What is the equivalent of GCC's #pragma GCC optimize("O0") or VS's #pragma optimize("", off) in Apple LLVM 5.0 compiler?
I need it to disable optimizations for just a section of code.
From a brief search it doesn't look like clang/llvm supports such a pragma at this time. If you don't want to turn off optimizations for an entire file I suggest factoring what you don't want optimized into a separate file and setting -O0 on it separately.
Actually there is now a way to do that by specifying an __attribute__ ((optnone)) to the function that wraps the code you don't want to be optimized.
For instance I'm using it to have a clear benchmark of an inline function
static void BM_notoptimizedfunction(benchmark::State& state) __attribute__ ((optnone)) {
// your code here won't be optimized by clang
}
And that's it !
Related
In existing code I saw #pragma once be used after header #includes
//Some_Header.h
#include "header1.h"
#include "header2.h"
#pragma once
//implementations
Instead of
//Some_Header.h
#pragma once
#include "header1.h"
#include "header2.h"
//implementations
I thought it always needed to be like the second example, does it matter where your #pragma once is defined or does the preprocessor pick it up anywhere in your file?
Edit
I know #pragma once is not part of the standard and include guards are but that is not my question.
#pragma once should be placed before any headers are included. Argument of #pragma directive is a subject to macro expansion. So content of included headers can alter the pragma behavior:
// whatever.hpp
...
#define once lol_no
// your_header.hpp
#include "whatever.hpp"
#pragma once // warning C4068: unknown pragma
#pragma once relates only to a file where it is placed. It matters to the compiler whether the file contains this pragma or not, and the position of it is unimportant. Therefore, the line with #pragma once can be placed anywhere in the file except for chunk of code that is excluded from compilation by the conditional preprocessor directives like #if, #ifdef and #ifndef. The excluded chunk of code is not parsed by the compiler and if it contains the preprocessor directive it has no effect.
Despite the fact that #pragma once can be placed at any line that is parsed by the compiler, I strongly recommend to follow the common practice and put #pragma once at the beginning of a header file.
Also, as #user7860670 mentioned, arguments of #pragma directive is a subject to macro expansion for MSVC compiler. But neither gcc nor clang support it:
example for msvc
example for gcc
example for clang
There isn't really a complete answer on this question covering more than the big 3 compilers, so here's my attempt at a more complete answer.
Summary
TL;DR: If you care for portability, put it before any #include or #define statements that may conflict with it (e.g. put it first in the header).
Supported by all major compilers ("the big 3" x86_64 compilers as well as intel, and embedded compilers)
Placement generally does not matter as long as it reaches the preprocessor (e.g. not blocked by an #if-branch)
Different compilers are opinionated on whether it should be first, without documenting what happens if not.
Most compilers already detect include guards and treat it as a pragma once anyway, making the benefit largely just not creating a unique guard name.
Below is a quick, summarized guide:
Compiler
Support
Documentation
Clang
Supported
GNU-compatible. Not documented, but code shows it as normal preprocessing
GCC
Supported
gcc pragmas
MSVC
Supported (1)
MSVC one pragma
Intel (ICC)
Supported (1)
Intel Compiler Reference Manual - Supported Pragmas
Intel (ICL)
Supported (1)
ICL uses MSVC front-end
Intel (ICX)
Supported
ICX is based on Clang
Texas Instruments
Supported (2)
Reference Manual 5.11.23
Texas Instruments (Clang)
Supported
This is a fork of Clang with all major features still in effect
ArmCC
Supported (3)
Compiler Docs for #pragma once
(1) - Supported, but is subject to macro expansion
(2) - Supported, but is documented to be expected at the beginning of the header.
(3) - Supported, but not recommended.
Details
GCC
From the GCC reference manual:
If #pragma once is seen when scanning a header file, that file will never be read again, no matter what. It is a less-portable alternative to using ‘#ifndef’ to guard the contents of header files against multiple inclusions.
(emphasis mine)
Scanning is done at preprocessing time, and so as long as the #pragma statement is visible to the preprocessor (not in an unreachable conditional block from #if), then it will take effect.
GCC's #pragma once is not affected by preprocessor substitution.
Live Example
Clang
Clang's reference manual doesn't actually specify #pragma once so far as I can tell, however Clang is meant to be compatible with most, if not all, GCC builtins and features.
Viewing the source code for Clang's preprocessor phase indicates what you would expect; it handles #pragma once during preprocessing (source)
void Preprocessor::HandlePragmaOnce(Token &OnceTok) {
...
// Mark the file as a once-only file now.
HeaderInfo.MarkFileIncludeOnce(getCurrentFileLexer()->getFileEntry());
}
Like GCC, the placement of the #pragma once does not matter if it's first, and is not affected by preprocessor substitution.
Live Example
MSVC
MSVC's documentation on #pragma once does not indicate where it belongs, just that it should be in a source (and has an example of it at the top).
As mentioned by others, when using #pragma once in MSVC, it is subject to preprocessor-expansion.
With Substitution
Live Example
Without Substitution
Live Example
Intel (CL-Based)
When using the Intel compiler on Windows, the compiler uses an MSVC Compatibility mode (ICL). Although it's not documented in the Supported Pragma, though it does appear to be supported.
The placement does not appear to matter as well so long as the preprocessor reaches it.
ICL's #pragma once is subject to the preprocessor-expansion issue that MSVC experiences.
Note: icl is not supported on compiler-explorer, so no example is available.
Intel (GNU-Based)
When using the Intel compiler on Linux or older macOS versions (ICC), the compiler uses a GNU compatibility mode.
Like above, it's not explicitly listed as a Supported Pragma, though it does appear to be supported in practice. The placement does not appear to matter as well so long as the preprocessor reaches it.
ICC's #pragma once is subject to the preprocessor-expansion issue that MSVC experiences.
With Substitution
Live Example
Without Substitution
Live Example
Intel (Clang-Based)
The newer Intel ICX NextGen compiler is based off of Clang / LLVM technology. Behaviourally, this matches what Clang does.
Unlike other Intel compilers, but like Clang, thisdoes not suffer from the preprocessor-expansion issue.
Live Example
Arm (armcc)
The armcc compiler advises against #pragma once, but does also provide an example of it existing after #define statements as an optional feature to work with inclusion guards.
Given the example, placement should likely not be an issue.
It's unclear where this will experience any preprocessor expansion.
Note: armcc is not supported on compiler-explorer, so no example is available.
Texas Instruments (TI ArmCL)
As mentioned in the reference manual, section 5.11.23:
This pragma should be used at the beginning of a header file that should only be included once. For example:
// hdr.h
#pragma once
#warn You will only see this message one time
struct foo
{
int member;
};
(Emphasis mine)
I haven't tested what happens if it's moved anywhere lower than the comment header, but the compiler only officially supports it at the beginning of the file.
I would suspect this should not matter, but cannot confirm.
It's unclear where this will experience any preprocessor expansion.
Note: tiarmcl (and other similar ti compilers) are not supported on compiler-explorer, so no example is available.
Texas Instruments (tiarmclang)
This is a fork of clang, so it behaves the same as clang does.
The #pragma once in this implementation can be effectively placed anywhere the preprocessor reaches, and does not deal with preprocessor substitution.
Note: tiarmclang is not supported on compiler-explorer, so no example is available.
I'm working on a compiler for a small language. Inside the compiler, I'm using the LLVM C++ API to generate llvm code, similar to the LLVM Kaleidoscope tutorial. So I'm using TheModule, TheContext, BasicBlocks,
and calls to Builder.Create...().
I can currently generate valid llvm code for arithmetic, control flow, and methods. However, I would also like my small language to support very simple OpenMP pragmas. For example,
#pragma omp parallel
{
print "Hello World"
}
I've tried writing a similar program in C++,
#include <iostream>
int main() {
#pragma omp parallel
{
std::cout << "Hi";
}
}
and generating llvm using clang++ -S -emit-llvm file.cpp -fopenmp. Along with the rest of the code, this generates the following lines which seem to implement the OpenMP functionality:
declare void #__kmpc_fork_call(%ident_t*, i32, void (i32*, i32*, ...)*, ...)
define internal void #.omp_outlined.(...)
From researching these statements, I found the Clang OpenMP API that contains calls like
OMPParallelDirective * OMPParallelDirective::Create(...)
I'm guessing this is what the Clang compiler uses to generate the statements above. However, it seems to be separate from the LLVM C++ API, as it doesn't reference TheContext, TheModule, etc...
So my question: Is there any way to leverage the Clang OpenMP API calls with my LLVM C++ API calls to generate the kmpc__fork_call and #.omp_outlined IR needed for parallel computation?
I did try decompiling the llvm generated from the C++ code back into LLVM C++ API code using llc -march=cpp file.bc ... but was unsuccessful.
The API you found operate on clang AST and are hardly usable outside clang. In fact, there are no OpenMP constructs at the LLVM IR level - everything is already lowered down to runtime calls, etc.
So, you'd really need to implement codegeneration for OpenMP by yourself emitting runtime calls as necessary (and per your language semantics).
I have some headers that fight with C++11 but I'd like to use features like initializer lists in my code. My research says that you can enable newer features at the command-line only, like:
g++ -std=c++11
What I'd really like is to put in my code:
#pragma CXX11_OFF
#include <old.hpp>
#pragma CXX11_ON
vector<int> v {1,2,3};
I haven't been able to find such a pragma. Does it exist?
No. Such a pragma does not exist.
You can find a list of all pragmas GCC supports in § 6.61 of the manual.
#pragma GCC diagnostic warning "-std=c++11"
This line adds a cpp 11 flag to compiler.
I don't know if there exists such a pragma. But turning C++11 on and off during the same compilation unit looks akward to me. Perhaps you should place the "old" part inside a dll and import it then in the "new" part. Then you can only compile the new part with c++11 enabled
Since such a pragma does not exist, my suggestion is to rely on make/cmake for picking the "right" files based on a compilation flag.
I have C++ code with OpenMP pragmas inside. I want to test this code both for multithread mode (with OpenMP) and in single thread mode (no OpenMP).
For now, to switch between modes I need to comment #pragma omp (or at least parallel).
What is the cleanest, or default, way to enable / disable OpenMP?
If you do not compile with -fopenmp option, you won't get the parallel code. You can do it with an appropiate define and makefile that generates all codes.
The OpenMP documentation says (only an example):
#ifdef _OPENMP
#include <omp.h>
#else
#define omp_get_thread_num() 0
#endif
See http://www.openmp.org/mp-documents/spec30.pdf (conditional compilation).
Look into the compiler manual for the switch that disables OpenMP. For GCC, OpenMP is disabled by default and enabled with the -fopenmp option.
Another option would be to run the code with the OMP_NUM_THREADS environment variable set to 1, though that is not exactly the same as compiling without OpenMP in the first place.
The way such things are usually handled (the general case) is with #defines and #ifdef:
In your header file:
#ifndef SINGLETHREADED
#pragma omp
#endif
When you compile, add -DSINGLETHREADED to disable OpenMP:
cc -DSINGLETHREADED <other flags go here> code.c
Using GCC, is it possible to specify a set of functions that are exempt from -Wframe-larger-than? (For example, main.)
GCC supplies you with pragmas for this purpose:
http://gcc.gnu.org/onlinedocs/gcc-4.5.2/gcc/Diagnostic-Pragmas.html#Diagnostic-Pragmas
Currently it won't do exactly what you want, since it seems to do it on a file by file basis, but in the next version of gcc (4.6), it appears as though it is context aware:
http://gcc.gnu.org/onlinedocs/gcc/Diagnostic-Pragmas.html#Diagnostic-Pragmas
This is a bit old, but I came across it looking for the same answer, so I thought I'd post my solution (found by trial and error):
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wframe-larger-than="
void func()
{ int x[2000] = {}; printf("%d",x[1]); }
#pragma GCC diagnostic pop
seems to work. For some reason, trying to use diagnostic warning did not work. It does not seem possible to change the stack size that generates the warning. Also, you need the = at the end. Maybe the next guy will find this and save themselves some time :). This is 4.6.2 (on an ARM cross compiler).
John
You can use the GCC diagnostic pragma.