Thread: assembly is slower !

I write a short _asm, but it is very slow:

//this is C++

Code:

double *pdb0 = pMatrix + ic;
double *pdb1 = pMatrix + i;
double dbT = ...;

//this piece of c code is replaced by _asm as bellow

Code:

for(j=0; j<iOrder; j++)   
{   
	*pdb0 = *pdb0 + dbT * *pdb1;   
	pdb0 += iOrder;
	pdb1 += iOrder;
}

Code:

_asm
{
	finit;
	mov		edi,	pdb0;		edi = pdb0
	mov		esi,	pdb1;		esi = pdb1

	mov		eax,	iOrder;		eax = const iOrder * 8
	shl		eax,	3;

	mov		ecx,	iOrder;
	fld		dbT;

Loop_Start:
	fld		qword ptr	[esi];	load *pd1
	fmul	        st(0),		st(1);

	fld		qword ptr	[edi];	load *pd0
	fadd;

	fstp	        qword ptr	[edi];	set to *pd0

	add		edi,	eax;		add iOrder*8;
	add		esi,	eax;		add iOrder*8;

	dec		ecx;
	jnz     Loop_Start;
}

above _asm works fine, but it is slower than C code.
why and how to improve it?

Thanks.

why is it so slow?

Originally Posted by mark9876

why is it so slow?

Because the compiler is better at coding in assembly.