Problem 114

Ways to fill a row of length 50 with red blocks of length ≥3 and black tiles.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 114
# Ways to fill a row of length 50 with red blocks of length ≥3 and black tiles.

extern {
    function calloc(n: i64, size: i64) -> ptr<void>
    function free(p: ptr<void>) -> void
}

function main() -> i32 {
    let n: i32 = 50
    let m: i32 = 3
    let a: ptr<i64> = calloc((n + 1) as i64, 8)
    if a == null { return 1 }
    a[0] = 1
    let mut i: i32 = 1
    while i <= n {
        # black tile at end
        a[i] = a[i - 1]
        # red block of length k>=m ending here
        let mut k: i32 = m
        while k <= i {
            if k == i {
                a[i] = a[i] + 1
            } else {
                # black before the block
                a[i] = a[i] + a[i - k - 1]
            }
            k = k + 1
        }
        i = i + 1
    }
    printf("%lld\n", a[n])
    free(a)
    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; }

int32_t main(void);



int32_t main(void) {
    int32_t n = 50;
    int32_t m = 3;
    int64_t* a = (int64_t*)(calloc(((int64_t)((n + 1))), 8));
    if (a == NULL) {
        return 1;
    }
    a[0] = 1;
    int32_t i = 1;
    while (i <= n) {
        a[i] = a[(i - 1)];
        int32_t k = m;
        while (k <= i) {
            if (k == i) {
                a[i] = (a[i] + 1);
            } else {
                a[i] = (a[i] + a[((i - k) - 1)]);
            }
            k = (k + 1);
        }
        i = (i + 1);
    }
    printf("%lld\n", a[n]);
    free(a);
    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 private @calloc(i64, i64) -> !llvm.ptr
  func.func private @free(!llvm.ptr) -> ()
  func.func @main() -> i32 {
    %0 = arith.constant 50 : i32
    %1 = arith.constant 3 : i32
    %3 = arith.constant 1 : i32
    %4 = arith.addi %0, %3 : i32
    %5 = arith.extsi %4 : i32 to i64
    %6 = arith.constant 8 : i32
    %7 = arith.extsi %6 : i32 to i64
    %2 = func.call @calloc(%5, %7) : (i64, i64) -> !llvm.ptr
    %8 = llvm.mlir.zero : !llvm.ptr
    %9 = llvm.icmp "eq" %2, %8 : !llvm.ptr
    cf.cond_br %9, ^bb0, ^bb1
    ^bb0:
      %10 = arith.constant 1 : i32
      func.return %10 : i32
    ^bb1:
      cf.br ^bb2
    ^bb2:
    %11 = arith.constant 1 : i32
    %12 = arith.constant 0 : i32
    %13 = arith.extsi %11 : i32 to i64
    %14 = arith.extsi %12 : i32 to i64
    %15 = llvm.getelementptr %2[%14] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %13, %15 : i64, !llvm.ptr
    %16 = arith.constant 1 : i32
    %17 = llvm.mlir.constant(1 : i64) : i64
    %18 = llvm.alloca %17 x i32 : (i64) -> !llvm.ptr
    llvm.store %16, %18 : i32, !llvm.ptr
    cf.br ^bb3
    ^bb3:
    %19 = llvm.load %18 : !llvm.ptr -> i32
    %20 = arith.cmpi sle, %19, %0 : i32
    cf.cond_br %20, ^bb4, ^bb5
    ^bb4:
      %22 = llvm.load %18 : !llvm.ptr -> i32
      %23 = arith.constant 1 : i32
      %24 = arith.subi %22, %23 : i32
      %25 = arith.extsi %24 : i32 to i64
      %26 = llvm.getelementptr %2[%25] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %21 = llvm.load %26 : !llvm.ptr -> i64
      %27 = llvm.load %18 : !llvm.ptr -> i32
      %28 = arith.extsi %27 : i32 to i64
      %29 = llvm.getelementptr %2[%28] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %21, %29 : i64, !llvm.ptr
      %30 = llvm.mlir.constant(1 : i64) : i64
      %31 = llvm.alloca %30 x i32 : (i64) -> !llvm.ptr
      llvm.store %1, %31 : i32, !llvm.ptr
      cf.br ^bb6
      ^bb6:
      %32 = llvm.load %31 : !llvm.ptr -> i32
      %33 = llvm.load %18 : !llvm.ptr -> i32
      %34 = arith.cmpi sle, %32, %33 : i32
      cf.cond_br %34, ^bb7, ^bb8
      ^bb7:
        %35 = llvm.load %31 : !llvm.ptr -> i32
        %36 = llvm.load %18 : !llvm.ptr -> i32
        %37 = arith.cmpi eq, %35, %36 : i32
        cf.cond_br %37, ^bb9, ^bb10
        ^bb9:
          %39 = llvm.load %18 : !llvm.ptr -> i32
          %40 = arith.extsi %39 : i32 to i64
          %41 = llvm.getelementptr %2[%40] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          %38 = llvm.load %41 : !llvm.ptr -> i64
          %42 = arith.constant 1 : i32
          %44 = arith.extsi %42 : i32 to i64
          %43 = arith.addi %38, %44 : i64
          %45 = llvm.load %18 : !llvm.ptr -> i32
          %46 = arith.extsi %45 : i32 to i64
          %47 = llvm.getelementptr %2[%46] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          llvm.store %43, %47 : i64, !llvm.ptr
          cf.br ^bb11
        ^bb10:
          %49 = llvm.load %18 : !llvm.ptr -> i32
          %50 = arith.extsi %49 : i32 to i64
          %51 = llvm.getelementptr %2[%50] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          %48 = llvm.load %51 : !llvm.ptr -> i64
          %53 = llvm.load %18 : !llvm.ptr -> i32
          %54 = llvm.load %31 : !llvm.ptr -> i32
          %55 = arith.subi %53, %54 : i32
          %56 = arith.constant 1 : i32
          %57 = arith.subi %55, %56 : i32
          %58 = arith.extsi %57 : i32 to i64
          %59 = llvm.getelementptr %2[%58] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          %52 = llvm.load %59 : !llvm.ptr -> i64
          %60 = arith.addi %48, %52 : i64
          %61 = llvm.load %18 : !llvm.ptr -> i32
          %62 = arith.extsi %61 : i32 to i64
          %63 = llvm.getelementptr %2[%62] : (!llvm.ptr, i64) -> !llvm.ptr, i64
          llvm.store %60, %63 : i64, !llvm.ptr
          cf.br ^bb11
        ^bb11:
        %64 = llvm.load %31 : !llvm.ptr -> i32
        %65 = arith.constant 1 : i32
        %66 = arith.addi %64, %65 : i32
        llvm.store %66, %31 : i32, !llvm.ptr
        cf.br ^bb6
      ^bb8:
      %67 = llvm.load %18 : !llvm.ptr -> i32
      %68 = arith.constant 1 : i32
      %69 = arith.addi %67, %68 : i32
      llvm.store %69, %18 : i32, !llvm.ptr
      cf.br ^bb3
    ^bb5:
    %70 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %72 = arith.extsi %0 : i32 to i64
    %73 = llvm.getelementptr %2[%72] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %71 = llvm.load %73 : !llvm.ptr -> i64
    %74 = llvm.call @printf(%70, %71) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%2) : (!llvm.ptr) -> ()
    %76 = arith.constant 0 : i32
    func.return %76 : i32
  }
}