Problem 327

Rooms of Doom: sum M(C,30) for C=3..40.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 327
# Rooms of Doom: sum M(C,30) for C=3..40.

function search(cards: i64, rooms: i64) -> i64 {
    let mut need: i64 = 1
    let transport: i64 = cards - 2
    let mut current: i64 = rooms
    while current > 0 {
        let mut consumed: i64 = 0
        if need >= cards {
            let mut moves: i64 = (need - cards) / transport
            if need - moves * transport >= cards {
                moves = moves + 1
            }
            need = need - moves * transport
            consumed = consumed + moves * cards
        }
        need = need + consumed + 1
        current = current - 1
    }
    return need
}

function main() -> i32 {
    let mut result: i64 = 0
    let mut c: i64 = 3
    while c <= 40 {
        result = result + search(c, 30)
        c = c + 1
    }
    printf("%lld\n", result)
    return 0
}

Generated C

#include <stdint.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

/* Flow runtime helpers */
typedef struct flow_temp_node { struct flow_temp_node* next; } flow_temp_node;
static flow_temp_node* flow_temp_head = NULL;
static int flow_temp_atexit_set = 0;
__attribute__((unused)) static void flow_temp_free_all(void) {
    while (flow_temp_head) {
        flow_temp_node* n = flow_temp_head;
        flow_temp_head = n->next;
        free(n);
    }
}
__attribute__((unused)) static void* flow_temp_alloc(size_t nbytes) {
    flow_temp_node* node = (flow_temp_node*)malloc(sizeof(flow_temp_node) + nbytes);
    if (!node) return NULL;
    node->next = flow_temp_head;
    flow_temp_head = node;
    if (!flow_temp_atexit_set) {
        flow_temp_atexit_set = 1;
        atexit(flow_temp_free_all);
    }
    return (void*)(node + 1);
}
#ifndef FLOW_DIAG
#define FLOW_DIAG(msg) fprintf(stderr, "%s", (msg))
#endif
#ifndef FLOW_LOG
#define FLOW_LOG(fmt, ...) printf(fmt, __VA_ARGS__)
#endif
#ifndef FLOW_LOG_EMPTY
#define FLOW_LOG_EMPTY(fmt) printf(fmt)
#endif
static char* flow_strcat(const char* a, const char* b) {
    size_t la = strlen(a ? a : ""), lb = strlen(b ? b : "");
    char* r = (char*)flow_temp_alloc(la + lb + 1);
    if (!r) return NULL;
    if (la) memcpy(r, a, la);
    if (lb) memcpy(r + la, b, lb);
    r[la + lb] = '\0';
    return r;
}

#define __flow_in_arr(arr, val) __extension__ ({ \
    int _found = 0; \
    size_t _n = sizeof(arr)/sizeof((arr)[0]); \
    for (size_t _i = 0; _i < _n; _i++) { \
        if ((arr)[_i] == (val)) { _found = 1; break; } \
    } _found; })

/* Unified fault handler (MISRA #279) — override with -DFLOW_FAULT_HANDLER=fn */
#ifndef FLOW_FAULT_HANDLER
__attribute__((unused)) static inline void flow_fault_handler(const char* msg) {
    fprintf(stderr, "flow: %s\n", msg ? msg : "fault");
    abort();
#if defined(__GNUC__) || defined(__clang__)
    __builtin_unreachable();
#endif
}
#else
#define flow_fault_handler FLOW_FAULT_HANDLER
#endif
#define flow_div_by_zero_handler() flow_fault_handler("division by zero")
#define flow_shift_ub_handler() flow_fault_handler("invalid shift (amount out of range or left-shift of negative)")

#ifndef FLOW_CHECKED_DIV
#define FLOW_CHECKED_DIV(L, R) (((R) != 0) ? ((L) / (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_MOD
#define FLOW_CHECKED_MOD(L, R) (((R) != 0) ? ((L) % (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHL
#define FLOW_CHECKED_SHL(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull)) && ((L) >= 0)) ? ((L) << (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHR
#define FLOW_CHECKED_SHR(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull))) ? ((L) >> (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif

#include <math.h>

void* _ui_state = NULL;

static inline float i32_to_f32(int32_t v) { return (float)v; }

/* Host stub for @gpu kernels (device codegen replaces this). */
static inline int32_t gpu_thread_id(void) { return 0; }

int64_t search_i64_i64(int64_t cards, int64_t rooms);
int32_t main(void);

int64_t search_i64_i64(int64_t cards, int64_t rooms) {
    int64_t need = 1;
    int64_t transport = (cards - 2);
    int64_t current = rooms;
    while (current > 0) {
        int64_t consumed = 0;
        if (need >= cards) {
            int64_t moves = FLOW_CHECKED_DIV(((need - cards)), (transport));
            if ((need - (moves * transport)) >= cards) {
                moves = (moves + 1);
            }
            need = (need - (moves * transport));
            consumed = (consumed + (moves * cards));
        }
        need = ((need + consumed) + 1);
        current = (current - 1);
    }
    return need;
}

int32_t main(void) {
    int64_t result = 0;
    int64_t c = 3;
    while (c <= 40) {
        result = (result + search_i64_i64(c, 30));
        c = (c + 1);
    }
    printf("%lld\n", result);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("%lld\n\00") {addr_space = 0 : i32} : !llvm.array<6 x i8>
  func.func @search(%arg0: i64, %arg1: i64) -> i64 {
    %0 = arith.constant 1 : i32
    %1 = arith.extsi %0 : i32 to i64
    %2 = llvm.mlir.constant(1 : i64) : i64
    %3 = llvm.alloca %2 x i64 : (i64) -> !llvm.ptr
    llvm.store %1, %3 : i64, !llvm.ptr
    %4 = arith.constant 2 : i32
    %6 = arith.extsi %4 : i32 to i64
    %5 = arith.subi %arg0, %6 : i64
    %7 = llvm.mlir.constant(1 : i64) : i64
    %8 = llvm.alloca %7 x i64 : (i64) -> !llvm.ptr
    llvm.store %arg1, %8 : i64, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %9 = llvm.load %8 : !llvm.ptr -> i64
    %10 = arith.constant 0 : i32
    %12 = arith.extsi %10 : i32 to i64
    %11 = arith.cmpi sgt, %9, %12 : i64
    cf.cond_br %11, ^bb1, ^bb2
    ^bb1:
      %13 = arith.constant 0 : i32
      %14 = arith.extsi %13 : i32 to i64
      %15 = llvm.mlir.constant(1 : i64) : i64
      %16 = llvm.alloca %15 x i64 : (i64) -> !llvm.ptr
      llvm.store %14, %16 : i64, !llvm.ptr
      %17 = llvm.load %3 : !llvm.ptr -> i64
      %18 = arith.cmpi sge, %17, %arg0 : i64
      cf.cond_br %18, ^bb3, ^bb4
      ^bb3:
        %19 = llvm.load %3 : !llvm.ptr -> i64
        %20 = arith.subi %19, %arg0 : i64
        %21 = arith.divsi %20, %5 : i64
        %22 = llvm.mlir.constant(1 : i64) : i64
        %23 = llvm.alloca %22 x i64 : (i64) -> !llvm.ptr
        llvm.store %21, %23 : i64, !llvm.ptr
        %24 = llvm.load %3 : !llvm.ptr -> i64
        %25 = llvm.load %23 : !llvm.ptr -> i64
        %26 = arith.muli %25, %5 : i64
        %27 = arith.subi %24, %26 : i64
        %28 = arith.cmpi sge, %27, %arg0 : i64
        cf.cond_br %28, ^bb6, ^bb7
        ^bb6:
          %29 = llvm.load %23 : !llvm.ptr -> i64
          %30 = arith.constant 1 : i32
          %32 = arith.extsi %30 : i32 to i64
          %31 = arith.addi %29, %32 : i64
          llvm.store %31, %23 : i64, !llvm.ptr
          cf.br ^bb8
        ^bb7:
          cf.br ^bb8
        ^bb8:
        %33 = llvm.load %3 : !llvm.ptr -> i64
        %34 = llvm.load %23 : !llvm.ptr -> i64
        %35 = arith.muli %34, %5 : i64
        %36 = arith.subi %33, %35 : i64
        llvm.store %36, %3 : i64, !llvm.ptr
        %37 = llvm.load %16 : !llvm.ptr -> i64
        %38 = llvm.load %23 : !llvm.ptr -> i64
        %39 = arith.muli %38, %arg0 : i64
        %40 = arith.addi %37, %39 : i64
        llvm.store %40, %16 : i64, !llvm.ptr
        cf.br ^bb5
      ^bb4:
        cf.br ^bb5
      ^bb5:
      %41 = llvm.load %3 : !llvm.ptr -> i64
      %42 = llvm.load %16 : !llvm.ptr -> i64
      %43 = arith.addi %41, %42 : i64
      %44 = arith.constant 1 : i32
      %46 = arith.extsi %44 : i32 to i64
      %45 = arith.addi %43, %46 : i64
      llvm.store %45, %3 : i64, !llvm.ptr
      %47 = llvm.load %8 : !llvm.ptr -> i64
      %48 = arith.constant 1 : i32
      %50 = arith.extsi %48 : i32 to i64
      %49 = arith.subi %47, %50 : i64
      llvm.store %49, %8 : i64, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %51 = llvm.load %3 : !llvm.ptr -> i64
    func.return %51 : i64
  }
  func.func @main() -> i32 {
    %52 = arith.constant 0 : i32
    %53 = arith.extsi %52 : i32 to i64
    %54 = llvm.mlir.constant(1 : i64) : i64
    %55 = llvm.alloca %54 x i64 : (i64) -> !llvm.ptr
    llvm.store %53, %55 : i64, !llvm.ptr
    %56 = arith.constant 3 : i32
    %57 = arith.extsi %56 : i32 to i64
    %58 = llvm.mlir.constant(1 : i64) : i64
    %59 = llvm.alloca %58 x i64 : (i64) -> !llvm.ptr
    llvm.store %57, %59 : i64, !llvm.ptr
    cf.br ^bb9
    ^bb9:
    %60 = llvm.load %59 : !llvm.ptr -> i64
    %61 = arith.constant 40 : i32
    %63 = arith.extsi %61 : i32 to i64
    %62 = arith.cmpi sle, %60, %63 : i64
    cf.cond_br %62, ^bb10, ^bb11
    ^bb10:
      %64 = llvm.load %55 : !llvm.ptr -> i64
      %66 = llvm.load %59 : !llvm.ptr -> i64
      %67 = arith.constant 30 : i32
      %68 = arith.extsi %67 : i32 to i64
      %65 = func.call @search(%66, %68) : (i64, i64) -> i64
      %69 = arith.addi %64, %65 : i64
      llvm.store %69, %55 : i64, !llvm.ptr
      %70 = llvm.load %59 : !llvm.ptr -> i64
      %71 = arith.constant 1 : i32
      %73 = arith.extsi %71 : i32 to i64
      %72 = arith.addi %70, %73 : i64
      llvm.store %72, %59 : i64, !llvm.ptr
      cf.br ^bb9
    ^bb11:
    %74 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %75 = llvm.load %55 : !llvm.ptr -> i64
    %76 = llvm.call @printf(%74, %75) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    %77 = arith.constant 0 : i32
    func.return %77 : i32
  }
}