static inline A0 asin(const A0& a0)
{
A0 sign, x, z;
// bf::tie(sign, x) = sign_and_abs(a0);
x = nt2::abs(a0);
sign = bitofsign(a0);
if ((x < single_constant<A0,0x38d1b717>())) return a0;
if ((x > One<A0>())) return Nan<A0>();
bool bx_larger_05 = (x > Half<A0>());
if (bx_larger_05)
{
z = Half<A0>()*oneminus(x);
x = sqrt(z);
}
else
{
z = sqr(x);
}
A0 z1 = madd(z, single_constant<A0,0x3d2cb352>(),
single_constant<A0,0x3cc617e3>());
z1 = madd(z1, z, single_constant<A0,0x3d3a3ec7>());
z1 = madd(z1, z, single_constant<A0,0x3d9980f6>());
z1 = madd(z1, z, single_constant<A0,0x3e2aaae4>());
z1 = madd(z1, z*x, x);
if(bx_larger_05)
{
z1 = z1+z1;
z1 = Pio_2<A0>()-z1;
}
return b_xor(z1, sign);
}
// compute the center and radius of the circumscribed circle defined by the three vertices
// i.e. the osculating circle of the three vertices
nv_scalar nv_find_circ_circle( vec3& center, const vec3& v1, const vec3& v2, const vec3& v3)
{
vec3 e0;
vec3 e1;
nv_scalar d1, d2, d3;
nv_scalar c1, c2, c3, oo_c;
sub(e0, v3, v1);
sub(e1, v2, v1);
dot(d1, e0, e1);
sub(e0, v3, v2);
sub(e1, v1, v2);
dot(d2, e0, e1);
sub(e0, v1, v3);
sub(e1, v2, v3);
dot(d3, e0, e1);
c1 = d2 * d3;
c2 = d3 * d1;
c3 = d1 * d2;
oo_c = nv_one / (c1 + c2 + c3);
mult(center,v1,c2 + c3);
madd(center,v2,c3 + c1);
madd(center,v3,c1 + c2);
center *= oo_c * nv_zero_5;
return nv_zero_5 * _sqrt((d1 + d2) * (d2 + d3) * (d3 + d1) * oo_c);
}
// compute the center and radius of the inscribed circle defined by the three vertices
nv_scalar nv_find_in_circle(vec3& center, const vec3& v1, const vec3& v2, const vec3& v3)
{
nv_scalar area = nv_area(v1, v2, v3);
// if the area is null
if (area < nv_eps)
{
center = v1;
return nv_zero;
}
nv_scalar oo_perim = nv_one / nv_perimeter(v1, v2, v3);
vec3 diff;
sub(diff, v2, v3);
mult(center, v1, nv_norm(diff));
sub(diff, v3, v1);
madd(center, v2, nv_norm(diff));
sub(diff, v1, v2);
madd(center, v3, nv_norm(diff));
center *= oo_perim;
return nv_two * area * oo_perim;
}
static inline A0 asin(const A0& a0)
{
A0 sign, x;
// bf::tie(sign, x) = sign_and_abs(a0);
x = abs(a0);
sign = bitofsign(a0);
A0 x_smaller_1e_4 = lt(x, single_constant<A0, 0x38d1b717>()); //1.0e-4f;
A0 x_larger_05 = gt(x, Half<A0>());
A0 x_else = b_or(x_smaller_1e_4, x_larger_05);
A0 a = b_and(x, x_smaller_1e_4);
A0 b = b_and(Half<A0>()*oneminus(x), x_larger_05);
A0 z = b_or(b_or(b_notand(x_else, sqr(x)), a), b);
x = b_notand(x_else, x);
a = b_and(sqrt(z), x_larger_05);
x = b_or(a, x);
A0 z1 = madd(z, single_constant<A0, 0x3d2cb352>(), single_constant<A0, 0x3cc617e3>());
z1 = madd(z1, z, single_constant<A0, 0x3d3a3ec7>());
z1 = madd(z1, z, single_constant<A0, 0x3d9980f6>());
z1 = madd(z1, z, single_constant<A0, 0x3e2aaae4>());
z1 = madd(z1, z*x, x);
z = select(x_smaller_1e_4, z, z1);
z1 = z+z;
z1 = Pio_2<A0>()-z1;
z = select(x_larger_05, z1, z);
return b_xor(z, sign);
}
mstrtoul(MINT *a, char *s, char **p, short int b)
{
MINT y, base;
int c, dectop, alphatop;
short qy;
int i;
mset(0,a);
MSET(b,&base);
y.len = 1;
y.val = &qy;
dectop = (b <= 10) ? '0' + b - 1 : '9';
if (b > 10) alphatop = 'a' + b - 10;
i=0;
while (isxdigit(c=s[i++])) {
if (isupper(c)) c = c - 'A' + 'a';
if (c >= '0' && c <= dectop) {
qy = c - '0';
mmult(a,&base,a);
if (qy != 0) madd(a,&y,a);
continue;
} if (b > 10 && (c >= 'a' && c <= alphatop)) {
qy = c - 'a' + 10;
mmult(a,&base,a);
madd(a,&y,a);
continue;
}
};
if (p!=NULL) (*p)=(char *)s+i-1;
}
static inline A0 kernel_atan(const A0& a0)
{
if (is_eqz(a0)) return Zero<A0>();
if (is_inf(a0)) return Pio_2<A0>();
A0 x = nt2::abs(a0);
A0 y;
if( x >single_constant<A0,0x401a827a>())//2.414213562373095 ) /* tan 3pi/8 */
{
y = Pio_2<A0>();
x = -rec(x);
}
else if( x > single_constant<A0,0x3ed413cd>()) //0.4142135623730950f ) /* tan pi/8 */
{
y = Pio_4<A0>();
x = minusone(x)/oneplus(x);
}
else
y = 0.0;
A0 z = sqr(x);
A0 z1 = madd(z, single_constant<A0,0x3da4f0d1>(),single_constant<A0,0xbe0e1b85>());
A0 z2 = madd(z, single_constant<A0,0x3e4c925f>(),single_constant<A0,0xbeaaaa2a>());
z1 = madd(z1, sqr(z), z2);
return add(y, madd(x, mul( z1, z), x));
// y +=
// ((( 8.05374449538e-2 * z
// - 1.38776856032E-1) * z
// + 1.99777106478E-1) * z
// - 3.33329491539E-1) * z * x
// + x;
}
static inline A0_n asin(const A0_n a0_n)
{
const A0 a0 = { a0_n };
A0 sign, x;
x = nt2::abs(a0);
sign = bitofsign(a0);
const bA0 x_smaller_1e_4 = lt(x, single_constant<A0, 0x38d1b717>()); //1.0e-4f;
const bA0 x_larger_05 = gt(x, Half<A0>());
const bA0 x_else = logical_or(x_smaller_1e_4, x_larger_05);
A0 a = if_else_zero(x_smaller_1e_4, x);
const A0 b = if_else_zero(x_larger_05, Half<A0>()*oneminus(x));
A0 z = b_or(b_or(if_zero_else(x_else, sqr(x)), a), b);
x = if_zero_else(x_else, x);
a = if_else_zero(x_larger_05, sqrt(z));
x = b_or(a, x);
A0 z1 = madd(z, single_constant<A0, 0x3d2cb352>(), single_constant<A0, 0x3cc617e3>());
z1 = madd(z1, z, single_constant<A0, 0x3d3a3ec7>());
z1 = madd(z1, z, single_constant<A0, 0x3d9980f6>());
z1 = madd(z1, z, single_constant<A0, 0x3e2aaae4>());
z1 = madd(z1, z*x, x);
z = select(x_smaller_1e_4, z, z1);
z1 = z+z;
z1 = Pio_2<A0>()-z1;
z = select(x_larger_05, z1, z);
return b_xor(z, sign);
}
void QuadMesh::interpolate(const RTCInterpolateArguments* const args)
{
unsigned int primID = args->primID;
float u = args->u;
float v = args->v;
RTCBufferType bufferType = args->bufferType;
unsigned int bufferSlot = args->bufferSlot;
float* P = args->P;
float* dPdu = args->dPdu;
float* dPdv = args->dPdv;
float* ddPdudu = args->ddPdudu;
float* ddPdvdv = args->ddPdvdv;
float* ddPdudv = args->ddPdudv;
unsigned int valueCount = args->valueCount;
/* calculate base pointer and stride */
assert((bufferType == RTC_BUFFER_TYPE_VERTEX && bufferSlot < numTimeSteps) ||
(bufferType == RTC_BUFFER_TYPE_VERTEX_ATTRIBUTE && bufferSlot <= vertexAttribs.size()));
const char* src = nullptr;
size_t stride = 0;
if (bufferType == RTC_BUFFER_TYPE_VERTEX_ATTRIBUTE) {
src = vertexAttribs[bufferSlot].getPtr();
stride = vertexAttribs[bufferSlot].getStride();
} else {
src = vertices[bufferSlot].getPtr();
stride = vertices[bufferSlot].getStride();
}
for (unsigned int i=0; i<valueCount; i+=4)
{
const vbool4 valid = vint4((int)i)+vint4(step) < vint4(int(valueCount));
const size_t ofs = i*sizeof(float);
const Quad& tri = quad(primID);
const vfloat4 p0 = vfloat4::loadu(valid,(float*)&src[tri.v[0]*stride+ofs]);
const vfloat4 p1 = vfloat4::loadu(valid,(float*)&src[tri.v[1]*stride+ofs]);
const vfloat4 p2 = vfloat4::loadu(valid,(float*)&src[tri.v[2]*stride+ofs]);
const vfloat4 p3 = vfloat4::loadu(valid,(float*)&src[tri.v[3]*stride+ofs]);
const vbool4 left = u+v <= 1.0f;
const vfloat4 Q0 = select(left,p0,p2);
const vfloat4 Q1 = select(left,p1,p3);
const vfloat4 Q2 = select(left,p3,p1);
const vfloat4 U = select(left,u,vfloat4(1.0f)-u);
const vfloat4 V = select(left,v,vfloat4(1.0f)-v);
const vfloat4 W = 1.0f-U-V;
if (P) {
vfloat4::storeu(valid,P+i,madd(W,Q0,madd(U,Q1,V*Q2)));
}
if (dPdu) {
assert(dPdu); vfloat4::storeu(valid,dPdu+i,select(left,Q1-Q0,Q0-Q1));
assert(dPdv); vfloat4::storeu(valid,dPdv+i,select(left,Q2-Q0,Q0-Q2));
}
if (ddPdudu) {
assert(ddPdudu); vfloat4::storeu(valid,ddPdudu+i,vfloat4(zero));
assert(ddPdvdv); vfloat4::storeu(valid,ddPdvdv+i,vfloat4(zero));
assert(ddPdudv); vfloat4::storeu(valid,ddPdudv+i,vfloat4(zero));
}
}
}
static inline A0 approx(const A0& x)
{
const A0 x2 = sqr(x);
A0 y1 = madd(Const<A0,0x3e5345fd>(), x2,Const<A0,0x3f95eceb>());
A0 y2 = madd(Const<A0,0x3f0ac229>(), x2,Const<A0,0x400237b4>());
y1 = madd(y1, x2, Const<A0,0x4029a924>());
y2 = madd(y2, x2, Const<A0,0x40135d8e>());
return oneplus(x*madd(x, y1, y2));
}
static inline A0 log2(const A0& a0)
{
A0 x, fe, x2, y;
kernel_log(a0, fe, x, x2, y);
y = madd(Mhalf<A0>(),x2, y);
// multiply log of fraction by log2(e)
A0 z = madd(x,single_constant<A0, 0x3ee2a8ed>(),mul(y,single_constant<A0, 0x3ee2a8ed>()));// 0.44269504088896340735992
A0 z1 = ((z+y)+x)+fe;
A0 y1 = a0-rec(abs(a0)); // trick to reduce selection testing
return seladd(is_inf(y1),b_or(z1, b_or(is_ltz(a0), is_nan(a0))),y1);
}
static inline A0 log(const A0& a0)
{
A0 x, fe, x2, y;
kernel_log(a0, fe, x, x2, y);
y = madd(fe, single_constant<A0, 0xb95e8083>(), y);
y = madd(Mhalf<A0>(), x2, y);
A0 z = x + y;
A0 y1 = a0-rec(abs(a0));// trick to reduce selection testing
A0 y2 = madd(single_constant<A0, 0x3f318000>(), fe, z);
y2 = if_nan_else(logical_or(is_ltz(a0), is_nan(a0)), y2);
return seladd(is_inf(y1), y2, y1);
}
static inline A0 log(const A0& a0)
{
typedef typename meta::strip<A0>::type stA0;
if (a0 == Inf<stA0>()) return a0;
if (is_eqz(a0)) return Minf<stA0>();
if (nt2::is_nan(a0)||is_ltz(a0)) return Nan<stA0>();
A0 x, fe, x2, y;
kernel_log(a0, fe, x, x2, y);
y = madd(fe, single_constant<stA0, 0xb95e8083>(), y);
y = madd(Mhalf<stA0>(), x2, y);
A0 z = x + y;
return madd(single_constant<stA0, 0x3f318000>(), fe, z);
}
void QuadMesh::interpolate(unsigned primID, float u, float v, RTCBufferType buffer, float* P, float* dPdu, float* dPdv, float* ddPdudu, float* ddPdvdv, float* ddPdudv, size_t numFloats)
{
/* test if interpolation is enabled */
#if defined(DEBUG)
if ((scene->aflags & RTC_INTERPOLATE) == 0)
throw_RTCError(RTC_INVALID_OPERATION,"rtcInterpolate can only get called when RTC_INTERPOLATE is enabled for the scene");
#endif
/* calculate base pointer and stride */
assert((buffer >= RTC_VERTEX_BUFFER0 && buffer < RTCBufferType(RTC_VERTEX_BUFFER0 + numTimeSteps)) ||
(buffer >= RTC_USER_VERTEX_BUFFER0 && buffer <= RTC_USER_VERTEX_BUFFER1));
const char* src = nullptr;
size_t stride = 0;
if (buffer >= RTC_USER_VERTEX_BUFFER0) {
src = userbuffers[buffer&0xFFFF].getPtr();
stride = userbuffers[buffer&0xFFFF].getStride();
} else {
src = vertices[buffer&0xFFFF].getPtr();
stride = vertices[buffer&0xFFFF].getStride();
}
for (size_t i=0; i<numFloats; i+=VSIZEX)
{
const vboolx valid = vintx((int)i)+vintx(step) < vintx(int(numFloats));
const size_t ofs = i*sizeof(float);
const Quad& tri = quad(primID);
const vfloatx p0 = vfloatx::loadu(valid,(float*)&src[tri.v[0]*stride+ofs]);
const vfloatx p1 = vfloatx::loadu(valid,(float*)&src[tri.v[1]*stride+ofs]);
const vfloatx p2 = vfloatx::loadu(valid,(float*)&src[tri.v[2]*stride+ofs]);
const vfloatx p3 = vfloatx::loadu(valid,(float*)&src[tri.v[3]*stride+ofs]);
const vboolx left = u+v <= 1.0f;
const vfloatx Q0 = select(left,p0,p2);
const vfloatx Q1 = select(left,p1,p3);
const vfloatx Q2 = select(left,p3,p1);
const vfloatx U = select(left,u,vfloatx(1.0f)-u);
const vfloatx V = select(left,v,vfloatx(1.0f)-v);
const vfloatx W = 1.0f-U-V;
if (P) {
vfloatx::storeu(valid,P+i,madd(W,Q0,madd(U,Q1,V*Q2)));
}
if (dPdu) {
assert(dPdu); vfloatx::storeu(valid,dPdu+i,select(left,Q1-Q0,Q0-Q1));
assert(dPdv); vfloatx::storeu(valid,dPdv+i,select(left,Q2-Q0,Q0-Q2));
}
if (ddPdudu) {
assert(ddPdudu); vfloatx::storeu(valid,ddPdudu+i,vfloatx(zero));
assert(ddPdvdv); vfloatx::storeu(valid,ddPdvdv+i,vfloatx(zero));
assert(ddPdudv); vfloatx::storeu(valid,ddPdudv+i,vfloatx(zero));
}
}
}
static inline void kernel_log(const A0& a0,
A0& fe,
A0& x,
A0& x2,
A0& y)
{
typedef typename meta::as_integer<A0, signed>::type int_type;
int_type e;
boost::fusion::tie(x, e) = fast_frexp(a0);
// std::cout << "x " << x << " e " << e << std::endl;
// bf::tie(x, e) = frexp(a0);
// std::cout << "x " << x << " e " << e << std::endl;
int_type x_lt_sqrthf = nt2::simd::native_cast<int_type>(gt(single_constant<A0, 0x3f3504f3>(),x));
e = e+x_lt_sqrthf;
x = x+b_and(x, x_lt_sqrthf)+single_constant<A0, 0xbf800000>();
x2 = sqr(x);
A0 y1 = madd(single_constant<A0, 0x3d9021bb>() ,x2,single_constant<A0, 0x3def251a>() );
A0 y2 = madd(single_constant<A0, 0xbdebd1b8>() ,x2,single_constant<A0, 0xbdfe5d4f>() );
y1 = madd(y1,x2,single_constant<A0, 0x3e11e9bf>() );
y2 = madd(y2,x2,single_constant<A0, 0xbe2aae50>() );
y1 = madd(y1,x2,single_constant<A0, 0x3e4cceac>() );
y2 = madd(y2,x2,single_constant<A0, 0xbe7ffffc>() );
y1 = madd(y1,x2,single_constant<A0, 0x3eaaaaaa>() );
y = madd(x,y2,y1)*x*x2;
fe = tofloat(e);
}
static inline void kernel_log(const A0& a0,
A0& fe,
A0& x,
A0& x2,
A0& y)
{
typedef typename meta::as_integer<A0, signed>::type int_type;
typedef typename meta::strip<A0>::type stA0;
int_type e;
boost::fusion::vector_tie(x, e) = fast_frexp(a0);
int_type x_lt_sqrthf = -(single_constant<stA0, 0x3f3504f3>() > x);
e += x_lt_sqrthf;
// if (x_lt_sqrthf) x+= x;
// x += single_constant<A0, 0xbf800000>();
x += b_and(x, x_lt_sqrthf)+single_constant<stA0,0xbf800000>();
x2 = sqr(x);
A0 y1 = madd(single_constant<stA0, 0x3d9021bb>() ,x2,single_constant<stA0, 0x3def251a>() );
A0 y2 = madd(single_constant<stA0, 0xbdebd1b8>() ,x2,single_constant<stA0, 0xbdfe5d4f>() );
y1 = madd(y1,x2,single_constant<stA0, 0x3e11e9bf>() );
y2 = madd(y2,x2,single_constant<stA0, 0xbe2aae50>() );
y1 = madd(y1,x2,single_constant<stA0, 0x3e4cceac>() );
y2 = madd(y2,x2,single_constant<stA0, 0xbe7ffffc>() );
y1 = madd(y1,x2,single_constant<stA0, 0x3eaaaaaa>() );
y = madd(x,y2,y1)*x*x2;
fe = tofloat(e);
}
int
main (int argc, char *argv[])
{
MINT *a, *b, *c, *d;
short h;
mp_set_memory_functions (NULL, NULL, NULL);
a = itom (123);
b = xtom ("DEADBEEF");
c = itom (0);
d = itom (0);
move (a, b);
madd (a, b, c);
msub (a, b, c);
mult (a, b, c);
mdiv (b, a, c, d);
sdiv (b, 2, c, &h);
msqrt (a, c, d);
pow (b, a, a, c);
rpow (a, 3, c);
gcd (a, b, c);
mcmp (a, b);
if (argc > 1)
{
min (c);
mout (a);
}
mtox (b);
mfree(a);
exit (0);
}
static inline A0 log2(const A0& a0)
{
typedef typename meta::strip<A0>::type stA0;
if (a0 == Inf<stA0>()) return a0;
if (is_eqz(a0)) return Minf<stA0>();
if (nt2::is_nan(a0)||is_ltz(a0)) return Nan<stA0>();
A0 x, fe, x2, y;
kernel_log(a0, fe, x, x2, y);
y = madd(Mhalf<stA0>(),x2, y);
// multiply log of fraction by log2(e)
A0 z = madd( x
, single_constant<stA0, 0x3ee2a8ed>()
, mul(y,single_constant<stA0, 0x3ee2a8ed>())// 0.44269504088896340735992
);
return ((z+y)+x)+fe;
}
vec3 & reflect(vec3& r, const vec3& n, const vec3& l)
{
nv_scalar n_dot_l;
n_dot_l = nv_two * dot(n_dot_l,n,l);
mult(r,l,-nv_one);
madd(r,n,n_dot_l);
return r;
}
static inline A0 log(const A0& a0)
{
A0 x, fe, x2, y;
kernel_log(a0, fe, x, x2, y);
y = madd(fe, single_constant<A0, 0xb95e8083>(), y);
y = madd(Mhalf<A0>(), x2, y);
A0 z = x + y;
// std::cout << "fe " << fe << std::endl;
// std::cout << "z " << z << std::endl;
// std::cout << "a0 " << a0 << std::endl;
// std::cout << "rec(a0) " << rec(a0) << std::endl;
A0 y1 = a0-rec(abs(a0));// trick to reduce selection testing
A0 y2 = madd(single_constant<A0, 0x3f318000>(), fe, z);
// std::cout << "y1 " << y1 << std::endl;
// std::cout << "y2 " << y2 << std::endl;
return seladd(is_inf(y1),b_or(y2, b_or(is_ltz(a0), is_nan(a0))),y1);
}
int main(int argc, const char * argv[]) {
testcout();
double c = madd(1.0, 2.0);
printf("1.0 + 2.0 = %f\n", c);
return 0;
}
static inline A0_n kernel_atan(const A0_n a0_n)
{
const A0 a0 = {a0_n};
const A0 x = nt2::abs(a0);
//here x is positive
const bA0 flag1 = lt(x, single_constant<A0, 0x401a827a>()); //tan3pio8);
const bA0 flag2 = logical_and(ge(x, single_constant<A0, 0x3ed413cd>()), flag1);
A0 yy = if_zero_else(flag1, Pio_2<A0>());
yy = select(flag2, Pio_4<A0>(), yy);
A0 xx = select(flag1, x, -rec(x));
xx = select(flag2, (minusone(x)/oneplus(x)),xx);
const A0 z = sqr(xx);
A0 z1 = madd(z, single_constant<A0, 0x3da4f0d1>(),single_constant<A0, 0xbe0e1b85>());
const A0 z2 = madd(z, single_constant<A0, 0x3e4c925f>(),single_constant<A0, 0xbeaaaa2a>());
z1 = madd(z1, sqr(z), z2);
return add(yy, madd(xx, mul( z1, z), xx));
}
static inline A0 atan(const A0& a0)
{
A0 x, sign;
x = nt2::abs(a0);
sign = bitofsign(a0);
// bf::tie(sign, x) = sign_and_abs(a0);
const A0 flag1 = lt(x, single_constant<A0, 0x401a827a>()); //tan3pio8);
const A0 flag2 = b_and(ge(x, single_constant<A0, 0x3ed413cd>()), flag1);
A0 yy = b_notand(flag1, Pio_2<A0>());
yy = select(flag2, Pio_4<A0>(), yy);
A0 xx = select(flag1, x, -rec(x));
xx = select(flag2, (minusone(x)/oneplus(x)),xx);
const A0 z = sqr(xx);
A0 z1 = madd(z, single_constant<A0, 0x3da4f0d1>(),single_constant<A0, 0xbe0e1b85>());
A0 z2 = madd(z, single_constant<A0, 0x3e4c925f>(),single_constant<A0, 0xbeaaaa2a>());
z1 = madd(z1, sqr(z), z2);
yy = add(yy, madd(xx, mul( z1, z), xx));
return b_xor(yy, sign);
}
static inline A0 sin_eval(const A0& z, const A0& x)
{
const A0 y1 = horner< NT2_HORNER_COEFF_T(stype, 6, (0x3de5d8fd1fcf0ec1ll,
0xbe5ae5e5a9291691ll,
0x3ec71de3567d4896ll,
0xbf2a01a019bfdf03ll,
0x3f8111111110f7d0ll,
0xbfc5555555555548ll) ) > (z);
return madd(mul(y1,z),x,x);
}
void mpolsub(MPOL *p, MPOL *q, MPOL *r)
{
register ip=0,iq=0,is=0;
MPOL s;
POL_ALLOC(&s,p->nterms + q->nterms);
while ((ip<p->nterms)&&(iq<q->nterms)) {
#if (! INTR)
(*PollPtr)();
#endif
switch ((*cmp_exp)(MEXPO(p,ip),MEXPO(q,iq))) {
case 1 : expocopy(MEXPO(p,ip),MEXPO(&s,is));
MCOPY(&(p->coefs[ip]),&(s.coefs[is]));
ip++;is++;break;
case -1 : expocopy(MEXPO(q,iq),MEXPO(&s,is));
MCOPY(&(q->coefs[iq]),&(s.coefs[is]));
mnegate(&(s.coefs[is]));
iq++;is++;break;
case 0 : MCOPY(&(q->coefs[iq]),&(s.coefs[is]));
mnegate(&(s.coefs[is]));
madd(&(p->coefs[ip]),&(s.coefs[is]),&(s.coefs[is]));
if (mtest(&(s.coefs[is]))) {
expocopy(MEXPO(p,ip),MEXPO(&s,is));
is++;
};
ip++;iq++;
};
};
while (ip<p->nterms) {
#if (! INTR)
(*PollPtr)();
#endif
expocopy(MEXPO(p,ip),MEXPO(&s,is));
MCOPY(&(p->coefs[ip]),&(s.coefs[is]));
ip++; is++;
};
while (iq<q->nterms) {
#if (! INTR)
(*PollPtr)();
#endif
expocopy(MEXPO(q,iq),MEXPO(&s,is));
MCOPY(&(q->coefs[iq]),&(s.coefs[is]));
mnegate(&(s.coefs[is]));
iq++; is++;
};
s.nterms = is;
if (is==0){
xfree((char *)s.coefs);
xfree((char *)s.expos);
};
mpolfree(r);
MPOLMOVEFREE(&s,r);
};
always_inline VecType vec_reciprocal_newton(VecType arg)
{
const VecType one = VecType::gen_one();
const VecType approx = fast_reciprocal(arg);
// One round of Newton-Raphson refinement
const VecType diff = one - approx * arg;
const VecType result = madd(diff, approx, approx);
return result;
}
static inline A0 asin(const A0& a0)
{
A0 x = nt2::abs(a0);
if ((x > One<A0>())) return Nan<A0>();
if ((x < Sqrteps<A0>())) return a0;
A0 zz;
if((x > double_constant<double,0x3fe4000000000000ll> ())) //0.625;
{
zz = oneminus(x);
const A0 vp = zz*horner< NT2_HORNER_COEFF_T(stype, 5,
(0x3f684fc3988e9f08ll,
0xbfe2079259f9290fll,
0x401bdff5baf33e6all,
0xc03991aaac01ab68ll,
0x403c896240f3081dll)
)>(zz)/
horner< NT2_HORNER_COEFF_T(stype, 5,
(0x3ff0000000000000ll,
0xc035f2a2b6bf5d8cll,
0x40626219af6a7f42ll,
0xc077fe08959063eell,
0x40756709b0b644bell)
)>(zz);
zz = sqrt(zz+zz);
A0 z = Pio_4<A0>()-zz;
zz = madd(zz, vp, double_constant<double,0xbc91a62633145c07ll>());
z = z-zz;
zz = z+Pio_4<A0>();
}
else
{
zz = sqr(x);
A0 z = zz*horner< NT2_HORNER_COEFF_T(stype, 6,
(0x3f716b9b0bd48ad3ll,
0xbfe34341333e5c16ll,
0x4015c74b178a2dd9ll,
0xc0304331de27907bll,
0x40339007da779259ll,
0xc020656c06ceafd5ll)
)>(zz)/
horner< NT2_HORNER_COEFF_T(stype, 6,
(0x3ff0000000000000ll,
0xc02d7b590b5e0eabll,
0x40519fc025fe9054ll,
0xc06265bb6d3576d7ll,
0x4061705684ffbf9dll,
0xc04898220a3607acll)
)>(zz);
zz = x*z+x;
}
return b_xor(bitofsign(a0), zz);
}
void main()
{
int a[M]={1,2,3,4,5,6,7,8,9,10};
int b[M]={1,1,1,1,1,1,1,1,1,1};
int c1[N][N],c2[N][N];
madd(a,b,c1);
mult(a,b,c2);
printf("a¾ØÕó:\n");disp1(a);
printf("b¾ØÕó:\n");disp1(b);
printf("a+b:\n");disp2(c1);
printf("a¡Áb:\n");disp2(c2);
printf("\n");
}
/* thread function */
void* threadfunc(void* arg) {
long long iterations = (long long)arg;
long long i;
if (sync == 'm') {
for (i = 0; i < iterations; i++) {
madd(&counter, 1);
}
for (i = 0; i < iterations; i++) {
madd(&counter, -1);
}
}
else if (sync == 's') {
for (i = 0; i < iterations; i++) {
sadd(&counter, 1);
}
for (i = 0; i < iterations; i++) {
sadd(&counter, -1);
}
}
else if (sync == 'c') {
for (i = 0; i < iterations; i++) {
cadd(&counter, 1);
}
for (i = 0; i < iterations; i++) {
cadd(&counter, -1);
}
}
else {
for (i = 0; i < iterations; i++) {
add(&counter, 1);
}
for (i = 0; i < iterations; i++) {
add(&counter, -1);
}
}
}
FN minvert(MINT *a, MINT *b, MINT *c)
{ MINT x, y, z, w, Anew, Aold;
int i = 0;
static MINT one;
static int oneinit = 1;
if (oneinit) {
oneinit = 0;
MSET(1,&one);
}
MINIT(&x);
MINIT(&y);
MINIT(&z);
MINIT(&w);
MINIT(&Aold);
MSET (1,&Anew);
mcopy(b, &x);
mcopy(a, &y);
/*
* Loop invariant:
*
* y = -1^i * Anew * a mod b
*/
while(mtest(&y) != 0)
{ mdiv(&x, &y, &w, &z);
mcopy(&Anew, &x);
mmult(&w, &Anew, &Anew);
madd(&Anew, &Aold, &Anew);
mmove(&x, &Aold);
mmove(&y, &x);
mmove(&z, &y);
i++;
}
if (mcmp(&one,&x)) {
mcopy(&one,c);
} else {
mmove(&Aold, c);
if( (i&01) == 0) msub(b, c, c);
}
MFREE(&x);
MFREE(&y);
MFREE(&z);
MFREE(&w);
MFREE(&Aold);
MFREE(&Anew);
}
static inline A0 atan(const A0& a0)
{
// static const A0 tanpio8 = double_constant<double, 0x3fda827999fcef31ll>();
if (is_eqz(a0)) return a0;
if (is_inf(a0)) return Pio_2<A0>()*sign(a0);
A0 x = nt2::abs(a0);
A0 y;
A0 flag = (x > double_constant<double,0x4003504f333f9de6ll>());
if (flag)
{
y = Pio_2<A0>();
x = -rec(x);
}
else if ((x <= double_constant<double,0x3fe51eb851eb851fll>()))
{
y = Zero<A0>();
}
else
{
y = Pio_4<A0>();
flag = Half<A0>();
x = minusone(x)/oneplus(x);
}
A0 z = sqr(x);
z = z*horner< NT2_HORNER_COEFF_T(stype, 5,
(0xbfec007fa1f72594ll,
0xc03028545b6b807all,
0xc052c08c36880273ll,
0xc05eb8bf2d05ba25ll,
0xc0503669fd28ec8ell)
)>(z)/
horner< NT2_HORNER_COEFF_T(stype, 6,
(0x3ff0000000000000ll,
0x4038dbc45b14603cll,
0x4064a0dd43b8fa25ll,
0x407b0e18d2e2be3bll,
0x407e563f13b049eall,
0x4068519efbbd62ecll)
)>(z);
z = madd(x, z, x);
static const A0 morebits = double_constant<double,0x3c91a62633145c07ll>();
z += flag * morebits;
y = y + z;
if( is_ltz(a0) ) y = -y;
return(y);
}
