AMD 250 manual 222, Code below Puts the Floating Point Sign Mask

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

The following functions use SSE and 3DNow! instructions to illustrate complex multiplication of streams of complex numbers x[] and y[] stored in a product stream prod[]. For these examples, assume that the sizes of x[] and y[] are even multiples of four.

Examples

Listing 25. Complex Multiplication of Streams of Complex Numbers (SSE)

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

;TO ASSEMBLE INTO *.obj DO THE FOLLOWING:

;ml.exe -coff -c cmplx_multiply_sse.asm

;

.586

.K3D

.XMM

_TEXT SEGMENT

PUBLIC _cmplx_multiply_sse

_cmplx_multiply_sse PROC NEAR ;==============================================================================

;INSTRUCTIONS BELOW SAVE THE REGISTER STATE WITH WHICH THIS ROUTINE WAS ENTERED

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

;WHILE THE REGISTERS BELOW MUST BE PRESERVED IF THE USER IS CHANGING THEM

push ebp

mov ebp, esp ;==============================================================================

;parameters passed into routine:

;[ebp+8] = ->x

;[ebp+12] = ->y

;[ebp+16] = num_cmplx_elem

;[ebp+20] = ->prod ;==============================================================================

push ebx		;	preserve contents in ebx,esi, and edi on stack
push	esi	;
push	edi	;

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

;THE CODE BELOW PUTS THE FLOATING POINT SIGN MASK

;[800000000000000800000000000000h]

;TO FLIP THE SIGN OF PACKED SINGLE PRECISION NUMBERS BY USING XORPS ;==============================================================================

mov eax, esp	; Copy stack		pointer into EAX.
mov ebx, 16
sub esp, 32	; Subtract 32 bytes from stack pointer.
and eax, 15	; AND old stack pointer address with 15 to
	;	determine	# of bytes the address is past a
	;	16-byte-aligned address.
sub ebx, eax	; EBX = # of		bytes above ESP to next
	;	16-byte-aligned address
mov edi, 0h	; EDI = 00000000h
mov esi, 80000000h	; EBX = 80000000h
shr ebx, 2	; EBX = # of		DWORDs past 16-byte-aligned address

222 Optimizing with SIMD Instructions Chapter 9