#pragma once typedef struct Mut { u8 type; usz ia; Arr* val; union { i32* ai32; f64* af64; u32* ac32; B* aB; }; } Mut; #define MAKE_MUT(N, IA) Mut N##_val; N##_val.type = el_MAX; N##_val.ia = (IA); Mut* N = &N##_val; static void mut_init(Mut* m, u8 n) { m->type = n; usz ia = m->ia; u8 ty; usz sz; // hack around inlining of the allocator too many times switch(n) { default: UD; case el_i32: ty = t_i32arr; sz = I32A_SZ(ia); break; case el_f64: ty = t_f64arr; sz = F64A_SZ(ia); break; case el_c32: ty = t_c32arr; sz = C32A_SZ(ia); break; case el_B:; HArr_p t = m_harrUp(ia); m->val = (Arr*)t.c; m->aB = t.c->a; return; } Arr* a = mm_alloc(sz, ty); a->ia = ia; m->val = a; switch(n) { default: UD; // gcc generates horrible code for this (which should just be two instructions), but that's what gcc does case el_i32: m->ai32 = ((I32Arr*)a)->a; break; case el_f64: m->af64 = ((F64Arr*)a)->a; break; case el_c32: m->ac32 = ((C32Arr*)a)->a; break; } } void mut_to(Mut* m, u8 n); static B mut_fv(Mut* m) { assert(m->type!=el_MAX); m->val->sh = &m->val->ia; B r = taga(m->val); srnk(r, 1); return r; } static B mut_fc(Mut* m, B x) { assert(m->type!=el_MAX); Arr* a = m->val; arr_shCopy(a, x); return taga(a); } static B mut_fcd(Mut* m, B x) { assert(m->type!=el_MAX); Arr* a = m->val; arr_shCopy(a, x); dec(x); return taga(a); } static Arr* mut_fp(Mut* m) { assert(m->type!=el_MAX); // has ia set return m->val; } static u8 el_or(u8 a, u8 b) { #define M(X) if(b==X) return a>X?a:X; switch (a) { default: UD; case el_c32: M(el_c32); return el_B; case el_i32: M(el_i32); M(el_f64); return el_B; case el_f64: M(el_i32); M(el_f64); return el_B; case el_B: return el_B; case el_MAX: return b; } #undef M } void mut_pfree(Mut* m, usz n); static void mut_set(Mut* m, usz ms, B x) { // consumes x; sets m[ms] to x again:; u8 nty; switch(m->type) { default: UD; case el_MAX: nty = isF64(x)? (q_i32(x)? el_i32 : el_f64) : (isC32(x)? el_c32 : el_B); goto change; case el_i32: { if (!q_i32(x)) { nty = isF64(x)? el_f64 : el_B; goto change; } m->ai32[ms] = o2iu(x); return; } case el_c32: { if (!isC32(x)) { nty = el_B; goto change; } m->ac32[ms] = o2cu(x); return; } case el_f64: { if (!isF64(x)) { nty = el_B; goto change; } m->af64[ms] = o2fu(x); return; } case el_B: { m->aB[ms] = x; return; } } change: mut_to(m, nty); goto again; } static void mut_setG(Mut* m, usz ms, B x) { // consumes; sets m[ms] to x, assumes the current type can store it switch(m->type) { default: UD; case el_i32: { assert(q_i32(x)); m->ai32[ms] = o2iu(x); return; } case el_c32: { assert(isC32(x)); m->ac32[ms] = o2cu(x); return; } case el_f64: { assert(isF64(x)); m->af64[ms] = o2fu(x); return; } case el_B: { m->aB[ms] = x; return; } } } static void mut_rm(Mut* m, usz ms) { // clears the object at position ms if (m->type == el_B) dec(m->aB[ms]); } static B mut_getU(Mut* m, usz ms) { switch(m->type) { default: UD; case el_i32: return m_i32(m->ai32[ms]); case el_c32: return m_c32(m->ac32[ms]); case el_f64: return m_f64(m->af64[ms]); case el_B: return m->aB[ms]; } } // doesn't consume; fills m[ms…ms+l] with x static void mut_fill(Mut* m, usz ms, B x, usz l) { again:; u8 nty; switch(m->type) { default: UD; case el_MAX: nty = isF64(x)? (q_i32(x)? el_i32 : el_f64) : (isC32(x)? el_c32 : el_B); goto change; case el_i32: { if (RARE(!q_i32(x))) { nty = isF64(x)? el_f64 : el_B; goto change; } i32* p = m->ai32+ms; i32 v = o2iu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_c32: { if (RARE(!isC32(x))) { nty = el_B; goto change; } u32* p = m->ac32+ms; u32 v = o2cu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_f64: { if (RARE(!isF64(x))) { nty = el_B; goto change; } f64* p = m->af64+ms; f64 v = o2fu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_B: { B* p = m->aB+ms; for (usz i = 0; i < l; i++) p[i] = x; if (isVal(x)) for (usz i = 0; i < l; i++) inc(x); return; } } change: mut_to(m, nty); goto again; } static void mut_fillG(Mut* m, usz ms, B x, usz l) { switch(m->type) { default: UD; case el_i32: { assert(q_i32(x)); i32* p = m->ai32+ms; i32 v = o2iu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_c32: { assert(isC32(x)); u32* p = m->ac32+ms; u32 v = o2cu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_f64: { assert(isF64(x)); f64* p = m->af64+ms; f64 v = o2fu(x); for (usz i = 0; i < l; i++) p[i] = v; return; } case el_B: { B* p = m->aB+ms; for (usz i = 0; i < l; i++) p[i] = x; if (isVal(x)) for (usz i = 0; i < l; i++) inc(x); return; } } } // expects x to be an array, each position must be written to precisely once // doesn't consume x static void mut_copy(Mut* m, usz ms, B x, usz xs, usz l) { // TODO try harder to not bump type assert(isArr(x)); u8 xt = v(x)->type; u8 xe = TIi(xt,elType); // printf("mut_%d[%d…%d] ← %s[%d…%d]\n", m->type, ms, ms+l, format_type(xt), xs, xs+l); fflush(stdout); again: switch(m->type) { default: UD; case el_MAX: goto change; case el_i32: { if (RARE(xt!=t_i32arr & xt!=t_i32slice)) goto change; i32* xp = i32any_ptr(x); memcpy(m->ai32+ms, xp+xs, l*4); return; } case el_c32: { if (RARE(xt!=t_c32arr & xt!=t_c32slice)) goto change; u32* xp = c32any_ptr(x); memcpy(m->ac32+ms, xp+xs, l*4); return; } case el_f64: { f64* xp; if (xt==t_f64arr) xp = f64arr_ptr(x); else if (xt==t_f64slice) xp = c(F64Slice,x)->a; else if (LIKELY(xt==t_i32arr|xt==t_i32slice)) { i32* xp = i32any_ptr(x); f64* rp = m->af64+ms; for (usz i = 0; i < l; i++) rp[i] = xp[i+xs]; return; } else goto change; memcpy(m->af64+ms, xp+xs, l*8); return; } case el_B: { B* mpo = m->aB+ms; B* xp; if (xt==t_harr) xp = harr_ptr(x); else if (xt==t_hslice) xp = c(HSlice,x)->a; else if (xt==t_fillarr) xp = c(FillArr,x)->a; else { BS2B xget = TIi(xt,get); for (usz i = 0; i < l; i++) mpo[i] = xget(x,i+xs); return; } memcpy(mpo, xp+xs, l*sizeof(B*)); for (usz i = 0; i < l; i++) inc(mpo[i]); return; } } change: mut_to(m, el_or(m->type, xe)); goto again; } static void mut_copyG(Mut* m, usz ms, B x, usz xs, usz l) { // mut_copy but x is guaranteed to be a subtype of m assert(isArr(x)); u8 xt = v(x)->type; switch(m->type) { default: UD; case el_i32: { i32* xp = i32any_ptr(x); memcpy(m->ai32+ms, xp+xs, l*4); return; } case el_c32: { u32* xp = c32any_ptr(x); memcpy(m->ac32+ms, xp+xs, l*4); return; } case el_f64: { f64* xp; if (xt==t_f64arr) xp = f64arr_ptr(x); else if (xt==t_f64slice) xp = c(F64Slice,x)->a; else { assert(TIi(xt,elType)==el_i32); i32* xp = i32any_ptr(x); f64* rp = m->af64+ms; for (usz i = 0; i < l; i++) rp[i] = xp[i+xs]; return; } memcpy(m->af64+ms, xp+xs, l*8); return; } case el_B: { B* mpo = m->aB+ms; B* xp; if (xt==t_harr) xp = harr_ptr(x); else if (xt==t_hslice) xp = c(HSlice,x)->a; else if (xt==t_fillarr) xp = c(FillArr,x)->a; else { BS2B xget = TIi(xt,get); for (usz i = 0; i < l; i++) mpo[i] = xget(x,i+xs); return; } memcpy(mpo, xp+xs, l*sizeof(B*)); for (usz i = 0; i < l; i++) inc(mpo[i]); return; } } } static B vec_join(B w, B x) { // consumes both usz wia = a(w)->ia; usz xia = a(x)->ia; usz ria = wia+xia; if (v(w)->refc==1) { u64 wsz = mm_size(v(w)); u8 wt = v(w)->type; if (wt==t_i32arr && fsizeof(I32Arr,a,i32,ria)ia = ria; memcpy(i32arr_ptr(w)+wia, i32any_ptr(x), xia*4); dec(x); return w; } if (wt==t_c32arr && fsizeof(C32Arr,a,u32,ria)ia = ria; memcpy(c32arr_ptr(w)+wia, c32any_ptr(x), xia*4); dec(x); return w; } if (wt==t_f64arr && fsizeof(F64Arr,a,f64,ria)ia = ria; memcpy(f64arr_ptr(w)+wia, f64any_ptr(x), xia*8); dec(x); return w; } if (wt==t_harr && fsizeof(HArr,a,B,ria)ia = ria; B* rp = harr_ptr(w)+wia; u8 xt = v(x)->type; u8 xe = TI(x,elType); if (xt==t_harr | xt==t_hslice | xt==t_fillarr) { B* xp = xt==t_harr? harr_ptr(x) : xt==t_hslice? c(HSlice, x)->a : fillarr_ptr(a(x)); memcpy(rp, xp, xia*sizeof(B)); for (usz i = 0; i < xia; i++) inc(rp[i]); } else if (xe==el_i32) { i32* xp = i32any_ptr(x); for (usz i = 0; i < xia; i++) rp[i] = m_i32(xp[i]); } else if (xe==el_c32) { u32* xp = c32any_ptr(x); for (usz i = 0; i < xia; i++) rp[i] = m_c32(xp[i]); } else if (xe==el_f64) { f64* xp = f64any_ptr(x); for (usz i = 0; i < xia; i++) rp[i] = m_f64(xp[i]); } else { BS2B xget = TI(x,get); for (usz i = 0; i < xia; i++) rp[i] = xget(x, i); } dec(x); return w; } } MAKE_MUT(r, ria); mut_init(r, el_or(TI(w,elType), TI(x,elType))); mut_copyG(r, 0, w, 0, wia); mut_copyG(r, wia, x, 0, xia); dec(w); dec(x); return mut_fv(r); } static inline bool inplace_add(B w, B x) { // fails if fills wouldn't be correct usz wia = a(w)->ia; usz ria = wia+1; if (v(w)->refc==1) { u64 wsz = mm_size(v(w)); u8 wt = v(w)->type; if (wt==t_i32arr && fsizeof(I32Arr,a,i32,ria)ia = ria; i32arr_ptr(w)[wia] = o2iu(x); return true; } if (wt==t_c32arr && fsizeof(C32Arr,a,u32,ria)ia = ria; c32arr_ptr(w)[wia] = o2cu(x); return true; } if (wt==t_f64arr && fsizeof(F64Arr,a,f64,ria)ia = ria; f64arr_ptr(w)[wia] = o2fu(x); return true; } if (wt==t_harr && fsizeof(HArr,a,B,ria)ia = ria; harr_ptr(w)[wia] = x; return true; } } return false; } B vec_addR(B w, B x); static B vec_add(B w, B x) { // consumes both; fills may be wrong if (inplace_add(w, x)) return w; return vec_addR(w, x); }