 8 bytes (sub esp, 8), then move edx and ecx to [esp] and [esp+4] respectively.
And then it can enter safely your _asm block. Now in its mind (if it has one), the picture is that [esp] is src0, [esp+4] is dst, [esp+8] is the 4 byte return address, and [esp+12] is src1. It no longer thinks about ecx and edx.
Thus your first instruction in the _asm block, mov ecx, [src0], should be interpreted as mov ecx, [esp], which is the same as
mov ecx, DWORD PTR _src0$[esp+8]
and the same for the other two instructions.
At this point, you might say, aha it's doing stupid things, I don't want it to waste time and space on that, is there a way?
Well there is a way - do not use inline assembly... it's convenient, but there is a compromise.
You can write the assembly function f in a .asm source file and public it. In the C/C++ code, declare it as extern 'C' f(...). Then, when you begin your assembly function f, you can play directly with your ecx and edx.
The compiler has decided to use a calling convention that uses "pass arguments in registers" aka __fastcall. This allows the compiler to pass some of the arguments in registers, instead of pushing onto stack, and this can reduce the overhead in the call, because moving from a variable to a register is faster than pushing onto the stack, and it's now already in a register when we get to the callee function, so no need to read it from the stack.
There is a lot more information about how calling conventions work on the web. The wikipedia article on x86 calling conventions is a good starting point.

Stack must be clean before function epilogue confusion

I'm studying assembly language from the book "Assembly Language Step-by-Step: Programming with Linux" by Jeff Dunteman, and have come across an interesting paragraph in the book which I'm most likely misunderstanding, hence would appreciate some clarification on:
"The stack must be clean before we destroy the stack frame and return control. This simply means that any temporary values that we may have pushed onto the stack during the program’s run must be gone. All that is left on the stack should be the caller’s EBP, EBX, ESI, and EDI values.
...
Once the stack is clean, to destroy the stack frame we must first pop the caller’s register values back into their registers, ensuring that the pops are in the correct order.
...
We restore the caller’s ESP by moving the value from EBP into ESP, and finally pop the caller’s EBP value off the stack."
Consider the following code generated from Visual Studio 2008:
int myFuncSum( int a, int b)
{
001B1020 push ebp
001B1021 mov ebp,esp
001B1023 push ecx <------------------
int c;
c = a + b;
001B1024 mov eax,dword ptr [ebp+8]
001B1027 add eax,dword ptr [ebp+0Ch]
001B102A mov dword ptr [ebp-4],eax
return c;
001B102D mov eax,dword ptr [ebp-4]
}
001B1030 mov esp,ebp
001B1032 pop ebp
001B1033 ret
The value of ecx (indicated), pushed to make space on the stack for my variable c, is, as far as I can see, only gone from the stack when we reset ESP; however, as quoted, the book states that the stack must be clean before we reset ESP. Can someone please clarify whether or not I am missing something?
The example from Visual Studio 2008 doesn't contradict the book. The book is covering the most elaborate case of a call. See the x86-32 Calling Convention as a cross-reference which spells it out with pictures.
In your example, there were no caller registers saved on the stack, so there are no pop instructions to be performed. This is part of the "clean up" that must occur before mov esp, ebp that the book is referring to. So more specifically, let's suppose the callee is saving si and di for the caller, then the prelude and postlude for the function might look like this:
push ebp ; save base pointer
mov ebp, esp ; setup stack frame in base pointer
sub esp, 4 ; reserve 4 bytes of local data
push si ; save caller's registers
push di
; do some stuff, reference 32-bit local variable -4(%ebp), etc
; Use si and di for our own purposes...
; clean up
pop di ; do the stack clean up
pop si ; restoring the caller's values
mov esp, ebp ; restore the stack pointer
pop ebp
ret
In your simple example, there were no saved caller registers, so no final pop instructions needed at the end.
Perhaps because it's simpler or faster, the compiler elected to do the following instruction in place of sub esp, 4:
push ecx
But the effect is the same: reserve 4 bytes for a local variable.
Notice the instruction:
push ebp
mov ebp,esp ; <<<<=== saves the stack base pointer
and the instruction:
mov esp,ebp ; <<<<<== restore the stack base pointer
pop ebp
So after this sequence the stack is clean again

Can anyone help me interpret this MSVC debug-mode disassembly from a simple Hello World?

I got the following simple C++ code:
#include <stdio.h>
int main(void)
{
::printf("\nHello,debugger!\n");
}
And from WinDbg, I got the following disassembly code:
SimpleDemo!main:
01111380 55 push ebp
01111381 8bec mov ebp,esp
01111383 81ecc0000000 sub esp,0C0h
01111389 53 push ebx
0111138a 56 push esi
0111138b 57 push edi
0111138c 8dbd40ffffff lea edi,[ebp-0C0h]
01111392 b930000000 mov ecx,30h
01111397 b8cccccccc mov eax,0CCCCCCCCh
0111139c f3ab rep stos dword ptr es:[edi]
0111139e 8bf4 mov esi,esp
011113a0 683c571101 push offset SimpleDemo!`string' (0111573c)
011113a5 ff15b0821101 call dword ptr [SimpleDemo!_imp__printf (011182b0)]
011113ab 83c404 add esp,4
011113ae 3bf4 cmp esi,esp
011113b0 e877fdffff call SimpleDemo!ILT+295(__RTC_CheckEsp) (0111112c)
011113b5 33c0 xor eax,eax
011113b7 5f pop edi
011113b8 5e pop esi
011113b9 5b pop ebx
011113ba 81c4c0000000 add esp,0C0h
011113c0 3bec cmp ebp,esp
011113c2 e865fdffff call SimpleDemo!ILT+295(__RTC_CheckEsp) (0111112c)
011113c7 8be5 mov esp,ebp
011113c9 5d pop ebp
011113ca c3 ret
I have some difficulties to fully understand it. What is the SimpleDemo!ILT things doing here?
What's the point of the instruction comparing ebp and esp at 011113c0?
Since I don't have any local variables in main() function, why there's still a sub esp,0C0h at the loacation of 01111383?
Many thanks.
Update 1
Though I still don't know what ILT means, but the __RTC_CheckESP is for runtime checks. These code can be elimiated by placing the following pragma before the main() function.
#pragma runtime_checks( "su", off )
Reference:
http://msdn.microsoft.com/en-us/library/8wtf2dfz.aspx
http://msdn.microsoft.com/en-us/library/6kasb93x.aspx
Update 2
The sub esp,0C0h instruction allocate another 0C0h bytes extra space on the stack. Then EAX is filled with 0xCCCCCCCC, this is 4 bytes, since ECX=30h, 4*30h=0C0h, so the instruction rep stos dword ptr es:[edi] fill exactly the extra spaces with 0xCC. But what is this extra space on stack for? Is this some kind of safe belt? Also I notice that if I turn off the runtime check as Update 1 shows, there's still such extra space on stack, though much smaller. And this space is not filled with 0xCC.
The assembly code without runtime check is like below:
SimpleDemo!main:
00231250 55 push ebp
00231251 8bec mov ebp,esp
00231253 83ec40 sub esp,40h <-- Still extra space allocated from stack, but smaller
00231256 53 push ebx
00231257 56 push esi
00231258 57 push edi
00231259 683c472300 push offset SimpleDemo!`string' (0023473c)
0023125e ff1538722300 call dword ptr [SimpleDemo!_imp__printf (00237238)]
00231264 83c404 add esp,4
00231267 33c0 xor eax,eax
00231269 5f pop edi
0023126a 5e pop esi
0023126b 5b pop ebx
0023126c 8be5 mov esp,ebp
0023126e 5d pop ebp
0023126f c3 ret
Most of the instructions are part of MSVC runtime checking, enabled by default for debug builds. Just calling printf and returning 0 in an optimized build takes much less code. (Godbolt compiler explorer). Other compilers (like GCC and clang) don't do as much stuff like stack-pointer comparison after calls, or poisoning stack memory with a recognizable 0xCC pattern to detect use-uninitialized, so their debug builds are like MSVC debug mode without its extra runtime checks.
I've annotated the assembler, hopefully that will help you a bit. Lines starting 'd' are debug code lines, lines starting 'r' are run time check code lines. I've also put in what I think a debug with no runtime checks version and release version would look like.
; The ebp register is used to access local variables that are stored on the stack,
; this is known as a stack frame. Before we start doing anything, we need to save
; the stack frame of the calling function so it can be restored when we finish.
push ebp
; These two instructions create our stack frame, in this case, 192 bytes
; This space, although not used in this case, is useful for edit-and-continue. If you
; break the program and add code which requires a local variable, the space is
; available for it. This is much simpler than trying to relocate stack variables,
; especially if you have pointers to stack variables.
mov ebp,esp
d sub esp,0C0h
; C/C++ functions shouldn't alter these three registers in 32-bit calling conventions,
; so save them. These are stored below our stack frame (the stack moves down in memory)
r push ebx
r push esi
r push edi
; This puts the address of the stack frame bottom (lowest address) into edi...
d lea edi,[ebp-0C0h]
; ...and then fill the stack frame with the uninitialised data value (ecx = number of
; dwords, eax = value to store)
d mov ecx,30h
d mov eax,0CCCCCCCCh
d rep stos dword ptr es:[edi]
; Stack checking code: the stack pointer is stored in esi
r mov esi,esp
; This is the first parameter to printf. Parameters are pushed onto the stack
; in reverse order (i.e. last parameter pushed first) before calling the function.
push offset SimpleDemo!`string'
; This is the call to printf. Note the call is indirect, the target address is
; specified in the memory address SimpleDemo!_imp__printf, which is filled in when
; the executable is loaded into RAM.
call dword ptr [SimpleDemo!_imp__printf]
; In C/C++, the caller is responsible for removing the parameters. This is because
; the caller is the only code that knows how many parameters were put on the stack
; (thanks to the '...' parameter type)
add esp,4
; More stack checking code - this sets the zero flag if the stack pointer is pointing
; where we expect it to be pointing.
r cmp esi,esp
; ILT - Import Lookup Table? This is a statically linked function which throws an
; exception/error if the zero flag is cleared (i.e. the stack pointer is pointing
; somewhere unexpected)
r call SimpleDemo!ILT+295(__RTC_CheckEsp))
; The return value is stored in eax by convention
xor eax,eax
; Restore the values we shouldn't have altered
r pop edi
r pop esi
r pop ebx
; Destroy the stack frame
r add esp,0C0h
; More stack checking code - this sets the zero flag if the stack pointer is pointing
; where we expect it to be pointing.
r cmp ebp,esp
; see above
r call SimpleDemo!ILT+295(__RTC_CheckEsp)
; This is the usual way to destroy the stack frame, but here it's not really necessary
; since ebp==esp
mov esp,ebp
; Restore the caller's stack frame
pop ebp
; And exit
ret
; Debug only, no runtime checks
push ebp
mov ebp,esp
d sub esp,0C0h
d lea edi,[ebp-0C0h]
d mov ecx,30h
d mov eax,0CCCCCCCCh
d rep stos dword ptr es:[edi]
push offset SimpleDemo!`string'
call dword ptr [SimpleDemo!_imp__printf]
add esp,4
xor eax,eax
mov esp,ebp
pop ebp
ret
; Release mode (The optimiser is clever enough to drop the frame pointer setup with no VLAs or other complications)
push offset SimpleDemo!`string'
call dword ptr [SimpleDemo!_imp__printf]
add esp,4
xor eax,eax
ret
Number one your code's main() is improperly formed. It doesn't return the int you promised it would return. Correcting this defect, we get:
#include
int main(int argc, char *argv[])
{
::printf("\nHello,debugger!\n");
return 0;
}
Additionally, any more, it is very strange to see #include <stdio.h> in a C++ program. I believe you want #include <cstdio>
In all cases, space must be made on the stack for arguments and for return values. main()'s return value requires stack space. main()s context to be saved during the call to printf() requires stack space. printf()'s arguments require stack space. printf()'s return value requires stack space. That's what the 0c0h byte stack frame is doing.
The first thing that happens is the incoming bas pointer is copied to the top of the stack. Then the new stack pointer is copied into the base pointer. We'll be checking later to be sure that the stack winds up back where it started from (because you have runtime checking turned on). Then we build the (0C0h bytes long) stack frame to hold our context and printf()'s arguments during the call to printf(). We jump to printf(). When we get back, we hop over the return value which you didn't check in your code (the only thing left on its frame) and make sure the stack after the call is in the same place it was before the call. We pop our context back off the stack. We then check that the final stack pointer matches the value we saved way up at the front. Then we pop the prior value of the base pointer off the very top of the stack and return.
That is code that is inserted by the compiler when you build with runtime checking (/RTC). Disable those options and it should be clearer. /GZ could also be causing this depending on your VS version.
For the record, I suspect that ILT means "Incremental Linking Thunk".
The way incremental linking (and Edit&Continue) works is the following: the linker adds a layer of indirection for every call via thunks which are grouped at the beginning of executable, and adds a huge reserved space after them. This way, when you're relinking the updated executable it can just put any new/changed code into the reserved area and patch only the affected thunks, without changing the rest of the code.
The 40 bytes is the worst case stack allocation for any called or subsequently called function. This is explained in glorious detail here.
What is this space reserved on the top of the stack for? First, space is created for any local variables. In this case, FunctionWith6Params() has two. However, those two local variables only account for 0x10 bytes. What’s the deal with the rest of the space created on the top of the stack?
On the x64 platform, when code prepares the stack for calling another function, it does not use push instructions to put the parameters on the stack as is commonly the case in x86 code. Instead, the stack pointer typically remains fixed for a particular function. The compiler looks at all of the functions the code in the current function calls, it finds the one with the maximum number of parameters, and then creates enough space on the stack to accommodate those parameters. In this example, FunctionWith6Params() calls printf() passing it 8 parameters. Since that is the called function with the maximum number of parameters, the compiler creates 8 slots on the stack. The top four slots on the stack will then be the home space used by any functions FunctionWith6Params() calls.

Register keyword in C++

What is difference between
int x=7;
and
register int x=7;
?
I am using C++.
register is a hint to the compiler, advising it to store that variable in a processor register instead of memory (for example, instead of the stack).
The compiler may or may not follow that hint.
According to Herb Sutter in "Keywords That Aren't (or, Comments by Another Name)":
A register specifier has the same
semantics as an auto specifier...
According to Herb Sutter, register is "exactly as meaningful as whitespace" and has no effect on the semantics of a C++ program.
In C++ as it existed in 2010, any program which is valid that uses the keywords "auto" or "register" will be semantically identical to one with those keywords removed (unless they appear in stringized macros or other similar contexts). In that sense the keywords are useless for properly-compiling programs. On the other hand, the keywords might be useful in certain macro contexts to ensure that improper usage of a macro will cause a compile-time error rather than producing bogus code.
In C++11 and later versions of the language, the auto keyword was re-purposed to act as a pseudo-type for objects which are initialized, which a compiler will automatically replace with the type of the initializing expression. Thus, in C++03, the declaration: auto int i=(unsigned char)5; was equivalent to int i=5; when used within a block context, and auto i=(unsigned char)5; was a constraint violation. In C++11, auto int i=(unsigned char)5; became a constraint violation while auto i=(unsigned char)5; became equivalent to auto unsigned char i=5;.
With today's compilers, probably nothing. Is was orginally a hint to place a variable in a register for faster access, but most compilers today ignore that hint and decide for themselves.
register is deprecated in C++11. It is unused and reserved in C++17.
Source: http://en.cppreference.com/w/cpp/keyword/register
Almost certainly nothing.
register is a hint to the compiler that you plan on using x a lot, and that you think it should be placed in a register.
However, compilers are now far better at determining what values should be placed in registers than the average (or even expert) programmer is, so compilers just ignore the keyword, and do what they wants.
The register keyword was useful for:
Inline assembly.
Expert C/C++ programming.
Cacheable variables declaration.
An example of a productive system, where the register keyword was required:
typedef unsigned long long Out;
volatile Out out,tmp;
Out register rax asm("rax");
asm volatile("rdtsc":"=A"(rax));
out=out*tmp+rax;
It has been deprecated since C++11 and is unused and reserved in C++17.
As of gcc 9.3, compiling using -std=c++2a, register produces a compiler warning, but it still has the desired effect and behaves identically to C's register when compiling without -O1–-Ofast optimisation flags in the respect of this answer. Using clang++-7 causes a compiler error however. So yes, register optimisations only make a difference on standard compilation with no optimisation -O flags, but they're basic optimisations that the compiler would figure out even with -O1.
The only difference is that in C++, you are allowed to take the address of the register variable which means that the optimisation only occurs if you don't take the address of the variable or its aliases (to create a pointer) or take a reference of it in the code (only on - O0, because a reference also has an address, because it's a const pointer on the stack, which, like a pointer can be optimised off the stack if compiling using -Ofast, except they will never appear on the stack using -Ofast, because unlike a pointer, they cannot be made volatile and their addresses cannot be taken), otherwise it will behave like you hadn't used register, and the value will be stored on the stack.
On -O0, another difference is that const register on gcc C and gcc C++ do not behave the same. On gcc C, const register behaves like register, because block-scope consts are not optimised on gcc. On clang C, register does nothing and only const block-scope optimisations apply. On gcc C, register optimisations apply but const at block-scope has no optimisation. On gcc C++, both register and const block-scope optimisations combine.
#include <stdio.h> //yes it's C code on C++
int main(void) {
const register int i = 3;
printf("%d", i);
return 0;
}
int i = 3;:
.LC0:
.string "%d"
main:
push rbp
mov rbp, rsp
sub rsp, 16
mov DWORD PTR [rbp-4], 3
mov eax, DWORD PTR [rbp-4]
mov esi, eax
mov edi, OFFSET FLAT:.LC0
mov eax, 0
call printf
mov eax, 0
leave
ret
register int i = 3;:
.LC0:
.string "%d"
main:
push rbp
mov rbp, rsp
push rbx
sub rsp, 8
mov ebx, 3
mov esi, ebx
mov edi, OFFSET FLAT:.LC0
mov eax, 0
call printf
mov eax, 0
mov rbx, QWORD PTR [rbp-8] //callee restoration
leave
ret
const int i = 3;
.LC0:
.string "%d"
main:
push rbp
mov rbp, rsp
sub rsp, 16
mov DWORD PTR [rbp-4], 3 //still saves to stack
mov esi, 3 //immediate substitution
mov edi, OFFSET FLAT:.LC0
mov eax, 0
call printf
mov eax, 0
leave
ret
const register int i = 3;
.LC0:
.string "%d"
main:
push rbp
mov rbp, rsp
mov esi, 3 //loads straight into esi saving rbx push/pop and extra indirection (because C++ block-scope const is always substituted immediately into the instruction)
mov edi, OFFSET FLAT:.LC0 // can't optimise away because printf only takes const char*
mov eax, 0 //zeroed: https://stackoverflow.com/a/6212755/7194773
call printf
mov eax, 0 //default return value of main is 0
pop rbp //nothing else pushed to stack -- more efficient than leave (rsp == rbp already)
ret
register tells the compiler to 1)store a local variable in a callee saved register, in this case rbx, and 2)optimise out stack writes if address of variable is never taken. const tells the compiler to substitute the value immediately (instead of assigning it a register or loading it from memory) and write the local variable to the stack as default behaviour. const register is the combination of these emboldened optimisations. This is as slimline as it gets.
Also, on gcc C and C++, register on its own seems to create a random 16 byte gap on the stack for the first local on the stack, which doesn't happen with const register.
Compiling using -Ofast however; register has 0 optimisation effect because if it can be put in a register or made immediate, it always will be and if it can't it won't be; const still optimises out the load on C and C++ but at file scope only; volatile still forces the values to be stored and loaded from the stack.
.LC0:
.string "%d"
main:
//optimises out push and change of rbp
sub rsp, 8 //https://stackoverflow.com/a/40344912/7194773
mov esi, 3
mov edi, OFFSET FLAT:.LC0
xor eax, eax //xor 2 bytes vs 5 for mov eax, 0
call printf
xor eax, eax
add rsp, 8
ret
Consider a case when compiler's optimizer has two variables and is forced to spill one onto stack. It so happened that both variables have the same weight to the compiler. Given there is no difference, the compiler will arbitrarily spill one of the variables. On the other hand, the register keyword gives compiler a hint which variable will be accessed more frequently. It is similar to x86 prefetch instruction, but for compiler optimizer.
Obviously register hints are similar to user-provided branch probability hints, and can be inferred from these probability hints. If compiler knows that some branch is taken often, it will keep branch related variables in registers. So I suggest caring more about branch hints, and forgetting about register. Ideally your profiler should communicate somehow with the compiler and spare you from even thinking about such nuances.