I've been looking around and I can only seem to find the option to omit the frame pointer.
Here are the compile flags that I'm using right now.
-Wall -Wextra -Wconversion -Wctor-dtor-privacy -Wnon-virtual-dtor -Wold-style-cast -g
Can someone tell me how to enable the frame pointer?
Note I'm using g++ version 4.6.6 on RHEL 6.2
Any optimization level (-O, -O2, -O3), enable -fomit-frame-pointer, but you can revert it with -fno-omit-frame-pointer. You can find more details here: http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
Related
I was trying some vectorisation after upgrading g++ from version 4.8.5 to 5.4.1. With this flags:
g++ particles-v3.cpp -o v3 -O3 -msse4.2 -mfpmath=sse -ftree-vectorizer-verbose=5 -ffast-math -m32 -march=native -std=c++11
While the same command gives over 4000 lines of detailed information about the vectorization with g++-4.8, with g++-5.4 it does not say anything.
Is there some major change in g++-5 that makes the -ftree-vectorizer-verbose=X unusable, or is there simply somethin wrong in the line? How to change it so that it works?
EDIT:
found out that using -fopt-info-vec-all gives exacty the info I wanted. Thus question solved.
I have a rather complex multithreaded code, which I compile using gcc 4.8.1. When compiling with
g++ -c file.cc -march=native -mfpmath=sse -mpreferred-stack-boundary=4
--param inline-unit-growth=50 -ggdb3 -Wall -Wextra -Winit-self
-O2 -fPIC -funroll-loops -fforce-addr -rdynamic
the code produced crashes with segfault (which I was unable to debug, but the address of a struct all of a sudden differs from what it was when constructed, in particular, is no longer aligned to 32bytes as required by the code but only to 8bytes).
When compiling with -O1 instead, the code works fine. I then added all the optimisation flags that make the difference between -O1 and -O2. (To this end, I created two files O1-opts and O2-opts via
g++ -march=native -mfpmath=sse -mpreferred-stack-boundary=4
--param inline-unit-growth=50 -ggdb3 -Wall -Wextra -Winit-self
-O1 -fPIC -funroll-loops -fforce-addr -rdynamic
-Q --help=optimizers > O1-opts
g++ -march=native -mfpmath=sse -mpreferred-stack-boundary=4
--param inline-unit-growth=50 -ggdb3 -Wall -Wextra -Winit-self
-O2 -fPIC -funroll-loops -fforce-addr -rdynamic
-Q --help=optimizers > O2-opts
when diff O1-opts O2-opts provides the option differences). When adding all the option differences to -O1, the code generated still does not crash. This puzzles me. So my question is: shouldn't this give exactly the same result as with -O2? (and also: what is the likely cause of my problem?)
The point is that the -O2 option not only sets different flags, but also enables additional optimizations in contrast to -O1.
The FAQ section of the GCC Wiki has an appropriate entry for this.
I installed Intel Compiler composer_xe_2013_sp1.3.174 on Linux. I am confused about the icc warnings. Feed icc with a simple program main.c as below:
int main(int argc, char **argv) {
int a = 1;
unsigned int b = -22;
if (b = a) {
}
}
I compiled the file with the command: icc -Wall main.c. Surprisingly, the command works silently without any warnings. Do I have to turn on the warnings switch on icc? Thanks
The Intel compiler doesn't really have good presets for warnings the way that gcc does (at least on Linux). The main warning option is -wn where n can be 0 to 5. The default is 1, and 4 & 5 have no meaning on Linux. It also supports some gcc options like -Wall and -Wextra. However:
-Wall is very minimalistic compared to gcc, as you found
-w2 and -w3 enable some useful diagnostics, but also a lot of spam remarks
-diag-disable:remark removes that spam but also a lot of useful diagnostics
In the end -w3 -diag-disable:remark is the best preset that icc has but it is still more minimalistic than gcc -Wall. Instead you need to either start with a minimal set of warnings and build up your own, or start with maximum diagnostics and disable any that get annoying using -wd.
Build Up
The main downside to the first approach is that Intel doesn't really document most of its warnings, so it is hard to know what is available to enable. However, it does support many of the GCC command line flags, so the GCC documentation is a good place to start. For example, here are settings that are relatively close to g++ -Wall, which is convenient if you want to develop with one and have a good chance of a clean build with the other:
icpc -Wall -Warray-bounds -Wchar-subscripts -Wcomment -Wenum-compare -Wformat -Wuninitialized -Wmaybe-uninitialized -Wmain -Wnarrowing -Wnonnull -Wparentheses -Wpointer-sign -Wreorder -Wreturn-type -Wsign-compare -Wsequence-point -Wtrigraphs -Wunused-function -Wunused-but-set-variable -Wunused-variable -Wwrite-strings
This isn't an exact match for gcc -Wall. There are differences between GCC's and ICC's implementation of the above warnings. I was also unable to find ICC options to match these GCC warnings:
-Wformat-contains-nul
-Wunused-label
-Wstrict-overflow
-Wvolatile-register-var
And I intentionally left these out, because the ICC version was much more spammy than GCC:
-Wstrict-aliasing So broad that any use of polymophism will cause this warning
-Wswitch Requires a default even if you have cases for all enumeration values
Trim Down
If GCC parity isn't a concern then the easiest way to learn what warnings ICC has is to enable them all, and then decide whether you like them or not. If you don't like a warning, you can disable it using it's diagnostics number, which often has more granularity that GCC's options do.
icpc -w3 -wd1418,2259
Here are some diagnostics that I have seen disabled in the past:
383: value copied to temporary, reference to temporary used
869: parameter "*" was never referenced
981: operands are evaluated in unspecified order
1418: external function definition with no prior declaration
1572: floating-point equality and inequality comparisons are unreliable
2259: non-pointer conversion may loose significant bits
11074: Inlining inhibited by limit max-size (or max-total-size)
11076: To get full report use -qopt-report=4 -qopt-report-phase ipo
161: disable warning for unrecognized pragmas
But I encourage you to start with them all on and pare down just the ones that are problematic for your code base.
Generally speaking, the best compilation options for a small program your developing is
-Wall -Wextra -std=c11 -pedantic
Contrary to the warning switch's name, Wall does not actually activate all warnings; you use both Wall and Wextra to get the majority of the important warnings.
-std switch sets the standard that the code uses; the most recent one is C11 therefore std=c11. Pedantic is a way to signal to the compiler that you want to write a program that doesn't use compiler-specific extensions. Pedantic requires the std switch and will emit warnings for any syntax, ect. that does not conform to the standard specified by std. Finally, if you want errors instead of warnings for usage of compiler-extension, use -pedantic-errors instead.*
(* - pedantic does not warn about the usage of non-standard libraries like conio.h)
Now if you compile the program with Wall Wextra std=c11 pedantic, you should get 1 warnings:
Warning: Line 4 - Suggest Parenthesis around truthy value (9 out of 10, this means you used = instead == in a comparison context).
If you fix that warning, you'll receive another warning:
Warning: Line 4 - Comparison between Signed and Unsigned Integer without a cast.
Adding an explicit cast or changing b to a normal int will solve this warning.
These days I am pretty happy using this target_options in CMake with icpc 2021.6.0 20220226, I collected them from several sources.
Any additions are welcome.
target_compile_options(
${TEST_EXE}
PRIVATE
$<$<COMPILE_LANG_AND_ID:CUDA,NVIDIA>:
...
>
$<$<AND:$<CXX_COMPILER_ID:GNU>,$<NOT:$<CUDA_COMPILER_ID:NVIDIA>>,$<NOT:$<CUDA_COMPILER_ID:Clang>>>:
...
>
$<$<OR:$<CXX_COMPILER_ID:Clang>,$<CXX_COMPILER_ID:AppleClang>,$<CUDA_COMPILER_ID:Clang>>:
...
>
$<$<CXX_COMPILER_ID:Intel>: # also IntelLLVM, XL (ibm), XLClang (ibm)
-Werror
-Wall
-Wextra
-diag-disable=remark
-diag-error:3846
-diag-disable=1011 # disables warning missing return at the end of non-void function
-wd161
-Wabi
-Warray-bounds
-Wcast-qual
-Wchar-subscripts
-Wcomment
-Wdeprecated
-Wenum-compare
-Wextra-tokens
-Wformat
-Wformat=2
-Wformat-security
-Wic-pointer
-Wnarrowing
-Wno-return-type
-Wnon-virtual-dtor
-Wnonnull
-Wmaybe-uninitialized
-Wmain
-Wmissing-declarations
-Wmissing-prototypes
-Wmultichar
-Woverloaded-virtual
-Woverflow
-Wparentheses
-Wpointer-arith
-Wpointer-sign
-Wreorder
-Wreturn-type
-Wsequence-point
-Wshadow
-Wsign-compare
-Wshorten-64-to-32
-Wno-strict-aliasing
-Wstrict-prototypes
-Wtrigraphs
-Wtype-limits
-Wuninitialized
-Wunused
-Wunused-but-set-variable
-Wunused-function
-Wunused-parameter
-Wunused-variable
-Wwrite-strings
>
$<$<OR:$<CXX_COMPILER_ID:PGI>,$<CXX_COMPILER_ID:NVHPC>>:
...
>
$<$<CXX_COMPILER_ID:MSVC>:
...
>
)
I just read a book which recommends enable the highest warning level in GCC. I just check the doc online, and found there are too much parameters. I want to enable the highest warning level, which parameter should I use?
And we use Boost heavily in our project.
Contrary to cl which has 4 levels, gcc only has a set of options that you can turn on or off.
As mentioned by others, the -Wall is the default, which turns on many warnings already. The -pedantic option adds a few more. And -Wextra yet another group...
But to really capture many warnings, you'll have to add many manually.
There is a set I like to use, although someone told me that some of those were contradictory, I find that list rather good for my development work:
-Werror -Wall -Wextra -pedantic -Wcast-align -Wcast-qual -Wctor-dtor-privacy -Wdisabled-optimization -Wformat=2 -Winit-self -Wlogical-op -Wmissing-include-dirs -Wnoexcept -Wold-style-cast -Woverloaded-virtual -Wredundant-decls -Wshadow -Wsign-promo -Wstrict-null-sentinel -Wstrict-overflow=5 -Wundef -Wno-unused -Wno-variadic-macros -Wno-parentheses -fdiagnostics-show-option
Note that I make use of -Werror because otherwise you get warnings and tend to ignore them. With -Werror, no more ignoring anything! Write pristine code and your software is much more likely to work as expected.
I guess you can use the -Wallswitch
I'm working on a large collaborative C++ project that is both developed and run on various flavors of Linux, OS X and Windows. We compile across these platforms with GCC, Visual Studio C++ and the Intel C++ compiler. As more and more people start developing code for the project, we're starting to see weird errors in compilation and runtime that are specific to particular compilers on particular operating systems. An example of this is implicit inclusion of headers that certain OS/compiler pairs seem to find for you, accidentally overloading a function from a base class in a derived class.
My goal is to make compilation on GCC more strict and catch more errors across all platforms so that we don't keep running into these problems. Here's my list of flags that I'm thinking about trying out for GCC that I've found via Google and the GCC man pages:
-Wall
-Wextra
-Winit-self
-Wold-style-cast
-Woverloaded-virtual
-Wuninitialized
-Wmissing-declarations
-Winit-self
-ansi
-pedantic
What are the other flags that people use to make GCC (and less importantly Visual Studio C++ and the Intel C++ Compiler) obey a stricter standard of the C++ language? Be specific about which compiler and version you're talking about, as some of these might not be implemented in all versions of all compilers.
Beside the pedantic-error that everyone else suggested, IMO, it's always good to run lint as part of your compile process.
There are some tools out there:
cpplint (free)
gimple lint
coverity
They will save a lot of your time.
You can make pedantic warnings into errors with -pedantic-errors. This will prevent developers from ignoring it. For that matter you could make all warnings into errors as well with -Werror although that can be counter productive in some cases (maybe not in yours though).
Overall, I think, as far as adhering to a strict standard goes, the -pedantic options are the most helpful.
Copy and paste the below line into your master CMake file. The below line comprises almost most useful compiler flags in order to test yourself stricter.
set(CMAKE_CXX_FLAGS "-O0 -fno-elide-constructors -pedantic-errors -ansi -Wextra -Wall -Winit-self -Wold-style-cast -Woverloaded-virtual -Wuninitialized -Wmissing-declarations -Winit-self -std=c++98")
If you don’t use CMake, just copy the flags in double quotes and send them to your compiler.
-pedantic-errors.
See more on gcc(1).
As well as -pendantic, you should also provide a -std switch. If you need a stricter compile then you should know what standard you are trying to conform to. Typically for current C++ this would be -std=c++98. (-ansi performs a similar function in C++ mode, but -std= is more explicit.)
In a similar situation we gave up and moved to the ACE framework, hiding the difference between platforms.
I wrote a blog post on this topic after researching several options. You also need to handle the cases where you are using other libraries, but they are not following strict compilation. Fortunately, there is an easy way to handle them as well. I have been using this extensively in all my projects.
In short, use the following compiler options to turn on very strict mode (below is what I put in CMakeLists.txt):
set(CMAKE_CXX_FLAGS "-std=c++11 -Wall -Wextra -Wstrict-aliasing -pedantic -fmax-errors=5 -Werror -Wunreachable-code -Wcast-align -Wcast-qual -Wctor-dtor-privacy -Wdisabled-optimization -Wformat=2 -Winit-self -Wlogical-op -Wmissing-include-dirs -Wnoexcept -Wold-style-cast -Woverloaded-virtual -Wredundant-decls -Wshadow -Wsign-promo -Wstrict-null-sentinel -Wstrict-overflow=5 -Wswitch-default -Wundef -Wno-unused -Wno-variadic-macros -Wno-parentheses -fdiagnostics-show-option ${CMAKE_CXX_FLAGS}")
You can read more about how to turn on and off this strict mode for specific portions of code here: http://shitalshah.com/p/how-to-enable-and-use-gcc-strict-mode-compilation/