int signal_energy_nodc(int *input,unsigned int length) {
int i;
int temp,temp2;
register __m64 mm0,mm1,mm2,mm3;
__m64 *in = (__m64 *)input;
#ifdef MAIN
short *printb;
#endif
mm0 = _m_pxor(mm0,mm0);
mm3 = _m_pxor(mm3,mm3);
for (i=0;i<length>>1;i++) {
mm1 = in[i];
mm2 = mm1;
mm1 = _m_pmaddwd(mm1,mm1);// SIMD complex multiplication
mm1 = _m_psradi(mm1,shift);
mm0 = _m_paddd(mm0,mm1);
// temp2 = mm0;
// printf("%d %d\n",((int *)&in[i])[0],((int *)&in[i])[1]);
// printb = (short *)&mm2;
// printf("mm2 %d : %d %d %d %d\n",i,printb[0],printb[1],printb[2],printb[3]);
}
/*
#ifdef MAIN
printb = (short *)&mm3;
printf("%d %d %d %d\n",printb[0],printb[1],printb[2],printb[3]);
#endif
*/
mm1 = mm0;
mm0 = _m_psrlqi(mm0,32);
mm0 = _m_paddd(mm0,mm1);
temp = _m_to_int(mm0);
temp/=length;
temp<<=shift; // this is the average of x^2
#ifdef MAIN
printf("E x^2 = %d\n",temp);
#endif
_mm_empty();
_m_empty();
return((temp>0)?temp:1);
}
bool BTri::isAbove3DNow(v2sf v0XY, v2sf v0Z1) const {
for (int i = 0; i < 3; i++){
v2sf planeXY = ((v2sf *) &edgePlanes[i])[0];
v2sf planeZD = ((v2sf *) &edgePlanes[i])[1];
v2sf dotXY = pfmul(planeXY, v0XY);
v2sf dotZD = pfmul(planeZD, v0Z1);
v2sf dot = pfacc(dotXY, dotZD);
dot = pfacc(dot, dot);
int d = _m_to_int(dot);
if (d < 0) return false;
}
return true;
}
bool BSP::isInOpenSpace3DNow(const vec3 &pos) const {
if (top != NULL){
SSENode *node = sseTop;
femms();
v2sf posXY, posZ1;
posXY = *(v2sf *) &pos.x;
posZ1.m64_f32[0] = pos.z;
posZ1.m64_f32[1] = 1.0f;
while (true){
v2sf planeXY = ((v2sf *) &node->tri.plane)[0];
v2sf planeZD = ((v2sf *) &node->tri.plane)[1];
v2sf dotXY = pfmul(planeXY, posXY);
v2sf dotZD = pfmul(planeZD, posZ1);
v2sf dot = pfacc(dotXY, dotZD);
dot = pfacc(dot, dot);
int d = _m_to_int(dot);
if (d > 0){
if (node->front){
node = node->front;
} else {
femms();
return true;
}
} else {
if (node->back){
node = node->back;
} else {
femms();
return false;
}
}
}
}
return false;
}
bool BNode::intersects3DNow(const vec4 &v0, const vec4 &v1, const vec4 &dir) const {
v2sf planeXY = ((v2sf *) &tri.plane)[0];
v2sf planeZD = ((v2sf *) &tri.plane)[1];
v2sf v0XY = ((v2sf *) &v0)[0];
v2sf v0Z1 = ((v2sf *) &v0)[1];
v2sf dotXY = pfmul(planeXY, v0XY);
v2sf dotZD = pfmul(planeZD, v0Z1);
v2sf dotD = pfacc(dotXY, dotZD);
dotD = pfacc(dotD, dotD);
int d = _m_to_int(dotD);
if (d > 0){
if (front != NULL && front->intersects3DNow(v0, v1, dir)) return true;
v2sf dotXY = pfmul(planeXY, ((v2sf *) &v1)[0]);
v2sf dotZD = pfmul(planeZD, ((v2sf *) &v1)[1]);
v2sf dot = pfacc(dotXY, dotZD);
dot = pfacc(dot, dot);
int d = _m_to_int(dot);
if (d < 0){
v2sf dirXY = ((v2sf *) &dir)[0];
v2sf dirZ0 = ((v2sf *) &dir)[1];
v2sf dotXY = pfmul(planeXY, dirXY);
v2sf dotZ0 = pfmul(planeZD, dirZ0);
v2sf dot = pfacc(dotXY, dotZ0);
dot = pfacc(dot, dot);
dot = pfrcp(dot);
dot = pfmul(dot, dotD);
dirXY = pfmul(dirXY, dot);
dirZ0 = pfmul(dirZ0, dot);
v0XY = pfsub(v0XY, dirXY);
v0Z1 = pfsub(v0Z1, dirZ0);
if (tri.isAbove3DNow(v0XY, v0Z1)){
return true;
}
if (back != NULL && back->intersects3DNow(v0, v1, dir)) return true;
}
} else {
if (back != NULL && back->intersects3DNow(v0, v1, dir)) return true;
v2sf dotXY = pfmul(planeXY, ((v2sf *) &v1)[0]);
v2sf dotZD = pfmul(planeZD, ((v2sf *) &v1)[1]);
v2sf dot = pfacc(dotXY, dotZD);
dot = pfacc(dot, dot);
int d = _m_to_int(dot);
if (d > 0){
v2sf dirXY = ((v2sf *) &dir)[0];
v2sf dirZ0 = ((v2sf *) &dir)[1];
v2sf dotXY = pfmul(planeXY, dirXY);
v2sf dotZ0 = pfmul(planeZD, dirZ0);
v2sf dot = pfacc(dotXY, dotZ0);
dot = pfacc(dot, dot);
dot = pfrcp(dot);
dot = pfmul(dot, dotD);
dirXY = pfmul(dirXY, dot);
dirZ0 = pfmul(dirZ0, dot);
v0XY = pfsub(v0XY, dirXY);
v0Z1 = pfsub(v0Z1, dirZ0);
if (tri.isAbove3DNow(v0XY, v0Z1)){
return true;
}
if (front != NULL && front->intersects3DNow(v0, v1, dir)) return true;
}
}
return false;
}
int test91(__m64 a) {
// CHECK: movd
return _m_to_int(a);
}
int main(int, char**)
{
__m64 a = _mm_setzero_si64();
a = _mm_shuffle_pi16(a, 0);
return _m_to_int(a);
}
int test_m_to_int(__m64 a) {
// CHECK-LABEL: test_m_to_int
// CHECK: extractelement <2 x i32>
return _m_to_int(a);
}
int signal_energy(int *input,unsigned int length) {
int i;
int temp,temp2;
register __m64 mm0,mm1,mm2,mm3;
__m64 *in = (__m64 *)input;
#ifdef MAIN
short *printb;
#endif
mm0 = _m_pxor(mm0,mm0);
mm3 = _m_pxor(mm3,mm3);
for (i=0;i<length>>1;i++) {
mm1 = in[i];
mm2 = mm1;
mm1 = _m_pmaddwd(mm1,mm1);
mm1 = _m_psradi(mm1,shift);// shift any 32 bits blocs of the word by the value shift
mm0 = _m_paddd(mm0,mm1);// add the two 64 bits words 4 bytes by 4 bytes
// temp2 = mm0;
// printf("%d %d\n",((int *)&temp2)[0],((int *)&temp2)[1]);
// printb = (short *)&mm2;
// printf("mm2 %d : %d %d %d %d\n",i,printb[0],printb[1],printb[2],printb[3]);
mm2 = _m_psrawi(mm2,shift_DC);
mm3 = _m_paddw(mm3,mm2);// add the two 64 bits words 2 bytes by 2 bytes
// printb = (short *)&mm3;
// printf("mm3 %d : %d %d %d %d\n",i,printb[0],printb[1],printb[2],printb[3]);
}
/*
#ifdef MAIN
printb = (short *)&mm3;
printf("%d %d %d %d\n",printb[0],printb[1],printb[2],printb[3]);
#endif
*/
mm1 = mm0;
mm0 = _m_psrlqi(mm0,32);
mm0 = _m_paddd(mm0,mm1);
temp = _m_to_int(mm0);
temp/=length;
temp<<=shift; // this is the average of x^2
// now remove the DC component
mm2 = _m_psrlqi(mm3,32);
mm2 = _m_paddw(mm2,mm3);
mm2 = _m_pmaddwd(mm2,mm2);
temp2 = _m_to_int(mm2);
temp2/=(length*length);
temp2<<=(2*shift_DC);
#ifdef MAIN
printf("E x^2 = %d\n",temp);
#endif
temp -= temp2;
#ifdef MAIN
printf("(E x)^2=%d\n",temp2);
#endif
_mm_empty();
_m_empty();
return((temp>0)?temp:1);
}