#include "../core.h" #include "../utils/each.h" #include #if SINGELI #define BCALL(N, X) N(b(X)) #define interp_f64(X) b(X).f #pragma GCC diagnostic push #pragma GCC diagnostic ignored "-Wunused-variable" #include "../singeli/gen/dyarith.c" #pragma GCC diagnostic pop #endif #define P2(N) { if(isArr(w)|isArr(x)) { \ SLOWIF((!isArr(w) || TI(w,elType)!=el_B) && (!isArr(x) || TI(x,elType)!=el_B)) SLOW2("arithd " #N, w, x); \ return arith_recd(N##_c2, w, x); \ }} #if TYPED_ARITH #define GC2f(SYMB, NAME, EXPR, EXTRA) B NAME##_c2(B t, B w, B x) { \ if (isF64(w) & isF64(x)) return m_f64(EXPR); \ EXTRA \ if (isArr(w)|isArr(x)) { B ow=w; B ox=x; \ if (isArr(w)&isArr(x) && RNK(w)==RNK(x)) { \ if (!eqShPart(SH(w), SH(x), RNK(w))) thrF(SYMB ": Expected equal shape prefix (%H ≡ ≢𝕨, %H ≡ ≢𝕩)", w, x); \ usz ia = IA(x); \ u8 we = TI(w,elType); \ u8 xe = TI(x,elType); \ if ((we==el_i32|we==el_f64)&(xe==el_i32|xe==el_f64)) { \ f64* rp; B r = m_f64arrc(&rp, x); \ if (we==el_i32) { B w,x/*shadow*/; i32* wp = i32any_ptr(ow); \ if (xe==el_i32) { i32* xp = i32any_ptr(ox); for (usz i = 0; i < ia; i++) {w.f=wp[i];x.f=xp[i];rp[i]=EXPR;} } \ else { f64* xp = f64any_ptr(ox); for (usz i = 0; i < ia; i++) {w.f=wp[i];x.f=xp[i];rp[i]=EXPR;} } \ } else { B w,x/*shadow*/; f64* wp = f64any_ptr(ow); \ if (xe==el_i32) { i32* xp = i32any_ptr(ox); for (usz i = 0; i < ia; i++) {w.f=wp[i];x.f=xp[i];rp[i]=EXPR;} } \ else { f64* xp = f64any_ptr(ox); for (usz i = 0; i < ia; i++) {w.f=wp[i];x.f=xp[i];rp[i]=EXPR;} } \ } \ decG(w); decG(x); return num_squeeze(r); \ } \ } else if (isF64(w)&isArr(x)) { usz ia = IA(x); \ u8 xe = TI(x,elType); f64*rp; \ if (xe==el_i32) { B r=m_f64arrc(&rp, x); i32*xp=i32any_ptr(x); \ for (usz i = 0; i < ia; i++) {B x/*shadow*/;x.f=xp[i];rp[i]=EXPR;} \ decG(x); return num_squeeze(r); \ } \ if (xe==el_f64) { B r=m_f64arrc(&rp, x); f64*xp=f64any_ptr(x); \ for (usz i = 0; i < ia; i++) {B x/*shadow*/;x.f=xp[i];rp[i]=EXPR;} \ decG(x); return num_squeeze(r); \ } \ } else if (isF64(x)&isArr(w)) { usz ia = IA(w); \ u8 we = TI(w,elType); f64*rp; \ if (we==el_i32) { B r=m_f64arrc(&rp, w); i32*wp=i32any_ptr(w); \ for (usz i = 0; i < ia; i++) {B w/*shadow*/;w.f=wp[i];rp[i]=EXPR;} \ decG(w); return num_squeeze(r); \ } \ if (we==el_f64) { B r=m_f64arrc(&rp, w); f64*wp=f64any_ptr(w); \ for (usz i = 0; i < ia; i++) {B w/*shadow*/;w.f=wp[i];rp[i]=EXPR;} \ decG(w); return num_squeeze(r); \ } \ } \ P2(NAME) \ } \ thrM(SYMB ": Unexpected argument types"); \ } #define PF(N) f64* N##p = f64any_ptr(N); #define PI8(N) i8* N##p = i8any_ptr (N); #define PI16(N) i16* N##p = i16any_ptr(N); #define PI32(N) i32* N##p = i32any_ptr(N); #define Ri8(A) i8* rp; B r=m_i8arrc (&rp, A); #define Ri16(A) i16* rp; B r=m_i16arrc(&rp, A); #define Ri32(A) i32* rp; B r=m_i32arrc(&rp, A); #define Rf64(A) f64* rp; B r=m_f64arrc(&rp, A); static NOINLINE u8 iMakeEq(B* w, B* x, u8 we, u8 xe) { B* p = weCHR_MAX)thrM("+: Invalid character"); return m_c32((u32)r); } if (isF64(w) & isC32(x)) { u64 r = (u64)(o2cu(x)+o2i64(w)); if(r>CHR_MAX)thrM("+: Invalid character"); return m_c32((u32)r); } },{ if (isArr(w)&isC32(x) || isC32(w)&isArr(x)) { if (isArr(w)) { B t=w;w=x;x=t; } if (TI(x,elType) == el_i32) { u32 wv = o2cu(w); i32* xp = i32any_ptr(x); usz xia = IA(x); u32* rp; B r = m_c32arrc(&rp, x); for (usz i = 0; i < xia; i++) { rp[i] = (u32)(xp[i]+(i32)wv); if (rp[i]>CHR_MAX) thrM("+: Invalid character"); // safe to only check this as wv already must be below CHR_MAX, which is less than U32_MAX/2 } decG(x); return r; } } }, 0, SI_AA, SI_AS, SI_SA) GC2i("-", sub, wv-xv, { if (isC32(w) & isF64(x)) { u64 r = (u64)((i32)o2cu(w)-o2i64(x)); if(r>CHR_MAX)thrM("-: Invalid character"); return m_c32((u32)r); } if (isC32(w) & isC32(x)) return m_f64((i32)(u32)w.u - (i32)(u32)x.u); },{ if (isArr(w) && TI(w,elType)==el_c32) { if (isC32(x)) { i32 xv = (i32)o2cu(x); u32* wp = c32any_ptr(w); usz wia = IA(w); i32* rp; B r = m_i32arrc(&rp, w); for (usz i = 0; i < wia; i++) rp[i] = (i32)wp[i] - xv; decG(w); return r; } if (isArr(x) && eqShape(w, x)) { u32* wp = c32any_ptr(w); usz wia = IA(w); if (TI(x,elType)==el_i32) { u32* rp; B r = m_c32arrc(&rp, w); i32* xp = i32any_ptr(x); for (usz i = 0; i < wia; i++) { rp[i] = (u32)((i32)wp[i] - (i32)xp[i]); if (rp[i]>CHR_MAX) thrM("-: Invalid character"); // safe - see add } decG(w); decG(x); return r; } } } }, 0, SI_AA, SI_AS, SI_SA) GC2i("¬", not, 1+wv-xv, { if (isC32(w) & isF64(x)) { u64 r = (u64)(1+(i32)o2cu(w)-o2i64(x)); if(r>CHR_MAX)thrM("¬: Invalid character"); return m_c32((u32)r); } if (isC32(w) & isC32(x)) return m_f64(1 + (i32)(u32)w.u - (i32)(u32)x.u); }, {}, 0, NO_SI_AA, REG_AS, REG_SA) GC2i("×", mul, wv*xv, {}, {}, 2, SI_AA, SI_AS, SI_SA) GC2i("∧", and, wv*xv, {}, {}, 2, NO_SI_AA, REG_AS, REG_SA) GC2i("∨", or , (wv+xv)-(wv*xv), {}, {}, 1, NO_SI_AA, REG_AS, REG_SA) GC2i("⌊", floor, wv>xv?xv:wv, {}, {}, 2, NO_SI_AA, REG_AS, REG_SA) // optimizer optimizes out the fallback mess GC2i("⌈", ceil , wv>xv?wv:xv, {}, {}, 1, NO_SI_AA, REG_AS, REG_SA) GC2f("÷", div , w.f/x.f, {}) GC2f("⋆", pow , pow(w.f, x.f), {}) GC2f("√", root , pow(x.f, 1.0/w.f), {}) GC2f("|", stile, pfmod(x.f, w.f), {}) GC2f("⋆⁼",log , log(x.f)/log(w.f), {}) #undef GC2i #undef GC2f #undef P2