Inline assembly in Linux
github.com
github.com
Recap: call/cc is the ability to save the current state of a running thread, then revert to that state at a later point in time. In other words, at any point in your program, you can say "Save the current stack." It's saved as a function. Later, whenever you call that function, the current stack is thrown out, and replaced with the saved stack.
This is very useful for a number of reasons. It's also a very rare feature to have in your language.
Neither node nor lua has support for this. The closest I've found is a Lua extension which adds coroutine.clone(). In principle, this is the solution. In practice, it has a number of limitations, such as restrictions on when you're allowed to call coroutine.clone(). (For example, if your stack looks like Lua -> C -> Lua, then it won't work.)
I tried to channel my inner Mike Pall and solve this problem once and for all, but I'm not Mike Pall, and this is really hard. I was hoping you might know of any possible solution which is (a) practical, (b) cross-platform, and (c) works in all cases.
Why post this here? Because this post happens to attract exactly the kind of people that might know a way forward. There must be a way. Apologies for the off-topic comment.
He mentioned a paper titled "Exceptional Continuations in JavaScript" [2], that describes the method he used in his implementation.
Maybe you should get in touch?
I don't think you can safely solve this in the general case. There is a key problem I don't think you can work around.
Say your stack looks like C(1) -> Lua -> C -> Lua. The outermost C frames might not know anything about Lua (they just use some library that uses Lua as a library). Say you try to take a snapshot of this stack to create a continuation. You probably just want to snapshot the Lua -> C -> Lua part, since that is the portion of the stack representing the execution of the Lua program.
Now say all these frames return. Then the main program calls Lua again, through through a slightly different code-path, and now you have C(2) -> Lua. Say the embedded Lua program decides to resume the continuation.
Now keep in mind that the C stack is not position-independent. The C stack can contain pointers to the C stack, so when you resume, you need your resumed stack to live at exactly the same address as last time it ran. But what if C(1) and C(2) are not exactly the same size? What if we called one extra function before calling Lua the second time? It is impossible to copy the continuation's C stack back into its original position. So it's impossible to resume the continuation.
You could try to snapshot the entire C stack to get around this, including the outermost C frames. But this would be most unexpected for the C program that is using the Lua interpreter. Lua is supposed to just be a regular C library: you call a function, it does things, and then returns. It wouldn't be acceptable that calling a Lua interpreter function like lua_call() backs your entire C program to a previous state just because the embedded Lua program used a fancy feature called continuations!
There are many other things that would make this tricky at best to get working, but I think the problem above really tanks the idea completely.
Say you want to resume continuation K, which has a stack of some size N.
The current thread has a stack of size M. If M >= N, everything is fine: you can safely overwrite the current stack with K's stack.
If M < N, recurse until M >= N.
You could try to snapshot the entire C stack to get around this, including the outermost C frames.
Indeed! This is a solution.
It wouldn't be acceptable...
I like doing unacceptable things in my programs. It's the best part of programming, really.
There are a lot of solid arguments against call/cc. I think the most persuasive argument in favor of call/cc is that you become more powerful. Whatever metric you use to measure power, call/cc will improve it: Smaller code, less time spent writing code, and you can even write algorithms that you otherwise would not be able to.
Personally, I want call/cc in order to be able to use choose and fail. It's the ability to write programs that are guaranteed to never call fail(). pg explains it well:
"For example, this is a perfectly legitimate nondeterministic algorithm for discovering whether you have a known ancestor called Igor:
Function Ig(n)
if name(n) = ‘Igor’
return n
if parents(n)
return Ig(choose(parents(n)))
fail
Call/cc makes this possible. There are a lot of fun things to do. The last few chapters of On Lisp show some particularly interesting sketches.
The size of the continuation's stack doesn't matter for the problem I described, it's the size of the stack "underneath" your continuation that matters (ie. C(1) and C(2) above).
If C(2) > C(1) there is no way to shrink C(2) such that the continuation's stack can be copied into the right place.
> I like doing unacceptable things in my programs. It's the best part of programming, really.
What you do in your programs is up to you! But nobody else is going to use a C library that messes with the execution state of its callers (unless that is the point of the library, which it isn't with Lua).
To create a continuation, we need to copy the entire stack, by definition. But "the stack" is just an array of bytes. It's all the bytes between the current stack pointer and the "root" stack frame. So to create a continuation, copy these bytes and stash them somewhere, then set up a longjmp target to the current instruction.
To apply a continuation, i.e. to restore the stack, we overwrite the current stack starting from the root frame. Then we longjmp to where the continuation was originally created.
It seems like this scheme should work in any situation, but perhaps I'm missing something?
loeg pointed out getcontext(3) / setcontext(3), which seems promising. It looks like a standard way to sidestep all of this bookkeeping. It appears to be a high-level interface to the operations described above.
Lua is just a language, though. It's not "for" anything in particular.
Imagine you have a C program like this:
#include <fancylib.h>
int main() {
for (int i = 0; i < 10; i++) {
printf("val[%d] = %d\n", i, fancylib_calculate(i));
}
fancylib_cleanup();
}
Now imagine that the Lua function run by fancylib decides to use continuations. When you call fancylib_calculate(0), it creates a continuation. And when you call fancylib_calculate(1), it decides to call the continuation.
If you restore the entire C stack to resume the Lua continuation, it will reset the loop in main() to i=0! Your program might end up printing val[0] over and over, in an infinite loop. This would be extremely surprising to you as the author of main(), because you were just trying to write a normal old for() loop. The Lua continuation should just restore the Lua-related stack, not the stack of the functions calling Lua!
That's the point of continuations, though. That's a feature, not a bug. When you create a continuation, you're saying "whatever happens after this, allow me to do it again at some later time." If the calling library happened to be in a loop, then the goal is to serialize that loop so that it can be invoked again, at a later point.
If you keep applying the continuation in a loop, then you'll get an infinite loop. But if you invoke the continuation once, (and if subsequent calls to fancylib_calculate() don't), then you'll get the ability to print
val[0] = 42
val[1] = 99
...
val[9] = 7
In fact, you gain the ability to prevent the program from terminating, in a controlled fashion. Since you have access to the continuation, you can choose to invoke it on the 10th call to fancylib_calculate(), up to 3 times in a row. That would produce output like:
val[0] = 42
val[1] = 99
...
val[9] = 7
val[0] = 42
val[1] = 99
...
val[9] = 7
val[0] = 42
val[1] = 99
...
val[9] = 7
Does that make sense? Apologies if we're talking past each other. I appreciate the patience.
If that's what you're after, then by all means implement that. :) But I think most people would expect the Lua interpreter state to be self-contained, and not to affect the state of the surrounding C execution environment.
Unless you are allowing arbitrary code execution, you are likely the one passing in scripts to the interpreter - these sorts of interactions should be well documented by the lua scripts themselves...and I don't recommend allowing arbitrary lua anyhow - os.syscall and friends say hi - need to block access to those carefully. Thus, we have two situations - either you know and trust the code not to do unexpected things (or to do them in an expected manner :) ), or you've set up your lua environment in such a way to disallow such calls, and it doesn't matter.
If you want to use this to do something that takes multiple invocations of lua_resume to complete without the calling Lua code being aware, you might be able to use lua_yieldk.
#define ASM_VEC_BYTE_COUNT_SET(vec, sum, mask, shuf) \
__asm volatile ("vpsrld $4, %[VEC], %[SUM]\n" \
"vpand %[MASK], %[VEC], %[VEC]\n" \
"vpand %[MASK], %[SUM], %[SUM]\n" \
"vpshufb %[VEC], %[SHUF], %[VEC]\n" \
"vpshufb %[SUM], %[SHUF], %[SUM]\n" \
"vpaddb %[VEC], %[SUM], %[SUM]\n" : \
/* rd/wr ymm */ [VEC] "+&x" (vec), \
/* write ymm */ [SUM] "=&x" (sum) : \
/* read ymm */ [MASK] "x" (mask), \
/* read ymm */ [SHUF] "x" (shuf))
2) If you are using the same assembly more than once in your program, declare your assembly within a #define macro, then use the macro in your code.
3) Use "__asm volatile". Declaring "volatile" is not required, but once you are writing inline assembly you usually know more than the compiler about where the block should go.
5) If you have multiple lines of assembly and output registers, you are almost always safer to use "+&" and "=&" for your constraint rather than just "+" or "=". Search for "early clobber" for details.
6) Strongly prefer single type constraints. The more flexibility you give the compiler, the more likely it will defeat your efforts at optimization. Use explicit memory addressing modes rather than "m". The modifier "c" is needed for the offset.
#define ASM_VEC_LOAD_OFFSET_MEM(off, mem, vec) \
__asm volatile ("vmovdqu %c[OFF](%[MEM]), %[VEC]\n" : \
/* destination */ [VEC] "=x" (vec) : \
/* byte offset */ [OFF] "i" (off), \
/* mem address */ [MEM] "r" (mem))
Another approach that's not quite there yet but is becoming more possible is to use https://www.cilkplus.org to annotate your C code to force automatic vectorization. It's native to ICC, built-in to GCC 5.0+, and available as an extension to Clang: https://news.ycombinator.com/item?id=11550250
5 - I can't think of any meaning early clobber has on an input+output constraint ("+")?
6 - there are many cases where you really do want to give the compiler flexibility in addressing modes. Unfortunately clang tends to ignore that and generate (reg) regardless.
7 - not really different than GPRs; you use "r" as the constraint then a modifier like "k" for the size.
I guess the lesson is that yeah gcc inline asm is powerful, but they try to leave it undocumented for a reason. Also, who stole number 4?
re 5: Barring compiler bugs, I think you'd be right if correctness was the only issue. But I'm pretty sure I've sometimes solved problems by adding it, although this may have been when working around the POPCNT bug that added a false dependency on the output. It also might have been when reading and writing a variable multiple times?
re 6: In theory, yes. But usually in these cases you should be writing intrinsics or straight C instead of inline assembly. The place where this comes up most for me is when I have two variables that use the same index, and I want to ensure "DEC/JNZ" fusion at the end of the loop. If I let the compiler choose, it will find a way to defeat me by incrementing both array addresses. The other case is when you explicitly want a store to use Port 7 for address generation, which only happens without an index register.
re 7: Yes, I just personally find it more confusing because "x" fits so well with "XMM", and thus it feels odd to use it when you want only a "YMM". Also, see here for problems with a Clang and %q[VEC]: http://stackoverflow.com/questions/34459803/in-gnu-c-inline-...
re 4: Oops, I forgot to renumber. I had another comment suggesting that one always use the "V" VEX prefix on vector commands and the explicit output register, but deleted it because it seemed off topic.
Of course this is not true for synchronization primitives and the like.
My goal is to "lock in" an established level of performance once I've achieved it, so that compiler upgrades or changes don't result in performance drops. I often compare the output of multiple compilers with a matrix of optimization flags, choose the best blocks from each, and then hand-optimize from there while cross-referencing Agner's handbooks with Likwid's performance reports. If I've chosen to use inline assembly, the chances that the compiler will succeed in further optimizing my code is very low.
I realize it's not a popular view, but I think that using volatile with __asm is usually the correct approach. If you don't need "volatile", you probably should be using an intrinsic instead. I think the alternative (which may in fact be the better solution) is dropping to straight assembly for the entire function or distributing binary code.
> Do not expect a sequence of asm statements to remain perfectly consecutive after compilation, even when you are using the volatile qualifier. If certain instructions need to remain consecutive in the output, put them in a single multi-instruction asm statement.
The second danger is that "asm volatile" hides incorrect operand specification. If you examine the assembly, you might get the wrong assembly, and adding "volatile" might fix it. However, the incorrect operand specification might cause problems in other parts of the code. These are harder to diagnose. Stack Overflow is littered with questions by people who specify asm operands wrong, add "volatile" to fix the assembly, but other things are still broken. My general procedure is to work with asm blocks at -O2 or higher without using volatile, and make sure I'm getting the desired results that way (unless I'm writing some synchronization primitives).
Yet it is just so damn easy to write larger, multi-statement asm blocks. With larger blocks, the intent of the programmer is clear. It becomes obvious to both the reader and to the compiler that the assembly should be emitted as-is, rather than moved or reordered.
Finally, you can often get the results you want with the auto-vectorizer, restrict, and __builtin_assume_aligned. Whenever that is possible I'd prefer it.
The asm {} blocks of PC compilers are so much developer friendly.
I rather use an external Assembler than GCC's inline syntax.
#define mb() asm volatile("mfence":::"memory")
#define rmb() asm volatile("lfence":::"memory")
#define wmb() asm volatile("sfence" ::: "memory")
[1]: https://github.com/andrewrk/zig/blob/master/std/linux_x86_64...