int main(int argc, const char * argv[]) {
add(1,2);
add(1,2);
PRINTMAX(12, 13);
PRINTMAX(12, 13);
printf("%d\n",MAXOFNUMBER(100, 200));
printf("*******************\n");
double sum = ADD(1.1, 2);//预处理 阶段 就会换成 1+2
printf("sum = %f\n",sum);
printf("%d\n",ADD(1, 2)*ADD(2, 3));//8 //1+2*2+3
printf("%d\n",ADD2(1, 2)*ADD2(2, 3));//(1+2)*(2+3)
printf("%d\n",MUL(3-1, 5-2));//(3-1*5-2)
printf("%d\n",MUL2(3-1, 5-2));//((3-1)*(5-2))
printf("*******************\n");
printf(kPath);
double r = 2.0;
double s = PI*r*r;
double c = 2*PI*r;
printf("s = %f c= %f\n",s,c);
return 0;
}
static WEBP_INLINE void PredictLineGradient(const uint8_t* pinput,
const uint8_t* ppred,
uint8_t* poutput, int stride,
int size) {
int w;
const v16i8 zero = { 0 };
while (size >= 16) {
v16u8 pred0, dst0;
v8i16 a0, a1, b0, b1, c0, c1;
const v16u8 tmp0 = LD_UB(ppred - 1);
const v16u8 tmp1 = LD_UB(ppred - stride);
const v16u8 tmp2 = LD_UB(ppred - stride - 1);
const v16u8 src0 = LD_UB(pinput);
ILVRL_B2_SH(zero, tmp0, a0, a1);
ILVRL_B2_SH(zero, tmp1, b0, b1);
ILVRL_B2_SH(zero, tmp2, c0, c1);
ADD2(a0, b0, a1, b1, a0, a1);
SUB2(a0, c0, a1, c1, a0, a1);
CLIP_SH2_0_255(a0, a1);
pred0 = (v16u8)__msa_pckev_b((v16i8)a1, (v16i8)a0);
dst0 = src0 - pred0;
ST_UB(dst0, poutput);
ppred += 16;
pinput += 16;
poutput += 16;
size -= 16;
}
for (w = 0; w < size; ++w) {
const int pred = ppred[w - 1] + ppred[w - stride] - ppred[w - stride - 1];
poutput[w] = pinput[w] - (pred < 0 ? 0 : pred > 255 ? 255 : pred);
}
}
MMath::MMath(): Methoded("math") {
// ^FUNC(expr)
#define ADDX(name, X) \
add_native_method(#name, Method::CT_STATIC, _##name, X, X)
#define ADD0(name) ADDX(name, 0)
#define ADD1(name) ADDX(name, 1)
#define ADD2(name) ADDX(name, 2)
ADD1(round); ADD1(floor); ADD1(ceiling);
ADD1(trunc); ADD1(frac);
ADD1(abs); ADD1(sign);
ADD1(exp);
ADD1(log); ADD1(log10);
ADD1(sin); ADD1(asin);
ADD1(cos); ADD1(acos);
ADD1(tan); ADD1(atan);
ADD1(degrees); ADD1(radians);
ADD1(sqrt);
ADD1(random);
// ^math:pow(x;y)
ADD2(pow);
// ^math:crypt[password;salt]
ADD2(crypt);
// ^math:md5[string]
ADD1(md5);
// ^math:sha1[string]
ADD1(sha1);
// ^math:digest[method;string|file;options]
add_native_method("digest", Method::CT_STATIC, _digest, 2, 3);
// ^math:crc32[string]
ADD1(crc32);
// ^math:uuid[]
ADD0(uuid);
// ^math:uid64[]
ADD0(uid64);
// ^math:convert[number](base-from;base-to)
add_native_method("convert", Method::CT_STATIC, _convert, 3, 3);
}
void C_RLibrary::initbuiltinape(void)
{
#define ADD(sym) extern C_RBASE * sym(char *desc); _add_dll(0,sym,"Builtin_" #sym, 0)
#define ADD2(sym,name) extern C_RBASE * sym(char *desc); _add_dll(0,sym,name, 0)
#ifdef LASER
ADD(RLASER_Cone);
ADD(RLASER_BeatHold);
ADD(RLASER_Line);
ADD(RLASER_Bren); // not including it for now
ADD(RLASER_Transform);
#else
ADD2(R_ChannelShift,"Channel Shift");
ADD2(R_ColorReduction,"Color Reduction");
ADD2(R_Multiplier,"Multiplier");
ADD2(R_VideoDelay,"Holden04: Video Delay");
ADD2(R_MultiDelay,"Holden05: Multi Delay");
#endif
#undef ADD
#undef ADD2
}
double tan(double x) {
#include "utan.h"
#include "utan.tbl"
int ux,i,n;
double a,da,a2,b,db,c,dc,c1,cc1,c2,cc2,c3,cc3,fi,ffi,gi,pz,s,sy,
t,t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,w,x2,xn,xx2,y,ya,yya,z0,z,zz,z2,zz2;
int p;
number num,v;
mp_no mpa,mpt1,mpt2;
#if 0
mp_no mpy;
#endif
int __branred(double, double *, double *);
int __mpranred(double, mp_no *, int);
/* x=+-INF, x=NaN */
num.d = x;
ux = num.i[HIGH_HALF];
if ((ux&0x7ff00000)==0x7ff00000) return x-x;
w=(x<ZERO) ? -x : x;
/* (I) The case abs(x) <= 1.259e-8 */
if (w<=g1.d) return x;
/* (II) The case 1.259e-8 < abs(x) <= 0.0608 */
if (w<=g2.d) {
/* First stage */
x2 = x*x;
t2 = x*x2*(d3.d+x2*(d5.d+x2*(d7.d+x2*(d9.d+x2*d11.d))));
if ((y=x+(t2-u1.d*t2)) == x+(t2+u1.d*t2)) return y;
/* Second stage */
c1 = x2*(a15.d+x2*(a17.d+x2*(a19.d+x2*(a21.d+x2*(a23.d+x2*(a25.d+
x2*a27.d))))));
EMULV(x,x,x2,xx2,t1,t2,t3,t4,t5)
ADD2(a13.d,aa13.d,c1,zero.d,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(a11.d,aa11.d,c1,cc1,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(a9.d ,aa9.d ,c1,cc1,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(a7.d ,aa7.d ,c1,cc1,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(a5.d ,aa5.d ,c1,cc1,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(a3.d ,aa3.d ,c1,cc1,c2,cc2,t1,t2)
MUL2(x2,xx2,c2,cc2,c1,cc1,t1,t2,t3,t4,t5,t6,t7,t8)
MUL2(x ,zero.d,c1,cc1,c2,cc2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(x ,zero.d,c2,cc2,c1,cc1,t1,t2)
if ((y=c1+(cc1-u2.d*c1)) == c1+(cc1+u2.d*c1)) return y;
return tanMp(x);
}
void
SECTION
__doasin(double x, double dx, double v[]) {
#include "doasin.h"
static const double
d5 = 0.22372159090911789889975459505194491E-01,
d6 = 0.17352764422456822913014975683014622E-01,
d7 = 0.13964843843786693521653681033981614E-01,
d8 = 0.11551791438485242609036067259086589E-01,
d9 = 0.97622386568166960207425666787248914E-02,
d10 = 0.83638737193775788576092749009744976E-02,
d11 = 0.79470250400727425881446981833568758E-02;
double xx,p,pp,u,uu,r,s;
double tc,tcc;
#ifndef DLA_FMS
double hx,tx,hy,ty,tp,tq;
#endif
/* Taylor series for arcsin for Double-Length numbers */
xx = x*x+2.0*x*dx;
p = ((((((d11*xx+d10)*xx+d9)*xx+d8)*xx+d7)*xx+d6)*xx+d5)*xx;
pp = 0;
MUL2(x,dx,x,dx,u,uu,tp,hx,tx,hy,ty,tq,tc,tcc);
ADD2(p,pp,c4.x,cc4.x,p,pp,r,s);
MUL2(p,pp,u,uu,p,pp,tp,hx,tx,hy,ty,tq,tc,tcc);
ADD2(p,pp,c3.x,cc3.x,p,pp,r,s);
MUL2(p,pp,u,uu,p,pp,tp,hx,tx,hy,ty,tq,tc,tcc);
ADD2(p,pp,c2.x,cc2.x,p,pp,r,s);
MUL2(p,pp,u,uu,p,pp,tp,hx,tx,hy,ty,tq,tc,tcc);
ADD2(p,pp,c1.x,cc1.x,p,pp,r,s);
MUL2(p,pp,u,uu,p,pp,tp,hx,tx,hy,ty,tq,tc,tcc);
MUL2(p,pp,x,dx,p,pp,tp,hx,tx,hy,ty,tq,tc,tcc);
ADD2(p,pp,x,dx,p,pp,r,s);
v[0]=p;
v[1]=pp; /* arcsin(x+dx)=v[0]+v[1] */
}
void vpx_fdct8x8_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v4i32 vec_w;
LD_SH8(input, stride, in0, in1, in2, in3, in4, in5, in6, in7);
ADD4(in0, in1, in2, in3, in4, in5, in6, in7, in0, in2, in4, in6);
ADD2(in0, in2, in4, in6, in0, in4);
vec_w = __msa_hadd_s_w(in0, in0);
vec_w += __msa_hadd_s_w(in4, in4);
out[0] = HADD_SW_S32(vec_w);
out[1] = 0;
}
static void shDrawPaintMesh(VGContext *c, SHVector2 *min, SHVector2 *max,
VGPaintMode mode, GLenum texUnit)
{
SHPaint *p;
SHVector2 pmin, pmax;
SHfloat K = 1.0f;
/* Pick the right paint */
if (mode == VG_FILL_PATH) {
p = (c->fillPaint ? c->fillPaint : &c->defaultPaint);
}else if (mode == VG_STROKE_PATH) {
p = (c->strokePaint ? c->strokePaint : &c->defaultPaint);
K = SH_CEIL(c->strokeMiterLimit * c->strokeLineWidth) + 1.0f;
}
/* We want to be sure to cover every pixel of this path so better
take a pixel more than leave some out (multisampling is tricky). */
SET2V(pmin, (*min)); SUB2(pmin, K,K);
SET2V(pmax, (*max)); ADD2(pmax, K,K);
/* Construct appropriate OpenGL primitives so as
to fill the stencil mask with select paint */
switch (p->type) {
case VG_PAINT_TYPE_LINEAR_GRADIENT:
shDrawLinearGradientMesh(p, min, max, mode, texUnit);
break;
case VG_PAINT_TYPE_RADIAL_GRADIENT:
shDrawRadialGradientMesh(p, min, max, mode, texUnit);
break;
case VG_PAINT_TYPE_PATTERN:
if (p->pattern != VG_INVALID_HANDLE) {
shDrawPatternMesh(p, min, max, mode, texUnit);
break;
}/* else behave as a color paint */
case VG_PAINT_TYPE_COLOR:
glColor4fv((GLfloat*)&p->color);
glBegin(GL_QUADS);
glVertex2f(pmin.x, pmin.y);
glVertex2f(pmax.x, pmin.y);
glVertex2f(pmax.x, pmax.y);
glVertex2f(pmin.x, pmax.y);
glEnd();
break;
}
}
static void hevc_addblk_4x4_msa(int16_t *coeffs, uint8_t *dst, int32_t stride)
{
uint32_t dst0, dst1, dst2, dst3;
v8i16 dst_r0, dst_l0, in0, in1;
v4i32 dst_vec = { 0 };
v16u8 zeros = { 0 };
LD_SH2(coeffs, 8, in0, in1);
LW4(dst, stride, dst0, dst1, dst2, dst3);
INSERT_W4_SW(dst0, dst1, dst2, dst3, dst_vec);
ILVRL_B2_SH(zeros, dst_vec, dst_r0, dst_l0);
ADD2(dst_r0, in0, dst_l0, in1, dst_r0, dst_l0);
CLIP_SH2_0_255(dst_r0, dst_l0);
dst_vec = (v4i32) __msa_pckev_b((v16i8) dst_l0, (v16i8) dst_r0);
ST4x4_UB(dst_vec, dst_vec, 0, 1, 2, 3, dst, stride);
}
void
SECTION
__dubsin (double x, double dx, double v[])
{
double r, s, c, cc, d, dd, d2, dd2, e, ee,
sn, ssn, cs, ccs, ds, dss, dc, dcc;
#ifndef DLA_FMS
double p, hx, tx, hy, ty, q;
#endif
mynumber u;
int4 k;
u.x = x + big.x;
k = u.i[LOW_HALF] << 2;
x = x - (u.x - big.x);
d = x + dx;
dd = (x - d) + dx;
/* sin(x+dx)=sin(Xi+t)=sin(Xi)*cos(t) + cos(Xi)sin(t) where t ->0 */
MUL2 (d, dd, d, dd, d2, dd2, p, hx, tx, hy, ty, q, c, cc);
sn = __sincostab.x[k]; /* */
ssn = __sincostab.x[k + 1]; /* sin(Xi) and cos(Xi) */
cs = __sincostab.x[k + 2]; /* */
ccs = __sincostab.x[k + 3]; /* */
/* Taylor series for sin ds=sin(t) */
MUL2 (d2, dd2, s7.x, ss7.x, ds, dss, p, hx, tx, hy, ty, q, c, cc);
ADD2 (ds, dss, s5.x, ss5.x, ds, dss, r, s);
MUL2 (d2, dd2, ds, dss, ds, dss, p, hx, tx, hy, ty, q, c, cc);
ADD2 (ds, dss, s3.x, ss3.x, ds, dss, r, s);
MUL2 (d2, dd2, ds, dss, ds, dss, p, hx, tx, hy, ty, q, c, cc);
MUL2 (d, dd, ds, dss, ds, dss, p, hx, tx, hy, ty, q, c, cc);
ADD2 (ds, dss, d, dd, ds, dss, r, s);
/* Taylor series for cos dc=cos(t) */
MUL2 (d2, dd2, c8.x, cc8.x, dc, dcc, p, hx, tx, hy, ty, q, c, cc);
ADD2 (dc, dcc, c6.x, cc6.x, dc, dcc, r, s);
MUL2 (d2, dd2, dc, dcc, dc, dcc, p, hx, tx, hy, ty, q, c, cc);
ADD2 (dc, dcc, c4.x, cc4.x, dc, dcc, r, s);
MUL2 (d2, dd2, dc, dcc, dc, dcc, p, hx, tx, hy, ty, q, c, cc);
ADD2 (dc, dcc, c2.x, cc2.x, dc, dcc, r, s);
MUL2 (d2, dd2, dc, dcc, dc, dcc, p, hx, tx, hy, ty, q, c, cc);
MUL2 (cs, ccs, ds, dss, e, ee, p, hx, tx, hy, ty, q, c, cc);
MUL2 (dc, dcc, sn, ssn, dc, dcc, p, hx, tx, hy, ty, q, c, cc);
SUB2 (e, ee, dc, dcc, e, ee, r, s);
ADD2 (e, ee, sn, ssn, e, ee, r, s); /* e+ee=sin(x+dx) */
v[0] = e;
v[1] = ee;
}
void __dubsin(Double x, Double dx, Double v[]) {
Double r,s,p,hx,tx,hy,ty,q,c,cc,d,dd,d2,dd2,e,ee,
sn,ssn,cs,ccs,ds,dss,dc,dcc;
#if 0
Double xx,y,yy,z,zz;
#endif
mynumber u;
int4 k;
u.x()=x+big.x();
k = u.i[LOW_HALF]<<2;
x=x-(u.x()-big.x());
d=x+dx;
dd=(x-d)+dx;
/* sin(x+dx)=sin(Xi+t)=sin(Xi)*cos(t) + cos(Xi)sin(t) where t ->0 */
MUL2(d,dd,d,dd,d2,dd2,p,hx,tx,hy,ty,q,c,cc);
sn=sincos.x(k); /* */
ssn=sincos.x(k+1); /* sin(Xi) and cos(Xi) */
cs=sincos.x(k+2); /* */
ccs=sincos.x(k+3); /* */
MUL2(d2,dd2,s7.x(),ss7.x(),ds,dss,p,hx,tx,hy,ty,q,c,cc); /* Taylor */
ADD2(ds,dss,s5.x(),ss5.x(),ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc); /* series */
ADD2(ds,dss,s3.x(),ss3.x(),ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc); /* for sin */
MUL2(d,dd,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,d,dd,ds,dss,r,s); /* ds=sin(t) */
MUL2(d2,dd2,c8.x(),cc8.x(),dc,dcc,p,hx,tx,hy,ty,q,c,cc); ;/* Taylor */
ADD2(dc,dcc,c6.x(),cc6.x(),dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc); /* series */
ADD2(dc,dcc,c4.x(),cc4.x(),dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc); /* for cos */
ADD2(dc,dcc,c2.x(),cc2.x(),dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc); /* dc=cos(t) */
MUL2(cs,ccs,ds,dss,e,ee,p,hx,tx,hy,ty,q,c,cc);
MUL2(dc,dcc,sn,ssn,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
SUB2(e,ee,dc,dcc,e,ee,r,s);
ADD2(e,ee,sn,ssn,e,ee,r,s); /* e+ee=sin(x+dx) */
v[0]=e;
v[1]=ee;
}
uint32_t vp10_avg_8x8_msa(const uint8_t *src, int32_t src_stride) {
uint32_t sum_out;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7;
v8u16 sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7;
v4u32 sum = { 0 };
LD_UB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
HADD_UB4_UH(src0, src1, src2, src3, sum0, sum1, sum2, sum3);
HADD_UB4_UH(src4, src5, src6, src7, sum4, sum5, sum6, sum7);
ADD4(sum0, sum1, sum2, sum3, sum4, sum5, sum6, sum7, sum0, sum2, sum4, sum6);
ADD2(sum0, sum2, sum4, sum6, sum0, sum4);
sum0 += sum4;
sum = __msa_hadd_u_w(sum0, sum0);
sum0 = (v8u16)__msa_pckev_h((v8i16)sum, (v8i16)sum);
sum = __msa_hadd_u_w(sum0, sum0);
sum = (v4u32)__msa_srari_w((v4i32)sum, 6);
sum_out = __msa_copy_u_w((v4i32)sum, 0);
return sum_out;
}
void vpx_fdct16x16_1_msa(const int16_t *input, int16_t *out, int32_t stride) {
int sum, i;
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v4i32 vec_w = { 0 };
for (i = 0; i < 4; ++i) {
LD_SH2(input, 8, in0, in1);
input += stride;
LD_SH2(input, 8, in2, in3);
input += stride;
LD_SH2(input, 8, in4, in5);
input += stride;
LD_SH2(input, 8, in6, in7);
input += stride;
ADD4(in0, in1, in2, in3, in4, in5, in6, in7, in0, in2, in4, in6);
ADD2(in0, in2, in4, in6, in0, in4);
vec_w += __msa_hadd_s_w(in0, in0);
vec_w += __msa_hadd_s_w(in4, in4);
}
sum = HADD_SW_S32(vec_w);
out[0] = (int16_t)(sum >> 1);
}
static void render_world_transform (Render* render, Ship* player)
{
COPY(render->cam, player->pos);
if (render->shaking) {
float hit = ship_hit(player);
Vec3 shake = {
shake_hit * RAND * player->radius * hit +
shake_vel * RAND * player->radius * player->vel[0] / MAX_VEL_X +
shake_thrust * RAND * player->radius * player->lefton +
shake_thrust * RAND * player->radius * player->righton +
shake_velZ * RAND * player->radius * player->vel[2] / MAX_VEL_Z,
shake_hit * RAND * player->radius * hit +
shake_vel * RAND * player->radius * player->vel[1] / MAX_VEL_Y +
shake_velZ * RAND * player->radius * player->vel[2] / MAX_VEL_Z,
shake_hit * RAND * player->radius * hit +
shake_vel * RAND * player->radius * player->vel[2] / MAX_VEL_Z
};
ADD(render->cam, shake);
ADDSCALE(render->cam, player->repulsion, hit * player->radius * 2);
}
ADD2(render->target, player->pos, player->lookAt);
//render->target[1]=render->target[1]*.5+player->pos[1]*.5;
//render->target[2]+=10;
GLfloat lightpos[] = {render->cam[0], render->cam[1], render->cam[2] + 1, 1.0f};
glLightfv(GL_LIGHT1, GL_POSITION, lightpos);
gluLookAt(
render->cam[0], render->cam[1], render->cam[2],
render->target[0], render->target[1], render->target[2],
sin(player->roll), cos(player->roll), 0
);
}
double
SECTION
__ieee754_atan2 (double y, double x)
{
int i, de, ux, dx, uy, dy;
static const int pr[MM] = { 6, 8, 10, 20, 32 };
double ax, ay, u, du, u9, ua, v, vv, dv, t1, t2, t3, t7, t8,
z, zz, cor, s1, ss1, s2, ss2;
#ifndef DLA_FMS
double t4, t5, t6;
#endif
number num;
static const int ep = 59768832, /* 57*16**5 */
em = -59768832; /* -57*16**5 */
/* x=NaN or y=NaN */
num.d = x;
ux = num.i[HIGH_HALF];
dx = num.i[LOW_HALF];
if ((ux & 0x7ff00000) == 0x7ff00000)
{
if (((ux & 0x000fffff) | dx) != 0x00000000)
return x + x;
}
num.d = y;
uy = num.i[HIGH_HALF];
dy = num.i[LOW_HALF];
if ((uy & 0x7ff00000) == 0x7ff00000)
{
if (((uy & 0x000fffff) | dy) != 0x00000000)
return y + y;
}
/* y=+-0 */
if (uy == 0x00000000)
{
if (dy == 0x00000000)
{
if ((ux & 0x80000000) == 0x00000000)
return 0;
else
return opi.d;
}
}
else if (uy == 0x80000000)
{
if (dy == 0x00000000)
{
if ((ux & 0x80000000) == 0x00000000)
return -0.0;
else
return mopi.d;
}
}
/* x=+-0 */
if (x == 0)
{
if ((uy & 0x80000000) == 0x00000000)
return hpi.d;
else
return mhpi.d;
}
/* x=+-INF */
if (ux == 0x7ff00000)
{
if (dx == 0x00000000)
{
if (uy == 0x7ff00000)
{
if (dy == 0x00000000)
return qpi.d;
}
else if (uy == 0xfff00000)
{
if (dy == 0x00000000)
return mqpi.d;
}
else
{
if ((uy & 0x80000000) == 0x00000000)
return 0;
else
return -0.0;
}
}
}
else if (ux == 0xfff00000)
{
if (dx == 0x00000000)
{
if (uy == 0x7ff00000)
{
if (dy == 0x00000000)
return tqpi.d;
}
else if (uy == 0xfff00000)
{
if (dy == 0x00000000)
return mtqpi.d;
}
else
{
if ((uy & 0x80000000) == 0x00000000)
return opi.d;
else
return mopi.d;
}
}
}
/* y=+-INF */
if (uy == 0x7ff00000)
{
if (dy == 0x00000000)
return hpi.d;
}
else if (uy == 0xfff00000)
{
if (dy == 0x00000000)
return mhpi.d;
}
/* either x/y or y/x is very close to zero */
ax = (x < 0) ? -x : x;
ay = (y < 0) ? -y : y;
de = (uy & 0x7ff00000) - (ux & 0x7ff00000);
if (de >= ep)
{
return ((y > 0) ? hpi.d : mhpi.d);
}
else if (de <= em)
{
if (x > 0)
{
if ((z = ay / ax) < TWOM1022)
return normalized (ax, ay, y, z);
else
return signArctan2 (y, z);
}
else
{
return ((y > 0) ? opi.d : mopi.d);
}
}
/* if either x or y is extremely close to zero, scale abs(x), abs(y). */
if (ax < twom500.d || ay < twom500.d)
{
ax *= two500.d;
ay *= two500.d;
}
/* Likewise for large x and y. */
if (ax > two500.d || ay > two500.d)
{
ax *= twom500.d;
ay *= twom500.d;
}
/* x,y which are neither special nor extreme */
if (ay < ax)
{
u = ay / ax;
EMULV (ax, u, v, vv, t1, t2, t3, t4, t5);
du = ((ay - v) - vv) / ax;
}
else
{
u = ax / ay;
EMULV (ay, u, v, vv, t1, t2, t3, t4, t5);
du = ((ax - v) - vv) / ay;
}
if (x > 0)
{
/* (i) x>0, abs(y)< abs(x): atan(ay/ax) */
if (ay < ax)
{
if (u < inv16.d)
{
v = u * u;
zz = du + u * v * (d3.d
+ v * (d5.d
+ v * (d7.d
+ v * (d9.d
+ v * (d11.d
+ v * d13.d)))));
if ((z = u + (zz - u1.d * u)) == u + (zz + u1.d * u))
return signArctan2 (y, z);
MUL2 (u, du, u, du, v, vv, t1, t2, t3, t4, t5, t6, t7, t8);
s1 = v * (f11.d + v * (f13.d
+ v * (f15.d + v * (f17.d + v * f19.d))));
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (u, du, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (u, du, s2, ss2, s1, ss1, t1, t2);
if ((z = s1 + (ss1 - u5.d * s1)) == s1 + (ss1 + u5.d * s1))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
i = (TWO52 + TWO8 * u) - TWO52;
i -= 16;
t3 = u - cij[i][0].d;
EADD (t3, du, v, dv);
t1 = cij[i][1].d;
t2 = cij[i][2].d;
zz = v * t2 + (dv * t2
+ v * v * (cij[i][3].d
+ v * (cij[i][4].d
+ v * (cij[i][5].d
+ v * cij[i][6].d))));
if (i < 112)
{
if (i < 48)
u9 = u91.d; /* u < 1/4 */
else
u9 = u92.d;
} /* 1/4 <= u < 1/2 */
else
{
if (i < 176)
u9 = u93.d; /* 1/2 <= u < 3/4 */
else
u9 = u94.d;
} /* 3/4 <= u <= 1 */
if ((z = t1 + (zz - u9 * t1)) == t1 + (zz + u9 * t1))
return signArctan2 (y, z);
t1 = u - hij[i][0].d;
EADD (t1, du, v, vv);
s1 = v * (hij[i][11].d
+ v * (hij[i][12].d
+ v * (hij[i][13].d
+ v * (hij[i][14].d
+ v * hij[i][15].d))));
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
if ((z = s2 + (ss2 - ub.d * s2)) == s2 + (ss2 + ub.d * s2))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
/* (ii) x>0, abs(x)<=abs(y): pi/2-atan(ax/ay) */
if (u < inv16.d)
{
v = u * u;
zz = u * v * (d3.d
+ v * (d5.d
+ v * (d7.d
+ v * (d9.d
+ v * (d11.d
+ v * d13.d)))));
ESUB (hpi.d, u, t2, cor);
t3 = ((hpi1.d + cor) - du) - zz;
if ((z = t2 + (t3 - u2.d)) == t2 + (t3 + u2.d))
return signArctan2 (y, z);
MUL2 (u, du, u, du, v, vv, t1, t2, t3, t4, t5, t6, t7, t8);
s1 = v * (f11.d
+ v * (f13.d
+ v * (f15.d + v * (f17.d + v * f19.d))));
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (u, du, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (u, du, s2, ss2, s1, ss1, t1, t2);
SUB2 (hpi.d, hpi1.d, s1, ss1, s2, ss2, t1, t2);
if ((z = s2 + (ss2 - u6.d)) == s2 + (ss2 + u6.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
i = (TWO52 + TWO8 * u) - TWO52;
i -= 16;
v = (u - cij[i][0].d) + du;
zz = hpi1.d - v * (cij[i][2].d
+ v * (cij[i][3].d
+ v * (cij[i][4].d
+ v * (cij[i][5].d
+ v * cij[i][6].d))));
t1 = hpi.d - cij[i][1].d;
if (i < 112)
ua = ua1.d; /* w < 1/2 */
else
ua = ua2.d; /* w >= 1/2 */
if ((z = t1 + (zz - ua)) == t1 + (zz + ua))
return signArctan2 (y, z);
t1 = u - hij[i][0].d;
EADD (t1, du, v, vv);
s1 = v * (hij[i][11].d
+ v * (hij[i][12].d
+ v * (hij[i][13].d
+ v * (hij[i][14].d
+ v * hij[i][15].d))));
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
SUB2 (hpi.d, hpi1.d, s2, ss2, s1, ss1, t1, t2);
if ((z = s1 + (ss1 - uc.d)) == s1 + (ss1 + uc.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
/* (iii) x<0, abs(x)< abs(y): pi/2+atan(ax/ay) */
if (ax < ay)
{
if (u < inv16.d)
{
v = u * u;
zz = u * v * (d3.d
+ v * (d5.d
+ v * (d7.d
+ v * (d9.d
+ v * (d11.d + v * d13.d)))));
EADD (hpi.d, u, t2, cor);
t3 = ((hpi1.d + cor) + du) + zz;
if ((z = t2 + (t3 - u3.d)) == t2 + (t3 + u3.d))
return signArctan2 (y, z);
MUL2 (u, du, u, du, v, vv, t1, t2, t3, t4, t5, t6, t7, t8);
s1 = v * (f11.d
+ v * (f13.d + v * (f15.d + v * (f17.d + v * f19.d))));
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (u, du, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (u, du, s2, ss2, s1, ss1, t1, t2);
ADD2 (hpi.d, hpi1.d, s1, ss1, s2, ss2, t1, t2);
if ((z = s2 + (ss2 - u7.d)) == s2 + (ss2 + u7.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
i = (TWO52 + TWO8 * u) - TWO52;
i -= 16;
v = (u - cij[i][0].d) + du;
zz = hpi1.d + v * (cij[i][2].d
+ v * (cij[i][3].d
+ v * (cij[i][4].d
+ v * (cij[i][5].d
+ v * cij[i][6].d))));
t1 = hpi.d + cij[i][1].d;
if (i < 112)
ua = ua1.d; /* w < 1/2 */
else
ua = ua2.d; /* w >= 1/2 */
if ((z = t1 + (zz - ua)) == t1 + (zz + ua))
return signArctan2 (y, z);
t1 = u - hij[i][0].d;
EADD (t1, du, v, vv);
s1 = v * (hij[i][11].d
+ v * (hij[i][12].d
+ v * (hij[i][13].d
+ v * (hij[i][14].d
+ v * hij[i][15].d))));
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
ADD2 (hpi.d, hpi1.d, s2, ss2, s1, ss1, t1, t2);
if ((z = s1 + (ss1 - uc.d)) == s1 + (ss1 + uc.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
/* (iv) x<0, abs(y)<=abs(x): pi-atan(ax/ay) */
if (u < inv16.d)
{
v = u * u;
zz = u * v * (d3.d
+ v * (d5.d
+ v * (d7.d
+ v * (d9.d + v * (d11.d + v * d13.d)))));
ESUB (opi.d, u, t2, cor);
t3 = ((opi1.d + cor) - du) - zz;
if ((z = t2 + (t3 - u4.d)) == t2 + (t3 + u4.d))
return signArctan2 (y, z);
MUL2 (u, du, u, du, v, vv, t1, t2, t3, t4, t5, t6, t7, t8);
s1 = v * (f11.d + v * (f13.d + v * (f15.d + v * (f17.d + v * f19.d))));
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (u, du, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (u, du, s2, ss2, s1, ss1, t1, t2);
SUB2 (opi.d, opi1.d, s1, ss1, s2, ss2, t1, t2);
if ((z = s2 + (ss2 - u8.d)) == s2 + (ss2 + u8.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
i = (TWO52 + TWO8 * u) - TWO52;
i -= 16;
v = (u - cij[i][0].d) + du;
zz = opi1.d - v * (cij[i][2].d
+ v * (cij[i][3].d
+ v * (cij[i][4].d
+ v * (cij[i][5].d + v * cij[i][6].d))));
t1 = opi.d - cij[i][1].d;
if (i < 112)
ua = ua1.d; /* w < 1/2 */
else
ua = ua2.d; /* w >= 1/2 */
if ((z = t1 + (zz - ua)) == t1 + (zz + ua))
return signArctan2 (y, z);
t1 = u - hij[i][0].d;
EADD (t1, du, v, vv);
s1 = v * (hij[i][11].d
+ v * (hij[i][12].d
+ v * (hij[i][13].d
+ v * (hij[i][14].d + v * hij[i][15].d))));
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
SUB2 (opi.d, opi1.d, s2, ss2, s1, ss1, t1, t2);
if ((z = s1 + (ss1 - uc.d)) == s1 + (ss1 + uc.d))
return signArctan2 (y, z);
return atan2Mp (x, y, pr);
}
int32_t vp8_mbuverror_msa(MACROBLOCK *mb)
{
BLOCK *be;
BLOCKD *bd;
int16_t *coeff_ptr, *dq_coeff_ptr;
int32_t err = 0;
uint32_t loop_cnt;
v8i16 coeff, coeff0, coeff1, coeff2, coeff3, coeff4;
v8i16 dq_coeff, dq_coeff2, dq_coeff3, dq_coeff4;
v4i32 diff0, diff1;
v2i64 err0, err1, err_dup0, err_dup1;
for (loop_cnt = 16; loop_cnt < 24; loop_cnt += 2)
{
be = &mb->block[loop_cnt];
bd = &mb->e_mbd.block[loop_cnt];
coeff_ptr = be->coeff;
dq_coeff_ptr = bd->dqcoeff;
coeff = LD_SH(coeff_ptr);
dq_coeff = LD_SH(dq_coeff_ptr);
coeff_ptr += 8;
dq_coeff_ptr += 8;
coeff2 = LD_SH(coeff_ptr);
dq_coeff2 = LD_SH(dq_coeff_ptr);
be = &mb->block[loop_cnt + 1];
bd = &mb->e_mbd.block[loop_cnt + 1];
coeff_ptr = be->coeff;
dq_coeff_ptr = bd->dqcoeff;
coeff3 = LD_SH(coeff_ptr);
dq_coeff3 = LD_SH(dq_coeff_ptr);
coeff_ptr += 8;
dq_coeff_ptr += 8;
coeff4 = LD_SH(coeff_ptr);
dq_coeff4 = LD_SH(dq_coeff_ptr);
ILVRL_H2_SH(coeff, dq_coeff, coeff0, coeff1);
HSUB_UH2_SW(coeff0, coeff1, diff0, diff1);
DOTP_SW2_SD(diff0, diff1, diff0, diff1, err0, err1);
ILVRL_H2_SH(coeff2, dq_coeff2, coeff0, coeff1);
HSUB_UH2_SW(coeff0, coeff1, diff0, diff1);
DPADD_SD2_SD(diff0, diff1, err0, err1);
err_dup0 = __msa_splati_d(err0, 1);
err_dup1 = __msa_splati_d(err1, 1);
ADD2(err0, err_dup0, err1, err_dup1, err0, err1);
err += __msa_copy_s_d(err0, 0);
err += __msa_copy_s_d(err1, 0);
ILVRL_H2_SH(coeff3, dq_coeff3, coeff0, coeff1);
HSUB_UH2_SW(coeff0, coeff1, diff0, diff1);
DOTP_SW2_SD(diff0, diff1, diff0, diff1, err0, err1);
ILVRL_H2_SH(coeff4, dq_coeff4, coeff0, coeff1);
HSUB_UH2_SW(coeff0, coeff1, diff0, diff1);
DPADD_SD2_SD(diff0, diff1, err0, err1);
err_dup0 = __msa_splati_d(err0, 1);
err_dup1 = __msa_splati_d(err1, 1);
ADD2(err0, err_dup0, err1, err_dup1, err0, err1);
err += __msa_copy_s_d(err0, 0);
err += __msa_copy_s_d(err1, 0);
}
return err;
}
/* routine computes the correctly rounded (to nearest) value of atan(x). */
double atan(double x) {
double cor,s1,ss1,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,u,u2,u3,
v,vv,w,ww,y,yy,z,zz;
#if 0
double y1,y2;
#endif
int i,ux,dx;
#if 0
int p;
#endif
static const int pr[M]={6,8,10,32};
number num;
#if 0
mp_no mpt1,mpx,mpy,mpy1,mpy2,mperr;
#endif
num.d = x; ux = num.i[HIGH_HALF]; dx = num.i[LOW_HALF];
/* x=NaN */
if (((ux&0x7ff00000)==0x7ff00000) && (((ux&0x000fffff)|dx)!=0x00000000))
return x+x;
/* Regular values of x, including denormals +-0 and +-INF */
u = (x<ZERO) ? -x : x;
if (u<C) {
if (u<B) {
if (u<A) { /* u < A */
return x; }
else { /* A <= u < B */
v=x*x; yy=x*v*(d3.d+v*(d5.d+v*(d7.d+v*(d9.d+v*(d11.d+v*d13.d)))));
if ((y=x+(yy-U1*x)) == x+(yy+U1*x)) return y;
EMULV(x,x,v,vv,t1,t2,t3,t4,t5) /* v+vv=x^2 */
s1=v*(f11.d+v*(f13.d+v*(f15.d+v*(f17.d+v*f19.d))));
ADD2(f9.d,ff9.d,s1,ZERO,s2,ss2,t1,t2)
MUL2(v,vv,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(f7.d,ff7.d,s1,ss1,s2,ss2,t1,t2)
MUL2(v,vv,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(f5.d,ff5.d,s1,ss1,s2,ss2,t1,t2)
MUL2(v,vv,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(f3.d,ff3.d,s1,ss1,s2,ss2,t1,t2)
MUL2(v,vv,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
MUL2(x,ZERO,s1,ss1,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(x,ZERO,s2,ss2,s1,ss1,t1,t2)
if ((y=s1+(ss1-U5*s1)) == s1+(ss1+U5*s1)) return y;
return atanMp(x,pr);
} }
else { /* B <= u < C */
i=(TWO52+TWO8*u)-TWO52; i-=16;
z=u-cij[i][0].d;
yy=z*(cij[i][2].d+z*(cij[i][3].d+z*(cij[i][4].d+
z*(cij[i][5].d+z* cij[i][6].d))));
t1=cij[i][1].d;
if (i<112) {
if (i<48) u2=U21; /* u < 1/4 */
else u2=U22; } /* 1/4 <= u < 1/2 */
else {
if (i<176) u2=U23; /* 1/2 <= u < 3/4 */
else u2=U24; } /* 3/4 <= u <= 1 */
if ((y=t1+(yy-u2*t1)) == t1+(yy+u2*t1)) return __signArctan(x,y);
z=u-hij[i][0].d;
s1=z*(hij[i][11].d+z*(hij[i][12].d+z*(hij[i][13].d+
z*(hij[i][14].d+z* hij[i][15].d))));
ADD2(hij[i][9].d,hij[i][10].d,s1,ZERO,s2,ss2,t1,t2)
MUL2(z,ZERO,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(hij[i][7].d,hij[i][8].d,s1,ss1,s2,ss2,t1,t2)
MUL2(z,ZERO,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(hij[i][5].d,hij[i][6].d,s1,ss1,s2,ss2,t1,t2)
MUL2(z,ZERO,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(hij[i][3].d,hij[i][4].d,s1,ss1,s2,ss2,t1,t2)
MUL2(z,ZERO,s2,ss2,s1,ss1,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(hij[i][1].d,hij[i][2].d,s1,ss1,s2,ss2,t1,t2)
if ((y=s2+(ss2-U6*s2)) == s2+(ss2+U6*s2)) return __signArctan(x,y);
return atanMp(x,pr);
}
}
int shDrawRadialGradientMesh(SHPaint *p, SHVector2 *min, SHVector2 *max,
VGPaintMode mode, GLenum texUnit)
{
SHint i, j;
float a, n;
SHfloat cx = p->radialGradient[0];
SHfloat cy = p->radialGradient[1];
SHfloat fx = p->radialGradient[2];
SHfloat fy = p->radialGradient[3];
float r = p->radialGradient[4];
float fcx, fcy, rr, C;
SHVector2 ux;
SHVector2 uy;
SHVector2 c, f;
SHVector2 cf;
SHMatrix3x3 *m;
SHMatrix3x3 mi;
SHint invertible;
SHVector2 corners[4];
SHVector2 fcorners[4];
SHfloat minOffset=0.0f;
SHfloat maxOffset=0.0f;
SHint maxI=0, maxJ=0;
SHfloat maxA=0.0f;
SHfloat startA=0.0f;
int numsteps = 100;
float step = 2*PI/numsteps;
SHVector2 tmin, tmax;
SHVector2 min1, max1, min2, max2;
/* Pick paint transform matrix */
SH_GETCONTEXT(0);
if (mode == VG_FILL_PATH)
m = &context->fillTransform;
else if (mode == VG_STROKE_PATH)
m = &context->strokeTransform;
/* Move focus into circle if outside */
SET2(cf, fx,fy);
SUB2(cf, cx,cy);
n = NORM2(cf);
if (n > r) {
DIV2(cf, n);
fx = cx + 0.995f * r * cf.x;
fy = cy + 0.995f * r * cf.y;
}
/* Precalculations */
rr = r*r;
fcx = fx - cx;
fcy = fy - cy;
C = fcx*fcx + fcy*fcy - rr;
/* Apply paint-to-user transformation
to focus and unit vectors */
SET2(f, fx, fy);
SET2(c, cx, cy);
SET2(ux, 1, 0);
SET2(uy, 0, 1);
ADD2(ux, cx, cy);
ADD2(uy, cx, cy);
TRANSFORM2(f, (*m));
TRANSFORM2(c, (*m));
TRANSFORM2(ux, (*m));
TRANSFORM2(uy, (*m));
SUB2V(ux, c); SUB2V(uy, c);
/* Boundbox corners */
SET2(corners[0], min->x, min->y);
SET2(corners[1], max->x, min->y);
SET2(corners[2], max->x, max->y);
SET2(corners[3], min->x, max->y);
/* Find inverse transformation (back to paint space) */
invertible = shInvertMatrix(m, &mi);
if (!invertible || r <= 0.0f) {
/* Fill boundbox with color at offset 1 */
SHColor *c = &p->stops.items[p->stops.size-1].color;
glColor4fv((GLfloat*)c); glBegin(GL_QUADS);
for (i=0; i<4; ++i) glVertex2fv((GLfloat*)&corners[i]);
glEnd();
return 1;
}
/*--------------------------------------------------------*/
/* Find min/max offset */
for (i=0; i<4; ++i) {
/* Transform to paint space */
SHfloat ax,ay, A,B,D,t, off;
TRANSFORM2TO(corners[i], mi, fcorners[i]);
SUB2(fcorners[i], fx, fy);
n = NORM2(fcorners[i]);
if (n == 0.0f) {
/* Avoid zero-length vectors */
off = 0.0f;
}else{
/* Distance from focus to circle at corner angle */
DIV2(fcorners[i], n);
ax = fcorners[i].x;
ay = fcorners[i].y;
A = ax*ax + ay*ay;
B = 2 * (fcx*ax + fcy*ay);
D = B*B - 4*A*C;
t = (-B + SH_SQRT(D)) / (2*A);
/* Relative offset of boundbox corner */
if (D <= 0.0f) off = 1.0f;
else off = n / t;
}
/* Find smallest and largest offset */
if (off < minOffset || i==0) minOffset = off;
if (off > maxOffset || i==0) maxOffset = off;
}
/* Is transformed focus inside original boundbox? */
if (f.x >= min->x && f.x <= max->x &&
f.y >= min->y && f.y <= max->y) {
/* Draw whole circle */
minOffset = 0.0f;
startA = 0.0f;
maxA = 2*PI;
}else{
/* Find most distant corner pair */
for (i=0; i<3; ++i) {
if (ISZERO2(fcorners[i])) continue;
for (j=i+1; j<4; ++j) {
if (ISZERO2(fcorners[j])) continue;
a = ANGLE2N(fcorners[i], fcorners[j]);
if (a > maxA || maxA == 0.0f)
{maxA=a; maxI=i; maxJ=j;}
}}
/* Pick starting angle */
if (CROSS2(fcorners[maxI],fcorners[maxJ]) > 0.0f)
startA = shVectorOrientation(&fcorners[maxI]);
else startA = shVectorOrientation(&fcorners[maxJ]);
}
/*---------------------------------------------------------*/
/* TODO: for minOffset we'd actually need to find minimum
of the gradient function when X and Y are substitued
with a line equation for each bound-box edge. As a
workaround we use 0.0f for now. */
minOffset = 0.0f;
step = PI/50;
numsteps = (SHint)SH_CEIL(maxA / step) + 1;
glActiveTexture(texUnit);
shSetGradientTexGLState(p);
glEnable(GL_TEXTURE_1D);
glBegin(GL_QUADS);
/* Walk the steps and draw gradient mesh */
for (i=0, a=startA; i<numsteps; ++i, a+=step) {
/* Distance from focus to circle border
at current angle (gradient space) */
float ax = SH_COS(a);
float ay = SH_SIN(a);
float A = ax*ax + ay*ay;
float B = 2 * (fcx*ax + fcy*ay);
float D = B*B - 4*A*C;
float t = (-B + SH_SQRT(D)) / (2*A);
if (D <= 0.0f) t = 0.0f;
/* Vectors pointing towards minimum and maximum
offset at current angle (gradient space) */
tmin.x = ax * t * minOffset;
tmin.y = ay * t * minOffset;
tmax.x = ax * t * maxOffset;
tmax.y = ay * t * maxOffset;
/* Transform back to user space */
min2.x = f.x + tmin.x * ux.x + tmin.y * uy.x;
min2.y = f.y + tmin.x * ux.y + tmin.y * uy.y;
max2.x = f.x + tmax.x * ux.x + tmax.y * uy.x;
max2.y = f.y + tmax.x * ux.y + tmax.y * uy.y;
/* Draw quad */
if (i!=0) {
glMultiTexCoord1f(texUnit, minOffset);
glVertex2fv((GLfloat*)&min1);
glVertex2fv((GLfloat*)&min2);
glMultiTexCoord1f(texUnit, maxOffset);
glVertex2fv((GLfloat*)&max2);
glVertex2fv((GLfloat*)&max1);
}
/* Save prev points */
min1 = min2;
max1 = max2;
}
glEnd();
glDisable(GL_TEXTURE_1D);
return 1;
}
static void fdct8x32_1d_row_even_4x(int16_t *input, int16_t *interm_ptr,
int16_t *out) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v4i32 vec0_l, vec1_l, vec2_l, vec3_l, vec4_l, vec5_l, vec6_l, vec7_l;
v4i32 vec0_r, vec1_r, vec2_r, vec3_r, vec4_r, vec5_r, vec6_r, vec7_r;
v4i32 tmp0_w, tmp1_w, tmp2_w, tmp3_w;
/* fdct32 even */
/* stage 2 */
LD_SH8(input, 8, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(input + 64, 8, in8, in9, in10, in11, in12, in13, in14, in15);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7,
in8, in9, in10, in11, in12, in13, in14, in15,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7,
in8, in9, in10, in11, in12, in13, in14, in15);
ST_SH8(vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, interm_ptr, 8);
ST_SH8(in8, in9, in10, in11, in12, in13, in14, in15, interm_ptr + 64, 8);
/* Stage 3 */
UNPCK_SH_SW(vec0, vec0_l, vec0_r);
UNPCK_SH_SW(vec1, vec1_l, vec1_r);
UNPCK_SH_SW(vec2, vec2_l, vec2_r);
UNPCK_SH_SW(vec3, vec3_l, vec3_r);
UNPCK_SH_SW(vec4, vec4_l, vec4_r);
UNPCK_SH_SW(vec5, vec5_l, vec5_r);
UNPCK_SH_SW(vec6, vec6_l, vec6_r);
UNPCK_SH_SW(vec7, vec7_l, vec7_r);
ADD4(vec0_r, vec7_r, vec1_r, vec6_r, vec2_r, vec5_r, vec3_r, vec4_r,
tmp0_w, tmp1_w, tmp2_w, tmp3_w);
BUTTERFLY_4(tmp0_w, tmp1_w, tmp2_w, tmp3_w, vec4_r, vec6_r, vec7_r, vec5_r);
ADD4(vec0_l, vec7_l, vec1_l, vec6_l, vec2_l, vec5_l, vec3_l, vec4_l,
vec0_r, vec1_r, vec2_r, vec3_r);
tmp3_w = vec0_r + vec3_r;
vec0_r = vec0_r - vec3_r;
vec3_r = vec1_r + vec2_r;
vec1_r = vec1_r - vec2_r;
DOTP_CONST_PAIR_W(vec4_r, vec6_r, tmp3_w, vec3_r, cospi_16_64,
cospi_16_64, vec4_r, tmp3_w, vec6_r, vec3_r);
FDCT32_POSTPROC_NEG_W(vec4_r);
FDCT32_POSTPROC_NEG_W(tmp3_w);
FDCT32_POSTPROC_NEG_W(vec6_r);
FDCT32_POSTPROC_NEG_W(vec3_r);
PCKEV_H2_SH(vec4_r, tmp3_w, vec6_r, vec3_r, vec4, vec5);
ST_SH2(vec5, vec4, out, 8);
DOTP_CONST_PAIR_W(vec5_r, vec7_r, vec0_r, vec1_r, cospi_24_64,
cospi_8_64, vec4_r, tmp3_w, vec6_r, vec3_r);
FDCT32_POSTPROC_NEG_W(vec4_r);
FDCT32_POSTPROC_NEG_W(tmp3_w);
FDCT32_POSTPROC_NEG_W(vec6_r);
FDCT32_POSTPROC_NEG_W(vec3_r);
PCKEV_H2_SH(vec4_r, tmp3_w, vec6_r, vec3_r, vec4, vec5);
ST_SH2(vec5, vec4, out + 16, 8);
LD_SH8(interm_ptr, 8, vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7);
SUB4(vec3, vec4, vec2, vec5, vec1, vec6, vec0, vec7, vec4, vec5, vec6, vec7);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 32);
ST_SH(in5, out + 56);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 40);
ST_SH(in5, out + 48);
LD_SH8(interm_ptr + 64, 8, in8, in9, in10, in11, in12, in13, in14, in15);
DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
ADD4(in8, vec3, in9, vec2, in14, vec5, in15, vec4, in0, vec1, vec6, in2);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 64);
ST_SH(in5, out + 120);
SUB2(in0, in1, in2, in3, in0, in2);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 72);
ST_SH(in5, out + 112);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
SUB4(in8, vec3, in15, vec4, in3, in2, in0, in1, in3, in0, vec2, vec5);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, in5, in4);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 80);
ST_SH(in5, out + 104);
ADD2(in3, in2, in0, in1, vec3, vec4);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, in4, in5);
FDCT_POSTPROC_2V_NEG_H(in4, in5);
ST_SH(in4, out + 96);
ST_SH(in5, out + 88);
}
void fdct8x16_1d_column(const int16_t *input, int16_t *tmp_ptr,
int32_t src_stride) {
v8i16 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 stp21, stp22, stp23, stp24, stp25, stp26, stp30;
v8i16 stp31, stp32, stp33, stp34, stp35, stp36, stp37;
v8i16 vec0, vec1, vec2, vec3, vec4, vec5, cnst0, cnst1, cnst4, cnst5;
v8i16 coeff = { cospi_16_64, -cospi_16_64, cospi_8_64, cospi_24_64,
-cospi_8_64, -cospi_24_64, cospi_12_64, cospi_20_64 };
v8i16 coeff1 = { cospi_2_64, cospi_30_64, cospi_14_64, cospi_18_64,
cospi_10_64, cospi_22_64, cospi_6_64, cospi_26_64 };
v8i16 coeff2 = {
-cospi_2_64, -cospi_10_64, -cospi_18_64, -cospi_26_64, 0, 0, 0, 0
};
LD_SH16(input, src_stride, in0, in1, in2, in3, in4, in5, in6, in7, in8, in9,
in10, in11, in12, in13, in14, in15);
SLLI_4V(in0, in1, in2, in3, 2);
SLLI_4V(in4, in5, in6, in7, 2);
SLLI_4V(in8, in9, in10, in11, 2);
SLLI_4V(in12, in13, in14, in15, 2);
ADD4(in0, in15, in1, in14, in2, in13, in3, in12, tmp0, tmp1, tmp2, tmp3);
ADD4(in4, in11, in5, in10, in6, in9, in7, in8, tmp4, tmp5, tmp6, tmp7);
FDCT8x16_EVEN(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp0, tmp1,
tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
ST_SH8(tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp_ptr, 32);
SUB4(in0, in15, in1, in14, in2, in13, in3, in12, in15, in14, in13, in12);
SUB4(in4, in11, in5, in10, in6, in9, in7, in8, in11, in10, in9, in8);
tmp_ptr += 16;
/* stp 1 */
ILVL_H2_SH(in10, in13, in11, in12, vec2, vec4);
ILVR_H2_SH(in10, in13, in11, in12, vec3, vec5);
cnst4 = __msa_splati_h(coeff, 0);
stp25 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst4);
cnst5 = __msa_splati_h(coeff, 1);
cnst5 = __msa_ilvev_h(cnst5, cnst4);
stp22 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst5);
stp24 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst4);
stp23 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst5);
/* stp2 */
BUTTERFLY_4(in8, in9, stp22, stp23, stp30, stp31, stp32, stp33);
BUTTERFLY_4(in15, in14, stp25, stp24, stp37, stp36, stp35, stp34);
ILVL_H2_SH(stp36, stp31, stp35, stp32, vec2, vec4);
ILVR_H2_SH(stp36, stp31, stp35, stp32, vec3, vec5);
SPLATI_H2_SH(coeff, 2, 3, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
stp26 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst0);
cnst0 = __msa_splati_h(coeff, 4);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
stp21 = DOT_SHIFT_RIGHT_PCK_H(vec2, vec3, cnst1);
BUTTERFLY_4(stp30, stp37, stp26, stp21, in8, in15, in14, in9);
ILVRL_H2_SH(in15, in8, vec1, vec0);
SPLATI_H2_SH(coeff1, 0, 1, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr);
cnst0 = __msa_splati_h(coeff2, 0);
cnst0 = __msa_ilvev_h(cnst1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 224);
ILVRL_H2_SH(in14, in9, vec1, vec0);
SPLATI_H2_SH(coeff1, 2, 3, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
ST_SH(in8, tmp_ptr + 128);
cnst1 = __msa_splati_h(coeff2, 2);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 96);
SPLATI_H2_SH(coeff, 2, 5, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
stp25 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
cnst1 = __msa_splati_h(coeff, 3);
cnst1 = __msa_ilvev_h(cnst0, cnst1);
stp22 = DOT_SHIFT_RIGHT_PCK_H(vec4, vec5, cnst1);
/* stp4 */
ADD2(stp34, stp25, stp33, stp22, in13, in10);
ILVRL_H2_SH(in13, in10, vec1, vec0);
SPLATI_H2_SH(coeff1, 4, 5, cnst0, cnst1);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 64);
cnst0 = __msa_splati_h(coeff2, 1);
cnst0 = __msa_ilvev_h(cnst1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 160);
SUB2(stp34, stp25, stp33, stp22, in12, in11);
ILVRL_H2_SH(in12, in11, vec1, vec0);
SPLATI_H2_SH(coeff1, 6, 7, cnst0, cnst1);
cnst1 = __msa_ilvev_h(cnst1, cnst0);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst1);
ST_SH(in8, tmp_ptr + 192);
cnst1 = __msa_splati_h(coeff2, 3);
cnst0 = __msa_ilvev_h(cnst0, cnst1);
in8 = DOT_SHIFT_RIGHT_PCK_H(vec0, vec1, cnst0);
ST_SH(in8, tmp_ptr + 32);
}
double
SECTION
__ieee754_log(double x) {
#define M 4
static const int pr[M]={8,10,18,32};
int i,j,n,ux,dx,p;
#if 0
int k;
#endif
double dbl_n,u,p0,q,r0,w,nln2a,luai,lubi,lvaj,lvbj,
sij,ssij,ttij,A,B,B0,y,y1,y2,polI,polII,sa,sb,
t1,t2,t7,t8,t,ra,rb,ww,
a0,aa0,s1,s2,ss2,s3,ss3,a1,aa1,a,aa,b,bb,c;
#ifndef DLA_FMS
double t3,t4,t5,t6;
#endif
number num;
mp_no mpx,mpy,mpy1,mpy2,mperr;
#include "ulog.tbl"
#include "ulog.h"
/* Treating special values of x ( x<=0, x=INF, x=NaN etc.). */
num.d = x; ux = num.i[HIGH_HALF]; dx = num.i[LOW_HALF];
n=0;
if (__builtin_expect(ux < 0x00100000, 0)) {
if (__builtin_expect(((ux & 0x7fffffff) | dx) == 0, 0))
return MHALF/ZERO; /* return -INF */
if (__builtin_expect(ux < 0, 0))
return (x-x)/ZERO; /* return NaN */
n -= 54; x *= two54.d; /* scale x */
num.d = x;
}
if (__builtin_expect(ux >= 0x7ff00000, 0))
return x+x; /* INF or NaN */
/* Regular values of x */
w = x-ONE;
if (__builtin_expect(ABS(w) > U03, 1)) { goto case_03; }
/*--- Stage I, the case abs(x-1) < 0.03 */
t8 = MHALF*w;
EMULV(t8,w,a,aa,t1,t2,t3,t4,t5)
EADD(w,a,b,bb)
/* Evaluate polynomial II */
polII = (b0.d+w*(b1.d+w*(b2.d+w*(b3.d+w*(b4.d+
w*(b5.d+w*(b6.d+w*(b7.d+w*b8.d))))))))*w*w*w;
c = (aa+bb)+polII;
/* End stage I, case abs(x-1) < 0.03 */
if ((y=b+(c+b*E2)) == b+(c-b*E2)) return y;
/*--- Stage II, the case abs(x-1) < 0.03 */
a = d11.d+w*(d12.d+w*(d13.d+w*(d14.d+w*(d15.d+w*(d16.d+
w*(d17.d+w*(d18.d+w*(d19.d+w*d20.d))))))));
EMULV(w,a,s2,ss2,t1,t2,t3,t4,t5)
ADD2(d10.d,dd10.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d9.d,dd9.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d8.d,dd8.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d7.d,dd7.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d6.d,dd6.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d5.d,dd5.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d4.d,dd4.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d3.d,dd3.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(d2.d,dd2.d,s2,ss2,s3,ss3,t1,t2)
MUL2(w,ZERO,s3,ss3,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
MUL2(w,ZERO,s2,ss2,s3,ss3,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(w,ZERO, s3,ss3, b, bb,t1,t2)
/* End stage II, case abs(x-1) < 0.03 */
if ((y=b+(bb+b*E4)) == b+(bb-b*E4)) return y;
goto stage_n;
/*--- Stage I, the case abs(x-1) > 0.03 */
case_03:
/* Find n,u such that x = u*2**n, 1/sqrt(2) < u < sqrt(2) */
n += (num.i[HIGH_HALF] >> 20) - 1023;
num.i[HIGH_HALF] = (num.i[HIGH_HALF] & 0x000fffff) | 0x3ff00000;
if (num.d > SQRT_2) { num.d *= HALF; n++; }
u = num.d; dbl_n = (double) n;
/* Find i such that ui=1+(i-75)/2**8 is closest to u (i= 0,1,2,...,181) */
num.d += h1.d;
i = (num.i[HIGH_HALF] & 0x000fffff) >> 12;
/* Find j such that vj=1+(j-180)/2**16 is closest to v=u/ui (j= 0,...,361) */
num.d = u*Iu[i].d + h2.d;
j = (num.i[HIGH_HALF] & 0x000fffff) >> 4;
/* Compute w=(u-ui*vj)/(ui*vj) */
p0=(ONE+(i-75)*DEL_U)*(ONE+(j-180)*DEL_V);
q=u-p0; r0=Iu[i].d*Iv[j].d; w=q*r0;
/* Evaluate polynomial I */
polI = w+(a2.d+a3.d*w)*w*w;
/* Add up everything */
nln2a = dbl_n*LN2A;
luai = Lu[i][0].d; lubi = Lu[i][1].d;
lvaj = Lv[j][0].d; lvbj = Lv[j][1].d;
EADD(luai,lvaj,sij,ssij)
EADD(nln2a,sij,A ,ttij)
B0 = (((lubi+lvbj)+ssij)+ttij)+dbl_n*LN2B;
B = polI+B0;
/* End stage I, case abs(x-1) >= 0.03 */
if ((y=A+(B+E1)) == A+(B-E1)) return y;
/*--- Stage II, the case abs(x-1) > 0.03 */
/* Improve the accuracy of r0 */
EMULV(p0,r0,sa,sb,t1,t2,t3,t4,t5)
t=r0*((ONE-sa)-sb);
EADD(r0,t,ra,rb)
/* Compute w */
MUL2(q,ZERO,ra,rb,w,ww,t1,t2,t3,t4,t5,t6,t7,t8)
EADD(A,B0,a0,aa0)
/* Evaluate polynomial III */
s1 = (c3.d+(c4.d+c5.d*w)*w)*w;
EADD(c2.d,s1,s2,ss2)
MUL2(s2,ss2,w,ww,s3,ss3,t1,t2,t3,t4,t5,t6,t7,t8)
MUL2(s3,ss3,w,ww,s2,ss2,t1,t2,t3,t4,t5,t6,t7,t8)
ADD2(s2,ss2,w,ww,s3,ss3,t1,t2)
ADD2(s3,ss3,a0,aa0,a1,aa1,t1,t2)
/* End stage II, case abs(x-1) >= 0.03 */
if ((y=a1+(aa1+E3)) == a1+(aa1-E3)) return y;
/* Final stages. Use multi-precision arithmetic. */
stage_n:
for (i=0; i<M; i++) {
p = pr[i];
__dbl_mp(x,&mpx,p); __dbl_mp(y,&mpy,p);
__mplog(&mpx,&mpy,p);
__dbl_mp(e[i].d,&mperr,p);
__add(&mpy,&mperr,&mpy1,p); __sub(&mpy,&mperr,&mpy2,p);
__mp_dbl(&mpy1,&y1,p); __mp_dbl(&mpy2,&y2,p);
if (y1==y2) return y1;
}
return y1;
}
static void fdct8x32_1d_row_odd_rd(int16_t *temp, int16_t *interm_ptr,
int16_t *out) {
v8i16 in16, in17, in18, in19, in20, in21, in22, in23;
v8i16 in24, in25, in26, in27, in28, in29, in30, in31;
v8i16 vec4, vec5;
in20 = LD_SH(temp + 32);
in21 = LD_SH(temp + 40);
in26 = LD_SH(temp + 80);
in27 = LD_SH(temp + 88);
DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
FDCT_POSTPROC_2V_NEG_H(in20, in21);
FDCT_POSTPROC_2V_NEG_H(in26, in27);
in18 = LD_SH(temp + 16);
in19 = LD_SH(temp + 24);
in28 = LD_SH(temp + 96);
in29 = LD_SH(temp + 104);
FDCT_POSTPROC_2V_NEG_H(in18, in19);
FDCT_POSTPROC_2V_NEG_H(in28, in29);
vec4 = in19 - in20;
ST_SH(vec4, interm_ptr + 32);
vec4 = in18 - in21;
ST_SH(vec4, interm_ptr + 88);
vec4 = in29 - in26;
ST_SH(vec4, interm_ptr + 64);
vec4 = in28 - in27;
ST_SH(vec4, interm_ptr + 56);
ADD4(in18, in21, in19, in20, in28, in27, in29, in26, in21, in20, in27, in26);
in22 = LD_SH(temp + 48);
in23 = LD_SH(temp + 56);
in24 = LD_SH(temp + 64);
in25 = LD_SH(temp + 72);
DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
FDCT_POSTPROC_2V_NEG_H(in22, in23);
FDCT_POSTPROC_2V_NEG_H(in24, in25);
in16 = LD_SH(temp);
in17 = LD_SH(temp + 8);
in30 = LD_SH(temp + 112);
in31 = LD_SH(temp + 120);
FDCT_POSTPROC_2V_NEG_H(in16, in17);
FDCT_POSTPROC_2V_NEG_H(in30, in31);
vec4 = in17 - in22;
ST_SH(vec4, interm_ptr + 40);
vec4 = in30 - in25;
ST_SH(vec4, interm_ptr + 48);
vec4 = in31 - in24;
ST_SH(vec4, interm_ptr + 72);
vec4 = in16 - in23;
ST_SH(vec4, interm_ptr + 80);
ADD4(in16, in23, in17, in22, in30, in25, in31, in24, in16, in17, in30, in31);
DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
ADD4(in16, in19, in17, in18, in30, in29, in31, in28, in27, in22, in21, in25);
DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
ADD2(in27, in26, in25, in24, in23, in20);
DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
ST_SH(vec5, out);
ST_SH(vec4, out + 120);
SUB2(in27, in26, in25, in24, in22, in21);
DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
ST_SH(vec5, out + 112);
ST_SH(vec4, out + 8);
SUB4(in17, in18, in16, in19, in31, in28, in30, in29, in23, in26, in24, in20);
DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
SUB2(in26, in27, in24, in25, in23, in20);
DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
ST_SH(vec4, out + 16);
ST_SH(vec5, out + 104);
ADD2(in26, in27, in24, in25, in22, in21);
DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
ST_SH(vec4, out + 24);
ST_SH(vec5, out + 96);
in20 = LD_SH(interm_ptr + 32);
in21 = LD_SH(interm_ptr + 88);
in27 = LD_SH(interm_ptr + 56);
in26 = LD_SH(interm_ptr + 64);
in16 = in20;
in17 = in21;
DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
in22 = LD_SH(interm_ptr + 40);
in25 = LD_SH(interm_ptr + 48);
in24 = LD_SH(interm_ptr + 72);
in23 = LD_SH(interm_ptr + 80);
SUB4(in23, in20, in22, in21, in25, in26, in24, in27, in28, in17, in18, in31);
DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
in16 = in28 + in29;
in19 = in31 + in30;
DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
ST_SH(vec5, out + 32);
ST_SH(vec4, out + 88);
SUB2(in28, in29, in31, in30, in17, in18);
DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
ST_SH(vec5, out + 40);
ST_SH(vec4, out + 80);
ADD4(in22, in21, in23, in20, in24, in27, in25, in26, in16, in29, in30, in19);
DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
SUB2(in29, in28, in30, in31, in16, in19);
DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
ST_SH(vec5, out + 72);
ST_SH(vec4, out + 48);
ADD2(in29, in28, in30, in31, in17, in18);
DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
ST_SH(vec4, out + 56);
ST_SH(vec5, out + 64);
}
static void fdct8x32_1d_row_even_rd(int16_t *temp, int16_t *out) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, temp0, temp1;
/* fdct32 even */
/* stage 2 */
LD_SH8(temp, 8, in0, in1, in2, in3, in4, in5, in6, in7);
LD_SH8(temp + 64, 8, in8, in9, in10, in11, in12, in13, in14, in15);
BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7,
in8, in9, in10, in11, in12, in13, in14, in15,
vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7,
in8, in9, in10, in11, in12, in13, in14, in15);
FDCT_POSTPROC_2V_NEG_H(vec0, vec1);
FDCT_POSTPROC_2V_NEG_H(vec2, vec3);
FDCT_POSTPROC_2V_NEG_H(vec4, vec5);
FDCT_POSTPROC_2V_NEG_H(vec6, vec7);
FDCT_POSTPROC_2V_NEG_H(in8, in9);
FDCT_POSTPROC_2V_NEG_H(in10, in11);
FDCT_POSTPROC_2V_NEG_H(in12, in13);
FDCT_POSTPROC_2V_NEG_H(in14, in15);
/* Stage 3 */
ADD4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, in0, in1, in2, in3);
temp0 = in0 + in3;
in0 = in0 - in3;
in3 = in1 + in2;
in1 = in1 - in2;
DOTP_CONST_PAIR(temp0, in3, cospi_16_64, cospi_16_64, temp1, temp0);
ST_SH(temp0, out);
ST_SH(temp1, out + 8);
DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
ST_SH(temp0, out + 16);
ST_SH(temp1, out + 24);
SUB4(vec3, vec4, vec2, vec5, vec1, vec6, vec0, vec7, vec4, vec5, vec6, vec7);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
ST_SH(temp0, out + 32);
ST_SH(temp1, out + 56);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
ST_SH(temp0, out + 40);
ST_SH(temp1, out + 48);
DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
ADD4(in8, vec3, in9, vec2, in14, vec5, in15, vec4, in0, vec1, vec6, in2);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
ST_SH(temp0, out + 64);
ST_SH(temp1, out + 120);
SUB2(in0, in1, in2, in3, in0, in2);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
ST_SH(temp0, out + 72);
ST_SH(temp1, out + 112);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
SUB4(in8, vec3, in15, vec4, in3, in2, in0, in1, in3, in0, vec2, vec5);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
ST_SH(temp0, out + 80);
ST_SH(temp1, out + 104);
ADD2(in3, in2, in0, in1, vec3, vec4);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
ST_SH(temp0, out + 96);
ST_SH(temp1, out + 88);
}
static void fdct8x32_1d_column_even_store(int16_t *input, int16_t *temp) {
v8i16 in0, in1, in2, in3, in4, in5, in6, in7;
v8i16 in8, in9, in10, in11, in12, in13, in14, in15;
v8i16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v8i16 temp0, temp1;
/* fdct even */
LD_SH4(input, 8, in0, in1, in2, in3);
LD_SH4(input + 96, 8, in12, in13, in14, in15);
BUTTERFLY_8(in0, in1, in2, in3, in12, in13, in14, in15,
vec0, vec1, vec2, vec3, in12, in13, in14, in15);
LD_SH4(input + 32, 8, in4, in5, in6, in7);
LD_SH4(input + 64, 8, in8, in9, in10, in11);
BUTTERFLY_8(in4, in5, in6, in7, in8, in9, in10, in11,
vec4, vec5, vec6, vec7, in8, in9, in10, in11);
/* Stage 3 */
ADD4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, in0, in1, in2, in3);
BUTTERFLY_4(in0, in1, in2, in3, temp0, in4, in1, in0);
DOTP_CONST_PAIR(temp0, in4, cospi_16_64, cospi_16_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp);
ST_SH(temp1, temp + 512);
DOTP_CONST_PAIR(in0, in1, cospi_24_64, cospi_8_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 256);
ST_SH(temp1, temp + 768);
SUB4(vec0, vec7, vec1, vec6, vec2, vec5, vec3, vec4, vec7, vec6, vec5, vec4);
DOTP_CONST_PAIR(vec6, vec5, cospi_16_64, cospi_16_64, vec5, vec6);
ADD2(vec4, vec5, vec7, vec6, vec0, vec1);
DOTP_CONST_PAIR(vec1, vec0, cospi_28_64, cospi_4_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 128);
ST_SH(temp1, temp + 896);
SUB2(vec4, vec5, vec7, vec6, vec4, vec7);
DOTP_CONST_PAIR(vec7, vec4, cospi_12_64, cospi_20_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 640);
ST_SH(temp1, temp + 384);
DOTP_CONST_PAIR(in13, in10, cospi_16_64, cospi_16_64, vec2, vec5);
DOTP_CONST_PAIR(in12, in11, cospi_16_64, cospi_16_64, vec3, vec4);
ADD4(in8, vec3, in9, vec2, in14, vec5, in15, vec4, in0, vec1, vec6, in2);
DOTP_CONST_PAIR(vec6, vec1, cospi_24_64, cospi_8_64, in1, in3);
ADD2(in0, in1, in2, in3, vec0, vec7);
DOTP_CONST_PAIR(vec7, vec0, cospi_30_64, cospi_2_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 64);
ST_SH(temp1, temp + 960);
SUB2(in0, in1, in2, in3, in0, in2);
DOTP_CONST_PAIR(in2, in0, cospi_14_64, cospi_18_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 576);
ST_SH(temp1, temp + 448);
SUB2(in9, vec2, in14, vec5, vec2, vec5);
DOTP_CONST_PAIR((-vec2), vec5, cospi_24_64, cospi_8_64, in2, in1);
SUB4(in8, vec3, in15, vec4, in3, in2, in0, in1, in3, in0, vec2, vec5);
DOTP_CONST_PAIR(vec5, vec2, cospi_22_64, cospi_10_64, temp1, temp0);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 320);
ST_SH(temp1, temp + 704);
ADD2(in3, in2, in0, in1, vec3, vec4);
DOTP_CONST_PAIR(vec4, vec3, cospi_6_64, cospi_26_64, temp0, temp1);
FDCT32_POSTPROC_2V_POS_H(temp0, temp1);
ST_SH(temp0, temp + 192);
ST_SH(temp1, temp + 832);
}
void __dubcos(double x, double dx, double v[]) {
double r,s,p,hx,tx,hy,ty,q,c,cc,d,dd,d2,dd2,e,ee,
sn,ssn,cs,ccs,ds,dss,dc,dcc;
#if 0
double xx,y,yy,z,zz;
#endif
mynumber u;
int4 k;
u.x=x+big.x;
k = u.i[LOW_HALF]<<2;
x=x-(u.x-big.x);
d=x+dx;
dd=(x-d)+dx; /* cos(x+dx)=cos(Xi+t)=cos(Xi)cos(t) - sin(Xi)sin(t) */
MUL2(d,dd,d,dd,d2,dd2,p,hx,tx,hy,ty,q,c,cc);
sn=sincos.x[k]; /* */
ssn=sincos.x[k+1]; /* sin(Xi) and cos(Xi) */
cs=sincos.x[k+2]; /* */
ccs=sincos.x[k+3]; /* */
MUL2(d2,dd2,s7.x,ss7.x,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,s5.x,ss5.x,ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,s3.x,ss3.x,ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
MUL2(d,dd,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,d,dd,ds,dss,r,s);
MUL2(d2,dd2,c8.x,cc8.x,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c6.x,cc6.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c4.x,cc4.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c2.x,cc2.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
MUL2(cs,ccs,ds,dss,e,ee,p,hx,tx,hy,ty,q,c,cc);
MUL2(dc,dcc,sn,ssn,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
MUL2(d2,dd2,s7.x,ss7.x,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,s5.x,ss5.x,ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,s3.x,ss3.x,ds,dss,r,s);
MUL2(d2,dd2,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
MUL2(d,dd,ds,dss,ds,dss,p,hx,tx,hy,ty,q,c,cc);
ADD2(ds,dss,d,dd,ds,dss,r,s);
MUL2(d2,dd2,c8.x,cc8.x,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c6.x,cc6.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c4.x,cc4.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(dc,dcc,c2.x,cc2.x,dc,dcc,r,s);
MUL2(d2,dd2,dc,dcc,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
MUL2(sn,ssn,ds,dss,e,ee,p,hx,tx,hy,ty,q,c,cc);
MUL2(dc,dcc,cs,ccs,dc,dcc,p,hx,tx,hy,ty,q,c,cc);
ADD2(e,ee,dc,dcc,e,ee,r,s);
SUB2(cs,ccs,e,ee,e,ee,r,s);
v[0]=e;
v[1]=ee;
}
static void shDrawPaintMesh(VGContext *c, SHVector2 *min, SHVector2 *max,
VGPaintMode mode, GLenum texUnit)
{
SHPaint *p;
SHVector2 pmin, pmax;
SHfloat K = 1.0f;
#ifdef ANDROIDVG
SHColor *color;
GLfloat v[6][2];
#endif
/* Pick the right paint */
if (mode == VG_FILL_PATH) {
p = (c->fillPaint ? c->fillPaint : &c->defaultPaint);
}else if (mode == VG_STROKE_PATH) {
p = (c->strokePaint ? c->strokePaint : &c->defaultPaint);
K = SH_CEIL(c->strokeMiterLimit * c->strokeLineWidth) + 1.0f;
}
/* We want to be sure to cover every pixel of this path so better
take a pixel more than leave some out (multisampling is tricky). */
SET2V(pmin, (*min)); SUB2(pmin, K,K);
SET2V(pmax, (*max)); ADD2(pmax, K,K);
/* Construct appropriate OpenGL primitives so as
to fill the stencil mask with select paint */
switch (p->type) {
case VG_PAINT_TYPE_LINEAR_GRADIENT:
shDrawLinearGradientMesh(p, min, max, mode, texUnit);
break;
case VG_PAINT_TYPE_RADIAL_GRADIENT:
shDrawRadialGradientMesh(p, min, max, mode, texUnit);
break;
case VG_PAINT_TYPE_PATTERN:
if (p->pattern != VG_INVALID_HANDLE) {
shDrawPatternMesh(p, min, max, mode, texUnit);
break;
}/* else behave as a color paint */
case VG_PAINT_TYPE_COLOR:
#ifdef ANDROIDVG
v[0][0] = pmin.x; v[0][1] = pmin.y;
v[1][0] = pmax.x; v[1][1] = pmin.y;
v[2][0] = pmax.x; v[2][1] = pmax.y;
v[3][0] = pmin.x; v[3][1] = pmin.y;
v[4][0] = pmax.x; v[4][1] = pmax.y;
v[5][0] = pmin.x; v[5][1] = pmax.y;
color = &p->color;
glColor4f(color->r, color->g, color->b, color->a);
glEnableClientState(GL_VERTEX_ARRAY);
glVertexPointer(2, GL_FLOAT, 0, v);
glDrawArrays(GL_TRIANGLES, 0, 6);
glDisableClientState(GL_VERTEX_ARRAY);
#else
glColor4fv((GLfloat*)&p->color);
glBegin(GL_QUADS);
glVertex2f(pmin.x, pmin.y);
glVertex2f(pmax.x, pmin.y);
glVertex2f(pmax.x, pmax.y);
glVertex2f(pmin.x, pmax.y);
glEnd();
#endif
break;
}
}
static void temporal_filter_apply_8size_msa(uint8_t *frame1_ptr,
uint32_t stride,
uint8_t *frame2_ptr,
int32_t strength_in,
int32_t filter_wt_in,
uint32_t *acc, uint16_t *cnt)
{
uint32_t row;
uint64_t f0, f1, f2, f3, f4, f5, f6, f7;
v16i8 frame1 = { 0 };
v16i8 frame2 = { 0 };
v16i8 frame3 = { 0 };
v16i8 frame4 = { 0 };
v16u8 frame_l, frame_h;
v8i16 frame2_0_h, frame2_1_h, mod0_h, mod1_h;
v8i16 diff0, diff1, cnt0, cnt1;
v4i32 const3, const16;
v4i32 filter_wt, strength;
v4i32 mod0_w, mod1_w, mod2_w, mod3_w;
v4i32 diff0_r, diff0_l, diff1_r, diff1_l;
v4i32 frame2_0, frame2_1, frame2_2, frame2_3;
v4i32 acc0, acc1, acc2, acc3;
filter_wt = __msa_fill_w(filter_wt_in);
strength = __msa_fill_w(strength_in);
const3 = __msa_ldi_w(3);
const16 = __msa_ldi_w(16);
for (row = 2; row--;)
{
LD2(frame1_ptr, stride, f0, f1);
frame1_ptr += (2 * stride);
LD2(frame2_ptr, 8, f2, f3);
frame2_ptr += 16;
LD2(frame1_ptr, stride, f4, f5);
frame1_ptr += (2 * stride);
LD2(frame2_ptr, 8, f6, f7);
frame2_ptr += 16;
LD_SW2(acc, 4, acc0, acc1);
LD_SW2(acc + 8, 4, acc2, acc3);
LD_SH2(cnt, 8, cnt0, cnt1);
INSERT_D2_SB(f0, f1, frame1);
INSERT_D2_SB(f2, f3, frame2);
INSERT_D2_SB(f4, f5, frame3);
INSERT_D2_SB(f6, f7, frame4);
ILVRL_B2_UB(frame1, frame2, frame_l, frame_h);
HSUB_UB2_SH(frame_l, frame_h, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
MUL4(mod0_w, const3, mod1_w, const3, mod2_w, const3, mod3_w, const3,
mod0_w, mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < const16);
diff0_l = (mod1_w < const16);
diff1_r = (mod2_w < const16);
diff1_l = (mod3_w < const16);
SUB4(const16, mod0_w, const16, mod1_w, const16, mod2_w, const16, mod3_w,
mod0_w, mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
mod2_w = diff1_r & mod2_w;
mod3_w = diff1_l & mod3_w;
MUL4(mod0_w, filter_wt, mod1_w, filter_wt, mod2_w, filter_wt, mod3_w,
filter_wt, mod0_w, mod1_w, mod2_w, mod3_w);
PCKEV_H2_SH(mod1_w, mod0_w, mod3_w, mod2_w, mod0_h, mod1_h);
ADD2(mod0_h, cnt0, mod1_h, cnt1, mod0_h, mod1_h);
ST_SH2(mod0_h, mod1_h, cnt, 8);
cnt += 16;
UNPCK_UB_SH(frame2, frame2_0_h, frame2_1_h);
UNPCK_SH_SW(frame2_0_h, frame2_0, frame2_1);
UNPCK_SH_SW(frame2_1_h, frame2_2, frame2_3);
MUL4(mod0_w, frame2_0, mod1_w, frame2_1, mod2_w, frame2_2, mod3_w,
frame2_3, mod0_w, mod1_w, mod2_w, mod3_w);
ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
mod0_w, mod1_w, mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
ST_SW2(mod2_w, mod3_w, acc + 8, 4);
acc += 16;
LD_SW2(acc, 4, acc0, acc1);
LD_SW2(acc + 8, 4, acc2, acc3);
LD_SH2(cnt, 8, cnt0, cnt1);
ILVRL_B2_UB(frame3, frame4, frame_l, frame_h);
HSUB_UB2_SH(frame_l, frame_h, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
MUL4(mod0_w, const3, mod1_w, const3, mod2_w, const3, mod3_w, const3,
mod0_w, mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < const16);
diff0_l = (mod1_w < const16);
diff1_r = (mod2_w < const16);
diff1_l = (mod3_w < const16);
SUB4(const16, mod0_w, const16, mod1_w, const16, mod2_w, const16, mod3_w,
mod0_w, mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
mod2_w = diff1_r & mod2_w;
mod3_w = diff1_l & mod3_w;
MUL4(mod0_w, filter_wt, mod1_w, filter_wt, mod2_w, filter_wt, mod3_w,
filter_wt, mod0_w, mod1_w, mod2_w, mod3_w);
PCKEV_H2_SH(mod1_w, mod0_w, mod3_w, mod2_w, mod0_h, mod1_h);
ADD2(mod0_h, cnt0, mod1_h, cnt1, mod0_h, mod1_h);
ST_SH2(mod0_h, mod1_h, cnt, 8);
cnt += 16;
UNPCK_UB_SH(frame4, frame2_0_h, frame2_1_h);
UNPCK_SH_SW(frame2_0_h, frame2_0, frame2_1);
UNPCK_SH_SW(frame2_1_h, frame2_2, frame2_3);
MUL4(mod0_w, frame2_0, mod1_w, frame2_1, mod2_w, frame2_2, mod3_w,
frame2_3, mod0_w, mod1_w, mod2_w, mod3_w);
ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
mod0_w, mod1_w, mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
ST_SW2(mod2_w, mod3_w, acc + 8, 4);
acc += 16;
}
}
static void fdct8x32_1d_column_odd_store(int16_t *input, int16_t *temp_ptr) {
v8i16 in16, in17, in18, in19, in20, in21, in22, in23;
v8i16 in24, in25, in26, in27, in28, in29, in30, in31, vec4, vec5;
in20 = LD_SH(input + 32);
in21 = LD_SH(input + 40);
in26 = LD_SH(input + 80);
in27 = LD_SH(input + 88);
DOTP_CONST_PAIR(in27, in20, cospi_16_64, cospi_16_64, in20, in27);
DOTP_CONST_PAIR(in26, in21, cospi_16_64, cospi_16_64, in21, in26);
in18 = LD_SH(input + 16);
in19 = LD_SH(input + 24);
in28 = LD_SH(input + 96);
in29 = LD_SH(input + 104);
vec4 = in19 - in20;
ST_SH(vec4, input + 32);
vec4 = in18 - in21;
ST_SH(vec4, input + 40);
vec4 = in29 - in26;
ST_SH(vec4, input + 80);
vec4 = in28 - in27;
ST_SH(vec4, input + 88);
in21 = in18 + in21;
in20 = in19 + in20;
in27 = in28 + in27;
in26 = in29 + in26;
LD_SH4(input + 48, 8, in22, in23, in24, in25);
DOTP_CONST_PAIR(in25, in22, cospi_16_64, cospi_16_64, in22, in25);
DOTP_CONST_PAIR(in24, in23, cospi_16_64, cospi_16_64, in23, in24);
in16 = LD_SH(input);
in17 = LD_SH(input + 8);
in30 = LD_SH(input + 112);
in31 = LD_SH(input + 120);
vec4 = in17 - in22;
ST_SH(vec4, input + 16);
vec4 = in16 - in23;
ST_SH(vec4, input + 24);
vec4 = in31 - in24;
ST_SH(vec4, input + 96);
vec4 = in30 - in25;
ST_SH(vec4, input + 104);
ADD4(in16, in23, in17, in22, in30, in25, in31, in24, in16, in17, in30, in31);
DOTP_CONST_PAIR(in26, in21, cospi_24_64, cospi_8_64, in18, in29);
DOTP_CONST_PAIR(in27, in20, cospi_24_64, cospi_8_64, in19, in28);
ADD4(in16, in19, in17, in18, in30, in29, in31, in28, in27, in22, in21, in25);
DOTP_CONST_PAIR(in21, in22, cospi_28_64, cospi_4_64, in26, in24);
ADD2(in27, in26, in25, in24, in23, in20);
DOTP_CONST_PAIR(in20, in23, cospi_31_64, cospi_1_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr);
ST_SH(vec4, temp_ptr + 960);
SUB2(in27, in26, in25, in24, in22, in21);
DOTP_CONST_PAIR(in21, in22, cospi_15_64, cospi_17_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 448);
ST_SH(vec4, temp_ptr + 512);
SUB4(in17, in18, in16, in19, in31, in28, in30, in29, in23, in26, in24, in20);
DOTP_CONST_PAIR((-in23), in20, cospi_28_64, cospi_4_64, in27, in25);
SUB2(in26, in27, in24, in25, in23, in20);
DOTP_CONST_PAIR(in20, in23, cospi_23_64, cospi_9_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec4, temp_ptr + 704);
ST_SH(vec5, temp_ptr + 256);
ADD2(in26, in27, in24, in25, in22, in21);
DOTP_CONST_PAIR(in21, in22, cospi_7_64, cospi_25_64, vec4, vec5);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec4, temp_ptr + 192);
ST_SH(vec5, temp_ptr + 768);
LD_SH4(input + 16, 8, in22, in23, in20, in21);
LD_SH4(input + 80, 8, in26, in27, in24, in25);
in16 = in20;
in17 = in21;
DOTP_CONST_PAIR(-in16, in27, cospi_24_64, cospi_8_64, in20, in27);
DOTP_CONST_PAIR(-in17, in26, cospi_24_64, cospi_8_64, in21, in26);
SUB4(in23, in20, in22, in21, in25, in26, in24, in27, in28, in17, in18, in31);
DOTP_CONST_PAIR(in18, in17, cospi_12_64, cospi_20_64, in29, in30);
ADD2(in28, in29, in31, in30, in16, in19);
DOTP_CONST_PAIR(in19, in16, cospi_27_64, cospi_5_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 832);
ST_SH(vec4, temp_ptr + 128);
SUB2(in28, in29, in31, in30, in17, in18);
DOTP_CONST_PAIR(in18, in17, cospi_11_64, cospi_21_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 320);
ST_SH(vec4, temp_ptr + 640);
ADD4(in22, in21, in23, in20, in24, in27, in25, in26, in16, in29, in30, in19);
DOTP_CONST_PAIR(-in16, in19, cospi_12_64, cospi_20_64, in28, in31);
SUB2(in29, in28, in30, in31, in16, in19);
DOTP_CONST_PAIR(in19, in16, cospi_19_64, cospi_13_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 576);
ST_SH(vec4, temp_ptr + 384);
ADD2(in29, in28, in30, in31, in17, in18);
DOTP_CONST_PAIR(in18, in17, cospi_3_64, cospi_29_64, vec5, vec4);
FDCT32_POSTPROC_2V_POS_H(vec5, vec4);
ST_SH(vec5, temp_ptr + 64);
ST_SH(vec4, temp_ptr + 896);
}
static void temporal_filter_apply_16size_msa(uint8_t *frame1_ptr,
uint32_t stride,
uint8_t *frame2_ptr,
int32_t strength_in,
int32_t filter_wt_in,
uint32_t *acc, uint16_t *cnt)
{
uint32_t row;
v16i8 frame1_0_b, frame1_1_b, frame2_0_b, frame2_1_b;
v16u8 frame_l, frame_h;
v16i8 zero = { 0 };
v8i16 frame2_0_h, frame2_1_h, mod0_h, mod1_h;
v8i16 diff0, diff1, cnt0, cnt1;
v4i32 const3, const16, filter_wt, strength;
v4i32 mod0_w, mod1_w, mod2_w, mod3_w;
v4i32 diff0_r, diff0_l, diff1_r, diff1_l;
v4i32 frame2_0, frame2_1, frame2_2, frame2_3;
v4i32 acc0, acc1, acc2, acc3;
filter_wt = __msa_fill_w(filter_wt_in);
strength = __msa_fill_w(strength_in);
const3 = __msa_ldi_w(3);
const16 = __msa_ldi_w(16);
for (row = 8; row--;)
{
frame1_0_b = LD_SB(frame1_ptr);
frame2_0_b = LD_SB(frame2_ptr);
frame1_ptr += stride;
frame2_ptr += 16;
frame1_1_b = LD_SB(frame1_ptr);
frame2_1_b = LD_SB(frame2_ptr);
LD_SW2(acc, 4, acc0, acc1);
LD_SW2(acc + 8, 4, acc2, acc3);
LD_SH2(cnt, 8, cnt0, cnt1);
ILVRL_B2_UB(frame1_0_b, frame2_0_b, frame_l, frame_h);
HSUB_UB2_SH(frame_l, frame_h, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
MUL4(mod0_w, const3, mod1_w, const3, mod2_w, const3, mod3_w, const3,
mod0_w, mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < const16);
diff0_l = (mod1_w < const16);
diff1_r = (mod2_w < const16);
diff1_l = (mod3_w < const16);
SUB4(const16, mod0_w, const16, mod1_w, const16, mod2_w, const16, mod3_w,
mod0_w, mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
mod2_w = diff1_r & mod2_w;
mod3_w = diff1_l & mod3_w;
MUL4(mod0_w, filter_wt, mod1_w, filter_wt, mod2_w, filter_wt, mod3_w,
filter_wt, mod0_w, mod1_w, mod2_w, mod3_w);
PCKEV_H2_SH(mod1_w, mod0_w, mod3_w, mod2_w, mod0_h, mod1_h)
ADD2(mod0_h, cnt0, mod1_h, cnt1, mod0_h, mod1_h);
ST_SH2(mod0_h, mod1_h, cnt, 8);
cnt += 16;
ILVRL_B2_SH(zero, frame2_0_b, frame2_0_h, frame2_1_h);
UNPCK_SH_SW(frame2_0_h, frame2_0, frame2_1);
UNPCK_SH_SW(frame2_1_h, frame2_2, frame2_3);
MUL4(mod0_w, frame2_0, mod1_w, frame2_1, mod2_w, frame2_2, mod3_w,
frame2_3, mod0_w, mod1_w, mod2_w, mod3_w);
ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
mod0_w, mod1_w, mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
ST_SW2(mod2_w, mod3_w, acc + 8, 4);
acc += 16;
LD_SW2(acc, 4, acc0, acc1);
LD_SW2(acc + 8, 4, acc2, acc3);
LD_SH2(cnt, 8, cnt0, cnt1);
ILVRL_B2_UB(frame1_1_b, frame2_1_b, frame_l, frame_h);
HSUB_UB2_SH(frame_l, frame_h, diff0, diff1);
UNPCK_SH_SW(diff0, diff0_r, diff0_l);
UNPCK_SH_SW(diff1, diff1_r, diff1_l);
MUL4(diff0_r, diff0_r, diff0_l, diff0_l, diff1_r, diff1_r, diff1_l,
diff1_l, mod0_w, mod1_w, mod2_w, mod3_w);
MUL4(mod0_w, const3, mod1_w, const3, mod2_w, const3, mod3_w, const3,
mod0_w, mod1_w, mod2_w, mod3_w);
SRAR_W4_SW(mod0_w, mod1_w, mod2_w, mod3_w, strength);
diff0_r = (mod0_w < const16);
diff0_l = (mod1_w < const16);
diff1_r = (mod2_w < const16);
diff1_l = (mod3_w < const16);
SUB4(const16, mod0_w, const16, mod1_w, const16, mod2_w, const16, mod3_w,
mod0_w, mod1_w, mod2_w, mod3_w);
mod0_w = diff0_r & mod0_w;
mod1_w = diff0_l & mod1_w;
mod2_w = diff1_r & mod2_w;
mod3_w = diff1_l & mod3_w;
MUL4(mod0_w, filter_wt, mod1_w, filter_wt, mod2_w, filter_wt, mod3_w,
filter_wt, mod0_w, mod1_w, mod2_w, mod3_w);
PCKEV_H2_SH(mod1_w, mod0_w, mod3_w, mod2_w, mod0_h, mod1_h);
ADD2(mod0_h, cnt0, mod1_h, cnt1, mod0_h, mod1_h);
ST_SH2(mod0_h, mod1_h, cnt, 8);
cnt += 16;
UNPCK_UB_SH(frame2_1_b, frame2_0_h, frame2_1_h);
UNPCK_SH_SW(frame2_0_h, frame2_0, frame2_1);
UNPCK_SH_SW(frame2_1_h, frame2_2, frame2_3);
MUL4(mod0_w, frame2_0, mod1_w, frame2_1, mod2_w, frame2_2, mod3_w,
frame2_3, mod0_w, mod1_w, mod2_w, mod3_w);
ADD4(mod0_w, acc0, mod1_w, acc1, mod2_w, acc2, mod3_w, acc3,
mod0_w, mod1_w, mod2_w, mod3_w);
ST_SW2(mod0_w, mod1_w, acc, 4);
ST_SW2(mod2_w, mod3_w, acc + 8, 4);
acc += 16;
frame1_ptr += stride;
frame2_ptr += 16;
}
}
/* routine computes the correctly rounded (to nearest) value of atan(x). */
double
atan (double x)
{
double cor, s1, ss1, s2, ss2, t1, t2, t3, t7, t8, t9, t10, u, u2, u3,
v, vv, w, ww, y, yy, z, zz;
#ifndef DLA_FMS
double t4, t5, t6;
#endif
int i, ux, dx;
static const int pr[M] = { 6, 8, 10, 32 };
number num;
num.d = x;
ux = num.i[HIGH_HALF];
dx = num.i[LOW_HALF];
/* x=NaN */
if (((ux & 0x7ff00000) == 0x7ff00000)
&& (((ux & 0x000fffff) | dx) != 0x00000000))
return x + x;
/* Regular values of x, including denormals +-0 and +-INF */
SET_RESTORE_ROUND (FE_TONEAREST);
u = (x < 0) ? -x : x;
if (u < C)
{
if (u < B)
{
if (u < A)
{
math_check_force_underflow_nonneg (u);
return x;
}
else
{ /* A <= u < B */
v = x * x;
yy = d11.d + v * d13.d;
yy = d9.d + v * yy;
yy = d7.d + v * yy;
yy = d5.d + v * yy;
yy = d3.d + v * yy;
yy *= x * v;
if ((y = x + (yy - U1 * x)) == x + (yy + U1 * x))
return y;
EMULV (x, x, v, vv, t1, t2, t3, t4, t5); /* v+vv=x^2 */
s1 = f17.d + v * f19.d;
s1 = f15.d + v * s1;
s1 = f13.d + v * s1;
s1 = f11.d + v * s1;
s1 *= v;
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (x, 0, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7,
t8);
ADD2 (x, 0, s2, ss2, s1, ss1, t1, t2);
if ((y = s1 + (ss1 - U5 * s1)) == s1 + (ss1 + U5 * s1))
return y;
return atanMp (x, pr);
}
}
else
{ /* B <= u < C */
i = (TWO52 + TWO8 * u) - TWO52;
i -= 16;
z = u - cij[i][0].d;
yy = cij[i][5].d + z * cij[i][6].d;
yy = cij[i][4].d + z * yy;
yy = cij[i][3].d + z * yy;
yy = cij[i][2].d + z * yy;
yy *= z;
t1 = cij[i][1].d;
if (i < 112)
{
if (i < 48)
u2 = U21; /* u < 1/4 */
else
u2 = U22;
} /* 1/4 <= u < 1/2 */
else
{
if (i < 176)
u2 = U23; /* 1/2 <= u < 3/4 */
else
u2 = U24;
} /* 3/4 <= u <= 1 */
if ((y = t1 + (yy - u2 * t1)) == t1 + (yy + u2 * t1))
return __signArctan (x, y);
z = u - hij[i][0].d;
s1 = hij[i][14].d + z * hij[i][15].d;
s1 = hij[i][13].d + z * s1;
s1 = hij[i][12].d + z * s1;
s1 = hij[i][11].d + z * s1;
s1 *= z;
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (z, 0, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, 0, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, 0, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, 0, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
if ((y = s2 + (ss2 - U6 * s2)) == s2 + (ss2 + U6 * s2))
return __signArctan (x, y);
return atanMp (x, pr);
}
}
else
{
if (u < D)
{ /* C <= u < D */
w = 1 / u;
EMULV (w, u, t1, t2, t3, t4, t5, t6, t7);
ww = w * ((1 - t1) - t2);
i = (TWO52 + TWO8 * w) - TWO52;
i -= 16;
z = (w - cij[i][0].d) + ww;
yy = cij[i][5].d + z * cij[i][6].d;
yy = cij[i][4].d + z * yy;
yy = cij[i][3].d + z * yy;
yy = cij[i][2].d + z * yy;
yy = HPI1 - z * yy;
t1 = HPI - cij[i][1].d;
if (i < 112)
u3 = U31; /* w < 1/2 */
else
u3 = U32; /* w >= 1/2 */
if ((y = t1 + (yy - u3)) == t1 + (yy + u3))
return __signArctan (x, y);
DIV2 (1, 0, u, 0, w, ww, t1, t2, t3, t4, t5, t6, t7, t8, t9,
t10);
t1 = w - hij[i][0].d;
EADD (t1, ww, z, zz);
s1 = hij[i][14].d + z * hij[i][15].d;
s1 = hij[i][13].d + z * s1;
s1 = hij[i][12].d + z * s1;
s1 = hij[i][11].d + z * s1;
s1 *= z;
ADD2 (hij[i][9].d, hij[i][10].d, s1, 0, s2, ss2, t1, t2);
MUL2 (z, zz, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][7].d, hij[i][8].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, zz, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][5].d, hij[i][6].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, zz, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][3].d, hij[i][4].d, s1, ss1, s2, ss2, t1, t2);
MUL2 (z, zz, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (hij[i][1].d, hij[i][2].d, s1, ss1, s2, ss2, t1, t2);
SUB2 (HPI, HPI1, s2, ss2, s1, ss1, t1, t2);
if ((y = s1 + (ss1 - U7)) == s1 + (ss1 + U7))
return __signArctan (x, y);
return atanMp (x, pr);
}
else
{
if (u < E)
{ /* D <= u < E */
w = 1 / u;
v = w * w;
EMULV (w, u, t1, t2, t3, t4, t5, t6, t7);
yy = d11.d + v * d13.d;
yy = d9.d + v * yy;
yy = d7.d + v * yy;
yy = d5.d + v * yy;
yy = d3.d + v * yy;
yy *= w * v;
ww = w * ((1 - t1) - t2);
ESUB (HPI, w, t3, cor);
yy = ((HPI1 + cor) - ww) - yy;
if ((y = t3 + (yy - U4)) == t3 + (yy + U4))
return __signArctan (x, y);
DIV2 (1, 0, u, 0, w, ww, t1, t2, t3, t4, t5, t6, t7, t8,
t9, t10);
MUL2 (w, ww, w, ww, v, vv, t1, t2, t3, t4, t5, t6, t7, t8);
s1 = f17.d + v * f19.d;
s1 = f15.d + v * s1;
s1 = f13.d + v * s1;
s1 = f11.d + v * s1;
s1 *= v;
ADD2 (f9.d, ff9.d, s1, 0, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f7.d, ff7.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f5.d, ff5.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (f3.d, ff3.d, s1, ss1, s2, ss2, t1, t2);
MUL2 (v, vv, s2, ss2, s1, ss1, t1, t2, t3, t4, t5, t6, t7, t8);
MUL2 (w, ww, s1, ss1, s2, ss2, t1, t2, t3, t4, t5, t6, t7, t8);
ADD2 (w, ww, s2, ss2, s1, ss1, t1, t2);
SUB2 (HPI, HPI1, s1, ss1, s2, ss2, t1, t2);
if ((y = s2 + (ss2 - U8)) == s2 + (ss2 + U8))
return __signArctan (x, y);
return atanMp (x, pr);
}
else
{
/* u >= E */
if (x > 0)
return HPI;
else
return MHPI;
}
}
}
}