compiler-optimization

问题 What is the best way to implement multiple versions of the same function that uses a specific CPU instructions if available (tested at run time), or falls back to a slower implementation if not? For example, x86 BMI2 provides a very useful PDEP instruction. How would I write a C code such that it tests BMI2 availability of the executing CPU on startup, and uses one of the two implementations -- one that uses _pdep_u64 call (available with -mbmi2 ), and another that does bit manipulation "by

问题 I read on a few sites that MSVC can actually emit say AVX instructions, when SSE2 architecture is used and detect the AVX support runtime. Is it true? I tested various loops that would definitely benefit from AVX/AVX2 support, but when run in debugger I couldn't really find any AVX instructions. When /arch:AVX is used, then it emits AVX instructions, but it of course crashes on CPUs that doesn't support it (tested), so no runtime detection either. I could use AVX intrinsics though and it

问题 I have some code like the following code block (I am not allowed to post the original code) inside a .cpp file that I think is being compiled by clang++ ( Ubuntu clang version 3.5.2-3ubuntu1 (tags/RELEASE_352/final) (based on LLVM 3.5.2) ). It looks like C code, because we are using GoogleTest for testing our C code. Anyways: size_t const SHIFT = 4; uint8_t var, var2; /* Omitted: Code that sets var to, say 00011000 (base 2) */ var2 = var; var = var << SHIFT >> SHIFT; // [1] result is 00011000

问题 I have some code like the following code block (I am not allowed to post the original code) inside a .cpp file that I think is being compiled by clang++ ( Ubuntu clang version 3.5.2-3ubuntu1 (tags/RELEASE_352/final) (based on LLVM 3.5.2) ). It looks like C code, because we are using GoogleTest for testing our C code. Anyways: size_t const SHIFT = 4; uint8_t var, var2; /* Omitted: Code that sets var to, say 00011000 (base 2) */ var2 = var; var = var << SHIFT >> SHIFT; // [1] result is 00011000

问题 Out of curiosity, I was wondering if there is any way that gcc does not optimize any function calls? In the generated assembly code, the printf function is replaced by putchar. This happens even with the default -O0 minimal optimization flag. #include <stdio.h> int main(void) { printf("a"); return 0; } (Godbolt showing GCC9 doing it, clang8 keeping it unchanged.) 回答1: Use -fno-builtin to disable all replacement and inlining of standard C functions with equivalents. Or use -fno-builtin

问题 Out of curiosity, I was wondering if there is any way that gcc does not optimize any function calls? In the generated assembly code, the printf function is replaced by putchar. This happens even with the default -O0 minimal optimization flag. #include <stdio.h> int main(void) { printf("a"); return 0; } (Godbolt showing GCC9 doing it, clang8 keeping it unchanged.) 回答1: Use -fno-builtin to disable all replacement and inlining of standard C functions with equivalents. Or use -fno-builtin

来源： https://stackoverflow.com/questions/53450282/data-alignment-to-enable-vectorization-efficient-cache-access

来源： https://stackoverflow.com/questions/53450282/data-alignment-to-enable-vectorization-efficient-cache-access

问题 When I want to do debugging of C or C++ programs, I've been taught to use -O0 to turn optimization OFF, and -ggdb to insert symbols into the executable which are optimized for using the GNU gdb debugger, which I use (or, you can use -glldb for LLVM/clang's lldb debugger, or just -g for general debugging symbols, but that won't be as good as -ggdb apparently...). However, I recently stumbled upon someone saying to use -Og (instead of -O0 ), and it caught me off-guard. Sure enough though, it's

问题 When I want to do debugging of C or C++ programs, I've been taught to use -O0 to turn optimization OFF, and -ggdb to insert symbols into the executable which are optimized for using the GNU gdb debugger, which I use (or, you can use -glldb for LLVM/clang's lldb debugger, or just -g for general debugging symbols, but that won't be as good as -ggdb apparently...). However, I recently stumbled upon someone saying to use -Og (instead of -O0 ), and it caught me off-guard. Sure enough though, it's