void vp9_fwht4x4_msa(const int16_t *input, int16_t *output,
int32_t src_stride) {
v8i16 in0, in1, in2, in3, in4;
LD_SH4(input, src_stride, in0, in1, in2, in3);
in0 += in1;
in3 -= in2;
in4 = (in0 - in3) >> 1;
SUB2(in4, in1, in4, in2, in1, in2);
in0 -= in2;
in3 += in1;
TRANSPOSE4x4_SH_SH(in0, in2, in3, in1, in0, in2, in3, in1);
in0 += in2;
in1 -= in3;
in4 = (in0 - in1) >> 1;
SUB2(in4, in2, in4, in3, in2, in3);
in0 -= in3;
in1 += in2;
SLLI_4V(in0, in1, in2, in3, 2);
TRANSPOSE4x4_SH_SH(in0, in3, in1, in2, in0, in3, in1, in2);
ST4x2_UB(in0, output, 4);
ST4x2_UB(in3, output + 4, 4);
ST4x2_UB(in1, output + 8, 4);
ST4x2_UB(in2, output + 12, 4);
}
static WEBP_INLINE void PredictLineInverse0(const uint8_t* src,
const uint8_t* pred,
uint8_t* dst, int length) {
v16u8 src0, pred0, dst0;
assert(length >= 0);
while (length >= 32) {
v16u8 src1, pred1, dst1;
LD_UB2(src, 16, src0, src1);
LD_UB2(pred, 16, pred0, pred1);
SUB2(src0, pred0, src1, pred1, dst0, dst1);
ST_UB2(dst0, dst1, dst, 16);
src += 32;
pred += 32;
dst += 32;
length -= 32;
}
if (length > 0) {
int i;
if (length >= 16) {
src0 = LD_UB(src);
pred0 = LD_UB(pred);
dst0 = src0 - pred0;
ST_UB(dst0, dst);
src += 16;
pred += 16;
dst += 16;
length -= 16;
}
for (i = 0; i < length; i++) {
dst[i] = src[i] - pred[i];
}
}
}
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);
}
}
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;
/* 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;
}
}
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;
}
static WEBP_INLINE void TrueMotion16x16(uint8_t* dst, const uint8_t* left,
const uint8_t* top) {
if (left != NULL) {
if (top != NULL) {
int j;
v8i16 d1, d2;
const v16i8 zero = { 0 };
const v8i16 TL = (v8i16)__msa_fill_h(left[-1]);
const v16u8 T = LD_UB(top);
ILVRL_B2_SH(zero, T, d1, d2);
SUB2(d1, TL, d2, TL, d1, d2);
for (j = 0; j < 16; j += 4) {
v16i8 t0, t1, t2, t3;
v8i16 r0, r1, r2, r3, r4, r5, r6, r7;
const v8i16 L0 = (v8i16)__msa_fill_h(left[j + 0]);
const v8i16 L1 = (v8i16)__msa_fill_h(left[j + 1]);
const v8i16 L2 = (v8i16)__msa_fill_h(left[j + 2]);
const v8i16 L3 = (v8i16)__msa_fill_h(left[j + 3]);
ADD4(d1, L0, d1, L1, d1, L2, d1, L3, r0, r1, r2, r3);
ADD4(d2, L0, d2, L1, d2, L2, d2, L3, r4, r5, r6, r7);
CLIP_SH4_0_255(r0, r1, r2, r3);
CLIP_SH4_0_255(r4, r5, r6, r7);
PCKEV_B4_SB(r4, r0, r5, r1, r6, r2, r7, r3, t0, t1, t2, t3);
ST_SB4(t0, t1, t2, t3, dst, BPS);
dst += 4 * BPS;
}
} else {
HorizontalPred16x16(dst, left);
}
} else {
if (top != NULL) {
VerticalPred16x16(dst, top);
} else {
const v16u8 out = (v16u8)__msa_fill_b(0x81);
STORE16x16(out, dst);
}
}
}
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_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_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 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 cave_model (Render* render, Cave* cave, int mode, float minimapoffset)
{
for (int i = 0; i < SEGMENT_COUNT-1; ++i) {
// aid bread-crumb track
if (render->aidtrack && mode == DISPLAYMODE_NORMAL && !(i&1)) {
glDisable(GL_LIGHTING);
glColor4f(0.5,0.5,1,1);
glBegin(GL_LINE_STRIP);
#define CRUMB_SIZE 0.1
glVertex3f(cave->centers[i][0]-CRUMB_SIZE,cave->centers[i][1]-CRUMB_SIZE,cave->centers[i][2]);
glVertex3f(cave->centers[i][0]+CRUMB_SIZE,cave->centers[i][1]+CRUMB_SIZE,cave->centers[i][2]);
glVertex3fv(cave->centers[i]);
glVertex3f(cave->centers[i][0]+CRUMB_SIZE,cave->centers[i][1]-CRUMB_SIZE,cave->centers[i][2]);
glVertex3f(cave->centers[i][0]-CRUMB_SIZE,cave->centers[i][1]+CRUMB_SIZE,cave->centers[i][2]);
glEnd();
}
if (render->lighting)
glEnable(GL_LIGHTING);
int i0 = (cave->i + i)%SEGMENT_COUNT;
if (cave->dirty[i0]) {
for (int mode = 0; mode < DISPLAYMODE_COUNT; ++mode) {
if (glIsList (render->gl_list[mode][i0]))
glDeleteLists (render->gl_list[mode][i0], 1);
render->gl_list[mode][i0] = 0;
}
cave->dirty[i0] = false;
}
if (render->gl_list[mode][i0] == 0) {
int id = render->gl_list[mode][i0] = i0 + render->list_start[mode];
glNewList (id, GL_COMPILE);
int i1 = (i0 + 1)%SEGMENT_COUNT;
if (mode == DISPLAYMODE_NORMAL) {
glBindTexture (GL_TEXTURE_2D,
#ifdef OUTSIDE_TEXTURE_FILE
cave->segs[0][0][2] < ROOM_LEN/2
? render->outside_texture_id :
#endif
render->wall_texture_id
);
glColor4f (1, 1, 1, 0.5);
}
glBegin (GL_QUAD_STRIP);
for (int k = 0; k <= SECTOR_COUNT; ++k) {
int k0 = k%SECTOR_COUNT;
if (mode == DISPLAYMODE_NORMAL) {
glTexCoord2f(
cave->segs[i0][k0][2]/SEGMENT_LEN/SEGMENT_COUNT,
(float)k/SECTOR_COUNT);
}
GLfloat thenormal[] = {0, 0, 0};
SUB2(thenormal, cave->centers[i0], cave->segs[i0][k0]);
NORM(thenormal);
glNormal3fv(thenormal);
glVertex3fv(cave->segs[i0][k0]);
if (mode == DISPLAYMODE_NORMAL) {
glTexCoord2f(
cave->segs[i1][k0][2]/SEGMENT_LEN/SEGMENT_COUNT,
(float)k/SECTOR_COUNT);
}
SUB2(thenormal, cave->centers[i1], cave->segs[i1][k0]);
NORM(thenormal);
glNormal3fv(thenormal);
glVertex3fv(cave->segs[i1][k0]);
}
glEnd();
glEndList();
}
if (mode == DISPLAYMODE_NORMAL) {
glEnable (GL_DEPTH_TEST);
glDisable (GL_BLEND);
glEnable (GL_TEXTURE_2D);
} else {
glDisable (GL_DEPTH_TEST);
glEnable (GL_BLEND);
glDisable (GL_TEXTURE_2D);
glDisable(GL_LIGHTING);
}
if (mode == DISPLAYMODE_MINIMAP) {
// apparently pow is a lot more complicated than what we need here.
#define FASTPOW6(a) a * a * a * a * a * a
float ioffset = minimapoffset - (int)(minimapoffset); // i wonder if this works at all?
float alpha = .12f - .12f * FASTPOW6(((SEGMENT_COUNT / 2.0f - i + ioffset) / SEGMENT_COUNT * 2.0f));
if(i > render->gauge * SEGMENT_COUNT)
glColor4f (1, 1, 1, alpha);
else
glColor4f (
huemap[i][0],
huemap[i][1],
huemap[i][2], alpha);
}
glCallList (render->gl_list[mode][i0]);
}
}
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);
}
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);
}
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 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);
}
/* 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;
}
}
}
}
