Hello,

I added a assembler file named my_strlen.S to my package, the contents of this file are below:

	.global _my_strlen			# Export the symbol _my_strlen
	.text						# code section

_my_strlen:
	
	#function prologue
	pushq %rbp
	movq %rsp, %rbp
	movq %rdi, %rsi 			# copy the string pointer to %rsi
	xorq %rax, %rax				# zero %rax to use as a counter

strlen_loop:
	cmpb $0, (%rsi)				# compare byte at %rsi with 0 (null terminator)
	je strlen_end				# if zero, end of string reached
	incq %rax					# increment counter
	incq %rsi					# move to next character
	jmp strlen_loop				# repeat loop

strlen_end:
	
	#function epilogue
	popq %rbp
	ret

Since this is AT&T assembler code, it compiled without any problem to a object file using right_click, build and compile option in TheIDE GUI. So far so good.

I then added a C++ file, which has the following content:

#include <iostream>

//declare the assembler function
extern "C" size_t my_strlen(const char* str);

int main(int argc, const char *argv[])
{
	const char* message = "Hello from Assembler";
	size_t length = my_strlen(message);
	std::cout << "Message: " << message << std::endl;
	std::cout << "Length: " << length << std::endl;
	return 0;
}

which can also be compiled using the same method as stated before.

Now my question: how do I link those two object files in the GUI? I suppose I have to go to Project menu item, then Custom build steps, but what do I enter in all these fields?

Thanks.

Hello Mirek,

Thanks for mentioning the verbose console check. The .S and .CPP files were correctly compiled in .O files, but the linking failed on the global export. Most stupid error: i forgot the underscore for my variable .

#include <iostream>

//declare the assembler function
extern "C" size_t _my_strlen(const char* str);

int main(int argc, const char *argv[])
{
	const char* message = "Hello from Assembler";
	size_t length = _my_strlen(message);
	std::cout << "Message: " << message << std::endl;
	std::cout << "Length: " << length << std::endl;
	return 0;
}

When you compile this all links up fine and runs perfectly!

So for those of you that want to include native assembler functions for SS2, AVX SIMD instructions, you can give it a try with this method
As Mirek mentioned, bit exotic, but I thought why not...

Just to be sure, are you aware there are intrinsics for these?

#include <iostream>
#include <immintrin.h>

int main(int argc, const char *argv[])
{
	alignas(32) float a[8] = {1.0f, 2.0f, 3.0f, 4.0f, 1.5f, 2.5f, 3.5f, 4.5f};
	alignas(32) float b[8] = {5.0f, 6.0f, 7.0f, 8.0f, 5.5f, 6.5f, 7.5f, 8.5f};
	
	//load data in AVX registers
	__m256 vec_a = _mm256_load_ps(a);
	__m256 vec_b = _mm256_load_ps(b);
	
	//multiply elements
	__m256 vec_mul = _mm256_mul_ps(vec_a, vec_b);
	
	//horizontal add to compute the sum of all elements
	__m256 temp = _mm256_hadd_ps(vec_mul, vec_mul);
	temp = _mm256_hadd_ps(temp, temp);
	
	//extract 128 lower bits and sum
	__m128 low = _mm256_castps256_ps128 (temp);
	__m128 high = _mm256_extractf128_ps (temp,1);
	__m128 sum = _mm_add_ps (low, high);
	
	//extract the final result
	float result = _mm_cvtss_f32 (sum);
	
	std::cout << "Dot product: " << result << std::endl;
	
	return 0;
}