AMD 250 manual 224

Software Optimization Guide for AMD64 Processors 25112 Rev. 3.06 September 2005

mulps xmm2, xmm6		; xmm2=[x1i*y1r,-x1i*y1i,x0i*y0r,-x0i*y0i]
mulps xmm3, xmm7		; xmm3=[x3i*y3r,-x3i*y3i,x2i*y2r,-x2i*y2i]
addps xmm0, xmm2		; xmm0=[x1r*y1i+x1i*y1r,x1r*y1r-x1i*y1i,
		;	x0r*y0i+x0i*y0r,x0r*y0r-x0i*y0i]
addps xmm1, xmm3		; xmm1=[x3r*y3i+x3i*y3r,x3r*y3r-x3i*y3i,
		;	x2r*y2i+x2i*y2r,x2r*y2r-x2i*y2i]
movntps	[eax+ecx*8], xmm0	; Stream XMM0 and XMM1 to representative
movntps	[eax+ecx*8+16], xmm1	;	memory address of prod[].
add	ecx, 4	; ECX = ECX + 4
jnz	eight_cmplx_prod_loop
sfence	; Finish all memory writes.

;==============================================================================

;INSTRUCTIONS BELOW RESTORE THE REGISTER STATE WITH WHICH THIS ROUTINE WAS

;ENTERED

;REGISTERS EAX, ECX, AND EDX ARE CONSIDERED VOLATILE AND ASSUMED TO BE CHANGED

;WHILE THE REGISTERS BELOW MUST BE PRESERVED IF THE USER IS CHANGING THEM add esp, 32

pop edi pop esi

pop ebx

mov esp, ebp pop ebp

;============================================================================== ret

_cmplx_multiply_sse ENDP _TEXT ENDS

END

Listing 26. Complex Multiplication of Streams of Complex Numbers (3DNow!™ Technology)

;cmplx_multiply_3dnow(float *x, float *y, int num_cmplx_elem, float *prod);

;TO ASSEMBLE INTO *.obj DO THE FOLLOWING:

;ml.exe -coff -c cmplx_multiply_3dnow.asm

;

.586

.K3D

.XMM

_TEXT SEGMENT

PUBLIC _cmplx_multiply_3dnow

;cmplx_multiply_3dnow(float *x, float *y, int num_cmplx_elem, float *prod);

;

;TO ASSEMBLE INTO *.obj DO THE FOLLOWING:

;ml.exe -coff -c cmplx_multiply_3dnow.asm

.586

.K3D

.XMM

_TEXT SEGMENT

PUBLIC _cmplx_multiply_3dnow

224 Optimizing with SIMD Instructions Chapter 9