From 52a48bfcd881602f1e84d39dd72b48fb93436e48 Mon Sep 17 00:00:00 2001 From: =?utf8?q?=C3=98yvind=20Kol=C3=A5s?= Date: Thu, 31 Aug 2017 06:19:09 +0200 Subject: [PATCH] babl: move trc functions inside babl-trc.c Use lolremez 9 degree polynominal approximations for 1.8 and 2.2 gamma. --- babl/babl-fish-path.c | 4 +- babl/babl-trc.c | 486 ++++++++++++++++++++++++++++++++++++++++++ babl/babl-trc.h | 372 -------------------------------- 3 files changed, 488 insertions(+), 374 deletions(-) diff --git a/babl/babl-fish-path.c b/babl/babl-fish-path.c index 4fe5ded..03e002f 100644 --- a/babl/babl-fish-path.c +++ b/babl/babl-fish-path.c @@ -431,7 +431,7 @@ static void prep_conversion (const Babl *babl) lut = matrixf + 9; for (i = 0; i < 256; i++) { - lut[i] = _babl_trc_to_linear (source_space->space.trc[0], i/255.0); + lut[i] = babl_trc_to_linear (source_space->space.trc[0], i/255.0); // XXX: should have green and blue luts as well } } @@ -698,7 +698,7 @@ babl_fish_path (const Babl *source, if (!done) { - if(0)babl_conversion_class_for_each (show_item, (void*)source->format.space); + if(1)babl_conversion_class_for_each (show_item, (void*)source->format.space); //babl_format_class_for_each (show_fmt, NULL); //babl_model_class_for_each (show_fmt, NULL); } diff --git a/babl/babl-trc.c b/babl/babl-trc.c index 3b32f74..6de0fd1 100644 --- a/babl/babl-trc.c +++ b/babl/babl-trc.c @@ -24,6 +24,492 @@ static BablTRC trc_db[MAX_TRCS]; + +static inline float babl_trc_lut_from_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + int entry = value * trc->lut_size + 0.5; + float ret = trc->inv_lut[ + (entry >= 0 && entry < trc->lut_size) ? + entry : + trc->lut_size-1]; + /* XXX: fixme, do linear interpolation */ + return ret; +} + +static inline float babl_trc_lut_to_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + int entry = value * trc->lut_size + 0.5; + float ret = trc->lut[ + (entry >= 0 && entry < trc->lut_size) ? + entry : + trc->lut_size-1]; + /* XXX: fixme, do linear interpolation */ + return ret; +} + +static inline float _babl_trc_linear (const Babl *trc_, float value) +{ + return 1.0; +} + +/* origin: FreeBSD /usr/src/lib/msun/src/e_powf.c, copied from musl */ +/* + * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. + */ +/* + * ==================================================== + * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. + * + * Developed at SunPro, a Sun Microsystems, Inc. business. + * Permission to use, copy, modify, and distribute this + * software is freely granted, provided that this notice + * is preserved. + * ==================================================== + */ + +//#include "libm.h" + +/* Get a 32 bit int from a float. */ +#define GET_FLOAT_WORD(w,d) \ +do { \ + union {float f; uint32_t i;} __u; \ + __u.f = (d); \ + (w) = __u.i; \ +} while (0) + +/* Set a float from a 32 bit int. */ +#define SET_FLOAT_WORD(d,w) \ +do { \ + union {float f; uint32_t i;} __u; \ + __u.i = (w); \ + (d) = __u.f; \ +} while (0) + + +static const float +bp[] = {1.0, 1.5,}, +dp_h[] = { 0.0, 5.84960938e-01,}, /* 0x3f15c000 */ +dp_l[] = { 0.0, 1.56322085e-06,}, /* 0x35d1cfdc */ +two24 = 16777216.0, /* 0x4b800000 */ +huge = 1.0e30, +tiny = 1.0e-30, +/* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */ +L1 = 6.0000002384e-01, /* 0x3f19999a */ +L2 = 4.2857143283e-01, /* 0x3edb6db7 */ +L3 = 3.3333334327e-01, /* 0x3eaaaaab */ +L4 = 2.7272811532e-01, /* 0x3e8ba305 */ +L5 = 2.3066075146e-01, /* 0x3e6c3255 */ +L6 = 2.0697501302e-01, /* 0x3e53f142 */ +P1 = 1.6666667163e-01, /* 0x3e2aaaab */ +P2 = -2.7777778450e-03, /* 0xbb360b61 */ +P3 = 6.6137559770e-05, /* 0x388ab355 */ +P4 = -1.6533901999e-06, /* 0xb5ddea0e */ +P5 = 4.1381369442e-08, /* 0x3331bb4c */ +lg2 = 6.9314718246e-01, /* 0x3f317218 */ +lg2_h = 6.93145752e-01, /* 0x3f317200 */ +lg2_l = 1.42860654e-06, /* 0x35bfbe8c */ +ovt = 4.2995665694e-08, /* -(128-log2(ovfl+.5ulp)) */ +cp = 9.6179670095e-01, /* 0x3f76384f =2/(3ln2) */ +cp_h = 9.6191406250e-01, /* 0x3f764000 =12b cp */ +cp_l = -1.1736857402e-04; /* 0xb8f623c6 =tail of cp_h */ +#if 0 +ivln2 = 1.4426950216e+00, /* 0x3fb8aa3b =1/ln2 */ +ivln2_h = 1.4426879883e+00, /* 0x3fb8aa00 =16b 1/ln2*/ +ivln2_l = 7.0526075433e-06; /* 0x36eca570 =1/ln2 tail*/ +#endif + +static inline float obabl_powf(float x, float y) +{ + float z,ax,z_h,z_l,p_h,p_l; + float y1,t1,t2,r,s,sn,t,u,v,w; + int32_t i,j,k,yisint,n; + int32_t hx,hy,ix,iy,is; + + GET_FLOAT_WORD(hx, x); + GET_FLOAT_WORD(hy, y); + ix = hx & 0x7fffffff; + iy = hy & 0x7fffffff; + +#if 0 + /* x**0 = 1, even if x is NaN */ + if (iy == 0) + return 1.0f; +#endif + /* 1**y = 1, even if y is NaN */ + if (hx == 0x3f800000) + return 1.0f; + /* NaN if either arg is NaN */ + if (ix > 0x7f800000 || iy > 0x7f800000) + return x + y; + + /* determine if y is an odd int when x < 0 + * yisint = 0 ... y is not an integer + * yisint = 1 ... y is an odd int + * yisint = 2 ... y is an even int + */ + yisint = 0; + if (hx < 0) { + if (iy >= 0x4b800000) + yisint = 2; /* even integer y */ + else if (iy >= 0x3f800000) { + k = (iy>>23) - 0x7f; /* exponent */ + j = iy>>(23-k); + if ((j<<(23-k)) == iy) + yisint = 2 - (j & 1); + } + } +#if 0 + /* special value of y */ + if (iy == 0x7f800000) { /* y is +-inf */ + if (ix == 0x3f800000) /* (-1)**+-inf is 1 */ + return 1.0f; + else if (ix > 0x3f800000) /* (|x|>1)**+-inf = inf,0 */ + return hy >= 0 ? y : 0.0f; + else /* (|x|<1)**+-inf = 0,inf */ + return hy >= 0 ? 0.0f: -y; + } + if (iy == 0x3f800000) /* y is +-1 */ + return hy >= 0 ? x : 1.0f/x; + if (hy == 0x40000000) /* y is 2 */ + return x*x; + if (hy == 0x3f000000) { /* y is 0.5 */ + if (hx >= 0) /* x >= +0 */ + return sqrtf(x); + } +#endif + + ax = fabsf(x); + /* special value of x */ + if (ix == 0x7f800000 || ix == 0 || ix == 0x3f800000) { /* x is +-0,+-inf,+-1 */ + z = ax; + if (hy < 0) /* z = (1/|x|) */ + z = 1.0f/z; + if (hx < 0) { + if (((ix-0x3f800000)|yisint) == 0) { + z = (z-z)/(z-z); /* (-1)**non-int is NaN */ + } else if (yisint == 1) + z = -z; /* (x<0)**odd = -(|x|**odd) */ + } + return z; + } + + sn = 1.0f; /* sign of result */ + if (hx < 0) { + if (yisint == 0) /* (x<0)**(non-int) is NaN */ + return (x-x)/(x-x); + if (yisint == 1) /* (x<0)**(odd int) */ + sn = -1.0f; + } + +#if 0 + /* |y| is huge */ + if (iy > 0x4d000000) { /* if |y| > 2**27 */ + /* over/underflow if x is not close to one */ + if (ix < 0x3f7ffff8) + return hy < 0 ? sn*huge*huge : sn*tiny*tiny; + if (ix > 0x3f800007) + return hy > 0 ? sn*huge*huge : sn*tiny*tiny; + /* now |1-x| is tiny <= 2**-20, suffice to compute + log(x) by x-x^2/2+x^3/3-x^4/4 */ + t = ax - 1; /* t has 20 trailing zeros */ + w = (t*t)*(0.5f - t*(0.333333333333f - t*0.25f)); + u = ivln2_h*t; /* ivln2_h has 16 sig. bits */ + v = t*ivln2_l - w*ivln2; + t1 = u + v; + GET_FLOAT_WORD(is, t1); + SET_FLOAT_WORD(t1, is & 0xfffff000); + t2 = v - (t1-u); + } else +#endif +{ + float s2,s_h,s_l,t_h,t_l; + n = 0; + /* take care subnormal number */ + if (ix < 0x00800000) { + ax *= two24; + n -= 24; + GET_FLOAT_WORD(ix, ax); + } + n += ((ix)>>23) - 0x7f; + j = ix & 0x007fffff; + /* determine interval */ + ix = j | 0x3f800000; /* normalize ix */ + if (j <= 0x1cc471) /* |x|>1) & 0xfffff000) | 0x20000000; + SET_FLOAT_WORD(t_h, is + 0x00400000 + (k<<21)); + t_l = ax - (t_h - bp[k]); + s_l = v*((u - s_h*t_h) - s_h*t_l); + /* compute log(ax) */ + s2 = s*s; + r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6))))); + r += s_l*(s_h+s); + s2 = s_h*s_h; + t_h = 3.0f + s2 + r; + GET_FLOAT_WORD(is, t_h); + SET_FLOAT_WORD(t_h, is & 0xfffff000); + t_l = r - ((t_h - 3.0f) - s2); + /* u+v = s*(1+...) */ + u = s_h*t_h; + v = s_l*t_h + t_l*s; + /* 2/(3log2)*(s+...) */ + p_h = u + v; + GET_FLOAT_WORD(is, p_h); + SET_FLOAT_WORD(p_h, is & 0xfffff000); + p_l = v - (p_h - u); + z_h = cp_h*p_h; /* cp_h+cp_l = 2/(3*log2) */ + z_l = cp_l*p_h + p_l*cp+dp_l[k]; + /* log2(ax) = (s+..)*2/(3*log2) = n + dp_h + z_h + z_l */ + t = (float)n; + t1 = (((z_h + z_l) + dp_h[k]) + t); + GET_FLOAT_WORD(is, t1); + SET_FLOAT_WORD(t1, is & 0xfffff000); + t2 = z_l - (((t1 - t) - dp_h[k]) - z_h); + } + + /* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */ + GET_FLOAT_WORD(is, y); + SET_FLOAT_WORD(y1, is & 0xfffff000); + p_l = (y-y1)*t1 + y*t2; + p_h = y1*t1; + z = p_l + p_h; + GET_FLOAT_WORD(j, z); + if (j > 0x43000000) /* if z > 128 */ + return sn*huge*huge; /* overflow */ + else if (j == 0x43000000) { /* if z == 128 */ + if (p_l + ovt > z - p_h) + return sn*huge*huge; /* overflow */ + } else if ((j&0x7fffffff) > 0x43160000) /* z < -150 */ // FIXME: check should be (uint32_t)j > 0xc3160000 + return sn*tiny*tiny; /* underflow */ + else if (j == 0xc3160000) { /* z == -150 */ + if (p_l <= z-p_h) + return sn*tiny*tiny; /* underflow */ + } + /* + * compute 2**(p_h+p_l) + */ + i = j & 0x7fffffff; + k = (i>>23) - 0x7f; + n = 0; + if (i > 0x3f000000) { /* if |z| > 0.5, set n = [z+0.5] */ + n = j + (0x00800000>>(k+1)); + k = ((n&0x7fffffff)>>23) - 0x7f; /* new k for n */ + SET_FLOAT_WORD(t, n & ~(0x007fffff>>k)); + n = ((n&0x007fffff)|0x00800000)>>(23-k); + if (j < 0) + n = -n; + p_h -= t; + } + t = p_l + p_h; + GET_FLOAT_WORD(is, t); + SET_FLOAT_WORD(t, is & 0xffff8000); + u = t*lg2_h; + v = (p_l-(t-p_h))*lg2 + t*lg2_l; + z = u + v; + w = v - (z - u); + t = z*z; + t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5)))); + r = (z*t1)/(t1-2.0f) - (w+z*w); + z = 1.0f - (r - z); + GET_FLOAT_WORD(j, z); + j += n<<23; + if ((j>>23) <= 0) /* subnormal output */ + z = scalbnf(z, n); + else + SET_FLOAT_WORD(z, j); + return sn*z; +} + +static inline float babl_powf (float x, float y) +{ + return expf (y * logf (x)); +} + +static inline float _babl_trc_gamma_to_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + return babl_powf (value, trc->gamma); +} + +static inline float _babl_trc_gamma_from_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + return babl_powf (value, trc->rgamma); +} + +#define POLY_DEGREE 9 + +static inline float _babl_trc_gamma_1_8_to_linear (const Babl *trc_, float x) +{ + if (x >= 0.01f && x <= 1.0f) + { +#if POLY_DEGREE==9 + float u = -2.9976517e+1f; + u = u * x + 1.2166704e+2f; + u = u * x + -2.0336221e+2f; + u = u * x + 1.8145976e+2f; + u = u * x + -9.4187362e+1f; + u = u * x + 2.9529147e+1f; + u = u * x + -6.013473f; + u = u * x + 1.8725695f; + u = u * x + 9.9435057e-3f; + return u * x + -3.0061697e-5f; +#elif POLY_DEGREE==10 + float u = 7.7739437e+1f; + u = u * x + -3.5356881e+2f; + u = u * x + 6.7648456e+2f; + u = u * x + -7.0946645e+2f; + u = u * x + 4.4648413e+2f; + u = u * x + -1.7452451e+2f; + u = u * x + 4.3126128e+1f; + u = u * x + -7.2036142f; + u = u * x + 1.9207626f; + u = u * x + 9.14659e-3f; + return u * x + -2.575215e-5f; +#endif + } + return babl_powf (x, 1.8); +} + +static inline float _babl_trc_gamma_1_8_from_linear (const Babl *trc_, float x) +{ + if (x >= 0.01f && x <= 1.0f) + { +#if POLY_DEGREE==9 + float u = 7.6987344e+2f; + u = u * x + -3.4624161e+3f; + u = u * x + 6.5169973e+3f; + u = u * x + -6.6683502e+3f; + u = u * x + 4.037826e+3f; + u = u * x + -1.4810227e+3f; + u = u * x + 3.2670293e+2f; + u = u * x + -4.3240358e+1f; + u = u * x + 4.6009555f; + return u * x + 3.6000385e-2f; +#elif POLY_DEGREE==10 + float u = -2.2269134e+3f; + u = u * x + 1.1122916e+4f; + u = u * x + -2.3689515e+4f; + u = u * x + 2.8091743e+4f; + u = u * x + -2.0332165e+4f; + u = u * x + 9.2757293e+3f; + u = u * x + -2.6697086e+3f; + u = u * x + 4.7661489e+2f; + u = u * x + -5.2577079e+1f; + u = u * x + 4.8359543f; + return u * x + 3.4288484e-2f; +#endif + } + return babl_powf (x, 1.0f/1.8f); +} + +static inline float _babl_trc_gamma_2_2_to_linear (const Babl *trc_, float x) +{ + if (x >= 0.01f && x <= 1.0f) + { +#if POLY_DEGREE==9 + float u = 1.4519824e+1f; + u = u * x + -5.6919938e+1f; + u = u * x + 9.1404232e+1f; + u = u * x + -7.7951568e+1f; + u = u * x + 3.8593448e+1f; + u = u * x + -1.1709313e+1f; + u = u * x + 2.594042f; + u = u * x + 4.716017e-1f; + u = u * x + -1.478392e-3f; + return u * x + 4.987805e-6f; +#elif POLY_DEGREE==10 + float u = -3.6278814e+1f; + u = u * x + 1.5984863e+2f; + u = u * x + -2.9493194e+2f; + u = u * x + 2.9678744e+2f; + u = u * x + -1.7840175e+2f; + u = u * x + 6.6563304e+1f; + u = u * x + -1.5968574e+1f; + u = u * x + 2.9224961f; + u = u * x + 4.5996165e-1f; + u = u * x + -1.306086e-3f; + return u * x + 4.1278131e-6f; +#endif + } + return babl_powf (x, 2.2f); +} + +static inline float _babl_trc_gamma_2_2_from_linear (const Babl *trc_, float x) +{ + if (x >= 0.01f && x <= 1.0f) + { +#if POLY_DEGREE==9 + float u = 1.0084733e+3f; + u = u * x + -4.5716932e+3f; + u = u * x + 8.6843755e+3f; + u = u * x + -8.9814293e+3f; + u = u * x + 5.5060078e+3f; + u = u * x + -2.0477969e+3f; + u = u * x + 4.5804658e+2f; + u = u * x + -6.0913328e+1f; + u = u * x + 5.8668695f; + return u * x + 7.1335777e-2f; +#elif POLY_DEGREE==10 + float u = -2.9423332e+3f; + u = u * x + 1.4803473e+4f; + u = u * x + -3.1791125e+4f; + u = u * x + 3.805972e+4f; + u = u * x + -2.7850114e+4f; + u = u * x + 1.2865342e+4f; + u = u * x + -3.7543241e+3f; + u = u * x + 6.7921578e+2f; + u = u * x + -7.5167916e+1f; + u = u * x + 6.239892f; + return u * x + 6.8539291e-2f; +#endif + } + return babl_powf (x, 1.0/2.2); +} + + +static inline float _babl_trc_srgb_to_linear (const Babl *trc_, float value) +{ + return babl_gamma_2_2_to_linear (value); +} + +static inline float _babl_trc_srgb_from_linear (const Babl *trc_, float value) +{ + return babl_linear_to_gamma_2_2f (value); +} + +static inline float _babl_trc_from_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + return trc->fun_from_linear (trc_, value); +} + +static inline float _babl_trc_to_linear (const Babl *trc_, float value) +{ + BablTRC *trc = (void*)trc_; + return trc->fun_to_linear (trc_, value); +} + + const Babl * babl_trc (const char *name) { diff --git a/babl/babl-trc.h b/babl/babl-trc.h index dbdf2fe..0c9c23a 100644 --- a/babl/babl-trc.h +++ b/babl/babl-trc.h @@ -47,378 +47,6 @@ typedef struct float (*fun_from_linear)(const Babl *trc_, float val); } BablTRC; - -static inline float babl_trc_lut_from_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - int entry = value * trc->lut_size + 0.5; - float ret = trc->inv_lut[ - (entry >= 0 && entry < trc->lut_size) ? - entry : - trc->lut_size-1]; - /* XXX: fixme, do linear interpolation */ - return ret; -} - -static inline float babl_trc_lut_to_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - int entry = value * trc->lut_size + 0.5; - float ret = trc->lut[ - (entry >= 0 && entry < trc->lut_size) ? - entry : - trc->lut_size-1]; - /* XXX: fixme, do linear interpolation */ - return ret; -} - -static inline float _babl_trc_linear (const Babl *trc_, float value) -{ - return 1.0; -} - -/* origin: FreeBSD /usr/src/lib/msun/src/e_powf.c, copied from musl */ -/* - * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com. - */ -/* - * ==================================================== - * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved. - * - * Developed at SunPro, a Sun Microsystems, Inc. business. - * Permission to use, copy, modify, and distribute this - * software is freely granted, provided that this notice - * is preserved. - * ==================================================== - */ - -//#include "libm.h" - -/* Get a 32 bit int from a float. */ -#define GET_FLOAT_WORD(w,d) \ -do { \ - union {float f; uint32_t i;} __u; \ - __u.f = (d); \ - (w) = __u.i; \ -} while (0) - -/* Set a float from a 32 bit int. */ -#define SET_FLOAT_WORD(d,w) \ -do { \ - union {float f; uint32_t i;} __u; \ - __u.i = (w); \ - (d) = __u.f; \ -} while (0) - - -static const float -bp[] = {1.0, 1.5,}, -dp_h[] = { 0.0, 5.84960938e-01,}, /* 0x3f15c000 */ -dp_l[] = { 0.0, 1.56322085e-06,}, /* 0x35d1cfdc */ -two24 = 16777216.0, /* 0x4b800000 */ -huge = 1.0e30, -tiny = 1.0e-30, -/* poly coefs for (3/2)*(log(x)-2s-2/3*s**3 */ -L1 = 6.0000002384e-01, /* 0x3f19999a */ -L2 = 4.2857143283e-01, /* 0x3edb6db7 */ -L3 = 3.3333334327e-01, /* 0x3eaaaaab */ -L4 = 2.7272811532e-01, /* 0x3e8ba305 */ -L5 = 2.3066075146e-01, /* 0x3e6c3255 */ -L6 = 2.0697501302e-01, /* 0x3e53f142 */ -P1 = 1.6666667163e-01, /* 0x3e2aaaab */ -P2 = -2.7777778450e-03, /* 0xbb360b61 */ -P3 = 6.6137559770e-05, /* 0x388ab355 */ -P4 = -1.6533901999e-06, /* 0xb5ddea0e */ -P5 = 4.1381369442e-08, /* 0x3331bb4c */ -lg2 = 6.9314718246e-01, /* 0x3f317218 */ -lg2_h = 6.93145752e-01, /* 0x3f317200 */ -lg2_l = 1.42860654e-06, /* 0x35bfbe8c */ -ovt = 4.2995665694e-08, /* -(128-log2(ovfl+.5ulp)) */ -cp = 9.6179670095e-01, /* 0x3f76384f =2/(3ln2) */ -cp_h = 9.6191406250e-01, /* 0x3f764000 =12b cp */ -cp_l = -1.1736857402e-04, /* 0xb8f623c6 =tail of cp_h */ -ivln2 = 1.4426950216e+00, /* 0x3fb8aa3b =1/ln2 */ -ivln2_h = 1.4426879883e+00, /* 0x3fb8aa00 =16b 1/ln2*/ -ivln2_l = 7.0526075433e-06; /* 0x36eca570 =1/ln2 tail*/ - -static inline float obabl_powf(float x, float y) -{ - float z,ax,z_h,z_l,p_h,p_l; - float y1,t1,t2,r,s,sn,t,u,v,w; - int32_t i,j,k,yisint,n; - int32_t hx,hy,ix,iy,is; - - GET_FLOAT_WORD(hx, x); - GET_FLOAT_WORD(hy, y); - ix = hx & 0x7fffffff; - iy = hy & 0x7fffffff; - -#if 0 - /* x**0 = 1, even if x is NaN */ - if (iy == 0) - return 1.0f; -#endif - /* 1**y = 1, even if y is NaN */ - if (hx == 0x3f800000) - return 1.0f; - /* NaN if either arg is NaN */ - if (ix > 0x7f800000 || iy > 0x7f800000) - return x + y; - - /* determine if y is an odd int when x < 0 - * yisint = 0 ... y is not an integer - * yisint = 1 ... y is an odd int - * yisint = 2 ... y is an even int - */ - yisint = 0; - if (hx < 0) { - if (iy >= 0x4b800000) - yisint = 2; /* even integer y */ - else if (iy >= 0x3f800000) { - k = (iy>>23) - 0x7f; /* exponent */ - j = iy>>(23-k); - if ((j<<(23-k)) == iy) - yisint = 2 - (j & 1); - } - } -#if 0 - /* special value of y */ - if (iy == 0x7f800000) { /* y is +-inf */ - if (ix == 0x3f800000) /* (-1)**+-inf is 1 */ - return 1.0f; - else if (ix > 0x3f800000) /* (|x|>1)**+-inf = inf,0 */ - return hy >= 0 ? y : 0.0f; - else /* (|x|<1)**+-inf = 0,inf */ - return hy >= 0 ? 0.0f: -y; - } - if (iy == 0x3f800000) /* y is +-1 */ - return hy >= 0 ? x : 1.0f/x; - if (hy == 0x40000000) /* y is 2 */ - return x*x; - if (hy == 0x3f000000) { /* y is 0.5 */ - if (hx >= 0) /* x >= +0 */ - return sqrtf(x); - } -#endif - - ax = fabsf(x); - /* special value of x */ - if (ix == 0x7f800000 || ix == 0 || ix == 0x3f800000) { /* x is +-0,+-inf,+-1 */ - z = ax; - if (hy < 0) /* z = (1/|x|) */ - z = 1.0f/z; - if (hx < 0) { - if (((ix-0x3f800000)|yisint) == 0) { - z = (z-z)/(z-z); /* (-1)**non-int is NaN */ - } else if (yisint == 1) - z = -z; /* (x<0)**odd = -(|x|**odd) */ - } - return z; - } - - sn = 1.0f; /* sign of result */ - if (hx < 0) { - if (yisint == 0) /* (x<0)**(non-int) is NaN */ - return (x-x)/(x-x); - if (yisint == 1) /* (x<0)**(odd int) */ - sn = -1.0f; - } - -#if 0 - /* |y| is huge */ - if (iy > 0x4d000000) { /* if |y| > 2**27 */ - /* over/underflow if x is not close to one */ - if (ix < 0x3f7ffff8) - return hy < 0 ? sn*huge*huge : sn*tiny*tiny; - if (ix > 0x3f800007) - return hy > 0 ? sn*huge*huge : sn*tiny*tiny; - /* now |1-x| is tiny <= 2**-20, suffice to compute - log(x) by x-x^2/2+x^3/3-x^4/4 */ - t = ax - 1; /* t has 20 trailing zeros */ - w = (t*t)*(0.5f - t*(0.333333333333f - t*0.25f)); - u = ivln2_h*t; /* ivln2_h has 16 sig. bits */ - v = t*ivln2_l - w*ivln2; - t1 = u + v; - GET_FLOAT_WORD(is, t1); - SET_FLOAT_WORD(t1, is & 0xfffff000); - t2 = v - (t1-u); - } else -#endif -{ - float s2,s_h,s_l,t_h,t_l; - n = 0; - /* take care subnormal number */ - if (ix < 0x00800000) { - ax *= two24; - n -= 24; - GET_FLOAT_WORD(ix, ax); - } - n += ((ix)>>23) - 0x7f; - j = ix & 0x007fffff; - /* determine interval */ - ix = j | 0x3f800000; /* normalize ix */ - if (j <= 0x1cc471) /* |x|>1) & 0xfffff000) | 0x20000000; - SET_FLOAT_WORD(t_h, is + 0x00400000 + (k<<21)); - t_l = ax - (t_h - bp[k]); - s_l = v*((u - s_h*t_h) - s_h*t_l); - /* compute log(ax) */ - s2 = s*s; - r = s2*s2*(L1+s2*(L2+s2*(L3+s2*(L4+s2*(L5+s2*L6))))); - r += s_l*(s_h+s); - s2 = s_h*s_h; - t_h = 3.0f + s2 + r; - GET_FLOAT_WORD(is, t_h); - SET_FLOAT_WORD(t_h, is & 0xfffff000); - t_l = r - ((t_h - 3.0f) - s2); - /* u+v = s*(1+...) */ - u = s_h*t_h; - v = s_l*t_h + t_l*s; - /* 2/(3log2)*(s+...) */ - p_h = u + v; - GET_FLOAT_WORD(is, p_h); - SET_FLOAT_WORD(p_h, is & 0xfffff000); - p_l = v - (p_h - u); - z_h = cp_h*p_h; /* cp_h+cp_l = 2/(3*log2) */ - z_l = cp_l*p_h + p_l*cp+dp_l[k]; - /* log2(ax) = (s+..)*2/(3*log2) = n + dp_h + z_h + z_l */ - t = (float)n; - t1 = (((z_h + z_l) + dp_h[k]) + t); - GET_FLOAT_WORD(is, t1); - SET_FLOAT_WORD(t1, is & 0xfffff000); - t2 = z_l - (((t1 - t) - dp_h[k]) - z_h); - } - - /* split up y into y1+y2 and compute (y1+y2)*(t1+t2) */ - GET_FLOAT_WORD(is, y); - SET_FLOAT_WORD(y1, is & 0xfffff000); - p_l = (y-y1)*t1 + y*t2; - p_h = y1*t1; - z = p_l + p_h; - GET_FLOAT_WORD(j, z); - if (j > 0x43000000) /* if z > 128 */ - return sn*huge*huge; /* overflow */ - else if (j == 0x43000000) { /* if z == 128 */ - if (p_l + ovt > z - p_h) - return sn*huge*huge; /* overflow */ - } else if ((j&0x7fffffff) > 0x43160000) /* z < -150 */ // FIXME: check should be (uint32_t)j > 0xc3160000 - return sn*tiny*tiny; /* underflow */ - else if (j == 0xc3160000) { /* z == -150 */ - if (p_l <= z-p_h) - return sn*tiny*tiny; /* underflow */ - } - /* - * compute 2**(p_h+p_l) - */ - i = j & 0x7fffffff; - k = (i>>23) - 0x7f; - n = 0; - if (i > 0x3f000000) { /* if |z| > 0.5, set n = [z+0.5] */ - n = j + (0x00800000>>(k+1)); - k = ((n&0x7fffffff)>>23) - 0x7f; /* new k for n */ - SET_FLOAT_WORD(t, n & ~(0x007fffff>>k)); - n = ((n&0x007fffff)|0x00800000)>>(23-k); - if (j < 0) - n = -n; - p_h -= t; - } - t = p_l + p_h; - GET_FLOAT_WORD(is, t); - SET_FLOAT_WORD(t, is & 0xffff8000); - u = t*lg2_h; - v = (p_l-(t-p_h))*lg2 + t*lg2_l; - z = u + v; - w = v - (z - u); - t = z*z; - t1 = z - t*(P1+t*(P2+t*(P3+t*(P4+t*P5)))); - r = (z*t1)/(t1-2.0f) - (w+z*w); - z = 1.0f - (r - z); - GET_FLOAT_WORD(j, z); - j += n<<23; - if ((j>>23) <= 0) /* subnormal output */ - z = scalbnf(z, n); - else - SET_FLOAT_WORD(z, j); - return sn*z; -} - -static inline float babl_powf (float x, float y) -{ - return expf (y * logf (x)); -} - -static inline float _babl_trc_gamma_to_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - return babl_powf (value, trc->gamma); -} - -static inline float _babl_trc_gamma_from_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - return babl_powf (value, trc->rgamma); -} - -static inline float _babl_trc_gamma_2_2_to_linear (const Babl *trc_, float value) -{ - return babl_powf (value, 2.2); -} - -static inline float _babl_trc_gamma_2_2_from_linear (const Babl *trc_, float value) -{ - return babl_powf (value, 1.0/2.2); -} - -static inline float _babl_trc_gamma_1_8_to_linear (const Babl *trc_, float value) -{ - return babl_powf (value, 1.8); -} - -static inline float _babl_trc_gamma_1_8_from_linear (const Babl *trc_, float value) -{ - return babl_powf (value, 1.0/1.8); -} - -static inline float _babl_trc_srgb_to_linear (const Babl *trc_, float value) -{ - return babl_gamma_2_2_to_linear (value); -} - -static inline float _babl_trc_srgb_from_linear (const Babl *trc_, float value) -{ - return babl_linear_to_gamma_2_2f (value); -} - -static inline float _babl_trc_from_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - return trc->fun_from_linear (trc_, value); -} - -static inline float _babl_trc_to_linear (const Babl *trc_, float value) -{ - BablTRC *trc = (void*)trc_; - return trc->fun_to_linear (trc_, value); -} - void babl_trc_class_init (void); -- 2.30.2