From 94868be24f610e0362b6388ee30a68cbaada07d4 Mon Sep 17 00:00:00 2001 From: dsh Date: Sat, 7 Nov 2026 12:23:42 +0800 Subject: [PATCH] Name the two geometry numbers the same way everywhere: cell_size, cell_argc The cell width was CELL_SIZE on the C++ side, and the searcher's CellBit count was CELL_ARGC. Both now match what the code and the docs around them already called them: * include/r8forge/r8asm.hh + src/r8asm.cc: CELL_SIZE -> cell_size (set_cell_size assigns this->cell_size, its parameter already had that name); * include/r8forge/r8veritas.hh: the template parameter CELL_ARGC -> cell_argc, 34 uses, so the searcher reads r8veritas; * AGENTS.md says which name means what, and why the name .r8asm programs see is still _CELL_SIZE: that one belongs to the macro library, not to the C++ side, and changing it would rewrite every macro in asm/; * the cross check entry now says which ROT-8-viz is which: the author's fork with --cell/--mem, and the upstream starcells/ROT-8 tree in the workspace, whose geometry is compile time cell_size/mem_size -- it is 11 / 259 too now, and `rot8_test ` against `r8forge_derive ... -1` still gives 1322 1786 1600 5 94 127 on the hash tape. All six self checks pass unchanged: 28729 character hash tape, synth_test 8/8, both verify_ tapes, byte_op_test 65536/65536 for clb_mul and the rest of the library, macro docs test green. --- AGENTS.md | 17 ++++++--- include/r8forge/r8asm.hh | 4 +-- include/r8forge/r8veritas.hh | 68 ++++++++++++++++++------------------ src/r8asm.cc | 14 ++++---- src/synth.cc | 2 +- 5 files changed, 57 insertions(+), 48 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index 7df2645..ba22b76 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -132,7 +132,12 @@ void rot8_rnz(rot8_iter *iter) { if ((*(iter->mem_now) & 1) == 1) { ... } } 曾经这里还写过第三条("多字节 cell 的位读不出来")—— 那是把 `biz` 当成 Brainfuck 的 整格零判断推出来的,**错的**,已删。 -## CELL_SIZE 与纸带长度是怎么用的 +## cell_size 与纸带长度是怎么用的 + +位宽叫 **`cell_size`**(一个 cell 多少 bit),纸带长度叫 `kDefaultMemoryCells`; +搜索器里"有多少个 CellBit 形参"叫 **`cell_argc`**(`r8veritas`)。这三个名字 +在两边的仓库里是统一的:C/C++ 侧 `cell_size`、搜索器侧 `cell_argc`、`.r8asm` 程序里看到的 +常量仍是 `_CELL_SIZE`(那是源语言的常量,属于宏库的公开词汇,不要跟着改)。 两个默认值集中在 **`include/r8forge/geometry.hh`**,现在是 **11 位 / 259 格**: @@ -167,9 +172,13 @@ inline constexpr uint32_t kDefaultCellMask = (1u << kDefaultCellBits) - 1u; `fibonacci` 8761 → **11959**。写死的长度断言要跟着改。 * **`output` 打印整个 cell 的值**,不再是一个字节;`out` 里出现 257、1088 这种数是对的 (`127+130 = 257`、`200*200 = 19*2048 + 1088`)。 -* **交叉验证要先对齐几何**:`ROT-8-viz` 现在默认也是 **11 位 / 259 格**(`--cell` / `--mem` - 可改)。几何一致时同一条纸带输出完全相同(`asm/test.r8asm` 两边都是 - `1322 1786 1600 5 94 127`);不一致时不是谁错了,是两台不同的机器。 +* **交叉验证要先对齐几何**:几何一致时同一条纸带输出完全相同(`asm/test.r8asm` 两边都是 + `1322 1786 1600 5 94 127`);不一致时不是谁错了,是两台不同的机器。作者那份 `ROT-8-viz` + (`C:\Users\peng0\git\ROT-8-viz`)默认 11 位 / 259 格,`--cell` / `--mem` 可改; + **workspace 里另有一份 `ROT-8-viz/`,那是 upstream `starcells/ROT-8` 的 `feat/memory-viz` + 分支**,几何是编译期常量(`cell_size` / `mem_size`,在 `include/rot8/rot8.h`),现在也是 + **11 位 / 259 格**。对答案的办法:`ROT-8-viz/build/Release/rot8_test.exe ` 把每个 + `OUT` 按行打出来,和 `r8forge_derive -1` 的 `out` 逐位比即可。 * 位宽变大还有个正面效果:`fibonacci` 不再在 256 处回绕,现在跑出 **13 项**(到 233)。 * **旧 tape 换到新位宽上不只是算错,还可能永远跑不完**:按 8 位写死的程序里 `%rep 8` 只转了 8 圈,在 11 位下 cell 没转回原位,"循环到某位清零"的条件于是永远不成立。 diff --git a/include/r8forge/r8asm.hh b/include/r8forge/r8asm.hh index c0b6bb4..97a84d9 100644 --- a/include/r8forge/r8asm.hh +++ b/include/r8forge/r8asm.hh @@ -104,8 +104,8 @@ private: class ast { public: ast_node root = {r8asm::ast_node_t::CODEBLOCK, "", {}}; - size_t CELL_SIZE; - // I know template class is the most elegant solution of CELL_SIZE + size_t cell_size; + // I know template class is the most elegant solution of cell_size // but I can't be bother to add that. // The assembler has no exceptions: anything that cannot be assembled diff --git a/include/r8forge/r8veritas.hh b/include/r8forge/r8veritas.hh index 737d69f..9533ef7 100644 --- a/include/r8forge/r8veritas.hh +++ b/include/r8forge/r8veritas.hh @@ -23,7 +23,7 @@ // Search strategy: // * only the input bits that influence the answer survive (the old // compress_stat_space), so a plane is a 2^K bit truth table instead of -// 2^CELL_ARGC, and the whole state is CELL_ARGC * 2^K bits +// 2^cell_argc, and the whole state is cell_argc * 2^K bits // * forward BFS keeps every reachable state of depth < d plus the block // that produced it // * the last block is not searched forward. A block clears its own @@ -56,10 +56,10 @@ #include #include -template class r8veritas { +template class r8veritas { public: using cell = uint8_t; - using stat = std::array; + using stat = std::array; using stat_ref = const stat *; using truthtable = std::pair; using op = std::function; @@ -85,7 +85,7 @@ public: // set_1. The inter-block hop from the previous sacrifice back to this one // is not in the model -- the emitter charges it afterwards. struct cost_options { - std::array addr{}; // cell address of each plane + std::array addr{}; // cell address of each plane int max_cost = 4096; // give up above this tape length size_t max_states = 600'000'000; unsigned threads = 0; // 0 = one worker per hardware thread @@ -113,16 +113,16 @@ public: std::string message; }; - std::array nicknames; - std::array tp_paths; + std::array nicknames; + std::array tp_paths; void initialize_nicknames() { - for (int i = 0; i < CELL_ARGC; i++) + for (int i = 0; i < cell_argc; i++) nicknames[i] = "%" + std::to_string(i + 1); } void initialize_tp_paths() { - for (int i = 0; i < CELL_ARGC; i++) + for (int i = 0; i < cell_argc; i++) tp_paths[i] = ".tp " + nicknames[i]; } @@ -135,7 +135,7 @@ public: truthtable generate_truthtable(const op &operation) { stat in{}, out{}; - for (size_t i = 0; i < (size_t(1) << CELL_ARGC); i++) { + for (size_t i = 0; i < (size_t(1) << cell_argc); i++) { in[i] = cell(i); out[i] = operation(cell(i)); } @@ -152,13 +152,13 @@ public: // starts from a canonical state. truthtable compress_stat_space(const truthtable &raw) { stat start = *raw.first; - for (int i = 0; i < CELL_ARGC; i++) { + for (int i = 0; i < cell_argc; i++) { bool influences = false; - for (size_t j = 0; j < (size_t(1) << CELL_ARGC); j++) + for (size_t j = 0; j < (size_t(1) << cell_argc); j++) if ((*raw.second)[j] != (*raw.second)[j ^ (size_t(1) << i)]) influences = true; if (!influences) - for (size_t j = 0; j < (size_t(1) << CELL_ARGC); j++) + for (size_t j = 0; j < (size_t(1) << cell_argc); j++) start[j] = cell(start[j] & ~(1u << i)); } return {keep(start), raw.second}; @@ -173,7 +173,7 @@ public: for (size_t j = 0; j < cur.size(); j++) if (cur[j] & (1u << i.sacrifice)) { cur[j] = cell(cur[j] & ~(1u << i.sacrifice)); - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (i.mask & (1 << b)) cur[j] = i.or_mode ? cell(cur[j] | (1u << b)) : cell(cur[j] ^ (1u << b)); @@ -189,7 +189,7 @@ public: // ---------------------------------------------------------------- output void output_stat(const stat &s) const { for (auto v : s) - std::cout << std::bitset(v).to_string() << " "; + std::cout << std::bitset(v).to_string() << " "; } void output_ins(const instruction &ins, std::ostream &os, @@ -199,7 +199,7 @@ public: if (ins.mask) { os << (ins.or_mode ? ", SET1: {" : ", FLIP: {"); bool first = true; - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (ins.mask & (1 << b)) { if (!first) os << ','; @@ -213,7 +213,7 @@ public: } os << tp_paths[ins.sacrifice] << "\n"; os << "biz\n"; - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (ins.mask & (1 << b)) { os << '\t' << tp_paths[b] << "\n"; os << '\t' << (ins.or_mode ? "set_1" : "flb") << "\n"; @@ -242,10 +242,10 @@ public: return r; }; - std::array bit_of{}; + std::array bit_of{}; int k = 0; - for (int i = 0; i < CELL_ARGC; i++) - for (size_t j = 0; j < (size_t(1) << CELL_ARGC); j++) + for (int i = 0; i < cell_argc; i++) + for (size_t j = 0; j < (size_t(1) << cell_argc); j++) if ((*raw.second)[j] != (*raw.second)[j ^ (size_t(1) << i)]) { bit_of[k++] = i; break; @@ -289,10 +289,10 @@ public: }; // which input bits matter, in compression order - std::array bit_of{}; + std::array bit_of{}; int k = 0; - for (int i = 0; i < CELL_ARGC; i++) { - for (size_t j = 0; j < (size_t(1) << CELL_ARGC); j++) + for (int i = 0; i < cell_argc; i++) { + for (size_t j = 0; j < (size_t(1) << cell_argc); j++) if ((*raw.second)[j] != (*raw.second)[j ^ (size_t(1) << i)]) { bit_of[k++] = i; break; @@ -336,10 +336,10 @@ private: // ---------------------------------------------------------------- engine // K relevant inputs: every plane is a 2^K bit truth table, the whole state - // is CELL_ARGC * 2^K bits packed into KEYW words. + // is cell_argc * 2^K bits packed into KEYW words. template struct engine { static constexpr int M = 1 << K; // bits per plane - static constexpr int KEYW = (CELL_ARGC * M + 63) / 64; + static constexpr int KEYW = (cell_argc * M + 63) / 64; static constexpr uint64_t PMASK = (M >= 64) ? ~uint64_t(0) : ((uint64_t(1) << (M & 63)) - 1); using key = std::array; @@ -364,7 +364,7 @@ private: key n = k; const uint64_t c = get(k, i.sacrifice); put(n, i.sacrifice, 0); - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (i.mask & (1 << b)) put(n, b, i.or_mode ? (get(k, b) | c) : (get(k, b) ^ c)); return n; @@ -456,7 +456,7 @@ private: for (size_t i = 0; i < insns.size(); i++) { const instruction &in = insns[i]; int lo = opt.addr[in.sacrifice], hi = lo, cnt = 0; - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (in.mask & (1 << b)) { cnt++; lo = std::min(lo, opt.addr[b]); @@ -470,7 +470,7 @@ private: // start = identity over the compressed inputs, goal = the target key start = E::zero(), goal = E::zero(); for (size_t j = 0; j < full.size(); j++) - for (int b = 0; b < CELL_ARGC; b++) { + for (int b = 0; b < cell_argc; b++) { if (j & (size_t(1) << b)) E::put(start, b, E::get(start, b) | (uint64_t(1) << j)); if ((*raw.second)[full[j]] & (1u << b)) @@ -675,7 +675,7 @@ private: // start = identity over the compressed inputs, goal = the target key start = E::zero(), goal = E::zero(); for (size_t j = 0; j < full.size(); j++) - for (int b = 0; b < CELL_ARGC; b++) { + for (int b = 0; b < cell_argc; b++) { if (j & (size_t(1) << b)) E::put(start, b, E::get(start, b) | (uint64_t(1) << j)); if ((*raw.second)[full[j]] & (1u << b)) @@ -794,10 +794,10 @@ private: const instruction &ins, uint64_t c, typename E::key pre, int b, F &f) { using key = typename E::key; - for (; b < CELL_ARGC; b++) + for (; b < cell_argc; b++) if (ins.mask & (1 << b)) break; - if (b == CELL_ARGC) + if (b == cell_argc) return f(pre); const uint64_t T = E::get(post, b); @@ -821,13 +821,13 @@ private: // ---------------------------------------------------------------- syntax std::vector build_instructions() const { std::vector out; - out.reserve(CELL_ARGC * (2 * (1 << (CELL_ARGC - 1)) + 1)); - for (int s = 0; s < CELL_ARGC; s++) { + out.reserve(cell_argc * (2 * (1 << (cell_argc - 1)) + 1)); + for (int s = 0; s < cell_argc; s++) { out.push_back({uint8_t(s), 0, false}); // pure sacrifice: clear plane s - for (int sub = 1; sub < (1 << (CELL_ARGC - 1)); sub++) { + for (int sub = 1; sub < (1 << (cell_argc - 1)); sub++) { uint8_t m = 0; int bit = 0; - for (int b = 0; b < CELL_ARGC; b++) + for (int b = 0; b < cell_argc; b++) if (b != s) m |= uint8_t(((sub >> bit++) & 1) << b); out.push_back({uint8_t(s), m, false}); diff --git a/src/r8asm.cc b/src/r8asm.cc index 39f72fb..a822c16 100644 --- a/src/r8asm.cc +++ b/src/r8asm.cc @@ -65,9 +65,9 @@ std::string peel(std::string &str) { // Peel the first word from string. #undef OPERATORS void r8asm::ast::set_cell_size(size_t cell_size) { - CELL_SIZE = cell_size; + this->cell_size = cell_size; // the names a .r8asm program sees stay as they were: _CELL_SIZE is the - // machine word width in bits (CELL_SIZE on the C++ side) + // machine word width in bits (cell_size on the C++ side) scopes = {{{"_CELL_SIZE", static_cast(cell_size)}, {"_CELL_MAX", static_cast((1 << cell_size) - 1)}, {"_ROL_STEP", static_cast(cell_size - 1)}}}; @@ -259,10 +259,10 @@ r8asm::ast_node r8asm::ast::expand_node_to_ins(const ast_node &node) { /* XOR children[0]: argument - WARNING: argument bigger than ((1< 5(0b101) - when define CELL_SIZE 2, only the last 2 bit will be noticed, + when define cell_size 2, only the last 2 bit will be noticed, and 5 means 0b01(1) here. then the node will be expanded like this: @@ -438,7 +438,7 @@ r8asm::ast_node r8asm::parse(const token_stream &stream) { filename += tp->value; // now tp points at RQUOTE - import_file(filename, main_tree.CELL_SIZE); + import_file(filename, main_tree.cell_size); // import_file has already said why if it failed; the error flag is // set, so this file will not be reported as assembled. } break; @@ -588,7 +588,7 @@ r8asm::ast_node r8asm::parse(const token_stream &stream) { main_tree.fail("%synth '" + name + "': cannot write the synthesised macro anywhere"); } else { - import_file(path, main_tree.CELL_SIZE); + import_file(path, main_tree.cell_size); std::remove(path.c_str()); } } @@ -894,7 +894,7 @@ bool r8asm::import_file(const std::string &filename, size_t cell_size) { return (false); } - // Before parsing: %include asks main_tree.CELL_SIZE for the width to hand + // Before parsing: %include asks main_tree.cell_size for the width to hand // to the included file, and that used to answer 0 because the size was // only set at the end of this function. main_tree.set_cell_size(cell_size); diff --git a/src/synth.cc b/src/synth.cc index efcefff..a82fae8 100644 --- a/src/synth.cc +++ b/src/synth.cc @@ -57,7 +57,7 @@ r8asm_synth_run(const std::vector &in, const std::vector &out, // ---- cache: the target table is what the search is given, so it is the // identity of the problem. Keying on it (rather than on the map or the stat // line the author wrote) means two blocks that describe the same table share - // one entry, and a table of a different CELL_SIZE can never collide with it + // one entry, and a table of a different cell_size can never collide with it // because the key covers every entry. // bump kModelVersion whenever the instruction model changes, otherwise a // table could be served a program that was optimal for the old model