static const R *t1fv_3(R *ri, R *ii, const R *W, stride ios, int m, int dist) { DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); int i; R *x; x = ri; BEGIN_SIMD(); for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 4)) { V T1, T3, T5, T6, T2, T4, T7, T8; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)])); T3 = BYTWJ(&(W[0]), T2); T4 = LD(&(x[WS(ios, 2)]), dist, &(x[0])); T5 = BYTWJ(&(W[TWVL * 2]), T4); T6 = VADD(T3, T5); ST(&(x[0]), VADD(T1, T6), dist, &(x[0])); T7 = VFNMS(LDK(KP500000000), T6, T1); T8 = VBYI(VMUL(LDK(KP866025403), VSUB(T5, T3))); ST(&(x[WS(ios, 2)]), VSUB(T7, T8), dist, &(x[0])); ST(&(x[WS(ios, 1)]), VADD(T7, T8), dist, &(x[WS(ios, 1)])); } END_SIMD(); return W; }
static void t1fuv_4(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(4, rs)) { V T1, T8, T3, T6, T7, T2, T5; T1 = LD(&(x[0]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T8 = BYTWJ(&(W[TWVL * 4]), T7); T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 2]), T2); T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T6 = BYTWJ(&(W[0]), T5); { V T4, T9, Ta, Tb; T4 = VSUB(T1, T3); T9 = VBYI(VSUB(T6, T8)); ST(&(x[WS(rs, 1)]), VSUB(T4, T9), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VADD(T4, T9), ms, &(x[WS(rs, 1)])); Ta = VADD(T1, T3); Tb = VADD(T6, T8); ST(&(x[WS(rs, 2)]), VSUB(Ta, Tb), ms, &(x[0])); ST(&(x[0]), VADD(Ta, Tb), ms, &(x[0])); } } } VLEAVE(); }
static void t1fuv_3(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP500000000, +0.500000000000000000000000000000000000000000000); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 4)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 4), MAKE_VOLATILE_STRIDE(rs)) { V T1, T2, T4; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T4 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); { V T3, T5, T8, T6, T7; T3 = BYTWJ(&(W[0]), T2); T5 = BYTWJ(&(W[TWVL * 2]), T4); T8 = VMUL(LDK(KP866025403), VSUB(T5, T3)); T6 = VADD(T3, T5); T7 = VFNMS(LDK(KP500000000), T6, T1); ST(&(x[0]), VADD(T1, T6), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VFMAI(T8, T7), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), VFNMSI(T8, T7), ms, &(x[0])); } } }
static const R *t1fv_6(R *ri, R *ii, const R *W, stride ios, int m, int dist) { DVK(KP500000000, +0.500000000000000000000000000000000000000000000); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); int i; R *x; x = ri; BEGIN_SIMD(); for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 10)) { V T4, Ti, Te, Tk, T9, Tj, T1, T3, T2; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(ios, 3)]), dist, &(x[WS(ios, 1)])); T3 = BYTWJ(&(W[TWVL * 4]), T2); T4 = VSUB(T1, T3); Ti = VADD(T1, T3); { V Tb, Td, Ta, Tc; Ta = LD(&(x[WS(ios, 4)]), dist, &(x[0])); Tb = BYTWJ(&(W[TWVL * 6]), Ta); Tc = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)])); Td = BYTWJ(&(W[0]), Tc); Te = VSUB(Tb, Td); Tk = VADD(Tb, Td); } { V T6, T8, T5, T7; T5 = LD(&(x[WS(ios, 2)]), dist, &(x[0])); T6 = BYTWJ(&(W[TWVL * 2]), T5); T7 = LD(&(x[WS(ios, 5)]), dist, &(x[WS(ios, 1)])); T8 = BYTWJ(&(W[TWVL * 8]), T7); T9 = VSUB(T6, T8); Tj = VADD(T6, T8); } { V Th, Tf, Tg, Tn, Tl, Tm; Th = VBYI(VMUL(LDK(KP866025403), VSUB(Te, T9))); Tf = VADD(T9, Te); Tg = VFNMS(LDK(KP500000000), Tf, T4); ST(&(x[WS(ios, 3)]), VADD(T4, Tf), dist, &(x[WS(ios, 1)])); ST(&(x[WS(ios, 1)]), VADD(Tg, Th), dist, &(x[WS(ios, 1)])); ST(&(x[WS(ios, 5)]), VSUB(Tg, Th), dist, &(x[WS(ios, 1)])); Tn = VBYI(VMUL(LDK(KP866025403), VSUB(Tk, Tj))); Tl = VADD(Tj, Tk); Tm = VFNMS(LDK(KP500000000), Tl, Ti); ST(&(x[0]), VADD(Ti, Tl), dist, &(x[0])); ST(&(x[WS(ios, 4)]), VADD(Tm, Tn), dist, &(x[0])); ST(&(x[WS(ios, 2)]), VSUB(Tm, Tn), dist, &(x[0])); } } END_SIMD(); return W; }
static void t1fuv_6(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP500000000, +0.500000000000000000000000000000000000000000000); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 10)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(rs)) { V T1, T2, Ta, Tc, T5, T7; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); { V T3, Tb, Td, T6, T8; T3 = BYTWJ(&(W[TWVL * 4]), T2); Tb = BYTWJ(&(W[TWVL * 6]), Ta); Td = BYTWJ(&(W[0]), Tc); T6 = BYTWJ(&(W[TWVL * 2]), T5); T8 = BYTWJ(&(W[TWVL * 8]), T7); { V Ti, T4, Tk, Te, Tj, T9; Ti = VADD(T1, T3); T4 = VSUB(T1, T3); Tk = VADD(Tb, Td); Te = VSUB(Tb, Td); Tj = VADD(T6, T8); T9 = VSUB(T6, T8); { V Tl, Tn, Tf, Th, Tm, Tg; Tl = VADD(Tj, Tk); Tn = VMUL(LDK(KP866025403), VSUB(Tk, Tj)); Tf = VADD(T9, Te); Th = VMUL(LDK(KP866025403), VSUB(Te, T9)); ST(&(x[0]), VADD(Ti, Tl), ms, &(x[0])); Tm = VFNMS(LDK(KP500000000), Tl, Ti); ST(&(x[WS(rs, 3)]), VADD(T4, Tf), ms, &(x[WS(rs, 1)])); Tg = VFNMS(LDK(KP500000000), Tf, T4); ST(&(x[WS(rs, 2)]), VFNMSI(Tn, Tm), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VFMAI(Tn, Tm), ms, &(x[0])); ST(&(x[WS(rs, 5)]), VFNMSI(Th, Tg), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VFMAI(Th, Tg), ms, &(x[WS(rs, 1)])); } } } } } VLEAVE(); }
static void t1fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP587785252, +0.587785252292473129168705954639072768597652438); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) { V Tc, Tg, Th, T5, Ta, Td; Tc = LD(&(x[0]), ms, &(x[0])); { V T2, T9, T4, T7; { V T1, T8, T3, T6; T1 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T2 = BYTWJ(&(W[0]), T1); T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[TWVL * 4]), T8); T3 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T4 = BYTWJ(&(W[TWVL * 6]), T3); T6 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T7 = BYTWJ(&(W[TWVL * 2]), T6); } Tg = VSUB(T2, T4); Th = VSUB(T7, T9); T5 = VADD(T2, T4); Ta = VADD(T7, T9); Td = VADD(T5, Ta); } ST(&(x[0]), VADD(Tc, Td), ms, &(x[0])); { V Ti, Tj, Tf, Tk, Tb, Te; Ti = VBYI(VFMA(LDK(KP951056516), Tg, VMUL(LDK(KP587785252), Th))); Tj = VBYI(VFNMS(LDK(KP587785252), Tg, VMUL(LDK(KP951056516), Th))); Tb = VMUL(LDK(KP559016994), VSUB(T5, Ta)); Te = VFNMS(LDK(KP250000000), Td, Tc); Tf = VADD(Tb, Te); Tk = VSUB(Te, Tb); ST(&(x[WS(rs, 1)]), VSUB(Tf, Ti), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VSUB(Tk, Tj), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 4)]), VADD(Ti, Tf), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VADD(Tj, Tk), ms, &(x[0])); } } } VLEAVE(); }
static void t1fv_5(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP618033988, +0.618033988749894848204586834365638117720309180); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 8)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 8), MAKE_VOLATILE_STRIDE(5, rs)) { V T1, T2, T9, T4, T7; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T9 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T4 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); { V T3, Ta, T5, T8; T3 = BYTWJ(&(W[0]), T2); Ta = BYTWJ(&(W[TWVL * 4]), T9); T5 = BYTWJ(&(W[TWVL * 6]), T4); T8 = BYTWJ(&(W[TWVL * 2]), T7); { V T6, Tg, Tb, Th; T6 = VADD(T3, T5); Tg = VSUB(T3, T5); Tb = VADD(T8, Ta); Th = VSUB(T8, Ta); { V Te, Tc, Tk, Ti, Td, Tj, Tf; Te = VSUB(T6, Tb); Tc = VADD(T6, Tb); Tk = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tg, Th)); Ti = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Th, Tg)); Td = VFNMS(LDK(KP250000000), Tc, T1); ST(&(x[0]), VADD(T1, Tc), ms, &(x[0])); Tj = VFNMS(LDK(KP559016994), Te, Td); Tf = VFMA(LDK(KP559016994), Te, Td); ST(&(x[WS(rs, 2)]), VFMAI(Tk, Tj), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VFNMSI(Tk, Tj), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 4)]), VFMAI(Ti, Tf), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VFNMSI(Ti, Tf), ms, &(x[WS(rs, 1)])); } } } } } VLEAVE(); }
static const R *t1fv_6(R *ri, R *ii, const R *W, stride ios, INT m, INT dist) { DVK(KP500000000, +0.500000000000000000000000000000000000000000000); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); INT i; R *x; x = ri; for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 10), MAKE_VOLATILE_STRIDE(ios)) { V T1, T2, Ta, Tc, T5, T7; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(ios, 3)]), dist, &(x[WS(ios, 1)])); Ta = LD(&(x[WS(ios, 4)]), dist, &(x[0])); Tc = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)])); T5 = LD(&(x[WS(ios, 2)]), dist, &(x[0])); T7 = LD(&(x[WS(ios, 5)]), dist, &(x[WS(ios, 1)])); { V T3, Tb, Td, T6, T8; T3 = BYTWJ(&(W[TWVL * 4]), T2); Tb = BYTWJ(&(W[TWVL * 6]), Ta); Td = BYTWJ(&(W[0]), Tc); T6 = BYTWJ(&(W[TWVL * 2]), T5); T8 = BYTWJ(&(W[TWVL * 8]), T7); { V Ti, T4, Tk, Te, Tj, T9; Ti = VADD(T1, T3); T4 = VSUB(T1, T3); Tk = VADD(Tb, Td); Te = VSUB(Tb, Td); Tj = VADD(T6, T8); T9 = VSUB(T6, T8); { V Tl, Tn, Tf, Th, Tm, Tg; Tl = VADD(Tj, Tk); Tn = VMUL(LDK(KP866025403), VSUB(Tk, Tj)); Tf = VADD(T9, Te); Th = VMUL(LDK(KP866025403), VSUB(Te, T9)); ST(&(x[0]), VADD(Ti, Tl), dist, &(x[0])); Tm = VFNMS(LDK(KP500000000), Tl, Ti); ST(&(x[WS(ios, 3)]), VADD(T4, Tf), dist, &(x[WS(ios, 1)])); Tg = VFNMS(LDK(KP500000000), Tf, T4); ST(&(x[WS(ios, 2)]), VFNMSI(Tn, Tm), dist, &(x[0])); ST(&(x[WS(ios, 4)]), VFMAI(Tn, Tm), dist, &(x[0])); ST(&(x[WS(ios, 5)]), VFNMSI(Th, Tg), dist, &(x[WS(ios, 1)])); ST(&(x[WS(ios, 1)]), VFMAI(Th, Tg), dist, &(x[WS(ios, 1)])); } } } } return W; }
static void q1fv_2(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms) { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(rs), MAKE_VOLATILE_STRIDE(vs)) { V T1, T2, T3, T4, T5, T6; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[0]), VSUB(T1, T2)); T4 = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)])); T5 = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)])); T6 = BYTWJ(&(W[0]), VSUB(T4, T5)); ST(&(x[WS(vs, 1)]), T3, ms, &(x[WS(vs, 1)])); ST(&(x[WS(vs, 1) + WS(rs, 1)]), T6, ms, &(x[WS(vs, 1) + WS(rs, 1)])); ST(&(x[0]), VADD(T1, T2), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VADD(T4, T5), ms, &(x[WS(rs, 1)])); } }
static const R *q1fv_2(R *ri, R *ii, const R *W, stride is, stride vs, INT m, INT dist) { INT i; R *x; x = ri; for (i = 0; i < m; i = i + VL, x = x + (VL * dist), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(is), MAKE_VOLATILE_STRIDE(vs)) { V T1, T2, T3, T4, T5, T6; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(is, 1)]), dist, &(x[WS(is, 1)])); T3 = BYTWJ(&(W[0]), VSUB(T1, T2)); T4 = LD(&(x[WS(vs, 1)]), dist, &(x[WS(vs, 1)])); T5 = LD(&(x[WS(vs, 1) + WS(is, 1)]), dist, &(x[WS(vs, 1) + WS(is, 1)])); T6 = BYTWJ(&(W[0]), VSUB(T4, T5)); ST(&(x[WS(vs, 1)]), T3, dist, &(x[WS(vs, 1)])); ST(&(x[WS(vs, 1) + WS(is, 1)]), T6, dist, &(x[WS(vs, 1) + WS(is, 1)])); ST(&(x[0]), VADD(T1, T2), dist, &(x[0])); ST(&(x[WS(is, 1)]), VADD(T4, T5), dist, &(x[WS(is, 1)])); } return W; }
static void t1fuv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP900968867, +0.900968867902419126236102319507445051165919162); DVK(KP222520933, +0.222520933956314404288902564496794759466355569); DVK(KP623489801, +0.623489801858733530525004884004239810632274731); DVK(KP781831482, +0.781831482468029808708444526674057750232334519); DVK(KP974927912, +0.974927912181823607018131682993931217232785801); DVK(KP433883739, +0.433883739117558120475768332848358754609990728); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(rs)) { V T1, Tg, Tj, T6, Ti, Tb, Tk, Tp, To; T1 = LD(&(x[0]), ms, &(x[0])); { V Td, Tf, Tc, Te; Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[TWVL * 4]), Tc); Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tf = BYTWJ(&(W[TWVL * 6]), Te); Tg = VADD(Td, Tf); Tj = VSUB(Tf, Td); } { V T3, T5, T2, T4; T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[0]), T2); T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T5 = BYTWJ(&(W[TWVL * 10]), T4); T6 = VADD(T3, T5); Ti = VSUB(T5, T3); } { V T8, Ta, T7, T9; T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T8 = BYTWJ(&(W[TWVL * 2]), T7); T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Ta = BYTWJ(&(W[TWVL * 8]), T9); Tb = VADD(T8, Ta); Tk = VSUB(Ta, T8); } ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); Tp = VBYI(VFMA(LDK(KP433883739), Ti, VFNMS(LDK(KP781831482), Tk, VMUL(LDK(KP974927912), Tj)))); To = VFMA(LDK(KP623489801), Tb, VFNMS(LDK(KP222520933), Tg, VFNMS(LDK(KP900968867), T6, T1))); ST(&(x[WS(rs, 4)]), VSUB(To, Tp), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)])); { V Tl, Th, Tn, Tm; Tl = VBYI(VFNMS(LDK(KP781831482), Tj, VFNMS(LDK(KP433883739), Tk, VMUL(LDK(KP974927912), Ti)))); Th = VFMA(LDK(KP623489801), Tg, VFNMS(LDK(KP900968867), Tb, VFNMS(LDK(KP222520933), T6, T1))); ST(&(x[WS(rs, 5)]), VSUB(Th, Tl), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), VADD(Th, Tl), ms, &(x[0])); Tn = VBYI(VFMA(LDK(KP781831482), Ti, VFMA(LDK(KP974927912), Tk, VMUL(LDK(KP433883739), Tj)))); Tm = VFMA(LDK(KP623489801), T6, VFNMS(LDK(KP900968867), Tg, VFNMS(LDK(KP222520933), Tb, T1))); ST(&(x[WS(rs, 6)]), VSUB(Tm, Tn), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VADD(Tm, Tn), ms, &(x[WS(rs, 1)])); } } }
static const R *q1fv_2(R *ri, R *ii, const R *W, stride is, stride vs, int m, int dist) { int i; R *x; x = ri; BEGIN_SIMD(); for (i = 0; i < m; i = i + VL, x = x + (VL * dist), W = W + (TWVL * 2)) { V T1, T2, T3, T4, T5, T6; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(is, 1)]), dist, &(x[WS(is, 1)])); T3 = BYTWJ(&(W[0]), VSUB(T1, T2)); T4 = LD(&(x[WS(vs, 1)]), dist, &(x[WS(vs, 1)])); T5 = LD(&(x[WS(vs, 1) + WS(is, 1)]), dist, &(x[WS(vs, 1) + WS(is, 1)])); T6 = BYTWJ(&(W[0]), VSUB(T4, T5)); ST(&(x[WS(vs, 1)]), T3, dist, &(x[WS(vs, 1)])); ST(&(x[WS(vs, 1) + WS(is, 1)]), T6, dist, &(x[WS(vs, 1) + WS(is, 1)])); ST(&(x[0]), VADD(T1, T2), dist, &(x[0])); ST(&(x[WS(is, 1)]), VADD(T4, T5), dist, &(x[WS(is, 1)])); } END_SIMD(); return W; }
static const R *t2fv_2(R *ri, R *ii, const R *W, stride ios, INT m, INT dist) { INT i; R *x; x = ri; for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(ios)) { V T1, T3, T2; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)])); T3 = BYTWJ(&(W[0]), T2); ST(&(x[WS(ios, 1)]), VSUB(T1, T3), dist, &(x[WS(ios, 1)])); ST(&(x[0]), VADD(T1, T3), dist, &(x[0])); } return W; }
static void t1fuv_2(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 2)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 2), MAKE_VOLATILE_STRIDE(2, rs)) { V T1, T3, T2; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[0]), T2); ST(&(x[WS(rs, 1)]), VSUB(T1, T3), ms, &(x[WS(rs, 1)])); ST(&(x[0]), VADD(T1, T3), ms, &(x[0])); } } VLEAVE(); }
static const R *t1fv_2(R *ri, R *ii, const R *W, stride ios, int m, int dist) { int i; R *x; x = ri; BEGIN_SIMD(); for (i = m; i > 0; i = i - VL, x = x + (VL * dist), W = W + (TWVL * 2)) { V T1, T3, T2; T1 = LD(&(x[0]), dist, &(x[0])); T2 = LD(&(x[WS(ios, 1)]), dist, &(x[WS(ios, 1)])); T3 = BYTWJ(&(W[0]), T2); ST(&(x[WS(ios, 1)]), VSUB(T1, T3), dist, &(x[WS(ios, 1)])); ST(&(x[0]), VADD(T1, T3), dist, &(x[0])); } END_SIMD(); return W; }
static void t1fuv_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP939692620, +0.939692620785908384054109277324731469936208134); DVK(KP296198132, +0.296198132726023843175338011893050938967728390); DVK(KP852868531, +0.852868531952443209628250963940074071936020296); DVK(KP173648177, +0.173648177666930348851716626769314796000375677); DVK(KP556670399, +0.556670399226419366452912952047023132968291906); DVK(KP766044443, +0.766044443118978035202392650555416673935832457); DVK(KP642787609, +0.642787609686539326322643409907263432907559884); DVK(KP663413948, +0.663413948168938396205421319635891297216863310); DVK(KP984807753, +0.984807753012208059366743024589523013670643252); DVK(KP150383733, +0.150383733180435296639271897612501926072238258); DVK(KP342020143, +0.342020143325668733044099614682259580763083368); DVK(KP813797681, +0.813797681349373692844693217248393223289101568); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 16)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 16), MAKE_VOLATILE_STRIDE(rs)) { V T1, T6, TA, Tt, Tf, Ts, Tw, Tn, Tv; T1 = LD(&(x[0]), ms, &(x[0])); { V T3, T5, T2, T4; T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[TWVL * 4]), T2); T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T5 = BYTWJ(&(W[TWVL * 10]), T4); T6 = VADD(T3, T5); TA = VMUL(LDK(KP866025403), VSUB(T5, T3)); } { V T9, Td, Tb, T8, Tc, Ta, Te; T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[0]), T8); Tc = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[TWVL * 12]), Tc); Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tb = BYTWJ(&(W[TWVL * 6]), Ta); Tt = VSUB(Td, Tb); Te = VADD(Tb, Td); Tf = VADD(T9, Te); Ts = VFNMS(LDK(KP500000000), Te, T9); } { V Th, Tl, Tj, Tg, Tk, Ti, Tm; Tg = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Th = BYTWJ(&(W[TWVL * 2]), Tg); Tk = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Tl = BYTWJ(&(W[TWVL * 14]), Tk); Ti = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Tj = BYTWJ(&(W[TWVL * 8]), Ti); Tw = VSUB(Tl, Tj); Tm = VADD(Tj, Tl); Tn = VADD(Th, Tm); Tv = VFNMS(LDK(KP500000000), Tm, Th); } { V Tq, T7, To, Tp; Tq = VBYI(VMUL(LDK(KP866025403), VSUB(Tn, Tf))); T7 = VADD(T1, T6); To = VADD(Tf, Tn); Tp = VFNMS(LDK(KP500000000), To, T7); ST(&(x[0]), VADD(T7, To), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VADD(Tp, Tq), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 6)]), VSUB(Tp, Tq), ms, &(x[0])); } { V TI, TB, TC, TD, Tu, Tx, Ty, Tr, TH; TI = VBYI(VSUB(VFNMS(LDK(KP342020143), Tv, VFNMS(LDK(KP150383733), Tt, VFNMS(LDK(KP984807753), Ts, VMUL(LDK(KP813797681), Tw)))), TA)); TB = VFNMS(LDK(KP642787609), Ts, VMUL(LDK(KP663413948), Tt)); TC = VFNMS(LDK(KP984807753), Tv, VMUL(LDK(KP150383733), Tw)); TD = VADD(TB, TC); Tu = VFMA(LDK(KP766044443), Ts, VMUL(LDK(KP556670399), Tt)); Tx = VFMA(LDK(KP173648177), Tv, VMUL(LDK(KP852868531), Tw)); Ty = VADD(Tu, Tx); Tr = VFNMS(LDK(KP500000000), T6, T1); TH = VFMA(LDK(KP173648177), Ts, VFNMS(LDK(KP296198132), Tw, VFNMS(LDK(KP939692620), Tv, VFNMS(LDK(KP852868531), Tt, Tr)))); ST(&(x[WS(rs, 7)]), VSUB(TH, TI), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), VADD(TH, TI), ms, &(x[0])); { V Tz, TE, TF, TG; Tz = VADD(Tr, Ty); TE = VBYI(VADD(TA, TD)); ST(&(x[WS(rs, 8)]), VSUB(Tz, TE), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VADD(TE, Tz), ms, &(x[WS(rs, 1)])); TF = VFMA(LDK(KP866025403), VSUB(TB, TC), VFNMS(LDK(KP500000000), Ty, Tr)); TG = VBYI(VADD(TA, VFNMS(LDK(KP500000000), TD, VMUL(LDK(KP866025403), VSUB(Tx, Tu))))); ST(&(x[WS(rs, 5)]), VSUB(TF, TG), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 4)]), VADD(TF, TG), ms, &(x[0])); } } } } VLEAVE(); }
static void t1fuv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP707106781, +0.707106781186547524400844362104849039284835938); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(rs)) { V T4, Tq, Tm, Tr, T9, Tt, Te, Tu, T1, T3, T2; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 6]), T2); T4 = VSUB(T1, T3); Tq = VADD(T1, T3); { V Tj, Tl, Ti, Tk; Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tj = BYTWJ(&(W[TWVL * 2]), Ti); Tk = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Tl = BYTWJ(&(W[TWVL * 10]), Tk); Tm = VSUB(Tj, Tl); Tr = VADD(Tj, Tl); } { V T6, T8, T5, T7; T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T6 = BYTWJ(&(W[0]), T5); T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); T8 = BYTWJ(&(W[TWVL * 8]), T7); T9 = VSUB(T6, T8); Tt = VADD(T6, T8); } { V Tb, Td, Ta, Tc; Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tb = BYTWJ(&(W[TWVL * 12]), Ta); Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[TWVL * 4]), Tc); Te = VSUB(Tb, Td); Tu = VADD(Tb, Td); } { V Ts, Tv, Tw, Tx; Ts = VADD(Tq, Tr); Tv = VADD(Tt, Tu); ST(&(x[WS(rs, 4)]), VSUB(Ts, Tv), ms, &(x[0])); ST(&(x[0]), VADD(Ts, Tv), ms, &(x[0])); Tw = VSUB(Tq, Tr); Tx = VBYI(VSUB(Tu, Tt)); ST(&(x[WS(rs, 6)]), VSUB(Tw, Tx), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VADD(Tw, Tx), ms, &(x[0])); { V Tg, To, Tn, Tp, Tf, Th; Tf = VMUL(LDK(KP707106781), VADD(T9, Te)); Tg = VADD(T4, Tf); To = VSUB(T4, Tf); Th = VMUL(LDK(KP707106781), VSUB(Te, T9)); Tn = VBYI(VSUB(Th, Tm)); Tp = VBYI(VADD(Tm, Th)); ST(&(x[WS(rs, 7)]), VSUB(Tg, Tn), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VADD(To, Tp), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VADD(Tg, Tn), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 5)]), VSUB(To, Tp), ms, &(x[WS(rs, 1)])); } } } }
static void t1fv_12(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 22)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 22), MAKE_VOLATILE_STRIDE(12, rs)) { V T1, TC, T6, T7, Ty, Tq, Tz, TA, T9, TD, Te, Tf, Tu, Tl, Tv; V Tw; { V T5, T3, T4, T2; T1 = LD(&(x[0]), ms, &(x[0])); T4 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T5 = BYTWJ(&(W[TWVL * 14]), T4); T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 6]), T2); TC = VSUB(T5, T3); T6 = VADD(T3, T5); T7 = VFNMS(LDK(KP500000000), T6, T1); } { V Tn, Tp, Tm, Tx, To; Tm = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tn = BYTWJ(&(W[0]), Tm); Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Ty = BYTWJ(&(W[TWVL * 16]), Tx); To = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Tp = BYTWJ(&(W[TWVL * 8]), To); Tq = VSUB(Tn, Tp); Tz = VADD(Tn, Tp); TA = VFNMS(LDK(KP500000000), Tz, Ty); } { V Td, Tb, T8, Tc, Ta; T8 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T9 = BYTWJ(&(W[TWVL * 10]), T8); Tc = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Td = BYTWJ(&(W[TWVL * 2]), Tc); Ta = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Tb = BYTWJ(&(W[TWVL * 18]), Ta); TD = VSUB(Td, Tb); Te = VADD(Tb, Td); Tf = VFNMS(LDK(KP500000000), Te, T9); } { V Ti, Tk, Th, Tt, Tj; Th = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Ti = BYTWJ(&(W[TWVL * 20]), Th); Tt = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Tu = BYTWJ(&(W[TWVL * 4]), Tt); Tj = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tk = BYTWJ(&(W[TWVL * 12]), Tj); Tl = VSUB(Ti, Tk); Tv = VADD(Tk, Ti); Tw = VFNMS(LDK(KP500000000), Tv, Tu); } { V Ts, TG, TF, TH; { V Tg, Tr, TB, TE; Tg = VSUB(T7, Tf); Tr = VADD(Tl, Tq); Ts = VFMA(LDK(KP866025403), Tr, Tg); TG = VFNMS(LDK(KP866025403), Tr, Tg); TB = VSUB(Tw, TA); TE = VSUB(TC, TD); TF = VFNMS(LDK(KP866025403), TE, TB); TH = VFMA(LDK(KP866025403), TE, TB); } ST(&(x[WS(rs, 1)]), VFNMSI(TF, Ts), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(TH, TG), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 11)]), VFMAI(TF, Ts), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 5)]), VFNMSI(TH, TG), ms, &(x[WS(rs, 1)])); } { V TS, TW, TV, TX; { V TQ, TR, TT, TU; TQ = VADD(T1, T6); TR = VADD(T9, Te); TS = VSUB(TQ, TR); TW = VADD(TQ, TR); TT = VADD(Tu, Tv); TU = VADD(Ty, Tz); TV = VSUB(TT, TU); TX = VADD(TT, TU); } ST(&(x[WS(rs, 9)]), VFNMSI(TV, TS), ms, &(x[WS(rs, 1)])); ST(&(x[0]), VADD(TW, TX), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VFMAI(TV, TS), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 6)]), VSUB(TW, TX), ms, &(x[0])); } { V TK, TO, TN, TP; { V TI, TJ, TL, TM; TI = VADD(T7, Tf); TJ = VADD(Tw, TA); TK = VSUB(TI, TJ); TO = VADD(TI, TJ); TL = VSUB(Tl, Tq); TM = VADD(TC, TD); TN = VMUL(LDK(KP866025403), VSUB(TL, TM)); TP = VMUL(LDK(KP866025403), VADD(TM, TL)); } ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0])); ST(&(x[WS(rs, 8)]), VFNMSI(TP, TO), ms, &(x[0])); ST(&(x[WS(rs, 10)]), VFNMSI(TN, TK), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0])); } } } VLEAVE(); }
static void t1fv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP216506350, +0.216506350946109661690930792688234045867850657); DVK(KP484122918, +0.484122918275927110647408174972799951354115213); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP509036960, +0.509036960455127183450980863393907648510733164); DVK(KP823639103, +0.823639103546331925877420039278190003029660514); DVK(KP587785252, +0.587785252292473129168705954639072768597652438); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 28)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 28), MAKE_VOLATILE_STRIDE(rs)) { V T1e, T7, TP, T12, T15, Tf, Tn, To, T1b, T1c, T1f, TQ, TR, TS, Tw; V TE, TF, TT, TU, TV; { V T1, T5, T3, T4, T2, T6; T1 = LD(&(x[0]), ms, &(x[0])); T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); T5 = BYTWJ(&(W[TWVL * 18]), T4); T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[TWVL * 8]), T2); T1e = VSUB(T5, T3); T6 = VADD(T3, T5); T7 = VADD(T1, T6); TP = VFNMS(LDK(KP500000000), T6, T1); } { V T9, Tq, Ty, Th, Te, T13, Tv, T10, TD, T11, Tm, T14; { V T8, Tp, Tx, Tg; T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[TWVL * 4]), T8); Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Tq = BYTWJ(&(W[TWVL * 10]), Tp); Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Ty = BYTWJ(&(W[TWVL * 16]), Tx); Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Th = BYTWJ(&(W[TWVL * 22]), Tg); } { V Tb, Td, Ta, Tc; Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Tb = BYTWJ(&(W[TWVL * 14]), Ta); Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[TWVL * 24]), Tc); Te = VADD(Tb, Td); T13 = VSUB(Td, Tb); } { V Ts, Tu, Tr, Tt; Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); Ts = BYTWJ(&(W[TWVL * 20]), Tr); Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tu = BYTWJ(&(W[0]), Tt); Tv = VADD(Ts, Tu); T10 = VSUB(Tu, Ts); } { V TA, TC, Tz, TB; Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); TA = BYTWJ(&(W[TWVL * 26]), Tz); TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); TC = BYTWJ(&(W[TWVL * 6]), TB); TD = VADD(TA, TC); T11 = VSUB(TC, TA); } { V Tj, Tl, Ti, Tk; Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tj = BYTWJ(&(W[TWVL * 2]), Ti); Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tl = BYTWJ(&(W[TWVL * 12]), Tk); Tm = VADD(Tj, Tl); T14 = VSUB(Tl, Tj); } T12 = VSUB(T10, T11); T15 = VSUB(T13, T14); Tf = VADD(T9, Te); Tn = VADD(Th, Tm); To = VADD(Tf, Tn); T1b = VADD(T13, T14); T1c = VADD(T10, T11); T1f = VADD(T1b, T1c); TQ = VFNMS(LDK(KP500000000), Te, T9); TR = VFNMS(LDK(KP500000000), Tm, Th); TS = VADD(TQ, TR); Tw = VADD(Tq, Tv); TE = VADD(Ty, TD); TF = VADD(Tw, TE); TT = VFNMS(LDK(KP500000000), Tv, Tq); TU = VFNMS(LDK(KP500000000), TD, Ty); TV = VADD(TT, TU); } { V TI, TG, TH, TM, TO, TK, TL, TN, TJ; TI = VMUL(LDK(KP559016994), VSUB(To, TF)); TG = VADD(To, TF); TH = VFNMS(LDK(KP250000000), TG, T7); TK = VSUB(Tw, TE); TL = VSUB(Tf, Tn); TM = VBYI(VFNMS(LDK(KP587785252), TL, VMUL(LDK(KP951056516), TK))); TO = VBYI(VFMA(LDK(KP951056516), TL, VMUL(LDK(KP587785252), TK))); ST(&(x[0]), VADD(T7, TG), ms, &(x[0])); TN = VADD(TI, TH); ST(&(x[WS(rs, 6)]), VSUB(TN, TO), ms, &(x[0])); ST(&(x[WS(rs, 9)]), VADD(TO, TN), ms, &(x[WS(rs, 1)])); TJ = VSUB(TH, TI); ST(&(x[WS(rs, 3)]), VSUB(TJ, TM), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 12)]), VADD(TM, TJ), ms, &(x[0])); } { V T16, T1m, T1u, T1h, T1o, T1a, T1p, TZ, T1t, T1l, T1d, T1g; T16 = VFNMS(LDK(KP509036960), T15, VMUL(LDK(KP823639103), T12)); T1m = VFMA(LDK(KP823639103), T15, VMUL(LDK(KP509036960), T12)); T1u = VBYI(VMUL(LDK(KP866025403), VADD(T1e, T1f))); T1d = VMUL(LDK(KP484122918), VSUB(T1b, T1c)); T1g = VFNMS(LDK(KP216506350), T1f, VMUL(LDK(KP866025403), T1e)); T1h = VSUB(T1d, T1g); T1o = VADD(T1d, T1g); { V T18, T19, TY, TW, TX; T18 = VSUB(TT, TU); T19 = VSUB(TQ, TR); T1a = VFNMS(LDK(KP587785252), T19, VMUL(LDK(KP951056516), T18)); T1p = VFMA(LDK(KP951056516), T19, VMUL(LDK(KP587785252), T18)); TY = VMUL(LDK(KP559016994), VSUB(TS, TV)); TW = VADD(TS, TV); TX = VFNMS(LDK(KP250000000), TW, TP); TZ = VSUB(TX, TY); T1t = VADD(TP, TW); T1l = VADD(TY, TX); } { V T17, T1i, T1r, T1s; ST(&(x[WS(rs, 5)]), VSUB(T1t, T1u), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 10)]), VADD(T1t, T1u), ms, &(x[0])); T17 = VSUB(TZ, T16); T1i = VBYI(VSUB(T1a, T1h)); ST(&(x[WS(rs, 8)]), VSUB(T17, T1i), ms, &(x[0])); ST(&(x[WS(rs, 7)]), VADD(T17, T1i), ms, &(x[WS(rs, 1)])); T1r = VSUB(T1l, T1m); T1s = VBYI(VADD(T1p, T1o)); ST(&(x[WS(rs, 11)]), VSUB(T1r, T1s), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 4)]), VADD(T1r, T1s), ms, &(x[0])); { V T1n, T1q, T1j, T1k; T1n = VADD(T1l, T1m); T1q = VBYI(VSUB(T1o, T1p)); ST(&(x[WS(rs, 14)]), VSUB(T1n, T1q), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VADD(T1n, T1q), ms, &(x[WS(rs, 1)])); T1j = VADD(TZ, T16); T1k = VBYI(VADD(T1a, T1h)); ST(&(x[WS(rs, 13)]), VSUB(T1j, T1k), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), VADD(T1j, T1k), ms, &(x[0])); } } } } }
static void t1fv_15(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP823639103, +0.823639103546331925877420039278190003029660514); DVK(KP910592997, +0.910592997310029334643087372129977886038870291); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP618033988, +0.618033988749894848204586834365638117720309180); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 28)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 28), MAKE_VOLATILE_STRIDE(rs)) { V Tq, Ty, Th, T1b, T10, Ts, TP, T7, Tu, TA, TC, Tj, Tk, TQ, Tf; { V T1, T4, T2, T9, Te; T1 = LD(&(x[0]), ms, &(x[0])); T4 = LD(&(x[WS(rs, 10)]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); { V T8, Tp, Tx, Tg; T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Tp = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Tx = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Tg = LD(&(x[WS(rs, 12)]), ms, &(x[0])); { V Tb, Td, Tr, T6, Tt, Tz, TB, Ti; { V T5, T3, Ta, Tc; Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); T5 = BYTWJ(&(W[TWVL * 18]), T4); T3 = BYTWJ(&(W[TWVL * 8]), T2); T9 = BYTWJ(&(W[TWVL * 4]), T8); Tq = BYTWJ(&(W[TWVL * 10]), Tp); Ty = BYTWJ(&(W[TWVL * 16]), Tx); Th = BYTWJ(&(W[TWVL * 22]), Tg); Tb = BYTWJ(&(W[TWVL * 14]), Ta); Td = BYTWJ(&(W[TWVL * 24]), Tc); Tr = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); T1b = VSUB(T5, T3); T6 = VADD(T3, T5); Tt = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); } Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); TB = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Ti = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Te = VADD(Tb, Td); T10 = VSUB(Td, Tb); Ts = BYTWJ(&(W[TWVL * 20]), Tr); TP = VFNMS(LDK(KP500000000), T6, T1); T7 = VADD(T1, T6); Tu = BYTWJ(&(W[0]), Tt); TA = BYTWJ(&(W[TWVL * 26]), Tz); TC = BYTWJ(&(W[TWVL * 6]), TB); Tj = BYTWJ(&(W[TWVL * 2]), Ti); Tk = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); } } TQ = VFNMS(LDK(KP500000000), Te, T9); Tf = VADD(T9, Te); } { V Tv, T13, TD, T14, Tl; Tv = VADD(Ts, Tu); T13 = VSUB(Tu, Ts); TD = VADD(TA, TC); T14 = VSUB(TC, TA); Tl = BYTWJ(&(W[TWVL * 12]), Tk); { V TT, Tw, T1d, T15, TU, TE, T11, Tm; TT = VFNMS(LDK(KP500000000), Tv, Tq); Tw = VADD(Tq, Tv); T1d = VADD(T13, T14); T15 = VSUB(T13, T14); TU = VFNMS(LDK(KP500000000), TD, Ty); TE = VADD(Ty, TD); T11 = VSUB(Tl, Tj); Tm = VADD(Tj, Tl); { V T19, TV, TK, TF, T1c, T12, TR, Tn; T19 = VSUB(TT, TU); TV = VADD(TT, TU); TK = VSUB(Tw, TE); TF = VADD(Tw, TE); T1c = VADD(T10, T11); T12 = VSUB(T10, T11); TR = VFNMS(LDK(KP500000000), Tm, Th); Tn = VADD(Th, Tm); { V T1g, T1e, T1m, T16, T18, TS, TL, To, T1f, T1u; T1g = VSUB(T1c, T1d); T1e = VADD(T1c, T1d); T1m = VFNMS(LDK(KP618033988), T12, T15); T16 = VFMA(LDK(KP618033988), T15, T12); T18 = VSUB(TQ, TR); TS = VADD(TQ, TR); TL = VSUB(Tf, Tn); To = VADD(Tf, Tn); T1f = VFNMS(LDK(KP250000000), T1e, T1b); T1u = VMUL(LDK(KP866025403), VADD(T1b, T1e)); { V T1o, T1a, TY, TO, TM, TG, TI, T1p, T1h, T1t, TX, TW; T1o = VFNMS(LDK(KP618033988), T18, T19); T1a = VFMA(LDK(KP618033988), T19, T18); TW = VADD(TS, TV); TY = VSUB(TS, TV); TO = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TK, TL)); TM = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TL, TK)); TG = VADD(To, TF); TI = VSUB(To, TF); T1p = VFNMS(LDK(KP559016994), T1g, T1f); T1h = VFMA(LDK(KP559016994), T1g, T1f); T1t = VADD(TP, TW); TX = VFNMS(LDK(KP250000000), TW, TP); { V T1q, T1s, T1k, T1i, T1l, TZ, TJ, TN, TH; ST(&(x[0]), VADD(T7, TG), ms, &(x[0])); TH = VFNMS(LDK(KP250000000), TG, T7); T1q = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1p, T1o)); T1s = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1p, T1o)); T1k = VMUL(LDK(KP951056516), VFMA(LDK(KP910592997), T1h, T1a)); T1i = VMUL(LDK(KP951056516), VFNMS(LDK(KP910592997), T1h, T1a)); ST(&(x[WS(rs, 10)]), VFMAI(T1u, T1t), ms, &(x[0])); ST(&(x[WS(rs, 5)]), VFNMSI(T1u, T1t), ms, &(x[WS(rs, 1)])); T1l = VFNMS(LDK(KP559016994), TY, TX); TZ = VFMA(LDK(KP559016994), TY, TX); TJ = VFNMS(LDK(KP559016994), TI, TH); TN = VFMA(LDK(KP559016994), TI, TH); { V T1n, T1r, T1j, T17; T1n = VFMA(LDK(KP823639103), T1m, T1l); T1r = VFNMS(LDK(KP823639103), T1m, T1l); T1j = VFNMS(LDK(KP823639103), T16, TZ); T17 = VFMA(LDK(KP823639103), T16, TZ); ST(&(x[WS(rs, 12)]), VFMAI(TM, TJ), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VFNMSI(TM, TJ), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 9)]), VFMAI(TO, TN), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 6)]), VFNMSI(TO, TN), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VFMAI(T1q, T1n), ms, &(x[0])); ST(&(x[WS(rs, 13)]), VFNMSI(T1q, T1n), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(T1s, T1r), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 8)]), VFNMSI(T1s, T1r), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VFMAI(T1k, T1j), ms, &(x[0])); ST(&(x[WS(rs, 11)]), VFNMSI(T1k, T1j), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 14)]), VFMAI(T1i, T17), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VFNMSI(T1i, T17), ms, &(x[WS(rs, 1)])); } } } } } } } } }
static void t1fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP923879532, +0.923879532511286756128183189396788286822416626); DVK(KP414213562, +0.414213562373095048801688724209698078569671875); DVK(KP707106781, +0.707106781186547524400844362104849039284835938); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) { V TO, Ta, TJ, TP, T14, Tq, T1i, T10, T1b, T1l, T13, T1c, TR, Tl, T15; V Tv; { V Tc, TW, T4, T19, T9, TD, TI, Tj, TZ, T1a, Te, Th, Tn, Tr, Tu; V Tp; { V T1, T2, T5, T7; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 12)]), ms, &(x[0])); { V Tz, TG, TB, TE; Tz = LD(&(x[WS(rs, 14)]), ms, &(x[0])); TG = LD(&(x[WS(rs, 10)]), ms, &(x[0])); TB = LD(&(x[WS(rs, 6)]), ms, &(x[0])); TE = LD(&(x[WS(rs, 2)]), ms, &(x[0])); { V Ti, TY, TX, Td, Tg, Tm, Tt, To; { V T3, T6, T8, TA, TH, TC, TF, Tb; Tb = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[TWVL * 14]), T2); T6 = BYTWJ(&(W[TWVL * 6]), T5); T8 = BYTWJ(&(W[TWVL * 22]), T7); TA = BYTWJ(&(W[TWVL * 26]), Tz); TH = BYTWJ(&(W[TWVL * 18]), TG); TC = BYTWJ(&(W[TWVL * 10]), TB); TF = BYTWJ(&(W[TWVL * 2]), TE); Tc = BYTWJ(&(W[0]), Tb); TW = VSUB(T1, T3); T4 = VADD(T1, T3); T19 = VSUB(T6, T8); T9 = VADD(T6, T8); Ti = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); TD = VADD(TA, TC); TY = VSUB(TA, TC); TI = VADD(TF, TH); TX = VSUB(TF, TH); } Td = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Tg = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Tm = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); Tj = BYTWJ(&(W[TWVL * 24]), Ti); Tt = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); To = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); TZ = VADD(TX, TY); T1a = VSUB(TY, TX); Te = BYTWJ(&(W[TWVL * 16]), Td); Th = BYTWJ(&(W[TWVL * 8]), Tg); Tn = BYTWJ(&(W[TWVL * 28]), Tm); Tr = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Tu = BYTWJ(&(W[TWVL * 20]), Tt); Tp = BYTWJ(&(W[TWVL * 12]), To); } } } { V Tf, T11, Tk, T12, Ts; TO = VADD(T4, T9); Ta = VSUB(T4, T9); TJ = VSUB(TD, TI); TP = VADD(TI, TD); Tf = VADD(Tc, Te); T11 = VSUB(Tc, Te); Tk = VADD(Th, Tj); T12 = VSUB(Th, Tj); Ts = BYTWJ(&(W[TWVL * 4]), Tr); T14 = VSUB(Tn, Tp); Tq = VADD(Tn, Tp); T1i = VFNMS(LDK(KP707106781), TZ, TW); T10 = VFMA(LDK(KP707106781), TZ, TW); T1b = VFNMS(LDK(KP707106781), T1a, T19); T1l = VFMA(LDK(KP707106781), T1a, T19); T13 = VFNMS(LDK(KP414213562), T12, T11); T1c = VFMA(LDK(KP414213562), T11, T12); TR = VADD(Tf, Tk); Tl = VSUB(Tf, Tk); T15 = VSUB(Tu, Ts); Tv = VADD(Ts, Tu); } } { V T1d, T16, TS, Tw, TU, TQ; T1d = VFMA(LDK(KP414213562), T14, T15); T16 = VFNMS(LDK(KP414213562), T15, T14); TS = VADD(Tq, Tv); Tw = VSUB(Tq, Tv); TU = VSUB(TO, TP); TQ = VADD(TO, TP); { V T1e, T1j, T17, T1m; T1e = VSUB(T1c, T1d); T1j = VADD(T1c, T1d); T17 = VADD(T13, T16); T1m = VSUB(T16, T13); { V TV, TT, TK, Tx; TV = VSUB(TS, TR); TT = VADD(TR, TS); TK = VSUB(Tw, Tl); Tx = VADD(Tl, Tw); { V T1h, T1f, T1o, T1k; T1h = VFMA(LDK(KP923879532), T1e, T1b); T1f = VFNMS(LDK(KP923879532), T1e, T1b); T1o = VFMA(LDK(KP923879532), T1j, T1i); T1k = VFNMS(LDK(KP923879532), T1j, T1i); { V T1g, T18, T1p, T1n; T1g = VFMA(LDK(KP923879532), T17, T10); T18 = VFNMS(LDK(KP923879532), T17, T10); T1p = VFMA(LDK(KP923879532), T1m, T1l); T1n = VFNMS(LDK(KP923879532), T1m, T1l); ST(&(x[WS(rs, 12)]), VFNMSI(TV, TU), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VFMAI(TV, TU), ms, &(x[0])); ST(&(x[0]), VADD(TQ, TT), ms, &(x[0])); ST(&(x[WS(rs, 8)]), VSUB(TQ, TT), ms, &(x[0])); { V TN, TL, TM, Ty; TN = VFMA(LDK(KP707106781), TK, TJ); TL = VFNMS(LDK(KP707106781), TK, TJ); TM = VFMA(LDK(KP707106781), Tx, Ta); Ty = VFNMS(LDK(KP707106781), Tx, Ta); ST(&(x[WS(rs, 1)]), VFNMSI(T1h, T1g), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 15)]), VFMAI(T1h, T1g), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(T1f, T18), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 9)]), VFNMSI(T1f, T18), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VFMAI(T1p, T1o), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 13)]), VFNMSI(T1p, T1o), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 11)]), VFMAI(T1n, T1k), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 5)]), VFNMSI(T1n, T1k), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 14)]), VFNMSI(TN, TM), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VFMAI(TN, TM), ms, &(x[0])); ST(&(x[WS(rs, 10)]), VFMAI(TL, Ty), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VFNMSI(TL, Ty), ms, &(x[0])); } } } } } } } } VLEAVE(); }
static void t2fv_32(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP831469612, +0.831469612302545237078788377617905756738560812); DVK(KP980785280, +0.980785280403230449126182236134239036973933731); DVK(KP668178637, +0.668178637919298919997757686523080761552472251); DVK(KP198912367, +0.198912367379658006911597622644676228597850501); DVK(KP923879532, +0.923879532511286756128183189396788286822416626); DVK(KP707106781, +0.707106781186547524400844362104849039284835938); DVK(KP414213562, +0.414213562373095048801688724209698078569671875); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 62)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(rs)) { V T26, T25, T1Z, T22, T1W, T2a, T2k, T2g; { V T4, T1z, T2o, T32, T2r, T3f, Tf, T1A, T34, T2L, T1D, TC, T33, T2O, T1C; V Tr, T2C, T3a, T2F, T3b, T1r, T21, T1k, T20, TQ, TM, TS, TL, T2t, TJ; V T10, T2u; { V Tt, T9, T2p, Te, T2q, TA, Tu, Tx; { V T1, T1x, T2, T1v; T1 = LD(&(x[0]), ms, &(x[0])); T1x = LD(&(x[WS(rs, 24)]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 16)]), ms, &(x[0])); T1v = LD(&(x[WS(rs, 8)]), ms, &(x[0])); { V T5, Tc, T7, Ta, T2m, T2n; T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 12)]), ms, &(x[0])); T7 = LD(&(x[WS(rs, 20)]), ms, &(x[0])); Ta = LD(&(x[WS(rs, 28)]), ms, &(x[0])); { V T1y, T3, T1w, T6, Td, T8, Tb, Ts, Tz; Ts = LD(&(x[WS(rs, 30)]), ms, &(x[0])); T1y = BYTWJ(&(W[TWVL * 46]), T1x); T3 = BYTWJ(&(W[TWVL * 30]), T2); T1w = BYTWJ(&(W[TWVL * 14]), T1v); T6 = BYTWJ(&(W[TWVL * 6]), T5); Td = BYTWJ(&(W[TWVL * 22]), Tc); T8 = BYTWJ(&(W[TWVL * 38]), T7); Tb = BYTWJ(&(W[TWVL * 54]), Ta); Tt = BYTWJ(&(W[TWVL * 58]), Ts); Tz = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T4 = VSUB(T1, T3); T2m = VADD(T1, T3); T1z = VSUB(T1w, T1y); T2n = VADD(T1w, T1y); T9 = VSUB(T6, T8); T2p = VADD(T6, T8); Te = VSUB(Tb, Td); T2q = VADD(Tb, Td); TA = BYTWJ(&(W[TWVL * 10]), Tz); } Tu = LD(&(x[WS(rs, 14)]), ms, &(x[0])); T2o = VADD(T2m, T2n); T32 = VSUB(T2m, T2n); Tx = LD(&(x[WS(rs, 22)]), ms, &(x[0])); } } { V Tv, To, Ty, Ti, Tj, Tm, Th; Th = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T2r = VADD(T2p, T2q); T3f = VSUB(T2q, T2p); Tf = VADD(T9, Te); T1A = VSUB(Te, T9); Tv = BYTWJ(&(W[TWVL * 26]), Tu); To = LD(&(x[WS(rs, 26)]), ms, &(x[0])); Ty = BYTWJ(&(W[TWVL * 42]), Tx); Ti = BYTWJ(&(W[TWVL * 2]), Th); Tj = LD(&(x[WS(rs, 18)]), ms, &(x[0])); Tm = LD(&(x[WS(rs, 10)]), ms, &(x[0])); { V T1f, T1h, T1a, T1c, T18, T2A, T2B, T1p; { V T15, T17, T1o, T1m; { V Tw, T2J, Tp, T2K, TB, Tk, Tn, T1n, T14, T16; T14 = LD(&(x[WS(rs, 31)]), ms, &(x[WS(rs, 1)])); T16 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); Tw = VSUB(Tt, Tv); T2J = VADD(Tt, Tv); Tp = BYTWJ(&(W[TWVL * 50]), To); T2K = VADD(TA, Ty); TB = VSUB(Ty, TA); Tk = BYTWJ(&(W[TWVL * 34]), Tj); Tn = BYTWJ(&(W[TWVL * 18]), Tm); T15 = BYTWJ(&(W[TWVL * 60]), T14); T17 = BYTWJ(&(W[TWVL * 28]), T16); T1n = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); { V T2M, Tl, T2N, Tq, T1l; T1l = LD(&(x[WS(rs, 23)]), ms, &(x[WS(rs, 1)])); T34 = VSUB(T2J, T2K); T2L = VADD(T2J, T2K); T1D = VFMA(LDK(KP414213562), Tw, TB); TC = VFNMS(LDK(KP414213562), TB, Tw); T2M = VADD(Ti, Tk); Tl = VSUB(Ti, Tk); T2N = VADD(Tn, Tp); Tq = VSUB(Tn, Tp); T1o = BYTWJ(&(W[TWVL * 12]), T1n); T1m = BYTWJ(&(W[TWVL * 44]), T1l); { V T1e, T1g, T19, T1b; T1e = LD(&(x[WS(rs, 27)]), ms, &(x[WS(rs, 1)])); T1g = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); T19 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T1b = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)])); T33 = VSUB(T2M, T2N); T2O = VADD(T2M, T2N); T1C = VFMA(LDK(KP414213562), Tl, Tq); Tr = VFNMS(LDK(KP414213562), Tq, Tl); T1f = BYTWJ(&(W[TWVL * 52]), T1e); T1h = BYTWJ(&(W[TWVL * 20]), T1g); T1a = BYTWJ(&(W[TWVL * 4]), T19); T1c = BYTWJ(&(W[TWVL * 36]), T1b); } } } T18 = VSUB(T15, T17); T2A = VADD(T15, T17); T2B = VADD(T1o, T1m); T1p = VSUB(T1m, T1o); } { V TG, TI, TZ, TX; { V T1i, T2E, T1d, T2D, TH, TY, TF; TF = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T1i = VSUB(T1f, T1h); T2E = VADD(T1f, T1h); T1d = VSUB(T1a, T1c); T2D = VADD(T1a, T1c); TH = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)])); TY = LD(&(x[WS(rs, 25)]), ms, &(x[WS(rs, 1)])); T2C = VADD(T2A, T2B); T3a = VSUB(T2A, T2B); TG = BYTWJ(&(W[0]), TF); { V TW, T1j, T1q, TP, TR, TK; TW = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); T2F = VADD(T2D, T2E); T3b = VSUB(T2E, T2D); T1j = VADD(T1d, T1i); T1q = VSUB(T1i, T1d); TI = BYTWJ(&(W[TWVL * 32]), TH); TZ = BYTWJ(&(W[TWVL * 48]), TY); TP = LD(&(x[WS(rs, 29)]), ms, &(x[WS(rs, 1)])); TX = BYTWJ(&(W[TWVL * 16]), TW); TR = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); TK = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); T1r = VFMA(LDK(KP707106781), T1q, T1p); T21 = VFNMS(LDK(KP707106781), T1q, T1p); T1k = VFMA(LDK(KP707106781), T1j, T18); T20 = VFNMS(LDK(KP707106781), T1j, T18); TQ = BYTWJ(&(W[TWVL * 56]), TP); TM = LD(&(x[WS(rs, 21)]), ms, &(x[WS(rs, 1)])); TS = BYTWJ(&(W[TWVL * 24]), TR); TL = BYTWJ(&(W[TWVL * 8]), TK); } } T2t = VADD(TG, TI); TJ = VSUB(TG, TI); T10 = VSUB(TX, TZ); T2u = VADD(TX, TZ); } } } } { V T2s, TT, T2x, T2P, T2Y, T2G, T37, T2v, T2w, TO, T2W, T30, T2U, TN, T2V; T2s = VSUB(T2o, T2r); T2U = VADD(T2o, T2r); TN = BYTWJ(&(W[TWVL * 40]), TM); TT = VSUB(TQ, TS); T2x = VADD(TQ, TS); T2P = VSUB(T2L, T2O); T2V = VADD(T2O, T2L); T2Y = VADD(T2C, T2F); T2G = VSUB(T2C, T2F); T37 = VSUB(T2t, T2u); T2v = VADD(T2t, T2u); T2w = VADD(TL, TN); TO = VSUB(TL, TN); T2W = VADD(T2U, T2V); T30 = VSUB(T2U, T2V); { V T3i, T3o, T36, T3r, T3h, T3j, T12, T1Y, TV, T1X, T3s, T3d, T2Q, T2H, T31; V T2Z; { V T35, T3g, T38, T2y, T11, TU; T35 = VADD(T33, T34); T3g = VSUB(T34, T33); T38 = VSUB(T2w, T2x); T2y = VADD(T2w, T2x); T11 = VSUB(TO, TT); TU = VADD(TO, TT); { V T3c, T39, T2X, T2z; T3c = VFNMS(LDK(KP414213562), T3b, T3a); T3i = VFMA(LDK(KP414213562), T3a, T3b); T3o = VFNMS(LDK(KP707106781), T35, T32); T36 = VFMA(LDK(KP707106781), T35, T32); T3r = VFNMS(LDK(KP707106781), T3g, T3f); T3h = VFMA(LDK(KP707106781), T3g, T3f); T39 = VFNMS(LDK(KP414213562), T38, T37); T3j = VFMA(LDK(KP414213562), T37, T38); T2X = VADD(T2v, T2y); T2z = VSUB(T2v, T2y); T12 = VFMA(LDK(KP707106781), T11, T10); T1Y = VFNMS(LDK(KP707106781), T11, T10); TV = VFMA(LDK(KP707106781), TU, TJ); T1X = VFNMS(LDK(KP707106781), TU, TJ); T3s = VSUB(T3c, T39); T3d = VADD(T39, T3c); T2Q = VSUB(T2G, T2z); T2H = VADD(T2z, T2G); T31 = VSUB(T2Y, T2X); T2Z = VADD(T2X, T2Y); } } { V Tg, T1U, TD, T1G, T13, T1s, T1H, T1B, T1V, T1E, T3k, T3p, T2e, T2f; Tg = VFMA(LDK(KP707106781), Tf, T4); T1U = VFNMS(LDK(KP707106781), Tf, T4); T3k = VSUB(T3i, T3j); T3p = VADD(T3j, T3i); { V T3v, T3t, T3e, T3m; T3v = VFNMS(LDK(KP923879532), T3s, T3r); T3t = VFMA(LDK(KP923879532), T3s, T3r); T3e = VFNMS(LDK(KP923879532), T3d, T36); T3m = VFMA(LDK(KP923879532), T3d, T36); { V T2R, T2T, T2I, T2S; T2R = VFNMS(LDK(KP707106781), T2Q, T2P); T2T = VFMA(LDK(KP707106781), T2Q, T2P); T2I = VFNMS(LDK(KP707106781), T2H, T2s); T2S = VFMA(LDK(KP707106781), T2H, T2s); ST(&(x[WS(rs, 24)]), VFNMSI(T31, T30), ms, &(x[0])); ST(&(x[WS(rs, 8)]), VFMAI(T31, T30), ms, &(x[0])); ST(&(x[0]), VADD(T2W, T2Z), ms, &(x[0])); ST(&(x[WS(rs, 16)]), VSUB(T2W, T2Z), ms, &(x[0])); { V T3u, T3q, T3l, T3n; T3u = VFMA(LDK(KP923879532), T3p, T3o); T3q = VFNMS(LDK(KP923879532), T3p, T3o); T3l = VFNMS(LDK(KP923879532), T3k, T3h); T3n = VFMA(LDK(KP923879532), T3k, T3h); ST(&(x[WS(rs, 4)]), VFMAI(T2T, T2S), ms, &(x[0])); ST(&(x[WS(rs, 28)]), VFNMSI(T2T, T2S), ms, &(x[0])); ST(&(x[WS(rs, 20)]), VFMAI(T2R, T2I), ms, &(x[0])); ST(&(x[WS(rs, 12)]), VFNMSI(T2R, T2I), ms, &(x[0])); ST(&(x[WS(rs, 22)]), VFNMSI(T3t, T3q), ms, &(x[0])); ST(&(x[WS(rs, 10)]), VFMAI(T3t, T3q), ms, &(x[0])); ST(&(x[WS(rs, 26)]), VFMAI(T3v, T3u), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VFNMSI(T3v, T3u), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VFMAI(T3n, T3m), ms, &(x[0])); ST(&(x[WS(rs, 30)]), VFNMSI(T3n, T3m), ms, &(x[0])); ST(&(x[WS(rs, 18)]), VFMAI(T3l, T3e), ms, &(x[0])); ST(&(x[WS(rs, 14)]), VFNMSI(T3l, T3e), ms, &(x[0])); T26 = VSUB(TC, Tr); TD = VADD(Tr, TC); } } } T1G = VFMA(LDK(KP198912367), TV, T12); T13 = VFNMS(LDK(KP198912367), T12, TV); T1s = VFNMS(LDK(KP198912367), T1r, T1k); T1H = VFMA(LDK(KP198912367), T1k, T1r); T1B = VFNMS(LDK(KP707106781), T1A, T1z); T25 = VFMA(LDK(KP707106781), T1A, T1z); T1V = VADD(T1C, T1D); T1E = VSUB(T1C, T1D); { V T1S, T1O, T1K, T1u, T1R, T1T, T1L, T1J; { V TE, T1M, T1I, T1N, T1t, T1Q, T1F, T1P, T28, T29; TE = VFMA(LDK(KP923879532), TD, Tg); T1M = VFNMS(LDK(KP923879532), TD, Tg); T1I = VSUB(T1G, T1H); T1N = VADD(T1G, T1H); T1t = VADD(T13, T1s); T1Q = VSUB(T1s, T13); T1F = VFMA(LDK(KP923879532), T1E, T1B); T1P = VFNMS(LDK(KP923879532), T1E, T1B); T28 = VFNMS(LDK(KP668178637), T1X, T1Y); T1Z = VFMA(LDK(KP668178637), T1Y, T1X); T1S = VFMA(LDK(KP980785280), T1N, T1M); T1O = VFNMS(LDK(KP980785280), T1N, T1M); T22 = VFMA(LDK(KP668178637), T21, T20); T29 = VFNMS(LDK(KP668178637), T20, T21); T1K = VFMA(LDK(KP980785280), T1t, TE); T1u = VFNMS(LDK(KP980785280), T1t, TE); T1R = VFNMS(LDK(KP980785280), T1Q, T1P); T1T = VFMA(LDK(KP980785280), T1Q, T1P); T1L = VFMA(LDK(KP980785280), T1I, T1F); T1J = VFNMS(LDK(KP980785280), T1I, T1F); T2e = VFNMS(LDK(KP923879532), T1V, T1U); T1W = VFMA(LDK(KP923879532), T1V, T1U); T2a = VSUB(T28, T29); T2f = VADD(T28, T29); } ST(&(x[WS(rs, 23)]), VFMAI(T1R, T1O), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 9)]), VFNMSI(T1R, T1O), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 25)]), VFNMSI(T1T, T1S), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(T1T, T1S), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 31)]), VFMAI(T1L, T1K), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VFNMSI(T1L, T1K), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 15)]), VFMAI(T1J, T1u), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 17)]), VFNMSI(T1J, T1u), ms, &(x[WS(rs, 1)])); } T2k = VFNMS(LDK(KP831469612), T2f, T2e); T2g = VFMA(LDK(KP831469612), T2f, T2e); } } } } { V T2i, T23, T2h, T27; T2i = VSUB(T22, T1Z); T23 = VADD(T1Z, T22); T2h = VFNMS(LDK(KP923879532), T26, T25); T27 = VFMA(LDK(KP923879532), T26, T25); { V T2c, T24, T2j, T2l, T2d, T2b; T2c = VFMA(LDK(KP831469612), T23, T1W); T24 = VFNMS(LDK(KP831469612), T23, T1W); T2j = VFMA(LDK(KP831469612), T2i, T2h); T2l = VFNMS(LDK(KP831469612), T2i, T2h); T2d = VFMA(LDK(KP831469612), T2a, T27); T2b = VFNMS(LDK(KP831469612), T2a, T27); ST(&(x[WS(rs, 21)]), VFNMSI(T2j, T2g), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 11)]), VFMAI(T2j, T2g), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 27)]), VFMAI(T2l, T2k), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 5)]), VFNMSI(T2l, T2k), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VFMAI(T2d, T2c), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 29)]), VFNMSI(T2d, T2c), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 19)]), VFMAI(T2b, T24), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 13)]), VFNMSI(T2b, T24), ms, &(x[WS(rs, 1)])); } } } } VLEAVE(); }
static void t1fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP587785252, +0.587785252292473129168705954639072768597652438); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP559016994, +0.559016994374947424102293417182819058860154590); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) { V Tr, TH, Tg, Tl, Tm, TA, TB, TJ, T5, Ta, Tb, TD, TE, TI, To; V Tq, Tp; To = LD(&(x[0]), ms, &(x[0])); Tp = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Tq = BYTWJ(&(W[TWVL * 8]), Tp); Tr = VSUB(To, Tq); TH = VADD(To, Tq); { V Td, Tk, Tf, Ti; { V Tc, Tj, Te, Th; Tc = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Td = BYTWJ(&(W[TWVL * 6]), Tc); Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tk = BYTWJ(&(W[0]), Tj); Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Tf = BYTWJ(&(W[TWVL * 16]), Te); Th = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Ti = BYTWJ(&(W[TWVL * 10]), Th); } Tg = VSUB(Td, Tf); Tl = VSUB(Ti, Tk); Tm = VADD(Tg, Tl); TA = VADD(Td, Tf); TB = VADD(Ti, Tk); TJ = VADD(TA, TB); } { V T2, T9, T4, T7; { V T1, T8, T3, T6; T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); T2 = BYTWJ(&(W[TWVL * 2]), T1); T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[TWVL * 4]), T8); T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); T4 = BYTWJ(&(W[TWVL * 12]), T3); T6 = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T7 = BYTWJ(&(W[TWVL * 14]), T6); } T5 = VSUB(T2, T4); Ta = VSUB(T7, T9); Tb = VADD(T5, Ta); TD = VADD(T2, T4); TE = VADD(T7, T9); TI = VADD(TD, TE); } { V Tn, Ts, Tt, Tx, Tz, Tv, Tw, Ty, Tu; Tn = VMUL(LDK(KP559016994), VSUB(Tb, Tm)); Ts = VADD(Tb, Tm); Tt = VFNMS(LDK(KP250000000), Ts, Tr); Tv = VSUB(T5, Ta); Tw = VSUB(Tg, Tl); Tx = VBYI(VFMA(LDK(KP951056516), Tv, VMUL(LDK(KP587785252), Tw))); Tz = VBYI(VFNMS(LDK(KP587785252), Tv, VMUL(LDK(KP951056516), Tw))); ST(&(x[WS(rs, 5)]), VADD(Tr, Ts), ms, &(x[WS(rs, 1)])); Ty = VSUB(Tt, Tn); ST(&(x[WS(rs, 3)]), VSUB(Ty, Tz), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VADD(Tz, Ty), ms, &(x[WS(rs, 1)])); Tu = VADD(Tn, Tt); ST(&(x[WS(rs, 1)]), VSUB(Tu, Tx), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 9)]), VADD(Tx, Tu), ms, &(x[WS(rs, 1)])); } { V TM, TK, TL, TG, TO, TC, TF, TP, TN; TM = VMUL(LDK(KP559016994), VSUB(TI, TJ)); TK = VADD(TI, TJ); TL = VFNMS(LDK(KP250000000), TK, TH); TC = VSUB(TA, TB); TF = VSUB(TD, TE); TG = VBYI(VFNMS(LDK(KP587785252), TF, VMUL(LDK(KP951056516), TC))); TO = VBYI(VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TC))); ST(&(x[0]), VADD(TH, TK), ms, &(x[0])); TP = VADD(TM, TL); ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VSUB(TP, TO), ms, &(x[0])); TN = VSUB(TL, TM); ST(&(x[WS(rs, 2)]), VADD(TG, TN), ms, &(x[0])); ST(&(x[WS(rs, 8)]), VSUB(TN, TG), ms, &(x[0])); } } } VLEAVE(); }
static void t2fv_8(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP707106781, +0.707106781186547524400844362104849039284835938); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 14)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 14), MAKE_VOLATILE_STRIDE(rs)) { V T1, T2, Th, Tj, T5, T7, Ta, Tc; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Th = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tj = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T5 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T7 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Ta = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); { V T3, Ti, Tk, T6, T8, Tb, Td; T3 = BYTWJ(&(W[TWVL * 6]), T2); Ti = BYTWJ(&(W[TWVL * 2]), Th); Tk = BYTWJ(&(W[TWVL * 10]), Tj); T6 = BYTWJ(&(W[0]), T5); T8 = BYTWJ(&(W[TWVL * 8]), T7); Tb = BYTWJ(&(W[TWVL * 12]), Ta); Td = BYTWJ(&(W[TWVL * 4]), Tc); { V Tq, T4, Tr, Tl, Tt, T9, Tu, Te, Tw, Ts; Tq = VADD(T1, T3); T4 = VSUB(T1, T3); Tr = VADD(Ti, Tk); Tl = VSUB(Ti, Tk); Tt = VADD(T6, T8); T9 = VSUB(T6, T8); Tu = VADD(Tb, Td); Te = VSUB(Tb, Td); Tw = VSUB(Tq, Tr); Ts = VADD(Tq, Tr); { V Tx, Tv, Tm, Tf; Tx = VSUB(Tu, Tt); Tv = VADD(Tt, Tu); Tm = VSUB(Te, T9); Tf = VADD(T9, Te); { V Tp, Tn, To, Tg; ST(&(x[WS(rs, 2)]), VFMAI(Tx, Tw), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VFNMSI(Tx, Tw), ms, &(x[0])); ST(&(x[0]), VADD(Ts, Tv), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VSUB(Ts, Tv), ms, &(x[0])); Tp = VFMA(LDK(KP707106781), Tm, Tl); Tn = VFNMS(LDK(KP707106781), Tm, Tl); To = VFNMS(LDK(KP707106781), Tf, T4); Tg = VFMA(LDK(KP707106781), Tf, T4); ST(&(x[WS(rs, 5)]), VFNMSI(Tp, To), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VFMAI(Tp, To), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(Tn, Tg), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VFNMSI(Tn, Tg), ms, &(x[WS(rs, 1)])); } } } } } } VLEAVE(); }
static void t1fv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP559016994, +0.559016994374947424102293417182819058860154590); DVK(KP250000000, +0.250000000000000000000000000000000000000000000); DVK(KP618033988, +0.618033988749894848204586834365638117720309180); DVK(KP951056516, +0.951056516295153572116439333379382143405698634); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) { V Td, TA, T4, Ta, Tk, TE, Tp, TF, TB, T9, T1, T2, Tb; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); { V Tg, Tn, Ti, Tl; Tg = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tn = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Ti = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Tl = LD(&(x[WS(rs, 6)]), ms, &(x[0])); { V T6, T8, T5, Tc; T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); { V T3, Th, To, Tj, Tm, T7; T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); T3 = BYTWJ(&(W[TWVL * 8]), T2); Th = BYTWJ(&(W[TWVL * 6]), Tg); To = BYTWJ(&(W[0]), Tn); Tj = BYTWJ(&(W[TWVL * 16]), Ti); Tm = BYTWJ(&(W[TWVL * 10]), Tl); T6 = BYTWJ(&(W[TWVL * 2]), T5); Td = BYTWJ(&(W[TWVL * 4]), Tc); T8 = BYTWJ(&(W[TWVL * 12]), T7); TA = VADD(T1, T3); T4 = VSUB(T1, T3); Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0])); Tk = VSUB(Th, Tj); TE = VADD(Th, Tj); Tp = VSUB(Tm, To); TF = VADD(Tm, To); } TB = VADD(T6, T8); T9 = VSUB(T6, T8); } } Tb = BYTWJ(&(W[TWVL * 14]), Ta); { V TL, TG, Tw, Tq, TC, Te; TL = VSUB(TE, TF); TG = VADD(TE, TF); Tw = VSUB(Tk, Tp); Tq = VADD(Tk, Tp); TC = VADD(Tb, Td); Te = VSUB(Tb, Td); { V TM, TD, Tv, Tf; TM = VSUB(TB, TC); TD = VADD(TB, TC); Tv = VSUB(T9, Te); Tf = VADD(T9, Te); { V TP, TN, TH, TJ, Tz, Tx, Tr, Tt, TI, Ts; TP = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TL, TM)); TN = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TM, TL)); TH = VADD(TD, TG); TJ = VSUB(TD, TG); Tz = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tv, Tw)); Tx = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tv)); Tr = VADD(Tf, Tq); Tt = VSUB(Tf, Tq); ST(&(x[0]), VADD(TA, TH), ms, &(x[0])); TI = VFNMS(LDK(KP250000000), TH, TA); ST(&(x[WS(rs, 5)]), VADD(T4, Tr), ms, &(x[WS(rs, 1)])); Ts = VFNMS(LDK(KP250000000), Tr, T4); { V TK, TO, Tu, Ty; TK = VFNMS(LDK(KP559016994), TJ, TI); TO = VFMA(LDK(KP559016994), TJ, TI); Tu = VFMA(LDK(KP559016994), Tt, Ts); Ty = VFNMS(LDK(KP559016994), Tt, Ts); ST(&(x[WS(rs, 8)]), VFNMSI(TN, TK), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VFNMSI(TP, TO), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0])); ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tu), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VFNMSI(Tx, Tu), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(Tz, Ty), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VFNMSI(Tz, Ty), ms, &(x[WS(rs, 1)])); } } } } } } VLEAVE(); }
static void q1fv_4(R *ri, R *ii, const R *W, stride rs, stride vs, INT mb, INT me, INT ms) { { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 6)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 6), MAKE_VOLATILE_STRIDE(8, rs), MAKE_VOLATILE_STRIDE(8, vs)) { V Tb, Tm, Tx, TI; { V Tc, T9, T3, TG, TA, TH, TD, Ta, T6, Td, Tn, To, Tq, Tr, Tf; V Tg; { V T1, T2, Ty, Tz, TB, TC, T4, T5; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Ty = LD(&(x[WS(vs, 3)]), ms, &(x[WS(vs, 3)])); Tz = LD(&(x[WS(vs, 3) + WS(rs, 2)]), ms, &(x[WS(vs, 3)])); TB = LD(&(x[WS(vs, 3) + WS(rs, 1)]), ms, &(x[WS(vs, 3) + WS(rs, 1)])); TC = LD(&(x[WS(vs, 3) + WS(rs, 3)]), ms, &(x[WS(vs, 3) + WS(rs, 1)])); T4 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T5 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); Tc = LD(&(x[WS(vs, 1)]), ms, &(x[WS(vs, 1)])); T9 = VADD(T1, T2); T3 = VSUB(T1, T2); TG = VADD(Ty, Tz); TA = VSUB(Ty, Tz); TH = VADD(TB, TC); TD = VSUB(TB, TC); Ta = VADD(T4, T5); T6 = VSUB(T4, T5); Td = LD(&(x[WS(vs, 1) + WS(rs, 2)]), ms, &(x[WS(vs, 1)])); Tn = LD(&(x[WS(vs, 2)]), ms, &(x[WS(vs, 2)])); To = LD(&(x[WS(vs, 2) + WS(rs, 2)]), ms, &(x[WS(vs, 2)])); Tq = LD(&(x[WS(vs, 2) + WS(rs, 1)]), ms, &(x[WS(vs, 2) + WS(rs, 1)])); Tr = LD(&(x[WS(vs, 2) + WS(rs, 3)]), ms, &(x[WS(vs, 2) + WS(rs, 1)])); Tf = LD(&(x[WS(vs, 1) + WS(rs, 1)]), ms, &(x[WS(vs, 1) + WS(rs, 1)])); Tg = LD(&(x[WS(vs, 1) + WS(rs, 3)]), ms, &(x[WS(vs, 1) + WS(rs, 1)])); } { V Tk, Te, Tv, Tp, Tw, Ts, Tl, Th, T7, TE, Tu, TF; ST(&(x[0]), VADD(T9, Ta), ms, &(x[0])); Tk = VADD(Tc, Td); Te = VSUB(Tc, Td); Tv = VADD(Tn, To); Tp = VSUB(Tn, To); Tw = VADD(Tq, Tr); Ts = VSUB(Tq, Tr); Tl = VADD(Tf, Tg); Th = VSUB(Tf, Tg); ST(&(x[WS(rs, 3)]), VADD(TG, TH), ms, &(x[WS(rs, 1)])); T7 = BYTWJ(&(W[0]), VFNMSI(T6, T3)); TE = BYTWJ(&(W[0]), VFNMSI(TD, TA)); { V Tt, Ti, Tj, T8; T8 = BYTWJ(&(W[TWVL * 4]), VFMAI(T6, T3)); ST(&(x[WS(rs, 2)]), VADD(Tv, Tw), ms, &(x[0])); Tt = BYTWJ(&(W[0]), VFNMSI(Ts, Tp)); ST(&(x[WS(rs, 1)]), VADD(Tk, Tl), ms, &(x[WS(rs, 1)])); Ti = BYTWJ(&(W[0]), VFNMSI(Th, Te)); Tj = BYTWJ(&(W[TWVL * 4]), VFMAI(Th, Te)); ST(&(x[WS(vs, 1)]), T7, ms, &(x[WS(vs, 1)])); ST(&(x[WS(vs, 1) + WS(rs, 3)]), TE, ms, &(x[WS(vs, 1) + WS(rs, 1)])); ST(&(x[WS(vs, 3)]), T8, ms, &(x[WS(vs, 3)])); Tu = BYTWJ(&(W[TWVL * 4]), VFMAI(Ts, Tp)); ST(&(x[WS(vs, 1) + WS(rs, 2)]), Tt, ms, &(x[WS(vs, 1)])); TF = BYTWJ(&(W[TWVL * 4]), VFMAI(TD, TA)); ST(&(x[WS(vs, 1) + WS(rs, 1)]), Ti, ms, &(x[WS(vs, 1) + WS(rs, 1)])); ST(&(x[WS(vs, 3) + WS(rs, 1)]), Tj, ms, &(x[WS(vs, 3) + WS(rs, 1)])); } Tb = BYTWJ(&(W[TWVL * 2]), VSUB(T9, Ta)); Tm = BYTWJ(&(W[TWVL * 2]), VSUB(Tk, Tl)); Tx = BYTWJ(&(W[TWVL * 2]), VSUB(Tv, Tw)); ST(&(x[WS(vs, 3) + WS(rs, 2)]), Tu, ms, &(x[WS(vs, 3)])); TI = BYTWJ(&(W[TWVL * 2]), VSUB(TG, TH)); ST(&(x[WS(vs, 3) + WS(rs, 3)]), TF, ms, &(x[WS(vs, 3) + WS(rs, 1)])); } } ST(&(x[WS(vs, 2)]), Tb, ms, &(x[WS(vs, 2)])); ST(&(x[WS(vs, 2) + WS(rs, 1)]), Tm, ms, &(x[WS(vs, 2) + WS(rs, 1)])); ST(&(x[WS(vs, 2) + WS(rs, 2)]), Tx, ms, &(x[WS(vs, 2)])); ST(&(x[WS(vs, 2) + WS(rs, 3)]), TI, ms, &(x[WS(vs, 2) + WS(rs, 1)])); } } VLEAVE(); }
static void t1fuv_9(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP939692620, +0.939692620785908384054109277324731469936208134); DVK(KP826351822, +0.826351822333069651148283373230685203999624323); DVK(KP879385241, +0.879385241571816768108218554649462939872416269); DVK(KP984807753, +0.984807753012208059366743024589523013670643252); DVK(KP666666666, +0.666666666666666666666666666666666666666666667); DVK(KP852868531, +0.852868531952443209628250963940074071936020296); DVK(KP907603734, +0.907603734547952313649323976213898122064543220); DVK(KP420276625, +0.420276625461206169731530603237061658838781920); DVK(KP673648177, +0.673648177666930348851716626769314796000375677); DVK(KP898197570, +0.898197570222573798468955502359086394667167570); DVK(KP347296355, +0.347296355333860697703433253538629592000751354); DVK(KP866025403, +0.866025403784438646763723170752936183471402627); DVK(KP439692620, +0.439692620785908384054109277324731469936208134); DVK(KP203604859, +0.203604859554852403062088995281827210665664861); DVK(KP152703644, +0.152703644666139302296566746461370407999248646); DVK(KP586256827, +0.586256827714544512072145703099641959914944179); DVK(KP968908795, +0.968908795874236621082202410917456709164223497); DVK(KP726681596, +0.726681596905677465811651808188092531873167623); DVK(KP500000000, +0.500000000000000000000000000000000000000000000); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 16)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 16), MAKE_VOLATILE_STRIDE(rs)) { V T1, T3, T5, T9, Th, Tb, Td, Tj, Tl, TD, T6; T1 = LD(&(x[0]), ms, &(x[0])); { V T2, T4, T8, Tg; T2 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); T8 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Tg = LD(&(x[WS(rs, 2)]), ms, &(x[0])); { V Ta, Tc, Ti, Tk; Ta = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); Ti = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Tk = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 4]), T2); T5 = BYTWJ(&(W[TWVL * 10]), T4); T9 = BYTWJ(&(W[0]), T8); Th = BYTWJ(&(W[TWVL * 2]), Tg); Tb = BYTWJ(&(W[TWVL * 6]), Ta); Td = BYTWJ(&(W[TWVL * 12]), Tc); Tj = BYTWJ(&(W[TWVL * 8]), Ti); Tl = BYTWJ(&(W[TWVL * 14]), Tk); } } TD = VSUB(T5, T3); T6 = VADD(T3, T5); { V Tt, Te, Tu, Tm, Tr, T7; Tt = VSUB(Tb, Td); Te = VADD(Tb, Td); Tu = VSUB(Tl, Tj); Tm = VADD(Tj, Tl); Tr = VFNMS(LDK(KP500000000), T6, T1); T7 = VADD(T1, T6); { V Tv, Tf, Ts, Tn; Tv = VFNMS(LDK(KP500000000), Te, T9); Tf = VADD(T9, Te); Ts = VFNMS(LDK(KP500000000), Tm, Th); Tn = VADD(Th, Tm); { V TG, TK, Tw, TJ, TF, TA, To, Tq; TG = VFNMS(LDK(KP726681596), Tt, Tv); TK = VFMA(LDK(KP968908795), Tv, Tt); Tw = VFNMS(LDK(KP586256827), Tv, Tu); TJ = VFNMS(LDK(KP152703644), Tu, Ts); TF = VFMA(LDK(KP203604859), Ts, Tu); TA = VFNMS(LDK(KP439692620), Tt, Ts); To = VADD(Tf, Tn); Tq = VMUL(LDK(KP866025403), VSUB(Tn, Tf)); { V TQ, TH, TL, TN, TB, Tp, Ty, TI, Tx; Tx = VFNMS(LDK(KP347296355), Tw, Tt); TQ = VFNMS(LDK(KP898197570), TG, TF); TH = VFMA(LDK(KP898197570), TG, TF); TL = VFMA(LDK(KP673648177), TK, TJ); TN = VFNMS(LDK(KP673648177), TK, TJ); TB = VFNMS(LDK(KP420276625), TA, Tu); ST(&(x[0]), VADD(T7, To), ms, &(x[0])); Tp = VFNMS(LDK(KP500000000), To, T7); Ty = VFNMS(LDK(KP907603734), Tx, Ts); TI = VFMA(LDK(KP852868531), TH, Tr); { V TO, TR, TM, TC, Tz, TP, TS, TE; TO = VFNMS(LDK(KP500000000), TH, TN); TR = VFMA(LDK(KP666666666), TL, TQ); TM = VMUL(LDK(KP984807753), VFNMS(LDK(KP879385241), TD, TL)); TC = VFNMS(LDK(KP826351822), TB, Tv); ST(&(x[WS(rs, 6)]), VFNMSI(Tq, Tp), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VFMAI(Tq, Tp), ms, &(x[WS(rs, 1)])); Tz = VFNMS(LDK(KP939692620), Ty, Tr); TP = VFMA(LDK(KP852868531), TO, Tr); TS = VMUL(LDK(KP866025403), VFMA(LDK(KP852868531), TR, TD)); ST(&(x[WS(rs, 8)]), VFMAI(TM, TI), ms, &(x[0])); ST(&(x[WS(rs, 1)]), VFNMSI(TM, TI), ms, &(x[WS(rs, 1)])); TE = VMUL(LDK(KP984807753), VFMA(LDK(KP879385241), TD, TC)); ST(&(x[WS(rs, 4)]), VFMAI(TS, TP), ms, &(x[0])); ST(&(x[WS(rs, 5)]), VFNMSI(TS, TP), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 7)]), VFMAI(TE, Tz), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 2)]), VFNMSI(TE, Tz), ms, &(x[0])); } } } } } } } VLEAVE(); }
static void t2fv_32(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP555570233, +0.555570233019602224742830813948532874374937191); DVK(KP831469612, +0.831469612302545237078788377617905756738560812); DVK(KP195090322, +0.195090322016128267848284868477022240927691618); DVK(KP980785280, +0.980785280403230449126182236134239036973933731); DVK(KP382683432, +0.382683432365089771728459984030398866761344562); DVK(KP923879532, +0.923879532511286756128183189396788286822416626); DVK(KP707106781, +0.707106781186547524400844362104849039284835938); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 62)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 62), MAKE_VOLATILE_STRIDE(rs)) { V T4, T1A, T2o, T32, Tf, T1v, T2r, T3f, TC, T1C, T2L, T34, Tr, T1D, T2O; V T33, T1k, T20, T2F, T3b, T1r, T21, T2C, T3a, TV, T1X, T2y, T38, T12, T1Y; V T2v, T37; { V T1, T1z, T3, T1x, T1y, T2, T1w, T2m, T2n; T1 = LD(&(x[0]), ms, &(x[0])); T1y = LD(&(x[WS(rs, 24)]), ms, &(x[0])); T1z = BYTWJ(&(W[TWVL * 46]), T1y); T2 = LD(&(x[WS(rs, 16)]), ms, &(x[0])); T3 = BYTWJ(&(W[TWVL * 30]), T2); T1w = LD(&(x[WS(rs, 8)]), ms, &(x[0])); T1x = BYTWJ(&(W[TWVL * 14]), T1w); T4 = VSUB(T1, T3); T1A = VSUB(T1x, T1z); T2m = VADD(T1, T3); T2n = VADD(T1x, T1z); T2o = VADD(T2m, T2n); T32 = VSUB(T2m, T2n); } { V T6, Td, T8, Tb; { V T5, Tc, T7, Ta; T5 = LD(&(x[WS(rs, 4)]), ms, &(x[0])); T6 = BYTWJ(&(W[TWVL * 6]), T5); Tc = LD(&(x[WS(rs, 12)]), ms, &(x[0])); Td = BYTWJ(&(W[TWVL * 22]), Tc); T7 = LD(&(x[WS(rs, 20)]), ms, &(x[0])); T8 = BYTWJ(&(W[TWVL * 38]), T7); Ta = LD(&(x[WS(rs, 28)]), ms, &(x[0])); Tb = BYTWJ(&(W[TWVL * 54]), Ta); } { V T9, Te, T2p, T2q; T9 = VSUB(T6, T8); Te = VSUB(Tb, Td); Tf = VMUL(LDK(KP707106781), VADD(T9, Te)); T1v = VMUL(LDK(KP707106781), VSUB(Te, T9)); T2p = VADD(T6, T8); T2q = VADD(Tb, Td); T2r = VADD(T2p, T2q); T3f = VSUB(T2q, T2p); } } { V Tt, TA, Tv, Ty; { V Ts, Tz, Tu, Tx; Ts = LD(&(x[WS(rs, 30)]), ms, &(x[0])); Tt = BYTWJ(&(W[TWVL * 58]), Ts); Tz = LD(&(x[WS(rs, 22)]), ms, &(x[0])); TA = BYTWJ(&(W[TWVL * 42]), Tz); Tu = LD(&(x[WS(rs, 14)]), ms, &(x[0])); Tv = BYTWJ(&(W[TWVL * 26]), Tu); Tx = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Ty = BYTWJ(&(W[TWVL * 10]), Tx); } { V Tw, TB, T2J, T2K; Tw = VSUB(Tt, Tv); TB = VSUB(Ty, TA); TC = VFMA(LDK(KP923879532), Tw, VMUL(LDK(KP382683432), TB)); T1C = VFNMS(LDK(KP923879532), TB, VMUL(LDK(KP382683432), Tw)); T2J = VADD(Tt, Tv); T2K = VADD(Ty, TA); T2L = VADD(T2J, T2K); T34 = VSUB(T2J, T2K); } } { V Ti, Tp, Tk, Tn; { V Th, To, Tj, Tm; Th = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Ti = BYTWJ(&(W[TWVL * 2]), Th); To = LD(&(x[WS(rs, 26)]), ms, &(x[0])); Tp = BYTWJ(&(W[TWVL * 50]), To); Tj = LD(&(x[WS(rs, 18)]), ms, &(x[0])); Tk = BYTWJ(&(W[TWVL * 34]), Tj); Tm = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Tn = BYTWJ(&(W[TWVL * 18]), Tm); } { V Tl, Tq, T2M, T2N; Tl = VSUB(Ti, Tk); Tq = VSUB(Tn, Tp); Tr = VFNMS(LDK(KP382683432), Tq, VMUL(LDK(KP923879532), Tl)); T1D = VFMA(LDK(KP382683432), Tl, VMUL(LDK(KP923879532), Tq)); T2M = VADD(Ti, Tk); T2N = VADD(Tn, Tp); T2O = VADD(T2M, T2N); T33 = VSUB(T2M, T2N); } } { V T15, T17, T1p, T1n, T1f, T1h, T1i, T1a, T1c, T1d; { V T14, T16, T1o, T1m; T14 = LD(&(x[WS(rs, 31)]), ms, &(x[WS(rs, 1)])); T15 = BYTWJ(&(W[TWVL * 60]), T14); T16 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); T17 = BYTWJ(&(W[TWVL * 28]), T16); T1o = LD(&(x[WS(rs, 23)]), ms, &(x[WS(rs, 1)])); T1p = BYTWJ(&(W[TWVL * 44]), T1o); T1m = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); T1n = BYTWJ(&(W[TWVL * 12]), T1m); { V T1e, T1g, T19, T1b; T1e = LD(&(x[WS(rs, 27)]), ms, &(x[WS(rs, 1)])); T1f = BYTWJ(&(W[TWVL * 52]), T1e); T1g = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); T1h = BYTWJ(&(W[TWVL * 20]), T1g); T1i = VSUB(T1f, T1h); T19 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T1a = BYTWJ(&(W[TWVL * 4]), T19); T1b = LD(&(x[WS(rs, 19)]), ms, &(x[WS(rs, 1)])); T1c = BYTWJ(&(W[TWVL * 36]), T1b); T1d = VSUB(T1a, T1c); } } { V T18, T1j, T2D, T2E; T18 = VSUB(T15, T17); T1j = VMUL(LDK(KP707106781), VADD(T1d, T1i)); T1k = VADD(T18, T1j); T20 = VSUB(T18, T1j); T2D = VADD(T1a, T1c); T2E = VADD(T1f, T1h); T2F = VADD(T2D, T2E); T3b = VSUB(T2E, T2D); } { V T1l, T1q, T2A, T2B; T1l = VMUL(LDK(KP707106781), VSUB(T1i, T1d)); T1q = VSUB(T1n, T1p); T1r = VSUB(T1l, T1q); T21 = VADD(T1q, T1l); T2A = VADD(T15, T17); T2B = VADD(T1n, T1p); T2C = VADD(T2A, T2B); T3a = VSUB(T2A, T2B); } } { V TG, TI, T10, TY, TQ, TS, TT, TL, TN, TO; { V TF, TH, TZ, TX; TF = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); TG = BYTWJ(&(W[0]), TF); TH = LD(&(x[WS(rs, 17)]), ms, &(x[WS(rs, 1)])); TI = BYTWJ(&(W[TWVL * 32]), TH); TZ = LD(&(x[WS(rs, 25)]), ms, &(x[WS(rs, 1)])); T10 = BYTWJ(&(W[TWVL * 48]), TZ); TX = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); TY = BYTWJ(&(W[TWVL * 16]), TX); { V TP, TR, TK, TM; TP = LD(&(x[WS(rs, 29)]), ms, &(x[WS(rs, 1)])); TQ = BYTWJ(&(W[TWVL * 56]), TP); TR = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); TS = BYTWJ(&(W[TWVL * 24]), TR); TT = VSUB(TQ, TS); TK = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); TL = BYTWJ(&(W[TWVL * 8]), TK); TM = LD(&(x[WS(rs, 21)]), ms, &(x[WS(rs, 1)])); TN = BYTWJ(&(W[TWVL * 40]), TM); TO = VSUB(TL, TN); } } { V TJ, TU, T2w, T2x; TJ = VSUB(TG, TI); TU = VMUL(LDK(KP707106781), VADD(TO, TT)); TV = VADD(TJ, TU); T1X = VSUB(TJ, TU); T2w = VADD(TL, TN); T2x = VADD(TQ, TS); T2y = VADD(T2w, T2x); T38 = VSUB(T2x, T2w); } { V TW, T11, T2t, T2u; TW = VMUL(LDK(KP707106781), VSUB(TT, TO)); T11 = VSUB(TY, T10); T12 = VSUB(TW, T11); T1Y = VADD(T11, TW); T2t = VADD(TG, TI); T2u = VADD(TY, T10); T2v = VADD(T2t, T2u); T37 = VSUB(T2t, T2u); } } { V T2W, T30, T2Z, T31; { V T2U, T2V, T2X, T2Y; T2U = VADD(T2o, T2r); T2V = VADD(T2O, T2L); T2W = VADD(T2U, T2V); T30 = VSUB(T2U, T2V); T2X = VADD(T2v, T2y); T2Y = VADD(T2C, T2F); T2Z = VADD(T2X, T2Y); T31 = VBYI(VSUB(T2Y, T2X)); } ST(&(x[WS(rs, 16)]), VSUB(T2W, T2Z), ms, &(x[0])); ST(&(x[WS(rs, 8)]), VADD(T30, T31), ms, &(x[0])); ST(&(x[0]), VADD(T2W, T2Z), ms, &(x[0])); ST(&(x[WS(rs, 24)]), VSUB(T30, T31), ms, &(x[0])); } { V T2s, T2P, T2H, T2Q, T2z, T2G; T2s = VSUB(T2o, T2r); T2P = VSUB(T2L, T2O); T2z = VSUB(T2v, T2y); T2G = VSUB(T2C, T2F); T2H = VMUL(LDK(KP707106781), VADD(T2z, T2G)); T2Q = VMUL(LDK(KP707106781), VSUB(T2G, T2z)); { V T2I, T2R, T2S, T2T; T2I = VADD(T2s, T2H); T2R = VBYI(VADD(T2P, T2Q)); ST(&(x[WS(rs, 28)]), VSUB(T2I, T2R), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VADD(T2I, T2R), ms, &(x[0])); T2S = VSUB(T2s, T2H); T2T = VBYI(VSUB(T2Q, T2P)); ST(&(x[WS(rs, 20)]), VSUB(T2S, T2T), ms, &(x[0])); ST(&(x[WS(rs, 12)]), VADD(T2S, T2T), ms, &(x[0])); } } { V T36, T3r, T3h, T3p, T3d, T3o, T3k, T3s, T35, T3g; T35 = VMUL(LDK(KP707106781), VADD(T33, T34)); T36 = VADD(T32, T35); T3r = VSUB(T32, T35); T3g = VMUL(LDK(KP707106781), VSUB(T34, T33)); T3h = VADD(T3f, T3g); T3p = VSUB(T3g, T3f); { V T39, T3c, T3i, T3j; T39 = VFMA(LDK(KP923879532), T37, VMUL(LDK(KP382683432), T38)); T3c = VFNMS(LDK(KP382683432), T3b, VMUL(LDK(KP923879532), T3a)); T3d = VADD(T39, T3c); T3o = VSUB(T3c, T39); T3i = VFNMS(LDK(KP382683432), T37, VMUL(LDK(KP923879532), T38)); T3j = VFMA(LDK(KP382683432), T3a, VMUL(LDK(KP923879532), T3b)); T3k = VADD(T3i, T3j); T3s = VSUB(T3j, T3i); } { V T3e, T3l, T3u, T3v; T3e = VADD(T36, T3d); T3l = VBYI(VADD(T3h, T3k)); ST(&(x[WS(rs, 30)]), VSUB(T3e, T3l), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VADD(T3e, T3l), ms, &(x[0])); T3u = VBYI(VADD(T3p, T3o)); T3v = VADD(T3r, T3s); ST(&(x[WS(rs, 6)]), VADD(T3u, T3v), ms, &(x[0])); ST(&(x[WS(rs, 26)]), VSUB(T3v, T3u), ms, &(x[0])); } { V T3m, T3n, T3q, T3t; T3m = VSUB(T36, T3d); T3n = VBYI(VSUB(T3k, T3h)); ST(&(x[WS(rs, 18)]), VSUB(T3m, T3n), ms, &(x[0])); ST(&(x[WS(rs, 14)]), VADD(T3m, T3n), ms, &(x[0])); T3q = VBYI(VSUB(T3o, T3p)); T3t = VSUB(T3r, T3s); ST(&(x[WS(rs, 10)]), VADD(T3q, T3t), ms, &(x[0])); ST(&(x[WS(rs, 22)]), VSUB(T3t, T3q), ms, &(x[0])); } } { V TE, T1P, T1I, T1Q, T1t, T1M, T1F, T1N; { V Tg, TD, T1G, T1H; Tg = VADD(T4, Tf); TD = VADD(Tr, TC); TE = VADD(Tg, TD); T1P = VSUB(Tg, TD); T1G = VFNMS(LDK(KP195090322), TV, VMUL(LDK(KP980785280), T12)); T1H = VFMA(LDK(KP195090322), T1k, VMUL(LDK(KP980785280), T1r)); T1I = VADD(T1G, T1H); T1Q = VSUB(T1H, T1G); } { V T13, T1s, T1B, T1E; T13 = VFMA(LDK(KP980785280), TV, VMUL(LDK(KP195090322), T12)); T1s = VFNMS(LDK(KP195090322), T1r, VMUL(LDK(KP980785280), T1k)); T1t = VADD(T13, T1s); T1M = VSUB(T1s, T13); T1B = VSUB(T1v, T1A); T1E = VSUB(T1C, T1D); T1F = VADD(T1B, T1E); T1N = VSUB(T1E, T1B); } { V T1u, T1J, T1S, T1T; T1u = VADD(TE, T1t); T1J = VBYI(VADD(T1F, T1I)); ST(&(x[WS(rs, 31)]), VSUB(T1u, T1J), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VADD(T1u, T1J), ms, &(x[WS(rs, 1)])); T1S = VBYI(VADD(T1N, T1M)); T1T = VADD(T1P, T1Q); ST(&(x[WS(rs, 7)]), VADD(T1S, T1T), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 25)]), VSUB(T1T, T1S), ms, &(x[WS(rs, 1)])); } { V T1K, T1L, T1O, T1R; T1K = VSUB(TE, T1t); T1L = VBYI(VSUB(T1I, T1F)); ST(&(x[WS(rs, 17)]), VSUB(T1K, T1L), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 15)]), VADD(T1K, T1L), ms, &(x[WS(rs, 1)])); T1O = VBYI(VSUB(T1M, T1N)); T1R = VSUB(T1P, T1Q); ST(&(x[WS(rs, 9)]), VADD(T1O, T1R), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 23)]), VSUB(T1R, T1O), ms, &(x[WS(rs, 1)])); } } { V T1W, T2h, T2a, T2i, T23, T2e, T27, T2f; { V T1U, T1V, T28, T29; T1U = VSUB(T4, Tf); T1V = VADD(T1D, T1C); T1W = VADD(T1U, T1V); T2h = VSUB(T1U, T1V); T28 = VFNMS(LDK(KP555570233), T1X, VMUL(LDK(KP831469612), T1Y)); T29 = VFMA(LDK(KP555570233), T20, VMUL(LDK(KP831469612), T21)); T2a = VADD(T28, T29); T2i = VSUB(T29, T28); } { V T1Z, T22, T25, T26; T1Z = VFMA(LDK(KP831469612), T1X, VMUL(LDK(KP555570233), T1Y)); T22 = VFNMS(LDK(KP555570233), T21, VMUL(LDK(KP831469612), T20)); T23 = VADD(T1Z, T22); T2e = VSUB(T22, T1Z); T25 = VADD(T1A, T1v); T26 = VSUB(TC, Tr); T27 = VADD(T25, T26); T2f = VSUB(T26, T25); } { V T24, T2b, T2k, T2l; T24 = VADD(T1W, T23); T2b = VBYI(VADD(T27, T2a)); ST(&(x[WS(rs, 29)]), VSUB(T24, T2b), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VADD(T24, T2b), ms, &(x[WS(rs, 1)])); T2k = VBYI(VADD(T2f, T2e)); T2l = VADD(T2h, T2i); ST(&(x[WS(rs, 5)]), VADD(T2k, T2l), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 27)]), VSUB(T2l, T2k), ms, &(x[WS(rs, 1)])); } { V T2c, T2d, T2g, T2j; T2c = VSUB(T1W, T23); T2d = VBYI(VSUB(T2a, T27)); ST(&(x[WS(rs, 19)]), VSUB(T2c, T2d), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 13)]), VADD(T2c, T2d), ms, &(x[WS(rs, 1)])); T2g = VBYI(VSUB(T2e, T2f)); T2j = VSUB(T2h, T2i); ST(&(x[WS(rs, 11)]), VADD(T2g, T2j), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 21)]), VSUB(T2j, T2g), ms, &(x[WS(rs, 1)])); } } } } VLEAVE(); }
static void t1fuv_7(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP900968867, +0.900968867902419126236102319507445051165919162); DVK(KP801937735, +0.801937735804838252472204639014890102331838324); DVK(KP974927912, +0.974927912181823607018131682993931217232785801); DVK(KP692021471, +0.692021471630095869627814897002069140197260599); DVK(KP554958132, +0.554958132087371191422194871006410481067288862); DVK(KP356895867, +0.356895867892209443894399510021300583399127187); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 12)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 12), MAKE_VOLATILE_STRIDE(7, rs)) { V T1, T2, T4, Te, Tc, T9, T7; T1 = LD(&(x[0]), ms, &(x[0])); T2 = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); T4 = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Te = LD(&(x[WS(rs, 4)]), ms, &(x[0])); Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T9 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); T7 = LD(&(x[WS(rs, 2)]), ms, &(x[0])); { V T3, T5, Tf, Td, Ta, T8; T3 = BYTWJ(&(W[0]), T2); T5 = BYTWJ(&(W[TWVL * 10]), T4); Tf = BYTWJ(&(W[TWVL * 6]), Te); Td = BYTWJ(&(W[TWVL * 4]), Tc); Ta = BYTWJ(&(W[TWVL * 8]), T9); T8 = BYTWJ(&(W[TWVL * 2]), T7); { V T6, Tk, Tg, Tl, Tb, Tm; T6 = VADD(T3, T5); Tk = VSUB(T5, T3); Tg = VADD(Td, Tf); Tl = VSUB(Tf, Td); Tb = VADD(T8, Ta); Tm = VSUB(Ta, T8); { V Th, Ts, Tp, Tu, Tn, Tx, Ti, Tt; Th = VFNMS(LDK(KP356895867), T6, Tg); Ts = VFMA(LDK(KP554958132), Tl, Tk); ST(&(x[0]), VADD(T1, VADD(T6, VADD(Tb, Tg))), ms, &(x[0])); Tp = VFNMS(LDK(KP356895867), Tb, T6); Tu = VFNMS(LDK(KP356895867), Tg, Tb); Tn = VFMA(LDK(KP554958132), Tm, Tl); Tx = VFNMS(LDK(KP554958132), Tk, Tm); Ti = VFNMS(LDK(KP692021471), Th, Tb); Tt = VMUL(LDK(KP974927912), VFMA(LDK(KP801937735), Ts, Tm)); { V Tq, Tv, To, Ty, Tj, Tr, Tw; Tq = VFNMS(LDK(KP692021471), Tp, Tg); Tv = VFNMS(LDK(KP692021471), Tu, T6); To = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tn, Tk)); Ty = VMUL(LDK(KP974927912), VFNMS(LDK(KP801937735), Tx, Tl)); Tj = VFNMS(LDK(KP900968867), Ti, T1); Tr = VFNMS(LDK(KP900968867), Tq, T1); Tw = VFNMS(LDK(KP900968867), Tv, T1); ST(&(x[WS(rs, 2)]), VFMAI(To, Tj), ms, &(x[0])); ST(&(x[WS(rs, 5)]), VFNMSI(To, Tj), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VFMAI(Tt, Tr), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 6)]), VFNMSI(Tt, Tr), ms, &(x[0])); ST(&(x[WS(rs, 3)]), VFMAI(Ty, Tw), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 4)]), VFNMSI(Ty, Tw), ms, &(x[0])); } } } } } } VLEAVE(); }
static void t1fv_16(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms) { DVK(KP923879532, +0.923879532511286756128183189396788286822416626); DVK(KP382683432, +0.382683432365089771728459984030398866761344562); DVK(KP707106781, +0.707106781186547524400844362104849039284835938); { INT m; R *x; x = ri; for (m = mb, W = W + (mb * ((TWVL / VL) * 30)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 30), MAKE_VOLATILE_STRIDE(16, rs)) { V TJ, T10, TD, T11, T1b, T1c, Ty, TK, T16, T17, T18, Tb, TN, T13, T14; V T15, Tm, TM, TG, TI, TH; TG = LD(&(x[0]), ms, &(x[0])); TH = LD(&(x[WS(rs, 8)]), ms, &(x[0])); TI = BYTWJ(&(W[TWVL * 14]), TH); TJ = VSUB(TG, TI); T10 = VADD(TG, TI); { V TA, TC, Tz, TB; Tz = LD(&(x[WS(rs, 4)]), ms, &(x[0])); TA = BYTWJ(&(W[TWVL * 6]), Tz); TB = LD(&(x[WS(rs, 12)]), ms, &(x[0])); TC = BYTWJ(&(W[TWVL * 22]), TB); TD = VSUB(TA, TC); T11 = VADD(TA, TC); } { V Tp, Tw, Tr, Tu, Ts, Tx; { V To, Tv, Tq, Tt; To = LD(&(x[WS(rs, 14)]), ms, &(x[0])); Tp = BYTWJ(&(W[TWVL * 26]), To); Tv = LD(&(x[WS(rs, 10)]), ms, &(x[0])); Tw = BYTWJ(&(W[TWVL * 18]), Tv); Tq = LD(&(x[WS(rs, 6)]), ms, &(x[0])); Tr = BYTWJ(&(W[TWVL * 10]), Tq); Tt = LD(&(x[WS(rs, 2)]), ms, &(x[0])); Tu = BYTWJ(&(W[TWVL * 2]), Tt); } T1b = VADD(Tp, Tr); T1c = VADD(Tu, Tw); Ts = VSUB(Tp, Tr); Tx = VSUB(Tu, Tw); Ty = VMUL(LDK(KP707106781), VSUB(Ts, Tx)); TK = VMUL(LDK(KP707106781), VADD(Tx, Ts)); } { V T2, T9, T4, T7, T5, Ta; { V T1, T8, T3, T6; T1 = LD(&(x[WS(rs, 15)]), ms, &(x[WS(rs, 1)])); T2 = BYTWJ(&(W[TWVL * 28]), T1); T8 = LD(&(x[WS(rs, 11)]), ms, &(x[WS(rs, 1)])); T9 = BYTWJ(&(W[TWVL * 20]), T8); T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)])); T4 = BYTWJ(&(W[TWVL * 12]), T3); T6 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)])); T7 = BYTWJ(&(W[TWVL * 4]), T6); } T16 = VADD(T2, T4); T17 = VADD(T7, T9); T18 = VSUB(T16, T17); T5 = VSUB(T2, T4); Ta = VSUB(T7, T9); Tb = VFNMS(LDK(KP923879532), Ta, VMUL(LDK(KP382683432), T5)); TN = VFMA(LDK(KP923879532), T5, VMUL(LDK(KP382683432), Ta)); } { V Td, Tk, Tf, Ti, Tg, Tl; { V Tc, Tj, Te, Th; Tc = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)])); Td = BYTWJ(&(W[0]), Tc); Tj = LD(&(x[WS(rs, 13)]), ms, &(x[WS(rs, 1)])); Tk = BYTWJ(&(W[TWVL * 24]), Tj); Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)])); Tf = BYTWJ(&(W[TWVL * 16]), Te); Th = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)])); Ti = BYTWJ(&(W[TWVL * 8]), Th); } T13 = VADD(Td, Tf); T14 = VADD(Ti, Tk); T15 = VSUB(T13, T14); Tg = VSUB(Td, Tf); Tl = VSUB(Ti, Tk); Tm = VFMA(LDK(KP382683432), Tg, VMUL(LDK(KP923879532), Tl)); TM = VFNMS(LDK(KP382683432), Tl, VMUL(LDK(KP923879532), Tg)); } { V T1a, T1g, T1f, T1h; { V T12, T19, T1d, T1e; T12 = VSUB(T10, T11); T19 = VMUL(LDK(KP707106781), VADD(T15, T18)); T1a = VADD(T12, T19); T1g = VSUB(T12, T19); T1d = VSUB(T1b, T1c); T1e = VMUL(LDK(KP707106781), VSUB(T18, T15)); T1f = VBYI(VADD(T1d, T1e)); T1h = VBYI(VSUB(T1e, T1d)); } ST(&(x[WS(rs, 14)]), VSUB(T1a, T1f), ms, &(x[0])); ST(&(x[WS(rs, 6)]), VADD(T1g, T1h), ms, &(x[0])); ST(&(x[WS(rs, 2)]), VADD(T1a, T1f), ms, &(x[0])); ST(&(x[WS(rs, 10)]), VSUB(T1g, T1h), ms, &(x[0])); } { V T1k, T1o, T1n, T1p; { V T1i, T1j, T1l, T1m; T1i = VADD(T10, T11); T1j = VADD(T1c, T1b); T1k = VADD(T1i, T1j); T1o = VSUB(T1i, T1j); T1l = VADD(T13, T14); T1m = VADD(T16, T17); T1n = VADD(T1l, T1m); T1p = VBYI(VSUB(T1m, T1l)); } ST(&(x[WS(rs, 8)]), VSUB(T1k, T1n), ms, &(x[0])); ST(&(x[WS(rs, 4)]), VADD(T1o, T1p), ms, &(x[0])); ST(&(x[0]), VADD(T1k, T1n), ms, &(x[0])); ST(&(x[WS(rs, 12)]), VSUB(T1o, T1p), ms, &(x[0])); } { V TF, TQ, TP, TR; { V Tn, TE, TL, TO; Tn = VSUB(Tb, Tm); TE = VSUB(Ty, TD); TF = VBYI(VSUB(Tn, TE)); TQ = VBYI(VADD(TE, Tn)); TL = VADD(TJ, TK); TO = VADD(TM, TN); TP = VSUB(TL, TO); TR = VADD(TL, TO); } ST(&(x[WS(rs, 7)]), VADD(TF, TP), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 15)]), VSUB(TR, TQ), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 9)]), VSUB(TP, TF), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 1)]), VADD(TQ, TR), ms, &(x[WS(rs, 1)])); } { V TU, TY, TX, TZ; { V TS, TT, TV, TW; TS = VSUB(TJ, TK); TT = VADD(Tm, Tb); TU = VADD(TS, TT); TY = VSUB(TS, TT); TV = VADD(TD, Ty); TW = VSUB(TN, TM); TX = VBYI(VADD(TV, TW)); TZ = VBYI(VSUB(TW, TV)); } ST(&(x[WS(rs, 13)]), VSUB(TU, TX), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 5)]), VADD(TY, TZ), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 3)]), VADD(TU, TX), ms, &(x[WS(rs, 1)])); ST(&(x[WS(rs, 11)]), VSUB(TY, TZ), ms, &(x[WS(rs, 1)])); } } } VLEAVE(); }